Structural and Functional Studies of ATP7B, the Copper(I)-Transporting P-type ATPase Implicated in
Wilson Disease
by
Negah Fatemi
A thesis submitted in conformity with the requirements for the degree of Doctor of Philosophy
Graduate Department of Biochemistry University of Toronto
© Copyright by Negah Fatemi 2011
ii
Structural and Functional Studies of ATP7B, the Copper(I)-Transporting P-type ATPase Implicated in Wilson Disease
Negah Fatemi
Doctor of Philosophy
Graduate Department of Biochemistry University of Toronto
2011
iii
Abstract
Copper is an integral component of key metabolic enzymes. Numerous physiological processes
depend on a fine balance between the biosynthetic incorporation of copper into proteins and the
export of excess copper from the cell. The homeostatic control of copper requires the activity of
the copper transporting ATPases (Cu-ATPases). In Wilson disease the disruption in the function
of the Cu-ATPase ATP7B results in the accumulation of excess copper and a marked deficiency
of copper-dependent enzymes. In this work, the structure of ATP7B has been modeled by
homology using the Ca-ATPase X-ray structure, enabling a mechanism of copper transport by
ATP7B to be proposed. The fourth transmembrane helix (TM4) of Ca-ATPase contains
conserved residues critical to cation binding and is predicted to correspond to TM6 of the
ATP7B homology model, containing the highly conserved CXXCPC motif. The interaction with
Cu(I) and the importance of the 3 cysteines in TM6 of ATP7B has been shown using model
peptides. ATP7B has a large cytoplasmic N-terminus comprised of six copper-binding domains
(WCBD1-6), each capable of binding one Cu(I). Protein-protein interactions between WCBDs
and the copper chaperone Atox1 has been shown, contrary to previous reports, to occur even in
the absence of copper. 15N relaxation measurements on the apo and Cu(I)-bound WCBD4-6
show that there is minimal change in the dynamic properties and the relative orientation of the
domains in the two states. The domain 4-5 linker remains flexible, and domain 5-6 is not a rigid
dimer, with flexibility between the domains. Copper transfer to and between WCBD1-6 likely
occurs via protein interactions facilitated by the flexibility of the domains with respect to each
other. The flexible linkers connecting the domains are important in giving the domains motional
freedom to interact with Atox1, to transfer copper to other domains, and finally to transfer copper
to the transmembrane site for transport across the membrane.
iv
Acknowledgments
This thesis represents not only my work at the bench; it is a milestone in more than one decade of
work at The Hospital for Sick Children. It is with immense gratitude that I acknowledge the
support and help of my supervisors Bibudhendra Sarkar and Julie Forman-Kay. They were my
mentors and friends, and I consider it an honor to have worked with them. Dr. Sarkar has been
supportive since the days I began working for him as a summer student. His devotion to his
research and his work with Scientists Without Borders, along with his zest for life and devotion
to his family, is what inspired me to enter graduate studies. He offered me a research project in
his laboratory and shared his vast knowledge of metal related disorders. And during the most
difficult times when writing this thesis, he gave me the moral support and the freedom I needed
to move forward. It has been truly an honor to be Dr. Sarkar’s last Ph.D. student.
Completion of this thesis would not have been possible had it not been for the phenomenal NMR
expertise of my co supervisor Julie Forman-Kay who gave me the opportunity to learn and apply
NMR spectroscopy to the characterization of the Wilson copper-binding domains. She not only
guided my research, but also supported me emotionally. She was an excellent role model of how
to balance a successful a career as a scientist conducting cutting edge research with being a
mother and having a family. The joy and enthusiasm she has for her research was contagious
and motivated me through the tough times of my Ph.D. work.
There have been many more people who have helped me. I wish to thank my supervisory
committee members David MacLennan and James Rini for their invaluable input over the years
and for reading my thesis. I wish to thank Ranjith Muhandiram for his patience when helping
me at the NMR centre and also for reading my NMR manuscript. Thank you to Lewis Kay for
providing invaluable feedback on my NMR manuscript, (and also for the magnet time!). I would
like to especially thank Dmitry Korzhnev for teaching me how to use his program DASHA for
the model-free interpretation of my relaxation data.
I am indebted to my two Moms at sickkids, Suree Narindrasorasak and Nira Jayanetti who
supported me and have been a source of love and energy ever since. Suree welcomed me to the
lab when I started in 1999 and taught me all about molecular biology, protein purification and
heavy metals (and gardening!). I would also like to acknowledge the guidance and enthusiasm
v
of Wilson disease expert Eve Roberts who I have had the pleasure to work alongside of when she
was on sabbatical in the Sarkar lab. I thank the many Sarkar lab members, Michael DiDonato,
Mike Tsay, Patrick Deschamps and Prasad Kulkarni, who supported me and who were also my
mentors in many ways. I would also like to thank them for the many memorable gastronomic
adventures that enriched my culinary taste buds!
I would like to thank the countless other people in the JFK lab who trained and helped me;
especially Hong Lin who has been an inspiration on how to make 'perfect protein preps', Rhea
Hudson for the molecular biology support, and Andrew Chong who was infinitely patient and
generous with assistance and advice on all fronts (from computer support to how to get your
toddler to go to sleep!). I am grateful for time spent with my office mates, Tanja Mittag, Jennifer
Baker and Voula Kannelis, and my bench mate Irina Bezsonova, for their friendship and for
sharing indispensable scientific, parenting and career advice.
Lastly, I would like to dedicate this thesis to my family for all their love and encouragement. To
my parents who have loved and supported me in all my endeavors, doostetan daram! And most
of all to my loving, supportive, encouraging, and patient husband Anthony who helped me
immensely during my Ph.D.. Thank you!
vi
Table of Contents
Acknowledgments.......................................................................................................................... iv
Table of Contents........................................................................................................................... vi
List of Figures .............................................................................................................................. xiii
List of Tables ............................................................................................................................... xvi
List of Abbreviations .................................................................................................................. xvii
Chapter 1......................................................................................................................................... 1
1 Introduction and Background..................................................................................................... 2
1.1 Copper Metabolism and Homeostasis ................................................................................ 2
1.1.1 Copper uptake and circulation ................................................................................ 2
1.1.2 Normal copper metabolism..................................................................................... 2
1.1.3 Intracellular copper trafficking ............................................................................... 3
1.2 Disorders of Copper Metabolism and Homeostasis............................................................ 5
1.2.1 Menkes disease ....................................................................................................... 5
1.2.2 Wilson disease ........................................................................................................ 5
1.2.2.1 Current treatments .................................................................................... 6
1.2.2.2 Animal models.......................................................................................... 6
1.3 The Copper-Transporting P-Type ATPase ATP7B............................................................ 7
1.3.1 Domain organization and important features.......................................................... 7
1.3.2 N-terminal copper-binding domains ....................................................................... 8
1.3.2.1 Role of the copper-binding domains ........................................................ 9
1.3.2.2 Copper binding properties of the N-terminal copper-binding domain ... 11
1.3.2.3 Zinc binding properties........................................................................... 13
1.3.3 Copper Transfer .................................................................................................... 13
1.3.3.1 Atox1 interaction, and regulation ........................................................... 13
vii
1.3.3.2 Copper transfer between copper-binding domains................................. 15
1.3.4 Nucleotide-binding / Phosphorylation domain .....................................................16
1.3.5 Transduction Channel and Substrate Specificity ..................................................17
1.3.5.1 Functional studies of WCBD by chimeric protein of ATP7B and ZntA........................................................................................................ 18
1.4 Thesis overview ................................................................................................................ 20
Chapter 2....................................................................................................................................... 21
2 Methodology and Optimization ............................................................................................... 22
2.1 Original WCBD4-6 construct ........................................................................................... 22
2.1.1 Limited Proteolysis of WCBD1-6 ........................................................................ 22
2.1.2 Cloning of WCBD4-6 into the GST fusion expression vector ............................. 23
2.1.3 Expression and purification steps ......................................................................... 23
2.1.4 Screening and optimization of reducing conditions.............................................. 25
2.1.5 Screening and optimization of detergents............................................................. 26
2.1.6 Screening and optimization of metal-binding....................................................... 27
2.1.7 Thermal denaturation of WCBD4-6 and WCBD6................................................ 28
2.2 Recloning of WCBD4-6, WCBD5-6 and WCBD6 with new boundaries ........................ 29
2.2.1 Expression and purification of the new protein constructs................................... 29
2.2.2 Optimization of expression and purification of WCBD4-6.................................. 30
2.2.3 Screening and optimization of reducing agents.................................................... 32
2.2.4 Screening and optimization of metal-binding....................................................... 33
2.2.5 Optimization of expression in minimal media...................................................... 33
2.2.6 Presence of 0-100 µM CuCl2 in the M9 media..................................................... 35
2.3 Optimal expression and purification conditions of WCBD4-6.........................................36
2.3.1 WCBD4-6 expressed as a GST fusion protein ..................................................... 36
2.3.2 WCBD4-6 expressed as a His(6)-SUMO fusion protein...................................... 37
viii
2.3.3 Copper content of WCBD4-6 NMR samples ....................................................... 38
2.3.3.1 Apo-WCBD4-6....................................................................................... 38
2.3.3.2 Cu(I)-WCBD4-6..................................................................................... 39
2.3.3.3 EPR spectroscopy for the detection of Cu(II) ........................................ 39
2.3.4 Optimization of NMR conditions ......................................................................... 41
2.3.5 Expression and purification of Atox1................................................................... 41
2.3.6 Production of WCBD1-6 for NMR analysis......................................................... 42
2.3.7 Functional complementation of the copA deficient E. coli strain, ∆LMG194 (copA::K), by ATP7B........................................................................................... 43
2.3.7.1 Cloning of ATP7B and N-terminal deletion mutants of ATP7B into the pBADMycHisC,P Expression Vector .............................................. 44
2.3.7.2 Expression of the WT and N-terminal deletion mutants of ATP7B in Bacterial Host ......................................................................................... 44
2.4 NMR spectroscopy in the structural characterization of proteins..................................... 45
2.4.1 Energy states and population distribution............................................................. 45
2.4.2 Chemical shift ....................................................................................................... 46
2.4.3 The HSQC spectrum............................................................................................. 47
2.4.4 Triple resonance assignment experiments ............................................................ 48
2.4.5 Chemical shift perturbations................................................................................. 50
2.4.6 Relaxation ............................................................................................................. 50
2.4.6.1 T1 and T2 relaxation................................................................................ 50
2.4.6.2 Heteronuclear NOE ................................................................................ 51
2.4.7 Model-free analysis............................................................................................... 51
2.4.7.1 Models of internal motions..................................................................... 52
Chapter 3....................................................................................................................................... 54
3 Insights into the Mechanism of Copper Transport by the Wilson and Menkes Disease Copper-Transporting ATPases................................................................................................. 55
3.1 Summary........................................................................................................................... 55
ix
3.2 Introduction....................................................................................................................... 55
3.3 Homology modelling ........................................................................................................ 56
3.4 Future direction and conclusion........................................................................................ 61
3.5 Probable structural implications for the Wilson and Menkes ATPases based on similarity to other members of the P-type ATPase family ...............................................62
3.6 Structural/functional aspects of copper-ATPases known to date .....................................63
3.7 Proposed mechanism of transport on the basis of other ATPases .................................... 64
Chapter 4....................................................................................................................................... 66
4 Characterization of the N-terminal Copper-Binding Domain of the Rat Copper-Transporting ATPase, ATP7B ................................................................................................. 67
4.1 Summary........................................................................................................................... 67
4.2 Introduction....................................................................................................................... 67
4.3 Materials and Methods...................................................................................................... 69
4.3.1 Construction of atp7b copper-binding region cDNA............................................ 69
4.3.2 Expression and purification .................................................................................. 70
4.3.3 Thermal denaturation analysis .............................................................................. 71
4.3.4 65Zn(II)-competition blotting ................................................................................ 71
4.3.5 Structural analysis by CD spectroscopy ............................................................... 71
4.3.6 Homology modelling of the N-terminal domains................................................. 72
4.4 Results............................................................................................................................... 72
4.4.1 Cloning, expression and purification .................................................................... 72
4.4.2 65Zn(II)-competition blotting ................................................................................ 74
4.4.3 Homology modelling of rCBD ............................................................................. 75
4.4.4 Structural analysis................................................................................................. 77
4.5 Discussion......................................................................................................................... 79
Chapter 5....................................................................................................................................... 84
x
5 Copper(I) Interaction with Model Peptides of WD6 and TM6 Domains of Wilson ATPase: Regulatory and Mechanistic Implications................................................................................ 85
5.1 Summary........................................................................................................................... 85
5.2 Introduction....................................................................................................................... 85
5.3 Materials and Methods...................................................................................................... 88
5.3.1 Homology modelling ............................................................................................ 88
5.3.2 Preparation of peptides ......................................................................................... 88
5.3.3 Assays ................................................................................................................... 89
5.3.3.1 Bicinchoninic acid assay ........................................................................ 89
5.3.3.2 Ellman assay........................................................................................... 89
5.3.3.3 Cu assay.................................................................................................. 89
5.3.4 CD spectroscopy ................................................................................................... 89
5.3.5 NMR spectroscopy................................................................................................ 90
5.4 Results............................................................................................................................... 90
5.4.1 Homology modelling ............................................................................................ 90
5.4.2 CD spectroscopy ................................................................................................... 91
5.4.3 1H NMR spectroscopy .......................................................................................... 99
5.5 Discussion....................................................................................................................... 103
Chapter 6..................................................................................................................................... 106
6 NMR Characterization of Copper-binding Domains 4-6 of ATP7B ..................................... 107
6.1 Summary......................................................................................................................... 107
6.2 Introduction..................................................................................................................... 107
6.3 Materials and Methods.................................................................................................... 110
6.3.1 Cloning, purification and NMR sample preparation........................................... 110
6.3.1.1 WCBD4-6............................................................................................. 110
6.3.1.2 Atox1 .................................................................................................... 111
xi
6.3.1.3 WCBD1-6............................................................................................. 111
6.3.2 Copper loading of protein samples ..................................................................... 112
6.3.3 NMR spectroscopy.............................................................................................. 113
6.3.3.1 Assignment experiments ...................................................................... 113
6.3.3.2 NMR Cu-titrations of WCBD4-6 ......................................................... 114
6.3.3.3 NMR titration of WCBD4-6 with Atox1.............................................. 114
6.3.3.4 NMR relaxation .................................................................................... 114
6.3.3.5 Model-free analysis .............................................................................. 115
6.4 Results............................................................................................................................. 116
6.4.1 Protein structure characterization ....................................................................... 116
6.4.2 Effect of Cu(I) binding on WCBD4-6 ................................................................ 117
6.4.3 The flexible linker maintains its flexibility in both apo and Cu(I)-bound states 120
6.4.4 Copper(I) binding does not change the relative orientation of domains............. 122
6.4.5 Protein-protein interactions with Atox1.............................................................. 126
6.4.6 Copper(I) binding to WCBD1-6 does not appear to induce a significant change in the relative mobility of the domains................................................... 129
6.5 Discussion....................................................................................................................... 129
Chapter 7..................................................................................................................................... 132
7 Summary and Future directions ............................................................................................. 133
7.1 Summary......................................................................................................................... 133
7.2 Future directions ............................................................................................................. 134
7.3 Conclusion ...................................................................................................................... 135
Bibliography ............................................................................................................................... 139
Appendices.................................................................................................................................. 155
A.1 The relaxation equations ................................................................................................ 155
A.2 Brownian rotational diffusion ........................................................................................ 155
xii
A.3 Models of overall rotation.............................................................................................. 156
A.3.1 Isotropic diffusion − diffusion as a sphere........................................................... 157
A.3.2 Anisotropic axially symmetric diffusion − diffusion as a spheroid ..................... 157
A.3.3 Anisotropic fully asymmetric diffusion − diffusion as an ellipsoid..................... 158
A.4 Models of internal motions ............................................................................................ 159
A.5 Statistical analysis, evaluation of model-free parameters and model selection ............. 160
A.6 1H, 15N, and 13C resonance assignments ........................................................................ 161
Copyright Acknowledgements.................................................................................................... 175
xiii
List of Figures
Figure 1.1: Normal copper transport pathways. ........................................................................... 3
Figure 1.2: Domain organization of P-type ATPases ATP7B and SERCA1a. ............................ 8
Figure 1.3: Proposed mechanism of copper transfer between Atox1 and a single copper-binding
domain........................................................................................................................................... 15
Figure 2.1: Limited proteolysis of WCBD1-6 using different proteases and different conditions.
....................................................................................................................................................... 22
Figure 2.2: pGEX-WCBD4-6 Expression Vector. .....................................................................23
Figure 2.3: SDS-PAGE showing the steps for protein purified from the original WCBD4-6
construct. ....................................................................................................................................... 24
Figure 2.4: The effect of 1 M BME on protein aggregation....................................................... 25
Figure 2.5: Anion Exchange FPLC to separate the two protein states in the presence of 1 M
BME.............................................................................................................................................. 26
Figure 2.6: Screening of various detergents to reduce protein aggregation. .............................. 27
Figure 2.7: Gel-shift pattern in response to metal binding ......................................................... 27
Figure 2.8: Thermal denaturation profile.................................................................................... 28
Figure 2.9: Recloning of new constructs of WCBD5, WCBD5-6 and WCBD4-6. ................... 29
Figure 2.10: The purification of the new GST-fusion constructs. .............................................. 30
Figure 2.11: Sequence comparison of the new and old N-terminal domain constructs. ............ 31
Figure 2.12: The purification of the new WCBD4-6 construct. ................................................. 31
Figure 2.13: Effect of reducing agents on aggregates of WCBD4-6.......................................... 32
Figure 2.14: Gel-shift pattern upon metal binding for the new construct of WCBD4-6. ........... 33
xiv
Figure 2.15: (A-D) Protein expression profiles of WCBD4-6 expressed using four different M9
media (Sarkar lab). (E and F) Comparison of the distribution of the protein between soluble and
urea extract fractions in LB media and Maniatis M9 minimal media. ......................................... 35
Figure 2.16: Effect of copper on protein expression. ................................................................. 35
Figure 2.17: Initial 1H-15N HSQC spectrum of WCBD4-6. ....................................................... 37
Figure 2.18: EPR spectra of 1 µM CuHis2 (A) and 1 µM CuSO4 (B) showing the characteristic
signal of Cu(II) centered at 3250 Gauss. Cu-loaded sample of WCBD4-6 does not have any
detectable amount of Cu(II) (C).................................................................................................... 40
Figure 2.19: NMR spectra of WCBD1-6A at 500 MHz............................................................. 42
Figure 2.20: Sequence similarity between ATP7B and CopA. .................................................. 43
Figure 2.21: The construction of the pBADMycHisC,P-ATP7B Expression Vector. ............... 44
Figure 2.22: 1H-15N HSQC spectrum of WCBD4-6................................................................... 47
Figure 2.23: Models of diffusion. ............................................................................................... 52
Figure 3.1: Homology model of ATP7B based on the known structure of SERCA1. ............... 59
Figure 3.2: The proposed catalytic cycle of copper transport by ATP7B based on models
proposed for classical ATPases and functional studies to date..................................................... 60
Figure 4.1: Purification of rCBD ................................................................................................ 73
Figure 4.2: Thermal denaturation of rCBD. ............................................................................... 73
Figure 4.3: Competition of rCBD with 65Zn(II) and transition metals. ...................................... 75
Figure 4.4: Structural alignment of the domains of rCBD. ........................................................76
Figure 4.5: CD spectroscopy of copper-reconstituted rCBD...................................................... 78
Figure 4.6: CD spectra extrapolations of rCBD. ........................................................................ 79
xv
Figure 4.7: 65Zn(II) and Cu(I)-competition experiments. ........................................................... 81
Figure 4.8: Near-UV CD spectra extrapolations as a function of moles of Cu(I) bound. .......... 83
Figure 5.1: The Wilson disease copper transporting ATPase..................................................... 86
Figure 5.2: Peptides used as models for the study of Wilson ATPase. ...................................... 87
Figure 5.3: Homology modelling of model peptides of the Wilson ATPase. ............................ 91
Figure 5.4: Changes in molar ellipticity at 198 nm of model peptides....................................... 93
Figure 5.5: CD spectroscopy of 2K8p and mutant peptides. ...................................................... 95
Figure 5.6: CD spectroscopy of 2K10p model peptide. .............................................................98
Figure 6.1: Secondary structure of apo WCBD4-6................................................................... 117
Figure 6.2: Cu(I) binding to WCBD4-6.................................................................................... 119
Figure 6.3: The effect of Cu(I) binding to WCBD4-6 on the flexible linker. .......................... 121
Figure 6.4: Dynamic properties of the WCBD4-6 residues in the apo and Cu(I)-bound states.
..................................................................................................................................................... 123
Figure 6.5: Rotational diffusion tensors of the apo and Cu(I)-bound domains 4, 5 and 6 in the
apo and Cu(I)-bound states. ........................................................................................................ 124
Figure 6.6: Model-free parameters for the local motion of the backbone amide N-H of domains
4, 5 and 6 in the apo state............................................................................................................ 125
Figure 6.7: Interaction of apo WCBD4-6 with apo and Cu(I) Atox1....................................... 127
Figure 6.8: WCBD1-6 spectra in apo and Cu(I)-bound states.................................................. 128
xvi
List of Tables
Table 2-1 Compositions of minimal media screened for WCBD4-6 protein expression. .......... 34
Table 5-1 1H NMR chemical shifts of 2K8p, 1s, 2s, and 3s that differ when Cu(I) is present in
solution at pH 8.00 and T = 298 K................................................................................................ 99
Table 5-2 Comparison of 1H NMR spectra at different [Cu(I)]:[3s] ratios at 298 K. ............... 102
Table A.6-1 1H, 15N, and 13C resonance assignments of human apoWCBD4-6 at 35 ºC, in 20
mM NaK-phosphate, 130 mM NaCl, 5mM DTT, pH 6.0. ......................................................... 161
xvii
List of Abbreviations
BCA bicinchoninic acid
BCS bathocuproinedisulfonic acid
BME β-mercaptoethanol
BSA bovine serum albumin
CT charge transfer
DMF N,N-dimethyl-formamide
DTT dithiothreitol (2,3-dihydroxy-1,4-dithiobutane)
EDTA ethylene diamine tetra-acetic acid
ESI-MS electrospray ionization mass spectroscopy
Fmoc 9-fluorenyloxycarbonyl-
GSH glutathione
HBTU O-benzotriazolylo-tetramethylo-isourinate
HPLC high performance liquid chromatography
HSQC heteronuclear single quantum correlation
LMCT ligand to metal charge transfer
NOESY nuclear Overhauser effect spectroscopy
PAL–PEG-PS polysterene–polyethylene glycol polymer support
PMSF phenylmethylsulfonyl fluoride
ROESY rotating Overhauser effect spectroscopy
SDS sodium dodecyl sulfate
TCA trichloroacetic acid
TFA trifluoroacetic acid
TGN trans-Golgi network
TIPS triisopropylsilane
TM6 transmembrane segment 6
TOCSY total correlation spectroscopy
TROSY transverse relaxation optimized spectroscopy
Tris Tris(hydroxymethyl) aminomethane
WCBD Wilson copper-binding domains
WD6 Wilson domain 6
1
Chapter 1
Introduction and Background
Much of this chapter is the contents of a review written for Environmental Health Perspectives,
110, Fatemi, N., & Sarkar, B. Molecular mechanism of copper transport in Wilson disease, 695-
698, 2002, with permission from The National Institute of Environmental Health Sciences.
2
1 Introduction and Background
1.1 Copper Metabolism and Homeostasis
1.1.1 Copper uptake and circulation
Copper is an essential trace element required by all living organisms for continued growth and
development (Roberts and Sarkar 2008). It plays a key role in cell metabolism as a cofactor for
numerous enzymes, such as ceruloplasmin, cytochrome c oxidase, superoxide dismutase,
dopamine-β-hydroxylase, tyrosinase, and lysyl oxidase (Cousins 1985). Although required in
trace amounts, excess copper has the ability to generate reactive oxygen species or free radicals
(Seth, Yang et al. 2004), which oxidize cellular components and disrupt cellular metabolism if
present in levels above the required amounts (Halliwell 1994). Therefore it is critical to tightly
regulate cellular copper concentrations. To avoid toxicity, there is a complex and sophisticated
network of copper-trafficking proteins including metallochaperones and copper-transporting
ATPases (Harrison, Jones et al. 2000; O'Halloran and Culotta 2000; Lutsenko, Efremov et al.
2002).
1.1.2 Normal copper metabolism
The pathway of copper from its point of entry in the intestine to its biliary excretion is depicted
in Figure 1.1. Copper first enters the circulation from the intestinal wall through the action of the
Menkes Cu-ATPase, also known as ATP7A (Chelly, Tumer et al. 1993; Mercer, Livingston et al.
1993; Vulpe, Levinson et al. 1993). Once in circulation it forms an exchangeable pool of copper,
the majority of which is complexed with albumin and histidine in the Cu(II) form (Sarkar 1999;
Roberts and Sarkar 2008). During uptake Cu(II) is reduced to Cu(I) by a hypothetical membrane
reductase and is absorbed by the cell through the action of a passive transporter (hCTR) (Zhou
and Gitschier 1997; Moller, Petersen et al. 2000; Lee, Pena et al. 2002; Petris 2004; Turski and
Thiele 2009) whose structure has been solved by electron crystallography to an in plane
resolution of 7Å (De Feo, Aller et al. 2009).
3
Figure 1.1: Normal copper transport pathways.
The location and role of the transporters and ligands in the uptake, circulation and intracellular distribution of copper beginning at the point of absorption from the intestine to its eventual delivery to various copper-dependent proteins.
1.1.3 Intracellular copper trafficking
In the cytoplasm copper may be complexed to a variety of ligands and copper chaperones, which
shuttle copper to various copper-dependent proteins (Singleton and Le Brun 2007). Atox1 (also
known as HAH1) is thought to be the copper chaperone, which delivers copper to ATP7B (Ccc2
in yeast), also known as the Wilson disease protein (Hamza, Schaefer et al. 1999; Larin, Mekios
et al. 1999; Hamza, Prohaska et al. 2003). The discovery of copper binding to protein disulfide
isomerase (PDI) (Narindrasorasak, Yao et al. 2003) and COMMD1 (Narindrasorasak, Kulkarni
et al. 2007) are important in view of the toxicity resulting from copper overload.
Immunohistochemical studies have shown that under steady-state conditions ATP7B is localized
primarily in the trans-Golgi network (TGN) (Hung, Suzuki et al. 1997; Schaefer, Roelofsen et al.
1999) where it transports copper from the cytoplasm to the lumen of the TGN to be incorporated
into the necessary proteins in the secretory pathway such as ceruloplasmin (Terada, Kawarada et
4
al. 1995; Terada, Aiba et al. 1999). Ceruloplasmin has ferroxidase activity and is the most
abundant copper-binding protein in the plasma (Linder 1991). In the human liver, ATP7B has
been localized predominantly in TGN vesicles and also in plasma membrane fractions enriched
in canalicular domains (Dijkstra, van den Berg et al. 1996). It has also been shown that the
ATPase translocates from the TGN to a cytoplasmic vesicular compartment in the vicinity of the
plasma membrane in response to elevated copper concentrations (Veldhuis, Gaeth et al. 2009).
In polarized hepatocytes, under high copper concentrations, ATP7B has been observed to
redistribute to the apical canalicular membrane where it is proposed to function in the direct
excretion of copper into bile (Hung, Suzuki et al. 1997; Schaefer, Roelofsen et al. 1999). This
continuous trafficking between the TGN and the plasma membrane has also been shown for the
Menkes disease protein (Yamaguchi, Heiny et al. 1996). Exactly how the cell manages such
tight regulation of copper and other essential yet potentially toxic trace metals is not fully
understood.
The protein interactions involving the Wilson copper-binding domains (WCBD) are of
considerable importance for understanding Cu-transport and regulation of the protein. The
interaction of WCBD(4-6) and four interacting partners are currently being studied. COMMD1
(copper metabolism gene MURR1 (mouse U2af1-rs1 region1) domain) belongs to a family of
multifunctional proteins that inhibit nuclear factor NF-kappaB. COMMD1 was implicated as a
regulator of copper metabolism by the discovery that a deletion of exon 2 of COMMD1 causes
copper toxicosis in Bedlington terriers (van De Sluis, Rothuizen et al. 2002). Human COMMD1
has been shown to interact with WCBD (Tao, Liu et al. 2003) and also shown by the Sarkar
group to bind Cu(II) in a 1:1 stoichiometry (Narindrasorasak, Kulkarni et al. 2007). The
dynactin subunit p62 (dynactin 4; DCTN4) was identified as an interacting partner in a yeast
two-hybrid screen of human liver cDNA library, and confirmed by co-immunoprecipitation from
mammalian cells. This interaction was shown to require Cu, the CXXC motifs, the region
between domain 4 and domain 6 of ATP7B but not the related ATP7A (Lim, Cater et al. 2006).
Since the dynactin complex is involved in vesicular trafficking it could be involved in the
copper-regulated trafficking of ATP7B. This interaction is a key step in the copper-induced
trafficking of ATP7B to vesicles travelling to the cell membrane and the subsequent removal of
excess copper into bile.
5
1.2 Disorders of Copper Metabolism and Homeostasis
1.2.1 Menkes disease
Menkes disease is an X-linked neurodegenerative disease caused by a global deficiency of
copper in the body as a result of the accumulation of copper in intestinal cells (Kodama 1993).
ATP7A, the Menkes disease gene is located on the X chromosome and encodes a P-type Cu-
ATPase (Bull, Thomas et al. 1993; Chelly, Tumer et al. 1993; Mercer, Livingston et al. 1993;
Tanzi, Petrukhin et al. 1993; Vulpe, Levinson et al. 1993). The Menkes disease protein, ATP7A,
is expressed in all tissues except liver.
1.2.2 Wilson disease
Wilson disease is a hepatic disease, which is often accompanied by neurological symptoms
(Sternlieb 1990) (Loudianos and Gitlin 2000). The Wilson disease gene ATP7B was localized to
the q14.3 band of chromosome 13 and cloned by two independent groups in 1993 (Bull, Thomas
et al. 1993; Tanzi, Petrukhin et al. 1993). The gene consists of 22 exons and encodes a copper-
transporting P-type ATPase (ATP7B) belonging to and sharing many of the features of the
cation-transporting P-type ATPase family (Petrukhin, Lutsenko et al. 1994). The Wilson disease
protein, ATP7B, is expressed mainly in the liver, and to some extent in the kidneys, brain and
placenta (Bull, Thomas et al. 1993; Tanzi, Petrukhin et al. 1993). It has been localized to the
trans-Golgi network (TGN) by immunohistochemical studies (Hung, Suzuki et al. 1997; Nagano,
Nakamura et al. 1998). Such studies have also shown the trafficking of the ATPase from the
TGN to cytoplasmic vesicles in response to an increase in copper concentration (Hung, Suzuki et
al. 1997; Schaefer, Roelofsen et al. 1999); this copper-dependent cycling of ATP7B probably
accounts for the biliary excretion of copper from the liver and correlates well with the Wilson
disease phenotype. Wilson disease is a hereditary disease that was first described by Dr. Kinnear
Wilson in 1912 (Wilson 1912). This autosomal recessive disorder of copper metabolism is
characterized by the toxic accumulation of copper in various tissues such as the liver, kidney,
brain, and placenta due to the lack of biliary excretion of copper from the body (Schilsky 1996),
resulting in the excessive accumulation of copper in the liver (Schaefer and Gitlin 1999).
Elevated urinary copper levels are observed, due to the accumulation of copper in the kidneys,
and impaired incorporation of copper into ceruloplasmin leads to lowered serum copper levels.
6
Increased liver copper concentrations are due to the deficient biliary excretion of copper from the
hepatocytes.
1.2.2.1 Current treatments
Chelation and zinc therapy are two treatments used for Wilson disease (Roberts and Schilsky
2003). Chelators such as D-penicillamine (Walshe 1956; Walshe 1999) and trientine (Walshe
1982; Scheinberg, Jaffe et al. 1987) are used to mobilize copper and facilitate its excretion from
the body through urine. Zinc is used to prevent copper uptake from the intestine into portal
circulation by inducing the synthesis of metallothionein (Hoogenraad 1998). Metallothionein
binds copper with high affinity and is subsequently eliminated in the feces as intestinal cells are
sloughed off (Brewer, Hill et al. 1983; Lipsky and Gollan 1987). These treatments, although
highly effective, are lifelong and can have severe side effects (Sherlock and Dooley 1993).
Currently the treatment of choice for Wilson disease is a combination therapy by trientine and
Zn. Liver transplantation may be the only hope for patients with acute liver failure, which
cannot be reversed with chelation or zinc therapy. Administration of copper-histidine is the only
therapy available for the treatment of Menkes disease (Sarkar, Lingertat-Walsh et al. 1993;
Tumer, Horn et al. 1996; Christodoulou, Danks et al. 1998).
1.2.2.2 Animal models
Many biochemical similarities of Wilson disease have been shown to exist in the Long Evans
Cinnamon (LEC) rat (Wu, Forbes et al. 1994; Terada and Sugiyama 1999). The rat orthologue
of ATP7B shows an amino acid sequence that is 80% similar to its human counterpart (Wu,
Forbes et al. 1994). Furthermore, elevated serum copper concentrations and decreased amounts
of ceruloplasmin establish the LEC rat as a suitable model for the study of Wilson disease. The
disease causing mutation is a deletion of 900 bp in the 3’ terminus of atp7b gene; the gene is
otherwise identical to the wild-type rat gene, even in the region encoding the N-terminus. The
rat orthologue of ATP7B has only five copper-binding motifs in its N-terminus and is missing
the region corresponding to the fourth copper-binding motif of the human ATP7B (Wu, Forbes
et al. 1994).
7
1.3 The Copper-Transporting P-Type ATPase ATP7B
ATP7B, and the closely related ATP7A, are both very similar to other P-type ATPases such as
the sarcoplasmic reticulum (SR) Ca-ATPase, SERCA1a (Toyoshima, Nakasako et al. 2000), and
the H-ATPase of Neurospora crassa (Scarborough 2000), which have been crystallized and their
structures solved. Sequence comparison between the Wilson and Menkes disease proteins
reveals that they have 56% amino acid sequence identity, which is most apparent in the core unit
common to all P-type ATPases (Bull, Thomas et al. 1993; Tanzi, Petrukhin et al. 1993;
Petrukhin, Lutsenko et al. 1994).
1.3.1 Domain organization and important features
The core unit is comprised of six transmembrane helices and two cytoplasmic loops,
encompassing the transduction, actuator, and the nucleotide-binding/phosphorylation domains.
The core elements are found in both the Wilson/Menkes Cu-ATPases and in the sarcoplasmic
reticulum Ca2+-ATPase (Figure 1.2). The distinctive feature of P-type ATPases is the formation
of a covalent phosphoenzyme intermediate in the catalytic cycle. The phosphorylation site is the
aspartic acid in the invariant sequence DKTG, which together with other conserved regions
defines the P-type ATPase family. Heavy metal P-type ATPases have an extra N-terminal pair
of transmembrane helices and an N-terminal cytoplasmic metal-binding domain containing
between one and six metal-binding motifs.
8
Figure 1.2: Domain organization of P-type ATPases ATP7B and SERCA1a.
SERCA1a the Ca2+-ATPase was used as a template to model ATP7B the Wilson disease copper-transporting ATPase. The enzymes contain core features and domains common to other P-type ATPases.
1.3.2 N-terminal copper-binding domains
The copper-transporting ATPases are highly conserved in evolutionarily diverse organisms, from
bacteria to yeast, drosophila, rat, mouse, sheep, human, etc. Although the motifs found in the
core of all heavy metal ATPases are highly conserved, there is great variation and an
evolutionary trend observed in the number of copper-binding domains within the N-terminal
domain of copper-transporting ATPases. This number increases from one or two copper-binding
domains in bacteria (Odermatt, Suter et al. 1993), yeast (Fu, Beeler et al. 1995) and plants
(Tabata, Kashiwagi et al. 1997) to three in C. elegans (Sambongi, Wakabayashi et al. 1997),
four in Drosophila (Adams, Celniker et al. 2000), five in rat (Wu, Forbes et al. 1994), and finally
six in ATP7A and ATP7B of humans (Bull, Thomas et al. 1993; Chelly, Tumer et al. 1993;
Mercer, Livingston et al. 1993; Tanzi, Petrukhin et al. 1993; Vulpe, Levinson et al. 1993). This
evolutionary trend in the increasing number of N-terminal copper-binding motifs may reflect the
need for greater copper buffering capacity or finer regulation of the Cu-ATPases in the more
highly evolved organisms.
9
The N-terminal copper-binding domain (WCBD) of ATP7B contains six repeats of the heavy
metal associated (HMA) domain with a βαββαβ fold. Each domain is approximately 70 amino
acids in length and contains one copy of a GMT/HCXXC motif which is positioned on an
exposed loop on the surface of the domain. This conserved metal-binding sequence motif is
also found in other heavy metal P-type ATPases and metallochaperones from a wide variety of
organisms (Nucifora, Chu et al. 1989; Odermatt, Suter et al. 1993; Phung, Ajlani et al. 1994; Ge,
Hiratsuka et al. 1995; Steele and Opella 1997; Gitschier, Moffat et al. 1998; Rosenzweig,
Huffman et al. 1999; Wimmer, Herrmann et al. 1999; Rensing, Fan et al. 2000; Banci, Bertini et
al. 2001; Banci, Bertini et al. 2002; Jones, Daly et al. 2003).
Despite the similarity in the Wilson and Menkes disease proteins (Bull, Thomas et al. 1993),
there are subtle differences in the N-terminal regions of the highly similar genes; for example,
although the coding regions of both ATP7A and ATP7B are virtually identical after the fifth
copper-binding domain (Tümer, Vural et al. 1995), the two are significantly different at the 5’
end. The coding region for domains 1-4 of the Wilson disease protein is contained in one exon
but in the Menkes disease protein span three exons (Dierick, Ambrosini et al. 1995). Also, in
ATP7B there is a large deletion corresponding to 78 amino acids between the regions coding for
metal-binding repeats 1 and 2.
1.3.2.1 Role of the copper-binding domains
Although the current structural information provides a good model for the transfer of copper
between a copper chaperone and copper-binding domains of a target ATPase, further structural
and biochemical studies are required to characterize the role of the multiple copper-binding
domains found in the N-terminus of these target ATPases. The Wilson and the Menkes copper-
binding N-terminal domains bind six copper atoms with the stoichiometry of one copper per
metal-binding repeat (DiDonato, Narindrasorasak et al. 1997; Lutsenko, Petrukhin et al. 1997),
and this binding occurs through a cooperative mechanism (DiDonato, Narindrasorasak et al.
1997; Jensen, Bonander et al. 1999). The bound copper is in the +1 oxidation state and is
coordinated by two cysteines in a distorted linear geometry (Ralle, Cooper et al. 1998; DiDonato,
Hsu et al. 2000). Circular dichroism (CD) spectroscopy has shown that the N-terminal domains
undergo secondary and tertiary conformational changes in response to copper-binding (Ralle,
Cooper et al. 1998; DiDonato, Hsu et al. 2000). Although it is not known just how the
10
conformational changes observed in the ATP7B N-terminal domain induced by copper-binding
affect the function of the ATPase, several possible roles have been suggested.
It has been suggested that copper-binding to the N-terminal domains may signal the protein to
traffic between the TGN and the plasma membrane in a copper-dependent manner (DiDonato,
Hsu et al. 2000). The Menkes and Wilson ATPases have been localized to the trans-Golgi
network (Petris, Mercer et al. 1996; Yamaguchi, Heiny et al. 1996) and have been shown to
traffic under high copper concentrations (Petris, Mercer et al. 1996; Goodyer, Jones et al. 1999;
Strausak, La Fontaine et al. 1999; Forbes and Cox 2000; Roelofsen, Wolters et al. 2000) to
vesicles which subsequently undergo exocytosis at the plasma membrane (Cater, La Fontaine et
al. 2006). From these studies it has been hypothesized that the metal binding domain not only
serves to bind copper for transport but also as a copper “sensor”. The domain could act as a
copper sensor if the binding of multiple metal atoms was able to induce a conformational change
in the domain. As the concentration of copper rises, the binding of additional metal atoms may
lead to a conformational change in the domain, which may then allow the Wilson disease protein
to be redistributed to pericanalicular vesicles (Cater, La Fontaine et al. 2006), where it could help
excrete excess copper into the bile. It has been suggested that copper-induced trafficking of
ATP7B does not require copper binding to the N-terminal metal-binding motifs or the CPC motif
(Cater, La Fontaine et al. 2007). However, one cannot rule out the involvement of the N-
terminal region in the copper-dependent trafficking of ATP7B, since this report was based on
mutations of copper binding sites in a phosphatase domain mutant background which is
constitutively localized to vesicular compartments.
In addition to its role in trafficking, a regulatory role has also been suggested for the N-terminal
domain. It has been shown that the N-terminus and the N/P domains interact, and that the
copper-bound N-terminal domain induces a conformational change in the N/P domain, altering
the protein-protein interaction (Tsivkovskii, MacArthur et al. 2001). The N-terminal copper-
binding domain has the ability to bind different transition metals with varying affinities
(DiDonato, Narindrasorasak et al. 1997), however, only copper can induce the aforementioned
conformational change in the N-terminus. XAS studies have revealed that zinc binds mostly to
nitrogen atoms in the N-terminus, not to the cysteine residues of the copper-binding motifs, and
the conformational changes induced by zinc-binding are completely different from those
observed for copper-binding (DiDonato, Zhang et al. 2002). These differences may serve as the
11
basis for the mechanism by which the ATPase discriminates between the different metals in vivo.
COMMD1 specifically interacts with the N-terminal region of ATP7B (Tao, Liu et al. 2003), and
although it does not affect the copper-induced trafficking of ATP7B, this interaction exerts a
regulatory effect by decreasing the stability of ATP7B (de Bie, van de Sluis et al. 2007).
Finally, the six domains of ATP7B although structurally similar, are different with respect to the
activity of the ATPase itself and in interactions with Atox1. Deletion mutagenesis studies on the
Wilson disease protein imply that domains 5 and 6, the repeats closest to the transmembrane
channel, cannot be replaced by the 4 most N-terminal repeats and are required for copper
delivery to the intramembrane copper-binding site(s) (Forbes, Hsi et al. 1999). In contrast
domains 1-4 have no effect on the affinity of the intramembrane sites but may regulate the access
of copper to these sites and modulate enzyme turnover (Huster and Lutsenko 2003). Similar
studies on the Menkes disease protein also indicate that its most N-terminal repeats are critical to
its function (Strausak, La Fontaine et al. 1999).
Atox1 was shown to interact in vivo with some, but not all, of the Wilson and Menkes N-
terminal repeats in a copper-dependent manner (Hamza, Schaefer et al. 1999; Larin, Mekios et
al. 1999; Walker, Huster et al. 2004; Achila, Banci et al. 2006; Banci, Bertini et al. 2009); this
suggests that the six domains of these two ATPases may differ in their overall structural
organization and function. Structural homology modelling of the six repeats of the Wilson
disease protein reveals that they have different electrostatic surfaces from one another, and,
therefore, are not all equally capable of interacting with the chaperone (Huffman and O'Halloran
2001). The chaperone may first selectively interact with a particular repeat, which may then
transfer copper to other repeats. The multiple copper bound repeats may interact differently with
chaperones when compared to a single copper-binding domain. Titration of domains 1-6 with
Cu(I)Atox1 (Banci, Bertini et al. 2009) demonstrated that there is a differential interaction with
domain 4 versus domains 5 and 6. A detailed understanding of the structure and function of this
ATPase will provide important insights into the mechanism of intracellular copper transport.
1.3.2.2 Copper binding properties of the N-terminal copper-binding domain
Wilson copper-binding domain, WCBD, is the 70 kDa N-terminal domain of ATP7B
encompassing all six metal-binding motifs. It binds six atoms of copper in the +1 oxidation
12
state. Immobilized metal affinity chromatography has shown that the WCBD is able to bind
different transition metals with varying affinities: Cu(II) > > Zn(II) > Ni(II) > Co(II) (DiDonato,
Narindrasorasak et al. 1997) reflecting the inability of the metal binding sites of the domains to
conform to the preferred ligation geometry of certain metals. The fusion protein could only be
released from the Cu(II) column using the cuprous chelator bathocupreinsulfonate (BCS) and the
formation of the orange (λmax= 480 nm) Cu(I)BCS2 complex. This suggests that not only is the
bound copper in the +1 oxidation state, but that Cu(II) atoms may be reduced to Cu(I) upon
binding to the domain.
Competition 65Zn(II)-blotting analysis (DiDonato, Narindrasorasak et al. 1997) showed that
Cu(I), Cd(II), Au(III), and Hg(II) were able to successfully compete with zinc for binding to the
domain. Copper was the strongest competitor and displayed a distinct cooperative binding
mechanism not observed with the other transition metals. Similar results have also been reported
for the metal binding properties of the MBP-fused metal domains of the Wilson and Menkes
ATPases (Lutsenko, Petrukhin et al. 1997).
Circular dichroism (CD) spectroscopy results show that copper binding induces conformational
changes in the WCBD (DiDonato, Hsu et al. 2000). The secondary structure region (200–270
nm) shows an increase in ellipticity upon binding of increasing amounts of copper, suggesting a
stabilization of secondary structures relative to the apo state and in agreement with the changes
observed in the aromatic region (250–350 nm). The greatest changes in the spectra occur
between the 2:1 and 4:1 copper-bound forms which were very similar in the secondary structure
region but significantly different in the tertiary structure region of the CD spectra. This may
reflect the cooperative copper binding, which was observed in competition 65Zn(II)-blotting
experiments (DiDonato, Narindrasorasak et al. 1997).
X-ray absorption spectroscopy (XAS) studies have provided information regarding the geometry
of the copper binding sites in WCBD (DiDonato, Hsu et al. 2000). The X-ray absorption near
edge structure (XANES) spectra display a characteristic feature of the 1s to 4p transition of Cu(I)
at 8,983 eV, verifying that copper bound to the WCBD is in the +1 state. The shape of the
XANES feature appears to be independent of copper stoichiometry, indicating that all the Cu(I)
sites are quite similar.
13
Extended X-ray absorption fine structure (EXAFS) data show that the first coordination sphere
consists of two sulfur atoms with a Cu–S distance of 2.17–2.19 Å, similar to the Cu–S bond
distance observed in Menkes disease protein and between the distances observed for trigonal and
linear Cu(I)–thiolate complexes (Ralle, Cooper et al. 1998). The intensity of the transition at
8,983 eV, which is indicative of the geometry around the copper atom, suggests that the Cu(I)
site in WCBD has an S-Cu-S angle between 120° and 180°. These observations provide
evidence that the copper atom is coordinated by two cysteines in a distorted linear geometry.
1.3.2.3 Zinc binding properties
Binding of increasing amounts of Zn(II) to WCBD gives rise to an overall loss in secondary
structure content relative to the native structure (DiDonato, Zhang et al. 2002), in contrast to
copper, which shows a progressive increase in secondary structure content as the domain binds
copper. The presence of positive ellipticity at 330 nm is indicative of the presence of disulfide
bonds, which would be expected in the absence of metal (Wingfield and Pain 1996). The
presence of disulfide bonds detectable at 330 nm in the 6:1 complex indicates that the bound
Zn(II) atoms may not be using cysteines as their primary ligands. CD results of the Zn(II)-
reconstituted WCBD support the XAS results and indicate that the ligation environment for
Zn(II) is very different from that for copper. This suggests that the conformational changes
induced by copper preclude the binding of Zn(II), but the converse is not true.
1.3.3 Copper Transfer
1.3.3.1 Atox1 interaction, and regulation
ATP7B, like other eukaryotic copper-ATPases, requires a metallochaperone, to supply it with
copper (Pufahl, Singer et al. 1997; Larin, Mekios et al. 1999; Lockhart and Mercer 2000;
Walker, Tsivkovskii et al. 2002) as the amount of free copper in the cytoplasm is extremely low
(Rae, Schmidt et al. 1999). Metallochaperones are soluble proteins that deliver copper ions to
specific target proteins via direct protein-protein interactions (Field, Luk et al. 2002). Together,
these copper transporters and copper-chaperones form an integrated network designed to regulate
the metabolism of copper such that there is less than one free atom of copper present within the
cell (Rae, Schmidt et al. 1999).
14
Atox1 is the metallochaperone that delivers copper to ATP7B. Several disease-associated
mutations in ATP7B have been shown to interfere with the interaction with Atox1 (Hamza,
Schaefer et al. 1999) suggesting that this interaction is integral to the function of ATP7B. Atox1
has been shown to directly transfer copper to ATP7B, and result in the stimulation of the
catalytic activity of ATP7B. Atox1 is also able to remove copper from ATP7B, and as a result
inhibit ATP7B (Walker, Tsivkovskii et al. 2002). Although the molecular details of this process
are still not fully understood, it is thought that Atox1 is able to regulate the functional activity of
ATP7B in this way.
Atox1 has the same overall fold as the individual copper-binding domains of ATP7B
(Wernimont, Huffman et al. 2000) and binds one Cu(I) (Hung, Casareno et al. 1998; Ralle,
Lutsenko et al. 2003). This, together with complementary charged surfaces has led to the
suggestion that Atox1 transfers copper to its counterparts via docking and ligand exchange
(Arnesano, Banci et al. 2002). This notion is supported by the crystallographic structure of two
Atox1 monomers sharing a single copper (Wernimont, Huffman et al. 2000). The N-terminal
region encompassing domains 1-4 interacts in a copper-dependent manner with Atox1 while the
domains 5-6 region does not, as shown in a yeast two-hybrid assay (Larin, Mekios et al. 1999).
Atox1, implicated as the metallochaperone for ATP7B, delivers copper ions to the WCBD
(Hung, Casareno et al. 1998; Hamza, Schaefer et al. 1999; Larin, Mekios et al. 1999). Atox1
itself has a copper-binding motif (Wernimont, Huffman et al. 2000) and is thought to specifically
interact through complementary electrostatic surfaces with the copper-binding motifs and
exchange copper (Walker, Huster et al. 2004). However, this may not be the only way by which
the ATP7B N-terminal domains obtain copper. Not all the domains found in the N-terminal
region of ATP7B possess the complementary electrostatic patches necessary for interaction with
Atox1 (Huffman and O'Halloran 2001), and the list of other possible copper-binding proteins is
growing. Many of these proteins themselves contain the CXXC motif or are associated in
complexes with proteins that contain the CXXC motif (She, Narindrasorasak et al. 2003; Smith,
She et al. 2004)
Domains 2 and 4 of ATP7B are likely the preferred site for Atox1 interaction and copper
transfer. It has been suggested that copper-binding to domain 2 works as a switch, which allows
access of the chaperone to the other copper binding sites in the N-terminal region (Walker,
15
Huster et al. 2004). Domains 2 and 4 appear to be acceptors of Cu(I) from Atox1 through
specific protein-protein interactions, following which the Cu(I) is somehow routed to domain 6
and then domain 5 prior to being transported across the membrane (Achila, Banci et al. 2006). In
a construct of domain 3-4, Atox1 forms a complex with domain 4 but not with domain 3 (Banci,
Bertini et al. 2008). It is interesting that, in single domain constructs, Atox1 interacts with only
some of the N-terminal copper-binding domains of ATP7B, however, in constructs where 5 of
the 6 CXXC metal binding motifs are mutated to SXXS Atox1 has the ability to deliver and
remove Cu(I) from all copper-binding domains of ATP7B (Yatsunyk and Rosenzweig 2007).
Recent titration work on domains 1-6 (Banci, Bertini et al. 2009) has confirmed that all six
domains can be metallated by Cu(I)Atox1, there is however, a differential interaction with
domain 4 versus domains 5 and 6. Transfer of Cu(I) between domain 1 and domain 4 has also
been observed (Bunce, Achila et al. 2006) supporting the notion of copper transfer amongst the
N-terminal domains.
1.3.3.2 Copper transfer between copper-binding domains
Structural/functional studies as well as the crystal and solution structures of single copper-
binding domains from copper chaperones and transporters have shed some light on the molecular
details of copper exchange between copper chaperones and their target molecules, and provided
support for the direct metal transfer mechanism first proposed by Pufahl et al. (Pufahl, Singer et
al. 1997).
Figure 1.3: Proposed mechanism of copper transfer between Atox1 and a single
copper-binding domain.
Atox1 is shown in blue and the target domain is shown in red. (Adapted from (Wernimont, Huffman et al. 2000)).
Since all Wilson and Menkes copper-binding domains are very similar in both the overall fold
and the metal-binding site to Atox1, the Atox1 dimer can be used as a model for the
16
chaperone/target protein complex. Metal exchange between the chaperone and the target domain
has been predicted to progress through discrete steps described in Figure 1.3 by Wernimont et al.
(Wernimont, Huffman et al. 2000), where the copper ion is initially bound by the chaperone in a
two-coordinate geometry, similar to the Hg(II)Atx1 structure (yeast homolog of Atox1)
(Rosenzweig, Huffman et al. 1999). Next the two domains contact one another primarily at the
metal-binding site. Protein-protein docking is stabilized by electrostatic interactions between a
group of structurally conserved positively charged residues on the surface of the chaperone and
negatively charged residues on the surface of the target domain. Upon docking, the first cysteine
in the GMT/HCXXC motif donates a third ligand to the Cu(I) ion, forming a structure similar to
the Hg(II)Atox1 structure (Rosenzweig, Huffman et al. 1999). The bond between Cu(I) and the
second cysteine of the chaperone then breaks, and a bond to the second cysteine of the target
domain is formed, similar to the Cu(I)Atox1 structure (Wernimont, Huffman et al. 2000) in
which the bond to the second cysteine of the chaperone is slightly longer. The Cu(I) is finally
coordinated by the two cysteine residues of the target domain, similar to the Ag(I)Mnk4 structure
(Gitschier, Moffat et al. 1998), and the chaperone molecule dissociates.
1.3.4 Nucleotide-binding / Phosphorylation domain
Copper binding to the N-terminal copper-binding domains of ATP7B also appear to affect the
nucleotide-binding/phosphorylation (N/P) domain thereby potentially having a regulatory effect
on the ATPase and subsequently on copper homeostasis. Copper binding to WCBD decreases
the interactions between the N-terminal region and the nucleotide-binding/phosphorylation
domain and affects the nucleotide binding properties of the N/P-domain, suggesting a
conformational difference in the N/P domain upon binding either apo- or Cu-bound WCBD.
(Tsivkovskii, MacArthur et al. 2001).
Copper also plays a role in the phosphorylation of WCBD. It has been suggested that binding of
copper to the metal binding domains of ATP7B induces a conformational change, exposing a
specific site for phosphorylation by a kinase (Vanderwerf, Cooper et al. 2001). There have been
several reports of the intracellular handling of copper being modulated by kinase-mediated
phosphorylation (Vanderwerf, Cooper et al. 2001; Cobbold, Ponnambalam et al. 2002;
Vanderwerf and Lutsenko 2002; Voskoboinik, Fernando et al. 2003; Lutsenko, Barnes et al.
2007). The phosphorylation is on serine residues (Vanderwerf and Lutsenko 2002), and is
17
distinct from the acyl-phosphate site at the invariant aspartate in the DKTGT motif. The kinase-
mediated phosphorylation site has been localized to the loop connecting metal binding domains 3
and 4 of ATP7B (Bartee, Ralle et al. 2009). Copper has been shown to be required for the PKA
regulatory phosphorylation of a serine residue in the N-terminal copper-binding domain of
Ccc2p, the yeast homolog of ATP7B (Valverde, Morin et al. 2008). These results suggest that
the removal of copper from the binding sites in the N-terminal or even in the transmembrane
domains (Solioz and Vulpe 1996; Lowe, Vieyra et al. 2004; Lutsenko, Barnes et al. 2007)
promotes a modification of the molecular environment of the phosphorylation site, impairing its
interaction with the kinase. This interaction may be modified by subtle changes in cytosolic
copper sensed by the N-terminal metal binding sites in the yeast Cu(I)-ATPase. The domain
dissociation may therefore expose new sites on either domain, making them available for
phosphorylation. This N-terminal region phosphorylation site is required for generating an
intramolecular signal that is transmitted to the conserved aspartate in the catalytic site and allows
the cycle to continue.
1.3.5 Transduction Channel and Substrate Specificity
Ca-ATPase (Toyoshima, Nakasako et al. 2000), Na,K-ATPase (Morth, Pedersen et al. 2007), and
H-ATPase (Pedersen, Buch-Pedersen et al. 2007) are three P-type ATPases for which a great
deal of structural information is available. In these P-type ATPases, transmembrane domains
M4, M5, and M6 form part of the channel and contain residues critical to cation binding. In
ATP7B, transmembrane domains TM6 and TM7 are predicted to correspond to M4 and M5 of P-
type ATPases and form part of the channel (Sweadner and Donnet 2001). In an experiment that
highlighted the central role of M4, the cation-binding specificity of the Na,K-ATPase was
successfully altered to that of H,K-ATPase by mutating residues within the channel (Mense,
Dunbar et al. 2000). TM6 of ATP7B corresponds to M4 of Ca-ATPase, and both transmembrane
domains contain a conserved proline residue found in all P-type ATPases. In the heavy metal
ATPases, highly conserved cysteine residues flank this proline residue to form a CPC motif.
Mammalian copper-transporting ATPases have an additional conserved cysteine, forming a
CXXCPC motif. Site-directed mutagenesis studies of the cysteine residues in the CPC motif
have revealed it to be essential for the copper transport function of the ATPase (Forbes and Cox
1998; Bissig, Wunderli-Ye et al. 2001). The CPC motif is predicted to be one of the copper-
binding sites in the channel. In addition, four amino acid residues in transmembrane segment 7
18
and 8 of the bacterial Cu-ATPase, CopA, were identified as being required for copper binding.
These residues correspond to Y1331, N1332, M1359, and S1362 in ATP7B (Mandal, Yang et al.
2004)
1.3.5.1 Functional studies of WCBD by chimeric protein of ATP7B and ZntA
There is a high degree of specificity for the metal ions that are transported by P-type ATPases.
Some common structural principles regarding ion selectivity have emerged from the crystal
structures of related P-type ATPases. The high similarity between the Ca2+, Na+ +- and H+,K+-
ATPase ion-binding sites suggests that ion selectivity is ultimately determined by a combination
of charge, coordination number, coordination geometry, and distances to the ligand groups.
These amino acids are often conserved in multiple clusters that form the pathway of the ion
through the transporter from the point of entry as it approaches the binding sites and finally to its
point of exit as it leaves the binding site. To further understand the fundamental mechanics of
substrate selectivity in this essential and diverse family of membrane transporters, it is important
to understand the structures of some of its other members, such as the heavy metal-transporting
ATPases.
ATP7A and ATP7B are copper-transporting ATPases. On the other hand, ZntA, another P-type
ATPase has been shown to be specific for Pb(II), Zn(II), Cd(II), and Hg(II) (Beard, Hashim et al.
1997; Rensing, Mitra et al. 1997; Rensing, Sun et al. 1998; Sharma, Rensing et al. 2000). In
addition to copper WCBD is able to bind six molar equivalents of zinc and undergo
conformational changes that are completely different from those induced by copper binding
(DiDonato, Zhang et al. 2002). Therefore, although the WCBD has the ability to bind several
different metals, the different conformations induced by different metals may allow the
transporter to differentiate between copper and other metals in vivo. The metal ion selectivity
and the role of the N-terminal domains in metal ion recognition by the transporter, was
demonstrated in an ATP7B/ZntA chimeric protein constructed by Hou et al. (Hou,
Narindrasorasak et al. 2001). The two chimeric proteins in which the N-terminal region of ZntA
is replaced with either the entire N-terminal domain of ATP7B or just the sixth metal-binding
motif of ATP7B (Hou, Narindrasorasak et al. 2001) show activity and confer resistance to
substrates of ZntA, but have no resistance or activity toward copper and silver, which are the
substrates of ATP7B. Although the N-terminal domain of ZntA is not essential for its activity or
19
metal selectivity, it is required for full catalytic activity and cannot be replaced by the N-terminal
domain of ATP7B which can also bind zinc (DiDonato, Zhang et al. 2002).
From these results it can be concluded that metal ion specificity is determined by the
transmembrane part of the ATPases and it appears that the amino-terminal domain cannot
override the intrinsic specificity of the core components of the ATPase, although the amino-
terminal domain interacts with other parts of the transporter in a metal ion specific manner. The
N-terminal domain of ATP7B cannot replace that of ZntA in restoring full catalytic activity.
It has been suggested that the full-length N-terminal domain of ATP7B may play a regulatory
role, such that metal ion binding releases an inhibitory interaction of this domain with the
hydrophilic ATPase domain (Tsivkovskii, MacArthur et al. 2001). In this case, one would
expect to observe copper-stimulated ATPase activity for WCBD1-6–ZntA chimera, but this is
not observed. Therefore, at least in the case of the WCBD1-6–ZntA chimera, the amino-terminal
domain does not play a regulatory role (Hou, Narindrasorasak et al. 2001).
While much is known, still many questions remain. One question is whether there is a structural
change in cooperative domain interactions upon Cu binding. We have used NMR to look at this
in chapter 6.
20
1.4 Thesis overview
In Chapter 2 I introduce and describe the methodology and optimization techniques used to
perform the experiments discussed in this thesis. I also discuss expression and purification steps
involved in the preparation of protein samples for NMR experiments. This chapter also provides
some NMR theory.
Chapter 3 describes the homology modelling of ATP7B using the crystal structure of the Ca2+-
ATPase as the template. Based on the similarity to other P-type ATPases, the model is used to
propose a catalytic cycle for the transport of copper that includes a role for the N-terminal
copper-binding domains (Fatemi and Sarkar 2002; Fatemi and Sarkar 2002).
Chapter 4 discusses the rat homologue of WCBD, called rCBD. Homology modelling is used to
reveal the presence of domain 4 which was previously thought to be absent in rCBD. The
presence of this domain helps to explain the similarity of the CD spectroscopy and 65Zn(II)-
blotting experiments results obtained on WCBD and rCBD (Tsay, Fatemi et al. 2004).
Copper binding by model peptides of ATP7B copper-binding domain 6 and transmembrane helix
6 are discussed in Chapter 5. Mutational analysis, CD spectroscopy, NMR spectroscopy and
homology modelling are used to probe the role of various residues in the model peptides. The
findings have been discussed with respect to copper trafficking and intermolecular interactions
(Myari, Hadjiliadis et al. 2004).
In Chapter 6 I discuss the NMR characterization of copper-binding domains 4-6 and 1-6 of
ATP7B. Experimental results pertaining to backbone assignment and dynamic characterization
of apo and copper loaded domains 4-6, as well as interactions with apo and Cu(I)-Atox1 are
discussed in the context of apo and copper loaded spectra of domains 1-6 (Fatemi, Korzhnev et
al. 2010).
Chapter 7 is a summary of the research results and a discussion of future experiments that can be
applied toward the further characterization of N-terminal copper-binding domains of ATP7B.
21
Chapter 2
Methodology and Optimization
22
2 Methodology and Optimization
This chapter describes the methodology and optimization techniques used to perform the
experiments discussed in this thesis, the expression and purification steps involved in the
preparation of protein samples for NMR experiments, and also some NMR theory.
2.1 Original WCBD4-6 construct
2.1.1 Limited Proteolysis of WCBD1-6 Some preliminary proteolytic digestions were performed on WDCu1-6 boundaries (provided by
Michael DiDonato). Figure 2.1 shows the different proteases and the conditions that were
employed in order to identify stable protein fragments suitable for structural analysis (circled in
red).
Figure 2.1: Limited proteolysis of WCBD1-6 using different proteases and different
conditions.
23
2.1.2 Cloning of WCBD4-6 into the GST fusion expression vector
A 5' primer was designed and used with an existing 3' primer (designed by Michael DiDonato) to
amplify the WCBD4-6 region (S342 – Q649) using PCR. This PCR product is 945 base pairs
and encodes a 32.6 kDa protein fragment. The PCR product and the pGEX-6P-2 vector were
both digested with the restriction enzymes BamHI and SalI (Figure 2.2).
Figure 2.2: pGEX-WCBD4-6 Expression Vector.
2.1.3 Expression and purification steps
E. coli BL21 cells were transformed with the expression vector. Clone #1,4,5 containing
WCBD4-6-pGEX-6P-2 was selected for expression following sequence analysis. The cells were
inoculated in LB media + 0.1 mg/mL ampicillin and allowed to grow at 37°C. At mid-log phase
(O.D.600 of 0.8), the cells were induced with IPTG for 4 h. Pre-induction and post-induction
samples indicate significant protein expression. The cells were pelleted by centrifugation and
resuspended in lysis buffer (50 mM Tris-HCl pH 8, 150 mM NaCl, 1 mM EDTA, 5 mM DTT,
10% glycerol, 1 mM PMSF, 4 mg lysozyme, 0.02% Triton X-100). The cell suspension was
incubated at room temperature for 20 min and lysed by freezing and thawing. The lysed cells
were centrifuged and the supernatant was saved as the soluble fraction (Sol). The pellet was
resuspended in solubilizing buffer (50 mM Tris-HCl pH 8, 6 M urea, 1 mM EDTA, 8 mM DTT)
and homogenized. The homogenate was centrifuged and the supernatant saved as the urea
extract (UE). This pellet was resuspended in a small volume of solubilizing buffer and saved as
the insoluble urea pellet (UP). SDS-PAGE indicated that the protein was found both in the
soluble fraction and the urea extract. The urea extract was dialyzed against refolding buffer for 4
h, the buffer changed and dialyzed overnight to remove the urea. The dialyzed fraction was
24
combined with the soluble fraction and applied to the GST affinity column. The column was
first eluted with GSH buffer (10 mM glutathione, 50 mM Tris-HCl) and then with TENDU
buffer (20 mM Tris-HCl pH 8, 1mM EDTA, 130mM NaCl, 5 mM DTT, 6 M urea). The pooled
fractions were refolded in refolding TED buffer (20 mM Tris-HCl pH 8, 1 mM EDTA, 5 mM
DTT) and applied to the Fast Q-Sepharose ion-exchange column equilibrated with the same
buffer. The protein was eluted with TEND buffer with an increasing NaCl gradient (50mM – 500
mM); the elution was monitored by measuring O.D.280 and SDS-PAGE. The Fast Q-Sepharose
column was not an effective step and was omitted in subsequent purifications (Figure 2.3).
Figure 2.3: SDS-PAGE showing the steps for protein purified from the original
WCBD4-6 construct.
The fusion protein was cleaved with 0.5 units of PreScission™ protease (Amersham Pharmacia
Biotech) per mg of fusion protein. Samples were taken at 24 h and 72 h post cleavage. SDS-
PAGE results show that the digestion was complete even after 24hrs. The digested protein
mixture was applied to a GST affinity column to separate the cleaved GST from the WCBD4-6
protein (Figure 2.3). The flow through contained WCBD4-6 and some GST which could not be
separated by repeating the GST affinity column step. Fast Q-Sepharose chromatography using a
NaCl gradient in either a TEND or TENDU buffer, was not an effective purification step. Size
exclusion chromatography using a G-50 gel filtration column (Figure 2.3) was also ineffective in
separating the GST or the high molecular weight impurities from WCBD4-6. Immobilized metal
ion affinity chromatography was performed using Zn(II). A zinc column was prepared by
charging a Chelating Sepharose Fast Flow resin (Amersham-Pharmacia Biotech) with 3 bed
volumes of 50 mM ZnSO4, washed thoroughly with 10 bed volumes of H2O and equilibrated
with 500 mM NaCl, 6M urea, 20 mM potassium phosphate buffer at pH 7. This buffer was also
25
prepared with pH 4, 4.5, 5, 5.5, 6, and 6.5. Bio-Gel P6 desalting column was used to change to
potassium phosphate buffer pH 7. The protein was loaded and the column was washed in a
stepwise manner with decreasing pH phosphate buffer. This column was very effective in
separating residual GST from the WCBD4-6 (Figure 2.3). The GST was in the flow through and
pH 7 wash, while WCBD4-6 eluted at pH 6, and appeared as a single band on SDS-PAGE. The
purified protein was concentrated to 30 mL ~2 mg/mL using an Amicon concentrator under
argon. It was then further concentrated to 3 mL, ~40mg/mL using Macrosep Centrifugal Device
(30K) at which point it still appeared as a single band on SDS-PAGE. The protein was refolded
in refolding buffer 20mM Tris-HCl pH 8, 10% glycerol, and 5mM DTT; it still showed a single
band on SDS-PAGE. The protein was finally dialyzed into argon purged 20mM Tris-HCl pH 8
buffer to maintain the reducing environment. This protein appeared highly aggregated on a
native gel (non-reducing, non-denaturing).
2.1.4 Screening and optimization of reducing conditions
When the protein was loaded onto the native gel with loading buffer containing 1 M BME, there
was only a slight reduction in protein aggregation. Reducing agents, DTT, BME and TCEP at
concentrations of 0.1, 0.25, 0.5, 0.8, 1 M were screened (not shown). Only the overnight
incubation with 1 M BME, resulted in the separation of the aggregates into two major bands on
native gel (Figure 2.4A). Interestingly overnight incubation with 100 mM DTT did not reduce
the aggregates; instead it appeared to enhance the degradation of the protein (not shown).
Separation of proteins on a native gel is based both on size and charge. Therefore the lower band
likely represents a monomer while the higher band is the dimer/multimer.
A) B)
Figure 2.4: The effect of 1 M BME on protein aggregation.
26
The two major bands observed on the native gel could successfully be separated using a Mono Q
anion exchange column, and with 20mM Tris-HCl pH 8, 1M BME, 0-250 mM NaCl gradient.
The lower MW band eluted from the column at 150-170 mM NaCl while the higher MW band
eluted later at ~190 mM NaCl ; the peak eluting at 240 mM is highly aggregated (Figure 2.5A).
Over time the protein aggregated irreversibly, (Figure 2.5B) could not be separated by
chromatography even in the presence of 1 M BME and appeared as a single band on a native gel.
Figure 2.5: Anion Exchange FPLC to separate the two protein states in the presence
of 1 M BME.
A) The two species can be separated into two peaks. B) Over time most of the protein has aggregated from the monomer to the multimeric form.
2.1.5 Screening and optimization of detergents
Three classes of detergents: Non-ionic (Triton X-100, Digitonin), Ionic (DOC, Na+ salt N-
Laurylsarcosine, SDS), and Zwitterionic (CHAS, Zwittergent) +/- 0.1 M BME) detergents were
tested. At 0.1% in final solution, none were able to reduce protein aggregation (Figure 2.6).
27
Figure 2.6: Screening of various detergents to reduce protein aggregation.
2.1.6 Screening and optimization of metal-binding
The effect of Cu(I), Zn(II), Cd(II), Ag(I) and Hg(I) ions, which have different coordination
geometries, on protein aggregation state were also investigated. Protein aggregation was
monitored both on SDS-PAGE (not shown) and native gels shown below in Figure 2.7.
A) B)
Figure 2.7: Gel-shift pattern in response to metal binding
A) WCBD6 shows a distinct gel shift pattern in response to the addition of different metals. B) WCBD4-6 shows a down shift in response to the addition of all metals.
The gel shift pattern observed upon metal binding is different for WCBD6 and WCBD4-6.
WCBD4-6 appears to bind all metals indiscriminately (Figure 2.7B). The native gel in Figure
28
2.7A shows that there is an upward band shift in WCBD6 in response to metal binding, which
may be an indication of a conformational change or dimerization. Cu, Zn, Ag and Cd all
produce the same upward band shift. Metal binding requires the presence of reducing agents to
reduce sulfhydryl groups to bind metal ions. The presence of excess BME in the sample buffer
causes all the upwardly shifted bands to disappear except for Zn and Cd bound forms. This may
be an indication that at high concentrations, BME can compete the metal ions out of its binding
site, except for Zn and Cd, which may be more tightly bound because they both coordinate in a
tetrahedral manner, whereas Ag and Cu coordinate linearly. The native gel shows two bands in
the presence of metals and reducing agents. The nonspecific metal-binding behavior of
WCBD4-6 with respect to different metals may indicate protein misfolding. This notion is
further supported by the results obtained from thermal denaturation studies.
2.1.7 Thermal denaturation of WCBD4-6 and WCBD6
Thermal denaturation of WCBD4-6 and WCBD6 was conducted from 25°C to 95°C and data
points were collected every 5°C (Figure 2.8). The WCBD4-6 is soluble in aqueous solutions but
has a tendency to form a rapidly aggregating and misfolded protein under non-denaturing
conditions, even in the presence of reducing agents. The thermal denaturation of WCBD4-6
supports this conclusion (Figure 2.8A). The CD spectra for thermal denaturation experiments
were collected and analyzed using a Jasco J-720 spectropolarimeter using a 0.1 cm path length
cylindrical CD cell.
A) B)
Figure 2.8: Thermal denaturation profile.
(A) The linear melting curve of WCBD4-6 is indicative of a misfolded protein. (B) The sigmoidal shaped curve of WCBD6 is indicative of a folded protein.
29
2.2 Recloning of WCBD4-6, WCBD5-6 and WCBD6 with new boundaries
New domain 4-6, domain 5-6 and domain 6 constructs were designed based on sequence
homology to other proteins whose structures have already been determined. These new
constructs share many of the expression and purification procedures and show greater than 95%
purity after a single step of GST-affinity chromatography.
2.2.1 Expression and purification of the new protein constructs
PCR products WCBD6, WCBD5-6 and WCBD4-6 (Figure 2.9A) were cloned into the
expression vector pGEX-WCBD(n) (Figure 2.9B) and were transformed into E. coli BL21 cells.
The GST-WCBD(n) fusion proteins were purified as before, with the slight modification that
only GSH buffer (10 mM glutathione, 50 mM Tris-HCl) was used to elute the fusion protein
from the GST affinity column. The cleavage by PreScission™ protease and separation of the
GST moiety was performed according to previously described protocol. Figure 2.10 shows the
purification of the new GST-fusion constructs.
A) B)
Figure 2.9: Recloning of new constructs of WCBD5, WCBD5-6 and WCBD4-6.
(A) PCR products of new constructs. (B) pGEX-WCBD(n) Expression Vector for the cloning of the new WCBD4-6, WCBD5-6 and WCBD6 as GST fusion proteins.
30
Figure 2.10: The purification of the new GST-fusion constructs.
2.2.2 Optimization of expression and purification of WCBD4-6
Despite the use of various chromatographic techniques, purification of WCBD4-6 ultimately
produced a rapidly aggregating and misfolded protein. Hence, the boundaries of the construct
were changed to reflect the boundaries of heavy metal associated (HMA) folds, producing a new
more stable WCBD4-6 suitable for NMR studies. Figure 2.11 depicts the boundaries of the new
and old constructs. The new domain 4-6 construct, is 29 kDa and spans residues T357-A632.
31
Original WCBD4-6 (pI=5.25, MW=32604.08, 308 residues, #342-649)
1 10 20 30 40 50 60
SSHSPGSPPRNQVQGTCSTTLIAIAGMTCASCVHSIEGMISQLEGVQQISVSLAEGTATV LYNPAVISPEELRAAIEDMGFEASVVSESCSTNPLGNHSAGNSMVQTTDGTPTSLQEVAP HTGRLPANHAPDILAKSPQSTRAVAPQKCFLQIKGMTCASCVSNIERNLQKEAGVLSVLV ALMAGKAEIKYDPEVIQPLEIAQFIQDLGFEAAVMEDYAGSDGNIELTITGMTCASCVHN IESKLTRTNGITYASVALATSKALVKFDPEIIGPRDIIKIIEEIGFHASLAQRNPNAHHL DHKMEIKQ
New WCBD4-6 (pI=4.77, MW=29137.28, 276 residues, #357-632)
1 10 20 30 40 50 60
TCSTTLIAIAGMTCASCVHSIEGMISQLEGVQQISVSLAEGTATVLYNPAVISPEELRAA IEDMGFEASVVSESCSTNPLGNHSAGNSMVQTTDGTPTSLQEVAPHTGRLPANHAPDILA KSPQSTRAVAPQKCFLQIKGMTCASCVSNIERNLQKEAGVLSVLVALMAGKAEIKYDPEV IQPLEIAQFIQDLGFEAAVMEDYAGSDGNIELTITGMTCASCVHNIESKLTRTNGITYAS VALATSKALVKFDPEIIGPRDIIKIIEEIGFHASLA
Figure 2.11: Sequence comparison of the new and old N-terminal domain constructs.
The sequences highlighted in yellow indicate the boundaries of a single copper-binding domain based on sequence homology to other metal binding domains. These metal binding domains all have a characteristic heavy metal associated (HMA), ferredoxin-like fold.
A) B) C)
Figure 2.12: The purification of the new WCBD4-6 construct.
BL21 and Codon Plus cells were screened for expression at 30 °C and 20 °C. Neither protein
expression nor solubility increased with Codon Plus cells at 30 °C or at 20 °C (Figure 2.12A and
B). However, proteins expressed at 18-20 °C were typically more stable during purification
(Figure 2.12C).
32
2.2.3 Screening and optimization of reducing agents
Figure 2.13 shows a comparison of the two constructs. Compared to the old construct the new
WCBD4-6 shows less aggregation and requires a lower concentration of reducing agents, 1 mM
DTT compared to up to 1 M BME.
A) B) C) D)
Figure 2.13: Effect of reducing agents on aggregates of WCBD4-6.
The old construct of WCBD4-6 (panels A and C) and new construct of WCBD4-6 (panels B and D) on denaturing gels (panels A and B) and non-denaturing gels (panels C and D).
On a non-reducing denaturing gel, the majority of the old construct is in the reduced monomeric
form, however there are some higher molecular weight complexes observed that are resistant to
SDS (Figure 2.13A). The protein was screened with DTT, BME and TCEP for optimization of
reducing agents, with DTT producing the least aggregation as observed on a non-denaturing
native gel (Figure 2.13D). Based on their molecular weights, these may be aggregated
multimeric forms such as dimers and trimers etc. On a non-reducing native gel the old construct
of WCBD4-6 appears to have a lot of aggregation yet one major band is visible, which is shifted
down upon the addition of excess reducing agent (Figure 2.13C). The new WCBD4-6 construct
requires only minimal reducing agents (1 mM DTT) to resolve into a prominent band (Figure
2.13D). It is likely that the upper band is the oxidized monomeric protein with disulfide bonds
intact, while the lower band represents the reduced monomeric form with reduced sulfhydryl
33
groups. On a reducing denaturing gel all WCBD4-6 is in the reduced monomeric form (Figure
2.13B).
2.2.4 Screening and optimization of metal-binding
The metal-binding characteristics of the new WCBD4-6 differs from the old construct, and
produces a more complete gel shift upon metal binding (Figure 2.14). It is of interest to note that
no gel shift is observed in the presence of copper. This is a surprising outcome since, it is
expected that Cu and Ag should behave similarly because they both have linear coordination
geometries. However, since copper is the biological ligand of this protein, it may possess a
unique binding interaction with the protein.
A) B)
Figure 2.14: Gel-shift pattern upon metal binding for the new construct of WCBD4-6.
The comparison of the old WCBD4-6 construct (A) and new WCBD4-6 construct (B). The new WCBD4-6 seems to discriminate between copper and the other metals in its gel-shift pattern.
2.2.5 Optimization of expression in minimal media
To perform NMR studies, the protein needs to be enriched with either 15N and/or 13C. Therefore
the protein has to be expressed in synthetic media enriched with 15N and 13C containing factors.
Table 2-1 lists the recipes for the minimal media that were screened for WCBD4-6 expression.
Figure 2.15A, B, C, and D shows the expression profile of these media. Subsequently the
standard recipe for M9 salts (#5) and minimal media based on Maniatis (Sambrook, Fritsch et al.
1989) was used for protein expression and purification. Figure 2.15E shows the expression
profile of WCBD4-6 in LB media in comparison to expression in Maniatis M9 minimal media
(Figure 2.15F).
34
Table 2-1 Compositions of minimal media screened for WCBD4-6 protein expression.
M9 Salts #1 #2 #3 #4 #5
Na2HPO4.7H2O 11.32 g 5.66 g 11.32 g 16.98 g 12.8 g
KH2PO4 3.0 g 1.5 g 3.0 g 13.0 g 3.0 g
K2HPO4 --- --- --- 10 g ---
K2SO4 --- --- --- 2.4 g ---
NaCl 0.5 g 0.5 g 0.5 g 0.5 g 0.5 g
1M MgSO4 2.0 mL 2.0 mL 1.0 mL 10.0 mL 1.0 mL
100 mM CaCl2 1.0 mL 1.0 mL 1.0 mL 1.0 mL 1.0 mL
10 mg/mL Thiamin 1.0 mL 1.0 mL 1.0 mL 1.0 mL 1.0 mL
10 mg/mL Biotin 1.0 mL 1.0 mL --- 1.0 mL
30% w/v Glucose 16.66 mL 6.66 mL 10.0 mL 26.66 mL 10.0 mL
15NH4Cl --- 1.0 g 1.0 g 1.1 g 1.0 g
(NH4)2SO4 1.0 g --- --- --- ---
1mM FeCl3.6H2O --- 1.0 mL --- --- ---
35
A) B) C) D) E) F)
Figure 2.15: (A-D) Protein expression profiles of WCBD4-6 expressed using four
different M9 media (Sarkar lab). (E and F) Comparison of the distribution of the protein
between soluble (Sol), urea extract (UE) and urea pellet (UP) fractions in LB media and
Maniatis M9 minimal media.
2.2.6 Presence of 0-100 µM CuCl2 in the M9 media
Figure 2.16: Effect of copper on protein expression.
The effect of increasing amounts of copper in the M9 media was investigated. BL21 cells were
grown in the presence of 0, 25, 50 and 100 µM CuCl2. Although SDS-PAGE showed that there
was protein expression in all concentrations of copper, overall, the addition of copper to the
media adversely affected protein expression (Figure 2.16).
36
2.3 Optimal expression and purification conditions of WCBD4-6
2.3.1 WCBD4-6 expressed as a GST fusion protein
The DNA sequence encoding WCBD4-6 (amino acids 357-632 of ATP7B) was amplified via
PCR and subcloned into a pGEX-6P-2 vector in E. coli BL21 (DE3) cells. Cells were grown at
37 ºC in M9 D2O minimal media, supplemented with 10 mg/L biotin, 10 mg/L thiamin, 0.3%
glucose (13C-glucose was used in samples prepared for backbone assignment experiments), and
0.1% 15NH4Cl. Expression of the fusion protein was induced at OD600 of 0.6-0.8 by the addition
of 0.1 mM IPTG for 18 h at 18 ºC. Cells were harvested at 8000 rpm for 15 min. Cell pellet was
resuspended in 1xPBS binding buffer containing 1 mM EDTA, 5 mM DTT and EDTA-free
Complete protease inhibitor mini-tablets (1 per 100 mL buffer), sonicated (pulsed) for 1 min five
times, and centrifuged for 30 min at 18 K. The residual pellets were resuspended in the same
buffer and sonicated once more. The pooled supernatants were filtered and applied to a GST
affinity column equilibrated with 1xPBS binding buffer. After 1hr incubation in the cold while
mixing, the column was washed extensively with the same buffer. The fusion protein was
digested on column with 40µL PreScission™ protease in 20 mL 1xPBS buffer while shaking at
4˚C overnight. The flow-though was collected and concentrated in denaturing buffer (6 M Gdn-
HCl, 50 mM NaPO4, 150 mM NaCl, 1 mM EDTA and 10 mM DTT). The concentrated digest
was applied to a superdex-200 Hi-load gel-filtration column equilibrated with denaturing buffer.
Fractions corresponding to WCBD4-6 were collected and concentrated in denaturing buffer. The
denatured protein was then refolded by rapid dilution to a final concentration of < 5 µM in argon
purged NMR sample buffer (20 mM Tris-HCl pH 7, 130mM NaCl and 1 mM DTT). Although
the First 15N purification yielded a 384.88 µM sample with 45.26 µM copper content, WCBD4-6
could be concentrated under argon to > 1 mM in the presence of 5 mM DTT without forming
aggregates. All buffers contained 5mM DTT and were purged with argon. All purification steps
were performed in a glovebox under an argon atmosphere. An initial 1H-15N HSQC
(heteronuclear single quantum correlation) spectrum of WCBD4-6 has indicated this protein to
be suitable and promising for further NMR studies (Figure 2.17).
37
Figure 2.17: Initial 1H-15N HSQC spectrum of WCBD4-6.
2.3.2 WCBD4-6 expressed as a His(6)-SUMO fusion protein
Although the new WCBD4-6 construct is much more pure than the old construct and requires
fewer steps to purify, the binding and elution of the WCBD4-6 GST fusion construct to the GST
affinity column was not efficient. This atypical binding suggests an altered interaction between
the GST moiety of the fusion protein and the glutathione-sepharose beads. Therefore WCBD4-6
was recloned into pET-SUMO for easier and quicker purification.
The DNA sequence encoding WCBD4-6 (amino acids 357-632 of ATP7B) was amplified via
PCR and subcloned into a pET SUMO vector in E. coli BL21 (DE3) cells. Cells were grown at
37 ºC in M9 D2O minimal media, supplemented with 10 mg/L biotin, 10 mg/L thiamin, 0.3%
glucose (13C-glucose was used in samples prepared for backbone assignment experiments), and
0.1% 15NH4Cl. Expression of the fusion protein was induced at OD600 of 0.6-0.8 by the addition
of 0.1 mM IPTG for 18 h at 18 ºC. Cells were harvested at 8000 rpm for 15 min. The cell pellet
was resuspended in Ni-column binding buffer A (20 mM Tris-HCl pH 8, 500 mM NaCl, 10 mM
imidazole, 2 mM DTT and EDTA-free Complete protease inhibitor mini-tablets 1 per 100 mL
buffer), sonicated (pulsed) for 1 min five times, and centrifuged for 30 min at 18 K. The residual
38
pellets were resuspended in the same buffer and sonicated once more. The pooled supernatants
were filtered and applied to a Ni2+ affinity column equilibrated with binding buffer A. After 1hr
incubation in the cold while mixing, the column was washed extensively with the same buffer.
The fusion protein was eluted with binding buffer A containing 500 mM imidazole. The eluate
was allowed to digest with SUMO protease (Ulp) in the cold overnight. The digest was
concentrated and applied to a superdex-75 Hi-load gel-filtration column equilibrated with NMR
sample buffer (20 mM NaK-phosphate pH 6, 130 mM NaCl and 5mM DTT). Fractions
corresponding to WCBD4-6 were collected and concentrated in argon-purged NMR sample
buffer.
2.3.3 Copper content of WCBD4-6 NMR samples
2.3.3.1 Apo-WCBD4-6
Protein expressed in minimal media purified from the soluble fraction typically contained 7-10%
Cu(I) and 3-6% Cu(I) when purified under denaturing conditions as measure using the Cu(I)-
assay (Brenner and Harris 1995), where 100% copper content would be equivalent to 3 copper
atoms per protein molecule. Treatment with the Cu(I)-chelator BCS reduces the copper content
of protein purified from the soluble fraction to 2-4 % Cu(I). Briefly, the protein is incubated
under non-denaturing conditions with DTT and excess BCS to produce the apo-protein. BCS is
removed by dialysis in argon-purged buffer A and gel-filtration in NMR sample buffer. The
copper content of the protein sample following this treatment was found to be similar to the
copper content of protein purified under denaturing conditions using 6M Gdn-HCl.
The apo state (0% copper) however, has not been so easy to attain. Even though great care has
been taken to exclude copper from all media and buffers during expression and purification, the
high affinity of the copper binding sites for copper is such that at the end of the purification there
is in the range of 7-10% copper present. This copper is bound in the form of Cu(I) as
demonstrated by EPR spectroscopy and its reaction with the Cu(I) specific chelating agent BCS.
Copper can not be released from its binding site by any means except by treatment with BCS
followed by extensive dialysis. Even after this treatment the copper content still remains at 2-
4%.
39
2.3.3.2 Cu(I)-WCBD4-6
The fully copper bound state of the protein has 3 moles of copper bound per mole of protein; all
of the 3 copper binding sites are bound in this case. The fully copper bound state can be
achieved either by the gradual titration of the protein to a maximally copper bound state, or
through the addition of excess copper followed by extensive dialysis. Copper-loading of
WCBD4-6 with the stoichiometry of 3:1 was achieved according to established protocols
(DiDonato, Hsu et al. 2000) using a 10-fold molar excess CuSO4 with DTT acting as a reductant.
Unbound copper was removed by dialysis in argon-purged buffer A and gel-filtration in NMR
sample buffer B. Alternatively, the apo-protein can be titrated with copper to produce the
copper-loaded protein. DTT concentration of the sample was gradually increased during the
copper titration from 5 mM to the excess concentration of 20 mM. All protein concentrations
were confirmed with amino acid analysis. Copper content of the samples was determined using
the copper-bicinchoninic acid assay (BCA assay) (Brenner and Harris 1995).
2.3.3.3 EPR spectroscopy for the detection of Cu(II)
Electron paramagnetic resonance (EPR) spectroscopy of the WCBD4-6 samples prepared for
NMR spectroscopy confirmed the absence of paramagnetic Cu(II) from these samples. EPR
spectroscopy is used for studying chemical species that have one or more unpaired electrons,
such as free radicals or complexes containing a transition metal ion. EPR is analogous to NMR,
except that in EPR it is electron spins that are excited instead of spins of atomic nuclei. This
technique is highly specific and limited to paramagnetic species, since ordinary solvents do not
give rise to EPR spectra, and was therefore ideal for detecting the presence of Cu(II). Cu(I)
(3d10) is EPR silent, while the single unpaired electron of Cu(II) (3d9) has a very distinctive
signal at 3250 Gauss (Figure 2.18A).
40
Figure 2.18: EPR spectra of 1 µM CuHis2 (A) and 1 µM CuSO4 (B) showing the
characteristic signal of Cu(II) centered at 3250 Gauss. Cu-loaded sample of WCBD4-6
does not have any detectable amount of Cu(II) (C).
41
Solution of 1 µM CuSO4 was used as a Cu(II) control and showed a signal at 3250 Gauss (Figure
2.18B). NMR samples were checked for the absence of Cu(II) by electron paramagnetic
resonance (EPR) spectroscopy to ensure that all copper in the protein sample was in the form of
Cu(I). Cu(I) is EPR silent while Cu(II) has a very distinctive signal at 3250 Gauss (Figure 2.18A
and B). The absence of the 3250 Gauss signal from the WCBD4-6 EPR spectrum (Figure 2.18C)
confirmed that the copper bound in the Cu-loaded samples is Cu(I) and that there are no
detectable levels of Cu(II) in the NMR samples.
2.3.4 Optimization of NMR conditions
1H-15N HSQC spectra were collected on 15N-labelled WCBD4-6 in order to determine optimal
conditions for the apo and copper bound states. Buffer, pH and temperature conditions were
screened to produce spectra with good resolution. HSQC spectra of protein titrated with
increasing amounts of copper showed the greatest resolution when collected at 30 °C in
phosphate buffer at pH 6. Although it is possible to concentrate WCBD4-6 under argon to > 1
mM in the presence of DTT, these samples were unstable and prone to precipitation; therefore
protein concentrations for NMR samples were typically 400-500 µM.
NMR samples were 450 - 500 µM concentration in 450 - 500 µL of NMR sample buffer B
containing 0.5 mM DSS, 0.05% azide, 10 µM benzamidine and 10% (v/v) D2O, and were sealed
in an argon-blanketed NMR tube. NMR samples were checked for the absence of Cu(II) by
electron paramagnetic resonance (EPR) spectroscopy to ensure that all copper in the protein
sample was in the form of Cu(I) (Figure 2.18C). Cu(I) is EPR silent while Cu(II) has a very
distinctive signal at 3250 Gauss (Figure 2.18A).
2.3.5 Expression and purification of Atox1
The pGEX-6P-2 plasmid containing the Atox1 was transformed into E. coli BL21 (DE3) cells,
and the protein was expressed and purified according to previously described protocol
(Narindrasorasak, Zhang et al. 2004) with a few changes. The GST fusion protein was digested
on columns according to a standard protocol with Pre-Scission protease. The flow-though
containing Atox1 was concentrated and, as a final purification step, it was applied to a superdex-
75 Hi-load gel-filtration column equilibrated with NMR sample buffer. Fractions corresponding
to Atox1 were collected and concentrated in argon-purged NMR sample buffer.
42
2.3.6 Production of WCBD1-6 for NMR analysis
Previous constructs beginning at Met1 and ending at Trp650 were not suitable for NMR, due to
problems with concentrating. Protein expressed in Origami cells, which have mutations in the
thioredoxin reductase (trxB) gene and are supposed to greatly enhance the expression of proteins
that require disulfide bond formation for protein folding, did not improve the problems
encountered during concentration.
Two new constructs with new N and C-terminal boundaries were designed. WCBD1-6A begins
at Met1 and ends at Asn635 and WCBD1-6B begins at Gln56 and ends at Asn635. Both
constructs were cloned in the pGEX-6P-2 vector to be expressed as an N-terminal GST fusion
protein. Figure 2.19A shows a TROSY spectrum of WCBD1-6A collected at 500 MHz at 35ºC
in NMR sample buffer (20 mM NaK-phosphate pH 6, 130 mM NaCl and 5mM DTT). Figure
2.19B shows an overlay of HSQC spectra of WCBD4-6 and WCBD1-6, with nearly complete
overlap of the peaks from domains 4-6 in both spectra. This strongly suggests that domains 4-6
have the same structure in both constructs, and that the conclusions drawn from NMR
characterization WCBD4-6 can be extended to WCBD1-6 also.
A) B)
Figure 2.19: NMR spectra of WCBD1-6A at 500 MHz.
(A) 1H-15N TROSY spectrum of WCBD1-6A at 500 MHz. (B) Overlay of 1H-15N HSQC spectra of WCBD1-6A in black and WCBD4-6 in blue at 500 MHz. The peaks from the WCBD4-6 spectrum overlap almost completely with the peaks from the WCBD1-6 spectrum, showing that domains 4-6 are identical in both constructs.
43
2.3.7 Functional complementation of the copA deficient E. coli strain, ∆LMG194 (copA::K), by ATP7B
CopA is a copper-transporting P-type ATPase found in E. coli LMG194, with a high degree of
homology to ATP7B (Figure 2.20). ∆LMG194 CopA::Kanamycin E. coli cells (provided by the
Rosen Lab) are CopA deletion mutant of LMG194 E. coli. Successful complementation of the
copA deficiency by ATP7B would be demonstrated by the cell’s ability to grow in the presence
of excess copper in the growth media. This would indicate the presence of a properly folded and
functional protein suitable for purification, leading to functional studies using everted membrane
vesicles and structural characterization. The rate of copper transport as a function of increasing
copper concentration can be measured using everted membrane vesicles. The rate of copper
transport may also be influenced by the kind of intracellular ligands delivering copper, such as
ATOX1, metallothionein, glutathione, or histidine. Successive mutation of each CXXC motif to
SXXS would also influence the rate of copper transport into the everted membrane vesicles.
Figure 2.20: Sequence similarity between ATP7B and CopA.
ATP7B from humans and CopA from E. coli are both copper transporting P-type ATPases that contain all the conserved sequences found in other P-type ATPases.
44
2.3.7.1 Cloning of ATP7B and N-terminal deletion mutants of ATP7B into the pBADMycHisC,P Expression Vector
Figure 2.21: The construction of the pBADMycHisC,P-ATP7B Expression Vector.
The pBADMycHisC vector suitable for transforming LMG194 E. coli was modified to include
the PreScission™ protease site to provide the option of removing the Myc-His moiety from the
C-terminus of the protein. Fragments I, II, IIIa and IIIb together spanning the entire length of
ATP7B, were ligated with pBADMycHisC,P vector cut with BglII and KpnI (Figure 2.21).
Sequencing and restriction enzyme analysis was used to confirm the correct ligation of fragments
I, II, IIIa and IIIb into the expression vector. Once the cloning of full length ATP7B into the
pBADMycHisC,P vector was achieved, the N-terminal was cut with appropriate enzymes and
ligated with each of the three N-terminal deletion fragments to produce ∆1-4-ATP7B, ∆1-5-
ATP7B and ∆1-6-ATP7B in the pBADMycHisC,P expression vector. The results were
confirmed by restriction enzyme analysis.
2.3.7.2 Expression of the WT and N-terminal deletion mutants of ATP7B in Bacterial Host
Full length ATP7B and three N-terminal copper-binding domain truncation mutants ∆1-4-
ATP7B, ∆1-5-ATP7B and ∆1-6-ATP7B have each been cloned in the pBADMycHisC,P
expression vector and used to transform ∆LMG194 CopA::Kanamycin E. coli cells. CopA in
pBADMycHisC has been used as a control to transform ∆LMG194 CopA::Kanamycin E. coli
cells.
45
Mini-expression screens of full length ATP7B in ∆ LMG194 CopA::Kanamycin with 0.02%,
and 0.2% arabinose showed no obvious induction; a band of 160 kDa would have been expected.
Western-blot with anti-His also did not indicate any expression. The NusA protein with His-tag
expressed from the pET-43.1a(+) vector in BL21 cells was used as the His-tagged control.
2.4 NMR spectroscopy in the structural characterization of proteins
Spectroscopy is based on the energy difference between energy states and the population
distribution of the states. The energy difference between energy states gives rise to the
frequency of the spectral peaks, with intensities that are proportional to and indicative of the
population difference of the states. In NMR spectroscopy there is a phenomenon called
relaxation, which influences both lineshapes and intensities of NMR signals. It provides
information about structure and dynamics of molecules. Therefore, an understanding of these
aspects is required to understand the basic principles of NMR spectroscopy.
2.4.1 Energy states and population distribution
The observable NMR signals come from all the nuclear spins in the presence of the magnetic
field. Nuclei with nonzero spin quantum numbers orient along specific directions with respect to
the magnetic field, and will rotate continuously or precess about the external magnetic field B0
due to the torque generated by the interaction of their nuclear angular momentum with the
magnetic field. For a spin ½ there are two allowed orientations; one pointing up and the other
pointing down with an angle of 54.7◦ relative to the magnetic field. There is energy associated
with each orientation state, or spin state, which is characterized by the intrinsic frequency of the
precession, the Larmor frequency ω0.
ω0 = –γ B0 ,
with γ being the nuclear gyromagnetic ratio, which has a specific value for a given isotope. The
bulk magnetization M0 results from the small population difference between the lower energy
state and the higher energy excited states. At equilibrium the bulk magnetization vector is
stationary along the magnetic field but the individual spin moments rotate about the axis of the
magnetic field. The energy difference between the two states can be described in terms of the
Larmor frequency,
46
∆E = (h/2π) γ B0 ,
where h is Planck’s constant. The ratio of the populations in the states is quantitatively described
by the Boltzmann equation:
Nhigher/Nlower = e – ∆E/kT ,
with T the temperature and k Boltzmann's constant. The spins in the lower energy state can be
excited to the higher energy state when energy corresponding to ∆E is applied. In NMR, this
energy is supplied by means of radiofrequency waves. As the excited population of spins
precess, an oscillating current can be detected. The population difference between spin states is
small, making NMR a low sensitivity technique. Sensitivity can be improved by using a
stronger magnetic field or higher sample temperature and by using uniformly isotopically
labelled samples because the natural abundance of many biologically relevant spin ½ nuclei is
low.
2.4.2 Chemical shift
All nuclear spins of the same isotope have the same intrinsic frequency, the Larmor frequency,
measured in megahertz (MHz). The magnetic field also affects the electrons, producing a small
local magnetic field which shields the nuclei from experiencing the full external magnetic field.
In a molecule each nucleus experiences a slightly different magnetic field that is influenced by
the small local fields produced by neighboring nuclei. The nuclei in a protein sample will all
resonate at slightly different frequencies and cause the dispersion of the signals in NMR spectra
due to shielding (σ) by their different chemical environments, described as:
ω = −γ B0 (1−σ) .
Because of the number of each type of nucleus present in a protein molecule, basic one-
dimensional (1D) spectra are crowded with overlapping signals making it impossible to analyze.
Therefore, multidimensional (2, 3 or 4D) experiments have been devised to deal with this
problem. To facilitate these experiments, it is desirable to isotopically label the protein with 13C
and 15N because the predominant naturally occurring isotope 12C is not NMR-active, and the
nuclear quadrupole moment of the predominant naturally occurring 14N isotope prevents high
resolution information from being obtained from this nitrogen isotope.
47
2.4.3 The HSQC spectrum
A protein contains protons and other magnetically active nuclei. 15N and 13C are of importance
especially in the structure determination of larger proteins (> 100 AA). These hetero nuclei
occur in low abundance and their gyromagnetic ratio is markedly lower than that of protons.
Therefore, two strategies are employed to obtain information from these nuclei: 1) Isotopic
enrichment of these nuclei in proteins and 2) enhancement of the signal to noise ratio by the use
of inverse NMR experiments in which the magnetization is transferred from protons to the hetero
nucleus. The HSQC (heteronuclear single quantum correlation) is the most important inverse
NMR experiment. As the name implies, the 1H-15N HSQC is a correlation of the nitrogen atom
and proton of an NH group, and each signal in a HSQC spectrum is a representation of a proton
that is bound to a nitrogen atom. Figure 2.22, shows a 1H-15N HSQC spectrum of WCBD4-6
recorded at 35ºC.
Figure 2.22: 1H-15N HSQC spectrum of WCBD4-6.
The HSQC or Heteronuclear single quantum correlation spectrum, as the name suggests, is the correlation spectra of the amide nitrogen and the proton directly bonded to it. Therefore each HSQC signal represents a single amino acid. The signals are well dispersed, indicative of a uniformly labelled folded protein. The overlapping signals in the central region may be from disordered regions such as the putative regulatory region.
48
A 2D heteronuclear single quantum correlation (HSQC) spectrum, where "heteronuclear" refers
to nuclei other than 1H, has one peak for each 1H bound to a heteronucleus. The 1H-15N HSQC
spectrum contains the signals of the HN protons in the protein backbone, and since there is only
one backbone HN per amino acid, each HSQC signal represents one single amino acid with the
exception of proline which has no amide-hydrogen due to the cyclic nature of its backbone. The
HSQC also contains signals from the NH2 groups of the side chains of Asn and Gln and of the
aromatic HN protons of Trp and His. There are a total of 16 aromatic residues; 4 Tyr, 6 Phe, and
6 His, in WCBD4-6. The 1H-15N HSQC is often referred to as the fingerprint of a protein
because each protein has a unique pattern of signal positions. The objective of this initial 1H-15N
HSQC spectrum was to asses the suitability of WCBD4-6 for NMR spectroscopy. This spectrum
is promising and shows that the protein is suitable for further NMR analysis. The protein is
uniformly labelled but many of the signals are overlapped, which may be due to the
heterogeneity of this particular sample with regard to disulfide bonds and copper content.
Another issue encountered in the study of proteins that are larger than 20 kDa, is that the
magnetization relaxes faster, which means there is less time to detect the signal. This in turn
causes the peaks to become broader and weaker, and eventually disappear. Two techniques have
been introduced to minimize the effects of relaxation: transverse relaxation optimized
spectroscopy (TROSY) and deuteration (2H labeling) of proteins. We have taken advantage of
both of these techniques in the study of WCBD4-6.
In order to analyze the nuclear magnetic resonance data, it is important to get a resonance
assignment for the protein: that is to find out which chemical shift in each dimension
corresponds to which atom. It is not possible to assign peaks to specific atoms from the
heteronuclear single quantum correlation alone; for this we require specific resonance
assignment experiments.
2.4.4 Triple resonance assignment experiments
These experiments are called 'triple resonance' because three different nuclei (15N, 13C, 1H) are
correlated. The most important advantage of the triple resonance spectra is that they separate
signals into three dimensions, therefore reducing the problem of spectral overlap. However, the
coordinates of clearly separated signals from different amino acids can sometimes be identical
('degenerate'), making it difficult to correctly assign the connectivities between amino acids in
49
the triple resonance spectra. Another advantage of triple resonance spectra is their high
sensitivity, due to the efficient and direct transfer of magnetization via the covalent chemical
bonds. This results in shorter transfer times and decreased losses due to relaxation compared to
homonuclear experiments. The disadvantage of all triple resonance experiments is the often
expensive preparation of doubly-labelled (15N, 13C) proteins. Triple-labelled (15N, 13C, 2H)
WCBD4-6 was produced for various resonance assignment experiments. Large proteins have
longer rotational correlation times (slower tumbling), and faster transverse relaxation, and
typically have broad peaks in their spectra. Increased transverse relaxation rate constants also
decrease overall sensitivity, because less magnetization survives through pulse sequence delays
to be detected by the receiver, making standard triple resonance experiments ineffective for
larger proteins. By deuterating and thus removing most of the protons, the relaxation properties
and the quality of the spectra of large proteins are improved. The NH groups are exchangeable,
and they will back-exchange to 1H when the protein is purified in normal aqueous solution. In
this way, many of the normal NH-based experiments can be carried out on triple-labelled protein.
In backbone assignment experiments, the protein is labelled with 13C and 15N, in order to transfer
magnetisation over the peptide bond, and thus connect different spin systems through bonds.
The experiments consist of a TROSY, where the 1H-15N plane is expanded in the 13C dimension.
There are six experiments which are typically performed in pairs as follows: HNCO /
HN(CA)CO, HNCA /HN(CO)CA, and the HNCACB / HN(CO)CACB (Sattler, Schleucher et al.
1999). For example the HN(CA)CO spectrum contains peaks at the chemical shifts of the
carbonyl carbons from the i and i-1 residues of the primary sequence. The HNCO only contains
the chemical shift of carbonyl carbons from the i-1 residue only. The HNCA and HN(CO)CA
works similarly, just with the alpha carbons rather than the carbonyls, and the HNCACB and the
HN(CO)CACB contains the alpha carbon and the beta carbon respectively. Some amino acid
types, such as Ser, Thr, Gly, and Pro, can be identified through their characteristic 13Cα and 13Cβ
values in these experiments. The (H)CC(CO)NH-TOCSY, simply referred to as the CCC-
TOCSY experiment was used to correlate sidechain carbon resonances with the 15N and 1HN of
the following residue to help in the identification of the residue type (Kanelis, Forman-Kay et al.
2001). All of these experiments were required for the backbone assignment of WCBD4-6.
50
2.4.5 Chemical shift perturbations
The HSQC or TROSY can be used to measure chemical shift perturbations (changes in peak
positions) to detect residues involved in ligand binding and protein-protein interaction. In fast
exchange only one set of peaks is observed throughout the titration, shifting in position as the
ratio of free:bound protein changes. Slow exchange (higher affinity binding) produces distinct
signals for free and bound states at intermediate titration points; by watching bound/free peak
intensities grow or weaken one can follow the binding reaction. Normally binding is observed as
either fast or slow exchange; but there is also intermediate exchange with broadening. Different
peaks in the same spectrum can display both of these regimes of exchange related to differences
in frequency and to rate of exchange (Booth, Keizer et al. 2002).
2.4.6 Relaxation
In addition to structures, NMR can yield information on the dynamics of various parts of the
protein. Measurements of relaxation times and diffusion tensors can provide information on
motions within a molecule, and on motions between protein domains. NMR relaxation occurs as
a result of local fluctuating magnetic fields within a molecule generated by molecular motions.
One of the types of motions that can be detected are those that occur on a fast time-scale (ps-ns).
Since nitrogen atoms are mainly found in the backbone of a protein, 15N relaxation times reflect
the motions of the backbone. This usually involves measuring relaxation times (T1 and T2) to
determine relaxation rates (R1= 1/T1 and R2 = 1/T2) and correlation times.
2.4.6.1 T1 and T2 relaxation
The process called T1 relaxation refers to the mean time for an individual nucleus to return to its
equilibrium state following excitation. T1 (longitudinal or spin-lattice relaxation time) is the time
constant used to describe the rate at which the Mz component of magnetization returns to
equilibrium (the Boltzmann distribution) after perturbation. T2 (transverse or spin-spin
relaxation time) is the time constant used to describe the rate at which the Mxy component of
magnetization returns to equilibrium (completely dephased, no coherence) after perturbation.
This is called T2 relaxation. In practice, the T2 time is the life time of the observed NMR signal,
the free induction decay, and defines the width of the NMR signal. Thus, a nucleus having a
large T2 time gives rise to a sharp signal, whereas nuclei with shorter T2 times give rise to more
broad signals. The T1 and T2 times are dependent on molecular motion.
51
2.4.6.2 Heteronuclear NOE
The steady-state heteronuclear nuclear Overhauser enhancement (NOE) is another parameter that
generates information on the motions of a protein molecule. Since NOE is related to distance
and molecular correlation time, it contains structural information. While traditional NOEs yields 1H-1H distance constraints, 1H-15N heteronuclear NOEs give information on the motion of the N-
H bond vector which is very sensitive to the correlation time of the molecule. Negative NOE
values correspond to flexible regions and positive values are associated with regions of lower
mobility.
2.4.7 Model-free analysis
The model-free analysis of NMR relaxation data, which is widely used for the study of protein
dynamics, consists of the separation of the Brownian rotational diffusion from internal motions
relative to the diffusion frame and the description of these internal motions by amplitude and
timescale. Through parametric restriction and the addition of the Rex parameter a number of
“model-free” models can be constructed. The model-free problem is often solved by initially
estimating the diffusion tensor. The model-free models are then optimized and the best model is
selected. Finally, the global model of all diffusion and model-free parameters is optimized.
Once the values of the relaxation rates and NOE are obtained, the overall rotational correlation
time is initially estimated from the R2/R1 ratio (Mandel, Akke et al. 1995), and used to obtain the
dynamic parameters (squared generalized order parameter S2, and rotational correlation time τR)
via a grid search by minimizing the error function (Mandel, Akke et al. 1995). The program
DASHA (Orekhov, Nolde et al. 1995) which uses the model-free approach of Lipari and Szabo
(Lipari and Szabo 1982) can be used to fit the 15N R1, R2 and heteronuclear NOE data to obtain
model-free parameters (S2, τe, and Rex) according to the extended model-free formalism using
minimization of the error function. S2 is the square of the Lipari and Szabo generalized order
parameter, τe is the effective correlation time, and Rex is conformational exchange rate. The
dynamic parameters S2, τe, and Rex can be plotted for each residue to show the distribution of
backbone and side-chain dynamics information. The order parameter reflects the amplitude of
the motion. A high S2 indicates that the motion of the bond vector is restricted. As S2 decreases,
the motion of the bond vector becomes less restricted and becomes completely isotropic as S2
approaches zero. In addition to S2, the internal correlation time τe characterizes the time scale of
52
that motion and is an indication of how fast the internal motion is. The conformational
exchange term reflects the existence of exchange processes on the micro- to millisecond time
scale.
Figure 2.23: Models of diffusion.
Models of diffusion are: spherical or isotropic diffusion, spheroidal or anisotropic axially symmetric diffusion, and ellipsoidal or anisotropic fully asymmetric diffusion.
2.4.7.1 Models of internal motions
Following the characterization of molecular overall rotation, the parameters of molecular
rotational diffusion tensors may be fixed and used in model-free analysis. 15N R1, R2 and 1H-15N
NOE data at two or more magnetic fields may be fitted using the program DASHA. In DASHA
the models of spectral density function including different number of parameters of internal
motions (shown in brackets) are:
model 1 = { S2 } ,
model 2 = { S2, Rex } ,
model 3 = { S2, τe } ,
model 4 = { S2, τe, Rex } and
model 5 = { Ss2, Sf
2, τs } ,
where S2 and Ss2, Sf
2 (S2 = Ss2, Sf
2) are generalized order parameters, τe, τs are correlation times
for the internal motions, and Rex is the exchange contribution to transverse relaxation due to
conformational exchange on µs-ms time-scales. The order parameters reflect angular amplitudes
of the internal motions of the amide bond vectors, while the correlation times reflect the time-
scale of the internal motions. In the original Lipari-Szabo approach order parameters S2 and
correlation times τe are used to characterize fast picosecond time-scale dynamics (Lipari and
Szabo 1982). The theory was extended to account for internal motions on two time-scales:
53
picosecond with the correlation time τf and order parameter Sf2, and sub-nanosecond with
correlation time τs and order parameter Ss2 (Clore, Szabo et al. 1990). Model selection is
performed based on the values of the χ2 target function obtained in fits of relaxation data; where
more complex models are selected if they lead to significant improvements in the fits (F-test
confidence level).
54
Chapter 3
Insights into the Mechanism of Copper Transport by the Wilson
and Menkes Disease Copper-Transporting ATPases
This chapter is adapted from publications in Inorganica Chimica Acta, 339, Fatemi, N., &
Sarkar, B. Insights into the mechanism of copper transport by the Wilson and Menkes disease
copper-transporting ATPase, 179-187, 2002, with permission from © Elsevier 2002, and Journal
of Bioenergetics and Biomembranes, 34, Fatemi, N., & Sarkar, B. Structural and Functional
Insights of Wilson Disease Copper-Transporting ATPase, 339-349, 2002, with permission from
© Springer Verlag 2002.
55
3 Insights into the Mechanism of Copper Transport by the Wilson and Menkes Disease Copper-Transporting ATPases
3.1 Summary
Wilson disease is an autosomal recessive disorder of copper metabolism. The gene for this
disorder has been cloned and identified to encode a copper-transporting ATPase (ATP7B), a
member of a large family of cation transporters, the P-type ATPases. In addition to the core
elements common to all P-type ATPases, the Wilson copper-transporting ATPase has a large
cytoplasmic N-terminus comprised of six heavy metal associated (HMA) domains, each of
which contains the copper-binding sequence motif GMT/HCXXC. Extensive studies addressing
the functional, regulatory, and structural aspects of heavy metal transport by heavy metal
transporters in general, have offered great insights into copper transport by Wilson copper-
transporting ATPase. The findings from these studies have been used together with homology
modelling of the Wilson disease copper-transporting ATPases based on the X-ray structure of the
sarcoplasmic reticulum (SR) calcium-ATPase, to present a hypothetical model of the mechanism
of copper transport by copper-transporting ATPases.
3.2 Introduction
Wilson disease was first described in 1912 by Dr. Kinnear Wilson as a familial, lethal
neurological disease accompanied by chronic liver disease leading to cirrhosis (Wilson 1912).
The disease is however, principally a disorder of hepatic copper disposition (Sass-Kortsak 1975;
Scheinberg and Sternlieb 1984; Danks 1995; Roberts and Cox 1998). Copper is not incorporated
into ceruloplasmin within the liver, and it is not excreted efficiently into bile. As a result, copper
accumulates in the liver and is eventually released into the blood stream and deposited in other
organs, notably the brain, kidneys, and cornea. This disease occurs worldwide with an average
incidence of one affected individual in 30,000. The Wilson disease gene (ATP7B) was initially
mapped to the q14.3 region of chromosome 13. Using this information as a road map, the gene
abnormal in Wilson disease was identified in 1993 (Bull, Thomas et al. 1993; Petrukhin, Fischer
et al. 1993; Tanzi, Petrukhin et al. 1993; Yamaguchi, Heiny et al. 1993). The Wilson disease
gene has been shown to span at least 80 kb of genomic DNA. Sequence analysis of the cDNA
56
indicates that it encodes a 1411 amino acid P-type ATPase (ATP7B) involved in the transport of
copper. It has an extended intracellular N-terminal segment with six GMT/HCXXC copper
binding motifs, eight membrane-spanning domains, three intracellular loops, and a short C-
terminal region. In this paper we present the current understanding and insights gained on the
structural and functional aspects of Wilson disease copper-transporting ATPase.
3.3 Homology modelling
Knowledge about the three-dimensional (3D) structure of ATP7B is critical to our understanding
of its function. However, the experimental determination of the 3D structure of this and other
such proteins is not always possible due to technical challenges. Models are useful when
experimental methods such as X-ray crystallography or NMR spectroscopy cannot be expected
to determine the structure of a protein in a reasonable time frame.
Homology modelling, often also called comparative protein modelling is presently the most
reliable and useful 3D structure prediction method. This method takes advantage of the solved
structures of closely related (homologous) proteins, to extrapolate the structure of a protein of
unknown structure. One of the most important assumptions made in homology modelling is that
the overall 3D structure of the target protein is similar to that of the related proteins, and that
regions with sequence homology have similar structure. It is assumed that conserved residues
within a family are conserved structurally, and residues involved in biological activity have
similar topology throughout the protein family (Fiser, Feig et al. 2002). Although some protein
domains adopt similar folds despite no significant sequence or functional similarity, it is
generally accepted that high sequence conservation is reflected by distinct structure similarity.
Homology modelling can be used as a powerful tool for the structural modelling of proteins,
particularly when the structures of other homologous proteins are available. In order to begin
modelling the structure of a desired protein one requires, the amino acid sequence of the target
protein, and the high-resolution structure of at least one related protein (template).
The first step in homology modelling is the identification and selection of structures that will
form the template for the target structure (model) (Marti-Renom, Stuart et al. 2000). We used
the 2.6 Å crystal structure of the calcium pump of sarcoplasmic reticulum, SERCA1a
(Toyoshima, Nakasako et al. 2000), as the template for modelling the core P-type ATPase
components of ATP7B (Fatemi and Sarkar 2002). The Na,K-ATPase (Rice, Young et al. 2001),
57
H-ATPase (Scarborough 2000), and L-2 haloacid dehalogenase (Hisano, Hata et al. 1996) belong
to the P-type ATPases family; these structures were used to guide the modelling of ATP7B. The
more recent structures of the Na,K-ATPase (Morth, Pedersen et al. 2007) and the H-ATPase
(Pedersen, Buch-Pedersen et al. 2007) were not available at the time of the development of this
model. Each one of the six copper-binding domains in the N-terminal region of the modeled
structure was based on the solution structure of the fourth metal-binding domain from the
Menkes copper-transporting ATPase (Gitschier, Moffat et al. 1998).
The 11 Å structure of the Na,K-ATPase has a high overall similarity to the E2 structure of the
Ca-ATPase (Rice, Young et al. 2001). Although there are several surface loops that do not align
well, the cytoplasmic domains are observed to have a similar arrangement. The structural
similarity between the Ca-ATPase phosphorylation domain and the catalytic domain of L-2
haloacid dehalogenase (Stokes and Green 2000), notably the conserved positioning of important
catalytic residues, suggests that the E1-to-E2 conformational change proposed for the Ca-
ATPase (Toyoshima, Nakasako et al. 2000) may take place in all P-type ATPases.
The automated alignments obtained from fold assignment are generally not accurate enough, and
need to be optimized by a specific sequence alignment method. Therefore, using gapped-
BLAST we aligned the sequence of ATP7B representing the copper-transporting ATPases with
the SERCA1a sequence (Fatemi and Sarkar 2002). For moderately similar sequences, as is the
case here, the initial alignment is often rough and the boundaries of nonconserved loops are not
very well defined. Therefore, the alignment was further corrected by hand in order to improve
on these weaknesses. The SWISS-MODEL and SWISS-PDB-viewer software programs (Guex
and Peitsch 1997) were used to generate a structural alignment of homologous regions of these
two transporters (Figure 3.1) (Fatemi and Sarkar 2002). Generally, sequences with low
homology to the template protein, less than 30% sequence identity (as is the case with ATP7B
and SERCA1a) are more difficult to model (Saqi, Russell et al. 1998; Rost 1999) but examples
of success do exist; as a rule of thumb, sequence similarity across the whole region is needed.
Because of the low target-template sequence identity, an “align-model-realign-remodel”
approach was used to optimize the model. Once the first model was computed, the alignment
was inspected and improved. Secondary structure prediction and experimental findings were
used in combination with mean force potential energy minimization (Sippl 1990; Sippl 1993) to
guide the model building procedure. This is a statistical method that evaluates the environment
58
of each residue and compares it to what is expected from a representative subset of proteins. The
modified alignment was then subjected to subsequent rounds of modelling and inspection, until
the model could no longer be improved any further.
The multiple domains of ATP7B, were modeled separately. Insertions and deletions were
incorporated into nonconserved regions where they were least likely to disrupt the overall
structure. Loop regions, which cannot be modeled by homology were not modeled. Unlike the
conserved regions, these loops are assumed to be flexible, and were largely disregarded in the
process of modelling the conserved domains and their spatial relation to each other. The end
result was a low resolution 3D model of the ATP7B structure, depicting the rough spatial
organization of the various domains and conserved sequence motifs with respect to each other.
The quality of the model determines the information that can be interpreted from it, thus it is
very important to be aware of the accuracy of the 3D model. A target-template sequence identity
of greater than 30% often produces medium to high accuracy models. A 3D model need not be
absolutely perfect or of high accuracy to be useful; but the accuracy level of the model does
determine its application (Jones, Taylor et al. 1992). Low-accuracy models display less than
30% sequence identity, and are likely to contain substantial modelling errors. However, such
models can still have the correct fold and may be used to predict the approximate biochemical
function of the modeled protein, or appraise the relatedness of the target and template proteins
(Sanchez and Sali 1997; Sanchez and Sali 1998).
59
Figure 3.1: Homology model of ATP7B based on the known structure of SERCA1.
The core of ATP7B encompassing the regions between TM3 and TM7 can be successfully modeled. Actuator domain (A); N-terminal metal-binding domain (M); nucleotide-binding domain (N); phosphorylation domain (P). The important and conserved residues are marked on ATP7B. TM1, TM2 and TM8 of ATP7B do not correspond to any of the transmembrane helices in SERCA1a and were not modeled. The six copper-binding domains in the N-terminal region of ATP7B each adopt a HMA, ferredoxin-like fold found in other metal-binding domains.
The low-resolution model of ATP7B structure generated by homology to the Ca-ATPase
structure (Figure 3.1) (Fatemi and Sarkar 2002), implies the presence of at least four
interconvertible phosphorylated and unphosphorylated conformations in the ATP-dependent
copper transport cycle of Cu-ATPases. Based on what is known about the structure and
mechanism of cation transport by other P-type ATPases, and based on functional studies on
copper and metal-transporting P-type ATPases, the homology model of ATP7B can be used to
propose a general mechanism for the ATP driven transport of copper by Cu-ATPases (Figure
3.2) (Fatemi and Sarkar 2002).
60
Figure 3.2: The proposed catalytic cycle of copper transport by ATP7B based on
models proposed for classical ATPases and functional studies to date.
The three cytosolic domains of P-type ATPases, the phosphorylation (P), the nucleotide-binding
(N) domain, and the actuator (A) domain each play very distinct roles during ion transport. The
catalytic site is comprised of the N and P domains. The copper-binding motifs found in the N-
terminus (M) probably serve as the initial binding sites for the copper ions prior to transport.
The N-terminal copper-binding domain has been shown to specifically bind and interact with the
nucleotide-binding domain of Cu-ATPase ATP7B but not with the nucleotide-binding domain of
the Na,K-ATPase (Tsivkovskii, MacArthur et al. 2001). In the ion-bound conformation the N
and P domains form an open jaw-like formation, and therefore require the closure of the N and P
domains to occur before phosphorylation can take place; domain closure is not dependent on the
occupation of the nucleotide-binding site. Also the A domain that appears to be involved in
transmission of conformational changes is largely dissociated from the main structure in the ion-
bound form. This implies that cation binding in the absence of nucleotide binding weakens the
interaction between the three domains. Copper binding to the N-terminus weakens its interaction
with the N-domain (Tsivkovskii, MacArthur et al. 2001), thereby causing them to dissociate.
The dissociation of the N-terminus from the N-domain restores the nucleotide binding affinity of
the N-domain, thereby allowing the binding of ATP to the nucleotide-binding site. Nucleotide
binding to the cation-bound state induces the association of the N- and P-domains with the A-
domain, closing the gap between the domains and bringing the conserved TGEA/S motif of the
A-domain close to the DKTG motif of the P-domain where phosphorylation takes place,
61
catalyzing hydrolysis of the bound phosphoryl group. These conformational changes are
transmitted to the translocation domain, perturbing the copper-binding site within the channel,
and changing its accessibility to cytosolic and lumenal spaces. In the Cu-ATPases,
transmembrane helices TM6, TM7, and TM8 correspond to M4, M5, and M6 of the Ca-ATPase
(Toyoshima, Nakasako et al. 2000), Na,K-ATPase (Rice, Young et al. 2001) and H-ATPase
(Scarborough 2000)). These transmembrane domains are associated with the transduction
channel and contain residues critical to cation binding. The central role of M4 has been
demonstrated in a clever experiment where, the cation binding specificity of the Na,K-ATPase
was altered to that of the H,K-ATPase by mutating residues within the channel (Mense, Dunbar
et al. 2000). M4 and M5 are predicted to correspond to transmembrane domains TM6 and TM7
of ATP7B, respectively (Sweadner and Donnet 2001), and both TM6 and M4 contain a
conserved proline residue found in all P-type ATPases. In heavy metal-transporting ATPases,
the conserved proline is flanked by a pair of cysteine residues to form the highly conserved CPC
motif. Mutational analysis has shown certain residues within the Cu-ATPases transmembrane
domains to be critical for copper transport function, in particular the CPC motif, which is
predicted to be one of the copper-binding sites within the channel (Forbes and Cox 1998; Bissig,
Wunderli-Ye et al. 2001; Voskoboinik, Greenough et al. 2001). Mammalian copper-transporting
ATPases have an additional conserved cysteine, forming a CXXCPC motif. We have
constructed a peptide corresponding to residues from TM6 of ATP7B and various C/S mutants of
this peptide, in our laboratory to further characterize copper binding to the CPC motif.
Preliminary CD results indicate that the peptide binds a single atom of copper, and that copper
binding induces secondary structure changes in the peptide (Myari, Hadjiliadis et al. 2004). The
hydrolysis of the phosphoenzyme requires the release of cation from the channel to the lumen.
The release of cation may cause the movement of the A domain and allow water to access and
hydrolyze the phosphoenzyme.
3.4 Future direction and conclusion
Although various aspects of copper metabolism have been studied for many years, relatively
little is known about the molecular mechanisms involved in the intracellular transport and
excretion of copper. We have used homology modelling in conjunction with structural and
functional studies to suggest a mechanism of copper transport by copper-transporting ATPases.
It is becoming increasingly evident with the advancements made in the field of structural
62
genomics, that homology modelling will continue to play an important role in 3D structure
prediction as more genomes are sequenced (Sali and Kuriyan 1999; Baker and Sali 2001). These
studies, however, are not likely to correctly predict features that do not occur in the template,
such as the structure of the N-terminal region, unique to heavy metal-transporting P-type
ATPases. Further studies are needed on the mechanism and sequence of binding of copper to the
N-terminal domains of the Wilson and Menkes disease proteins and how they affect their
structure, and the intracellular localization and trafficking of these transporters. The
identification of features within the transduction channel that confer metal ion selectivity is
another exciting area to explore. These questions are only a few of the many aspects of this
unique group of copper-transporting ATPases that are yet to be answered. As we gather more
structural information, kinetic and functional data, it will be possible to use these in conjunction
with homology modelling to construct a more detailed picture of the copper transport process
through the ATPase.
3.5 Probable structural implications for the Wilson and Menkes ATPases based on similarity to other members of the P-type ATPase family
Recently the 11 Å structure of the Na,K-ATPase has revealed that the overall shape is very
similar to the E2 structure of the Ca-ATPase (Rice, Young et al. 2001). Although there are
several surface loops that do not align well, the cytoplasmic domains are observed to have a
similar arrangement. Also, the structural similarity between the P domain to the catalytic domain
of L-2 haloacid dehalogenase (Stokes and Green 2000), notably the conserved positioning of
important catalytic residues, suggests that the E1-to-E2 conformational change proposed for the
Ca-ATPase (Toyoshima, Nakasako et al. 2000) is probably valid for all P-type ATPases.
Using gapped-BLAST we aligned the sequence of ATP7B representing the copper-transporting
ATPases with the SERCA1a sequence. This sequence alignment is used with SWISS-MODEL
and SWISS-PDB-viewer (Guex and Peitsch 1997) to generate a structural alignment of
homologous regions of the transporters (Figure 3.1).
The above observation leads us to predict that the cycle of ATP-dependent copper transport, also
involves at least four interconvertible phosphorylated and unphosphorylated conformations that
have been suggested for other P-type ATPases. Briefly, an increase in cytoplasmic copper
63
concentration could saturate the high-affinity copper-binding sites to form E1ATP·Cu. This
would activate the formation of a phosphoenzyme with ATP, occluding Cu within the protein to
form the high-energy intermediate E1P·Cu. In a rate-limiting step to generate E2P, the
phosphoenzyme would lose both its ability to re-phosphorylate ADP and its high affinity for
copper, and opens its gate to lumenal spaces. Water then enters the catalytic site and hydrolyzes
the phosphorylated aspartic acid residue to regenerate the E2 ATPase.
In a generic model, the three cytosolic domains of P-type ATPases, the P, N and A domains each
play very distinct roles during ion transport. The catalytic site is comprised of the N and P
domains. In the ion-bound conformation the N and P domains form an open jaw like formation,
and therefore requiring the closure of the N and P domains to occur before phosphorylation can
take place; domain closure is not dependent on the occupation of the nucleotide-binding site.
Also the A domain which appears to be involved in transmission of conformational changes, is
largely dissociated from the main structure in the ion-bound form. This implies that cation-
binding in the absence of nucleotide-binding weakens the interaction between the three domains.
Closure of the gap between the A domain and the N and P domains brings the conserved
TGEA/S sequence on the A domain closer to the phosphorylation site. These conformational
changes are conveyed to the translocation domain, and induce the binding of the cation. The
initial binding is then followed by the perturbation of the site, changing its accessibility to
cytosolic and lumenal spaces. The hydrolysis of the phosphoenzyme requires the release of
cation to the lumenal space. The release of cation may result in the movement of the A domain
and allow water to access and hydrolyze the phosphoenzyme.
3.6 Structural/functional aspects of copper-ATPases known to date
During the past decade since the discovery of the Menkes and Wilson disease genes, the studies
of many research groups have revealed much about how these Cu-ATPases work. The unique
metal-binding cytosolic N-terminal domain of these ATPases has been the focus of much
research. The structures of apo and different metal bound forms of single copper-binding
domains from various organisms have been well characterized and this has led to the proposal of
possible steps and intermediates that may be involved in the process of copper transfer from
chaperone to target domain. However, what remains unknown are the protein-protein
64
interactions and overall structure adopted by these copper-binding motifs, which are often found
in multiple copies at the N-terminus of the target ATPases, and how this affects copper transfer
between chaperone and target domains. Further structural and biochemical studies are required
to characterize the structure of the N-terminus of the target ATPases containing multiple copper-
binding domains.
As previously mentioned, in addition to the three conserved domains A, N and P, Cu-ATPases
possess a cytosolic copper-binding domain at their N-terminus. Many functional studies have
targeted the elucidation of the role of this domain in ATP-energized copper transport. The N-
terminal copper-binding domain seems to regulate two aspects of the Cu-ATPase. Although the
N-terminus has been variously shown to be involved in copper transport, it is not critical for
copper transport (Fan, Grass et al. 2001), or for conferring metal ion specificity (Hou,
Narindrasorasak et al. 2001). In Cu-ATPases with multiple copper-binding domains, the ones
closest to the transmembrane region appear to be functionally more important than the ones
closest to the amino terminus (Forbes, Hsi et al. 1999). Therefore, the role of the N-terminal
copper-binding domains may be to increase the overall catalytic rate of the transporter by
increasing the rate of metal ion binding (Mitra and Sharma 2001). The N-terminus also appears
to play a role in the cellular trafficking of the ATPase from the TGN to the plasma membrane in
response to increasing cytoplasmic copper concentrations. It has been shown that the presence
of at least one copper-binding domain close to the membrane channel is necessary for copper-
induced redistribution of the transporter (Forbes, Hsi et al. 1999; Strausak, La Fontaine et al.
1999). Copper-induced conformational changes observed in the N-terminus have been suggested
as a mechanism for this cellular trafficking (DiDonato, Hsu et al. 2000). Protein/protein
interactions or changes in the global conformation of the transporter may render a motif
accessible to the components of the membrane protein sorting machinery.
3.7 Proposed mechanism of transport on the basis of other ATPases
Based on what is known about the structure and mechanism of cation transport by other P-type
ATPases, and based on functional studies on copper and metal-transporting P-type ATPases in
general, we propose a scenario for the ATP driven copper transport by Cu-ATPases. This
catalytic cycle is predicted to involve the following steps and intermediates:
65
The copper-binding motifs found in the N-terminus probably serve as the initial binding sites for
the copper ions prior to transport. The N-terminal copper-binding domain has been shown to
specifically bind to and interact with the N domain of Cu-ATPase ATP7B, but not with the N
domain of the Na,K-ATPase (Tsivkovskii, MacArthur et al. 2001). Copper binding to the N-
terminus weakens the interaction between the two domains (Tsivkovskii, MacArthur et al. 2001),
thereby causing them to dissociate, but not before a copper ion is donated to the high affinity site
within the transmembrane channel. In the Cu-ATPases, the transduction domain is predicted to
consist of transmembrane helices TM6, TM7 and TM8 (corresponding to M4, M5 and M6 of
SERCA1a). Mutational analysis has shown certain residues within these transmembrane
domains to be critical for copper transport function, in particular the CPC motif (Forbes and Cox
1998; Bissig, Wunderli-Ye et al. 2001; Voskoboinik, Greenough et al. 2001). Preliminary CD
spectral studies of TM6 in this laboratory indicate that the conserved and crucial CPC site of
TM6 does bind a single copper atom and the flanking residues undergo secondary structural
changes upon copper-binding (unpublished observations).
The dissociation of the N-terminus from the N domain restores the nucleotide-binding affinity of
the N domain, thereby allowing the binding of ATP to the nucleotide-binding site. Nucleotide-
binding to the cation-bound state induces the association of the N and P domains with the A
domain, closing the gap between the domains and bringing the conserved TGEA/S motif of the
A domain close to the DKTG motif of the P domain where phosphorylation and hydrolysis of the
phosphoryl group takes place. These conformational changes are transmitted to the translocation
domain, perturbing the copper-binding site, changing its accessibility to cytosolic and lumenal
spaces. The release of cation may result in the movement of the A domain and allow water to
access and hydrolyze the phosphoenzyme.
66
Chapter 4
Characterization of the N-terminal Copper-Binding Domain of the
Rat Copper-Transporting ATPase, ATP7B
This chapter is from Biochimica et Biophysica Acta, 1688, Tsay, M.J., Fatemi, N.,
Narindrasorasak, S., Forbes, J.R., & Sarkar, B. Identification of the “missing domain” of the rat
copper-transporting ATPase, ATP7B: insight into the structural and metal binding characteristics
of its N-terminal copper-binding domain, 78-85, 2004, with permission from © Elsevier 2004.
This work has been supported by the Canadian Institute of Health Research (Grant MOP-1800).
Sample preparations, thermal denaturation, 65Zn(II)-competition blotting and CD spectroscopy
experiments were performed by Mike Tsay; corresponding portions in the Material and Methods
section were written by Mike Tsay. I wrote all other sections of this paper and performed the
homology modelling.
67
4 Characterization of the N-terminal Copper-Binding Domain of the Rat Copper-Transporting ATPase, ATP7B
4.1 Summary
Wilson disease is an autosomal disorder of copper transport caused by mutations in the ATP7B
gene encoding a copper-transporting P-type ATPase. The Long Evans Cinnamon (LEC) rat is an
established animal model for Wilson disease. We have used structural homology modelling of
the N-terminal copper-binding region of the rat ATP7B protein (rCBD) to reveal the presence of
a domain, the fourth domain (rD4), which was previously thought to be missing from rCBD.
Although the CXXC motif is absent from rD4, the overall fold is preserved. Using a wide range
of techniques, rCBD is shown to undergo metal-induced secondary and tertiary structural
changes similar to WCBD. Competition 65Zn(II)-blot experiments with rCBD demonstrate a
binding cooperativity unique to Cu(I). Far-UV circular dichroism (CD) spectra suggest
significant secondary structural transformation occurring when 2–3 molar equivalents of Cu(I) is
added. Near-UV CD spectra, which indicate tertiary structural transformation, show a
proportional decrease in rCBD disulfide bonds upon the incremental addition of Cu(I), and a
maximum 5:1 Cu(I) to protein ratio. The similarity of these results to those obtained for the
Wilson disease N-terminal copper-binding region (WCBD), which has six copper-binding
domains, suggests that the metal-dependent conformational changes observed in both proteins
may be largely determined by the protein–protein interactions taking place between the heavy
metal-associated (HMA) domains, and remain largely unaffected by the absence of one of the six
CXXC copper-binding sites.
4.2 Introduction
Copper is an essential trace element necessary for the proper functioning of many biological
processes (DiDonato and Sarkar 1997). However, an excess or deficiency of copper in the body
can lead to a diseased state, thus necessitating careful regulation, transport and storage of this
metal (Danks 1995). Wilson disease is an autosomal recessive disorder in which copper
homeostasis is severely affected and is characterized by the inability of hepatocytes to
incorporate copper into ceruloplasmin or to effectively efflux copper from the liver (Cox 1999).
These features of Wilson disease are caused by mutations in ATP7B, a gene encoding a
68
transmembrane P-type ATPase (ATP7B) found in the liver (Bull, Thomas et al. 1993; Tanzi,
Petrukhin et al. 1993). In Wilson disease, dysfunctional ATP7B compromises biliary copper
clearance, causing copper to accumulate in the liver. Eventually, an overload of metal leads to
hepatic damage and the release of non-ceruloplasmin bound copper, which ultimately
accumulates in peripheral organs such as the brain, cornea and kidney (Danks 1995).
Many biochemical similarities of Wilson disease have been shown to exist in the Long Evans
Cinnamon (LEC) rat (Wu, Forbes et al. 1994; Terada and Sugiyama 1999). The rat orthologue
of ATP7B shows an amino acid sequence that is 80% similar to its human counterpart (Wu,
Forbes et al. 1994). Furthermore, elevated serum copper concentrations and decreased amounts
of ceruloplasmin establish the LEC rat as a suitable model for the study of Wilson disease. The
disease causing mutation is a deletion of 900 bp in the 3’ terminus of the Atp7b gene; the gene is
otherwise identical to the wild-type rat gene, even in the region encoding the N-terminus.
Over recent years, the N-termini of ATP7B and ATP7A have received considerable attention,
with many groups hypothesizing that the function of this domain is vital for copper transport as
well as a key component of translocation processes (Forbes, Hsi et al. 1999; Goodyer, Jones et
al. 1999; Strausak, La Fontaine et al. 1999; Vanderwerf, Cooper et al. 2001). It is useful to
compare characteristics of rat and human ATP7B because of their close similarities, yet having a
difference involving the fourth domain of their N-terminal copper-binding region. Despite the
extensive literature, work on comparative metal-induced secondary and tertiary structural
changes of the N-terminus of these transporters is still incomplete. Here we have combined a
wide range of techniques to probe the structure of the rat copper-transporting ATPase N-terminal
copper-binding region (rCBD) and to compare it to the Wilson disease N-terminal copper-
binding region (WCBD). We use homology modelling to reveal the presence of a domain in
rCBD, which was previously thought to be absent from the rat ATP7B (Wu, Forbes et al. 1994).
We show that this domain, the fourth domain (rD4) of rCBD, has retained the typical fold
associated with HMA domains, even though it is missing the CXXC (where X is any amino acid)
Cu(I)-binding site. We establish that rCBD and WCBD both contain six domains, but unlike the
six HMA domains found in WCBD, only five of the domains in rCBD are HMA domains
containing the amino acid sequence CXXC. Our 65Zn(II)-competition blotting and circular
dichroism (CD) spectroscopic techniques reveal extensive similarities between the metal-binding
properties and metal-induced conformation changes observed in rCBD and WCBD. We propose
69
that the conformational changes observed in the N-terminal copper-binding region are triggered
initially by cooperative metal binding to the domains, and then driven to completion by favorable
protein–protein contacts established among some of the domains, and depend to a lesser extent
on metal binding.
4.3 Materials and Methods
4.3.1 Construction of atp7b copper-binding region cDNA
The cDNA fragment was generated by the polymerase chain reaction (PCR) directly from a
clonal phage lysate (clone 7; (Wu, Forbes et al. 1994)), containing the entire atp7b coding region
cDNA. PCR was carried out in a 50-µl volume using 1 unit of pfu polymerase (Stratagene), 2
mM each of dATP, dCTP, dGTP, and dTTP, 1.5 mM magnesium chloride, 50 ng of each
oligonucleotide primer and template cDNA in the buffer supplied by the manufacturer. The
reaction conditions were 5 min at 95 ºC for initial denaturation of template cDNA followed by
20 cycles of 30 s 95 ºC denaturation, 30 s 58 ºC annealing, and 3 min 72 ºC extension. The
primers used were as follows 5’-ACTGGGATCCATGCCTGAACAGGAGAGAAAG-3’ and 5’-
ACTGGTCGACTCACTGT-TTTATTTCCGTCTTGTG-3’. These primers incorporate a 5’
BamHI, and a 3’ SalI restriction endonuclease site to enable cloning of the amplified cDNA
fragment. Following PCR amplification, the cDNA fragment was purified by agarose gel
electrophoresis, and recovered from the excised gel slice using a QiaQuick Gel purification kit
(Qiagen) according to the manufacturer’s protocol. The purified cDNA fragment was double-
digested with BamHI and SalI restriction endonucleases (New England Biolabs) then subjected
to a further round of gel purification as described above. This cDNA fragment was ligated into
pUC19 vector (Pharmacia) double-digested with BamHI and SalI restriction endonucleases (New
England Biolabs) and gel-purified as described above. Ligations were carried out at 4 ºC
overnight using T4 DNA ligase (New England Biolabs) in buffer supplied by the manufacturer.
Ligated DNA was transformed into XL-1 Blue E. coli (Stratagene) by electroporation (BioRad).
Plasmid DNA was prepared from carbenicillin selected E. coli clones using Qiagen Miniprep
Spin columns according to the manufacturer’s protocol. The rCBD cDNA was excised out of the
pUC19 vector and cloned into a glutathione-S-transferase (GST) fusion expression vector,
pGEX-6P-2 (Amersham Pharmacia Biotech Inc.) to form the construct pGEX-rCBD.
70
4.3.2 Expression and purification
Expression and purification were performed as previously described with modifications
(DiDonato, Narindrasorasak et al. 1997). Briefly, pGEX-rCBD was amplified using E. coli
DH5α cells to retain plasmid integrity. The vector sequence was confirmed through Sanger
dideoxy method (DNA Sequencing Facility, The Centre for Applied Genomics, The Hospital for
Sick Children). The construct was then isolated by Miniprep (Qiagen) from DH5a cells and
transformed into E. coli strain BL21(DE3) cells for increased expression. Cells were grown in
normal LB media and induced at mid-log phase for 3.5 h with 0.1 mM isopropyl-1-thio-h-D-
galactopyranoside (IPTG). Cells were collected, centrifuged for 15 min at 5000 rpm (Beckman
Jm-10) and submitted to one to two freeze thaw cycles as necessary in lysis buffer (25 mM Tris–
HCl, pH 8.0, 130 mM NaCl, 1% Triton X-100, 1 mM EDTA, 10% glycerol, 1 mM PMSF, and
0.15 mg/ml lysozyme). The cells were ultracentrifuged (Beckman L-90K) at 118,000 g for 45
min. The supernatant liquid was applied to a GST-affinity column (Amersham Pharmacia), the
pellet was homogenized in solubilizing buffer (25 mM Tris–HCl, pH 8.0, 6 M urea, 130 mM
NaCl, 1 mM EDTA, 1 mM dithiothreitol (DTT)), and ultracentrifuged as before. Solubilization
was repeated as necessary. The solubilization supernatant from all centrifugations was dialyzed
extensively against 25 mM Tris–HCl, pH 8.0, 10% glycerol, 1 mM EDTA, 1 mM DTT, 1%
Triton X-100 (BioRad) and then applied to a GST-affinity column. The GST-fusion protein was
eluted using buffer containing 25 mM Tris–HCl pH 8.0, 6 M urea, 80 mM NaCl, 1 mM EDTA,
and 1 mM DTT. The quality of fractions was confirmed by SDS-polyacrylamide gel
electrophoresis (SDS-PAGE). Further purifications were performed on Sepharose Fast-Q anion
exchange column and eluted using the same buffer but with a NaCl gradient from 80 to 200 mM.
Purified fusion protein was then dialyzed into refolding buffer (elution buffer without urea) and
cleaved using 0.3 units/ml PreScission™ protease. rCBD was purified from uncleaved fusion
protein and GST by GST affinity column. Protein quantitation was performed by BCA assay
(Pierce). Protein was stripped of metal when necessary using a method previously described for
the preparation of WCBD (DiDonato, Zhang et al. 2002).
71
4.3.3 Thermal denaturation analysis
In order to confirm proper folding after various steps of purification, samples were examined by
thermal denaturation analysis using CD spectroscopic analysis carried out at 222 nm from 25 to
95 ºC in 25 mM Tris–Acetate pH 8.0. All spectra were corrected against buffer and noise
reduced. Spectra were collected and analyzed using an AVIV 62A DS spectrometer with a 0.1
cm rectangular CD cell. Data points were collected every 2 ºC with intermediary equilibration
times of 1 min.
4.3.4 65Zn(II)-competition blotting
Purified rCBD was electrophoresed on a 10% SDS-PAGE gel, transferred to nitrocellulose
membrane (BioRad) in 10 mM 3-[cyclohexylamino]-1-propanesulfonic acid (CAPS) (Sigma),
pH 11.0 at 300 mA for 1 h 45 min. The membrane was then cut into strips and soaked in a
reducing binding buffer (100 mM Tris–HCl, pH 7.2, 50 mM NaCl, 1 mM DTT) for 2 h.
Individual strips were placed in incubation tray wells, briefly washed in binding buffer without
DTT. Strips were incubated with 65Zn(II) to a final concentration of ~ 4.2 µM in 3 ml of binding
buffer without DTT with various concentrations of competitor metal for 30 min. When copper
was the competitor, DTT was included to reduce Cu(II) to Cu(I). Strips were washed twice for
15 min in the same buffer and then exposed to Kodak Biomax MR film for 16–24 h at -70 ºC.
Autoradiographs of nitrocellulose blots were over-exposed to reveal the lower limit of binding
and binding trends. Therefore, competition results are reported according to Zn(II) intensity, and
due to the design of the experiment, are only interpreted qualitatively.
4.3.5 Structural analysis by CD spectroscopy
Separate aliquots of apo-rCBD (29.4 µM) were combined with various concentrations of copper
(0–300 µM) in the form of CuSO4 with 1 mM DTT. The solutions were equilibrated for 30 min,
by which time it was assumed that all metals were reduced and bound to available sites. Excess
metal and DTT were dialyzed out in argon-purged buffer (50 mM Tris–Acetate pH 8.0). DTT
was included to reduce protein and Cu(II) to Cu(I) but necessarily dialyzed out because of
interference experienced in the near-UV range. Buffer was purged with argon to maintain
reduced conditions. CD spectra of Cu(I) samples were collected and analyzed using a Jasco J-
720 spectropolarimeter. Far-UV CD spectroscopic analyses were employed to qualitatively
monitor the change in structure as increasing amounts of metal are bound. The effect of metal
72
binding on tertiary structure was observed by near-UV spectroscopic analysis that measures the
specific aromatic residues and disulfide content of the protein, factors which are likely to be
dependent on the intrinsic properties of rCBD. Secondary structure (far-UV, 300–190 nm)
analysis was performed in a 0.1 cm path length cylindrical CD cell. Tertiary structure (near-UV,
400–250 nm) changes were recorded using a 2 cm path length cylindrical CD cell.
4.3.6 Homology modelling of the N-terminal domains
The rCBD was subjected to homology modelling based on the solved structures of other HMA
domains. SWISS-MODEL was used to model the HMA domains of the rCBD and SWISS-
PDB-viewer was used to generate a pictorial representation of the domains (Peitsch 1995;
Peitsch 1996; Guex and Peitsch 1997).
4.4 Results
4.4.1 Cloning, expression and purification
We have cloned the DNA segment coding for the rat N-terminal copper-binding region into a
GST pGEX-6P-2 vector. Subsequent expression of this plasmid and cleavage of the fusion
protein with PreScission™ protease resulted in free rCBD. rCBD was purified from uncleaved
fusion protein and GST protein by GST affinity and anion exchange columns. A typical
purification process is shown in Figure 4.1. The purity of the final product, rCBD protein, was
determined to be ~ 95% pure by 10% SDS-PAGE. Because at least part of the purification
process involved denaturation by urea, thermal denaturation was employed to confirm refolding.
The thermal denaturation curve of purified rCBD at 222 nm displays a sigmoid shape (Figure
4.2), indicating that the protein has a compact conformation typical of folded structures and
native proteins. This was also supported by earlier investigation of WCBD showing that the CD
profile of protein that had undergone denaturation and refolding during purification was similar
to protein purified under non-denaturing conditions (DiDonato, Zhang et al. 2002).
73
Figure 4.1: Purification of rCBD
Standard Marker (A). GST-rCBD expression in BL21(DE3) cells without IPTG (B) and following induction with IPTG (C). Initial purification by GST affinity chromatography (D) and then by ion exchange chromatography (E). Purified GST-rCBD was cleaved with PreScission™ protease (P). rCBD was purified from GST and GST-rCBD on a GST affinity column (G).
Figure 4.2: Thermal denaturation of rCBD.
The CD spectrum of thermal denaturation of rCBD (1mg/mL) at 222 nm shows a sigmoidal curve, typical of a folded protein.
74
4.4.2 65Zn(II)-competition blotting
The binding of certain transition metals was investigated by the use of a 65Zn(II)-blot assay; a
summary of results is given in Figure 4.3. Zinc(II) has previously been shown to bind to the
metal-binding domain of ATP7B (DiDonato, Narindrasorasak et al. 1997). Figure 4.3A shows
that various transition metals were able to successfully bind to the protein and displace 65Zn(II)
while others did not. rCBD shows an affinity for Cu(I)>Ag(I)>Hg(II)>Au(III)>Cd(II)>Cr(III)
>Fe(III). As the concentration of Cu(I), Ag(I), Hg(II) and Au(III) was increased, 65Zn(II) was
displaced in a linear fashion (Figure 4.3A). Co(II), Mg(II), and Ca(II) did not appear to be able
to compete with 65Zn(II) at all, while Cd(II), Cr(III) and Fe(III) competed to a lesser extent. A
more detailed analysis of copper as the competitor to 65Zn(II) reveals a sigmoidal pattern,
characteristic of a cooperative interaction of metal with the protein (Figure 4.3B).
75
Figure 4.3: Competition of rCBD with 65Zn(II) and transition metals.
rCBD (10 µg/lane) was electrophoresed on 10% SDS-PAGE and transferred to nitrocellulose. Nitrocellulose strips of rCBD were incubated with 65Zn(II) and one of a variety of transition metals as described in section 4.3.4. Zn intensity refers to the digital quantitation of autoradiograph bands and should not be considered equivalent to binding affinity. Autoradiographs were overexposed until binding patterns appeared; comparisons of various metal binding should be made relative to other metals (not to the baseline) and only qualitatively. Digital quantitation of autoradiographs revealed the relation in (A) between 65Zn(II) and ♦Au(III); �Co(II); �Cd(II); +Cr(III); •Fe(III); -Hg(II) and in B) between 65Zn(II) and ♦Cu(I).
4.4.3 Homology modelling of rCBD
The common belief has been that rat ATP7B contains five HMA domains while human ATP7B
contains six; this has been considered to be the key difference between these two copper
transporting P-type ATPases. Our homology modelling of the rCBD has revealed an additional
76
domain, which had not been immediately obvious from the primary sequence alone (Figure
4.4A). Due to the absence of the CXXC sequence motif from this region, which corresponds to
the fourth domain in WCBD, rCBD was considered to be missing this domain (Wu, Forbes et al.
1994), and remained undetected until now. The sequence alignment of the domains within rCBD
reveals that although only five contain the CXXC Cu(I)-binding sites (Figure 4.4B), the six
domains not only share significant sequence similarity and conservation of hydrophobic residues,
but more importantly they all possess the fold typical of HMA domains (Figure 4.4C).
Figure 4.4: Structural alignment of the domains of rCBD.
A) Homology model of rD4 show that it contains the classic HMA fold. B) Primary sequence alignment of the rCBD outlining the region containing the CXXC motif. Identical or highly similar residues are in bold letters. The degree of similarity is denoted by (*) identical; (:) strongly similar; and (.) weakly similar. C) Structural alignment of the six domains in rCBD reveals that although only five contain the CXXC copper-binding motif, all six have the same tertiary structure, typical of HMA domains.
77
4.4.4 Structural analysis
Molar ellipticity at 220 nm (Figure 4.5A) shows that the secondary structure of rCBD increases
in direct relation to the addition of Cu(I). Above a ratio of 2–3 mol of Cu(I) per mole of rCBD,
no significant changes in the secondary structure are observed. Near-UV CD was employed to
further characterize the metal binding characteristics of Cu(I) reconstituted rCBD. At 260 nm,
the molar ellipticity increases proportionally to an incremental addition of Cu(I) until 2 mol of
Cu(I) is added. Upon the addition of 2 and 3 mol of copper per mole of protein, there is a radical
change in the direction of the spectra indicating a critical change in the environment surrounding
aromatic residues (Figure 4.5B). Extrapolation from this wavelength clearly demonstrates that
major tertiary structural changes are occurring between 2 and 3 mol of Cu(I) per mole of rCBD
(Figure 4.6A). Although the aromatic residues clearly reflect the structural changes of rCBD,
disulfide spectra may more clearly elucidate events at the individual binding sites. Molar
ellipticities at 240-350 nm are responsible for a broad disulfide signal; however, because of
interferences by tryptophan and tyrosine, disulfide signals are only measurable when they make a
significant contribution to the signal and when measured at above 320 nm (Wingfield and Pain
1996). An analysis of CD spectra at an absorbance of 330 nm shows that the molar
ellipticity/residue decreases in direct relation to the addition of Cu(I) to apo rCBD but does not
change beyond 5 mol Cu(I) per mole of protein (Figure 4.6B).
78
Figure 4.5: CD spectroscopy of copper-reconstituted rCBD.
Far-UV spectra of copper-reconstituted rCBD showing the change in secondary structure in relation to apo with increasing amounts of copper bound (A). No significant change in secondary structure is observed beyond 2-3 molar ratio of copper bound to rCBD. Near-UV spectra of copper-reconstituted rCBD showing a dramatic change in the directions of the spectra, when the copper increased from 2 mol of copper bound to 3 mol of copper bound to rCBD (B).
79
Figure 4.6: CD spectra extrapolations of rCBD.
Extrapolation of near-UV spectra as a function of number of moles of Cu(I) bond rCBD at (♦) 260, (■) 290, and (▲) 330 nm. CD extrapolation of different ratios of Cu(I) bound to rCBD at (▲) 330 nm (B).
4.5 Discussion
This report describes the structural and metal-binding characteristics of the rCBD, in light of the
discovery of the presence of the fourth HMA-like domain, which was previously thought to be
missing in rCBD. The ability of different transition metals to compete with 65Zn(II) and to bind
to rCBD confirms earlier reports of multi-metal binding to copper-transporting ATPases
(DiDonato, Narindrasorasak et al. 1997). Indeed, our experiments show that metals other than
copper are binding to rCBD; this is not surprising considering that the CXXC site in the metal-
binding domains is ubiquitously found across species in different kinds of metal transporting
proteins such as in the E. coli Zn(II) transporter, ZntA (Rensing, Mitra et al. 1997) or the
80
mercury-transporting protein MerP (Powlowski and Sahlman 1999). rCBD’s diverse metal
affinity implies that in vivo ATP7B is able to discriminate among these metals by a mechanism
other than binding ability alone. Copper(I) and Ag(I) (data not shown) are the only metals
among those capable of binding to rCBD that displayed cooperativity upon binding. Considering
that ATP7B is a known copper-transporting protein, and that Cu(I) (Petris, Mercer et al. 1996;
Goodyer, Jones et al. 1999; Strausak, La Fontaine et al. 1999; Voskoboinik, Strausak et al. 1999;
Forbes and Cox 2000; Roelofsen, Wolters et al. 2000) and Ag(I) (Petris, Mercer et al. 1996) are
the only two metals capable of inducing the kind of metal dependent trafficking that has become
the hallmark of this class of proteins, it is reasonable that the interaction of Cu(I) and Ag(I) with
rCBD would induce a response distinct from those of other transition metals. Previous studies
have shown that the most likely configuration for Cu(I) bound to HMA domains is that of a bent
2-coordinate linear system (Ralle, Cooper et al. 1998; DiDonato, Hsu et al. 2000). It is
interesting to point out that despite rCBD having one less CXXC copper-binding site than
WCBD, they both display a similar pattern of cooperativity upon Cu(I) binding (Figure 4.7)
(DiDonato, Narindrasorasak et al. 1997). This indicates that the tertiary structure of the rCBD
may be an important factor in the cooperative metal binding character displayed. It appears that
the absence of the CXXC site from the fourth domain does not alter the cooperative binding of
metal to the remaining domains, as long as the HMA fold is present, and the important tertiary
protein-protein contacts between the HMA domains are maintained.
This raises an interesting question: if a conformational change is a necessary precursor for the
translocation of the protein from the trans-Golgi network to the plasma membrane (Terada, Aiba
et al. 1999; Forbes and Cox 2000) as already hypothesized (DiDonato, Hsu et al. 2000;
DiDonato, Zhang et al. 2002), is this cooperativity of copper binding related to the copper
induced translocation?
CD spectroscopy was used to monitor the conformational changes in the rCBD upon the binding
of Cu(I). The changes observed in the far-UV CD spectra are indicative of changes in secondary
structure, and the changes observed in the near-UV spectra are indicative of changes in the
environment of aromatic residues and of tertiary structure in general. When apo rCBD is slowly
titrated with Cu(I), significant changes occur in the secondary structure and aromatic regions of
the CD spectra, most distinctive is the transition from the apo to the 1:1 state (Figure 4.5A and
B). In vivo this can translate to a transition from low Cu(I) to elevated Cu(I) conditions,
81
inducing a conformational change in the rCBD and influencing the function of ATP7B. These
results may point to the conformational change in the N-terminal region as a possible trigger of
the translocation event to the plasma membrane; this is also suspected to be the case from the
results of the mutational studies on ATP7B (Forbes, Hsi et al. 1999) and ATP7A (Goodyer,
Jones et al. 1999; Strausak, La Fontaine et al. 1999).
Figure 4.7: 65Zn(II) and Cu(I)-competition experiments.
Comparison of 65Zn(II)-competition blotting experiments show extensive similarities between the metal-binding properties and metal-induced conformation changes observed in rCBD and WCBD.
The current literature tends to support two hypotheses on the function of the metal-binding
region of the copper-transporting ATPases. It is commonly accepted that although not necessary
for copper-transport across the membrane, the metal-binding region plays a central role in the
copper-dependent trafficking and regulation of copper-transporting ATPases. In general, Cu(I)
is believed to bind to the HMA domains and then transport across the membrane, likely by a
channel formed by the eight transmembrane helices present in the ATPases. The metal-binding
region may also serve a second function, that of a metal ‘‘sensor’’ (Vulpe, Levinson et al. 1993;
Petris, Mercer et al. 1996; DiDonato, Narindrasorasak et al. 1997; Cobine, George et al. 2000).
The gradual change in secondary structure, as observed by far-UV CD spectra, supports a model
where Cu(I) binds to sensor HMA domains and induces conformational change.
82
The results in the near-UV/tertiary structure region support results from both far-UV experiments
and 65Zn(II) competition assays. The changes in molar ellipticity/residue in response to the
addition of copper to rCBD, at 260, 290 and 330 nm (Figure 4.6A) are similar in pattern to those
observed for WCBD (Figure 4.8) (DiDonato, Zhang et al. 2002). The molar ellipticity/residue at
330 nm decreases linearly both in the case of rCBD and WCBD but does not change beyond 5
and 6 mol of copper for each respectively (Figure 4.8). Considering that only five of the six
domains of the rCBD contain the CXXC copper-binding site, and are likely to have similar
metal-binding affinities due to sequence and structural similarities, this result is in agreement
with the results obtained from structural homology modelling and provides further evidence that
copper binds at the CXXC motif and that there is only one copper binding per CXXC containing
HMA domain. Also, these results suggest that the ratio of fully saturated rCBD is 5 mol of
copper bound per mole of rCBD.
It should be noted that although rCBD contains five metal-binding motifs and the WCBD
contains six, the breakpoint for structural change in rCBD (Figure 4.6A) is similar to that found
in WCBD (Figure 4.8) (DiDonato, Zhang et al. 2002) despite the missing CXXC Cu(I)-binding
site. This observation is supported by the homology modelling of the rCBD, which indicates the
presence of six domains having the HMA domain fold (Figure 4.4). Since rD4 lacks the CXXC
metal-binding motif (Figure 4.4A), the fourth CXXC does not appear to be crucial for
cooperative binding copper binding to rCBD or structural/conformational changes as long as the
overall HMA domain fold is present. In addition, the lack of this site does not appear to affect
the multi-metal binding affinity of rCBD observed in competition studies (Figure 4.3), similar to
studies with WCBD (Figure 4.7) (DiDonato, Narindrasorasak et al. 1997). Another possibility
may be that the rD4 may not serve a vital function in copper transport or that the loss of the
CXXC sequence motif is an evolutionary adaptation that enables copper to bind or transport
more efficiently. The similarity in the pattern of binding as observed in the WCBD (DiDonato,
Zhang et al. 2002) and in rCBD tends to imply that the structural response to metal binding will
also be similar. The fact that both the WCBD and the rCBD have an equal number of
structurally homologous HMA domains is in support of this argument. However, since there are
no studies confirming the order in which copper binds to ATP7B or related proteins, it is difficult
to make definitive conclusions about the specific role that HMA domain 4 may play in the rat
and the human ATP7B.
83
Figure 4.8: Near-UV CD spectra extrapolations as a function of moles of Cu(I) bound.
Despite the fact that the rat domain can only bind five equivalents of copper compared to the human’s maximum of six, we could observe similar changes in the CD spectra of both and a sharp change in molar ellipticity per residue for both proteins occurring between the addition of 2-3 molar equivalents of copper.
It has been known for some time that not all of the copper-binding domains of the N-terminal
region are required for the copper-transport function of the ATPase (Forbes, Hsi et al. 1999;
Strausak, La Fontaine et al. 1999), and that the conformational changes in the N-terminal copper-
binding region in response to metal binding may have functional consequences for the ATPase
(Goodyer, Jones et al. 1999; Strausak, La Fontaine et al. 1999; Vanderwerf, Cooper et al. 2001).
Our studies propose that the metal-dependent conformational changes observed in the N-terminal
region of these ATPases may not require the presence of a CXXC motif on every domain, as
long as the domain has the proper shape, which is an HMA-domain-like fold, and as long as
some of them contain the CXXC metal binding sites. The breakpoint in the binding trend,
observed at around 2–3 mol of copper in the near-UV CD spectra (Figure 4.8), may well
correspond to the breakpoint determined by 65Zn(II)-blots (Figure 4.7). This suggests that at a
certain point during the cooperative binding of Cu(I), protein – protein interactions among the
HMA domains begin to dominate the conformational changes in rCBD, while metal binding to
the HMA domains makes a smaller contribution.
84
Chapter 5
Copper(I) Interaction with Model Peptides of WD6 and TM6
Domains of Wilson ATPase: Regulatory and Mechanistic
Implications
This chapter is from Journal of Inorganic Biochemistry, 98, Myari, A., Hadjiliadis, N., Fatemi,
N., & Sarkar, B. Copper(I) interaction with model peptides of WD6 and TM6 domains of
Wilson ATPase: regulatory and mechanistic implications, 1483-1494, 2004, with permission
from © Elsevier 2004. This work was supported by a grant (MOP-1800) from the Canadian
Institute of Health Research. I performed the homology modelling and wrote the introduction,
the corresponding experimental, results and discussion sections. Preparation of the peptides, the
assays, CD and NMR spectroscopy and the corresponding written sections were the work of
Alexandra Myari, a student of N. Hadjiliadis.
85
5 Copper(I) Interaction with Model Peptides of WD6 and TM6 Domains of Wilson ATPase: Regulatory and Mechanistic Implications
5.1 Summary
With the aim to investigate the mechanism of Cu(I) transport by Wilson ATPase (ATP7B), we
have studied the interaction of the peptides 2K10p (CH3CO–Lys–Gly–Met–Thr–Cys–Ala–Ser–
Cys–Val–His–Asn–Lys–CONH2), and 2K8p (CH3CO–Lys–Leu–Cys–Ile–Ala–Cys–Pro–Cys–
Ser–Lys–CONH2), part of the sixth metal binding domain (WD6) and the sixth transmembrane
segment (TM6) of Wilson ATPase, respectively, by means of CD, NMR spectroscopy and
homology modelling. In addition, interaction of Cu(I) with 2K8p mutants where each of the
three cysteines has been sequentially replaced with a serine, have been studied with the same
methods, in order to understand the role of each cysteine in copper binding. Results from CD
and NMR experiments show that the 2K8p peptide binds a single atom of copper and that copper
binding induces secondary conformational changes in the peptide. The three cysteine thiolates
present in the 2K8p peptide sequence act mainly as bridging ligands for Cu(I) binding.
Significant shifts of proline protons upon interaction with Cu(I) have been observed for all
peptides, implying a possible role of proline in facilitating Cu(I) binding. These findings have
been further discussed with respect to the molecular basis of copper trafficking and
intermolecular interactions.
5.2 Introduction
Wilson disease is a genetic disorder of copper metabolism characterized by the toxic
accumulation of copper in the liver and extrahepatic sites. The gene responsible for Wilson
disease, located on chromosome 13, was identified in 1993 (Bull, Thomas et al. 1993; Petrukhin,
Fischer et al. 1993; Tanzi, Petrukhin et al. 1993) and found to encode a copper transporting P-
type ATPase (ATP7B). ATP7B is localized in the trans-Golgi network of hepatocytes under
low copper conditions, redistributes to cytoplasmic vesicles when cells are exposed to elevated
copper levels, and then recycles back to the trans-Golgi network when copper is removed
(Terada, Schilsky et al. 1998; Forbes and Cox 2000; Sarkar 2000).
86
Figure 5.1: The Wilson disease copper transporting ATPase.
The predicted structure of the Wilson disease copper transporting ATPase contains general
features conserved among members of the P-type ATPase family. These are the TGEA motif
(actuator domain), the DKTGT and TGDN motifs (phosphorylation domain) and the sequence
MXGDGXNDXP found in the hinge region that connects the phosphorylation domain to the
transmembrane segment. ATP7B is further classified as a heavy metal transporting P-type
ATPase since it contains six metal-binding motifs, GMTCXXC, at the N-terminus of the
molecule, the CPC motif in the sixth intramembrane region (TM6), the SEHPL motif in the
nucleotide binding domain and eight transmembrane segments (Terada, Schilsky et al. 1998;
Sarkar 2000) (Figure 5.1). Studies of site-directed mutagenesis have shown that domains closest
to the transmembrane segment, domain six in particular, appear to be the most important for
copper transport across the membrane (Forbes, Hsi et al. 1999).
One of the earliest structures to be made available for a heavy metal associated domain was that
of the mercury binding protein MerP (Steele and Opella 1997). The highly stable and unique
conformation adopted by a mercury bound 18-residue linear peptide of the metal-binding CXXC
loop of MerP is similar to the conformation of the loop in the context of the 72-residue mercury
bound protein (Veglia, Porcelli et al. 2000). These peptides’ ability to mimic the metal-binding
loop highlights their utility as important tools to model metal-binding sites in metal protein
interaction studies. The 2K10p peptide was synthesized, containing the residues
87
GMTCASCVHN from the 6th N-terminal copper-binding domain of ATP7B in order to study the
interaction of metal-binding sites with Cu(I).
Another potentially important site for copper ligation resides in the transmembrane domain. The
fourth transmembrane helix of P-type ATPases contains a conserved proline that is important in
ion translocation (Sweadner and Donnet 2001). This helix in the Ca-ATPase (Toyoshima,
Nakasako et al. 2000), Na,K-ATPase (Rice, Young et al. 2001) and H-ATPase (Scarborough
2000), is predicted to correspond to transmembrane helix six of ATP7B. This transmembrane
domain is associated with the transduction channel and contains residues critical to cation
binding. In heavy metal-transporting ATPases a pair of cysteine residues flanks the conserved
proline to form the highly conserved CPC motif. Mutations of the CPC residues result in non-
functional proteins (Hung, Suzuki et al. 1997; Forbes and Cox 1998; Yoshimizu, Omote et al.
1998) that are unable to redistribute in response to copper (Forbes and Cox 2000; Petris,
Voskoboinik et al. 2002). Mammalian copper-transporting ATPases have an additional
conserved cysteine, forming a CXXCPC motif.
The construction of chimeric proteins consisting of the ZntA (zinc transporting ATPase) core
domain and the amino-terminal domain of ATP7B resulted in the maintenance of the ability to
transport zinc, but not copper (Hou, Narindrasorasak et al. 2001), thus illustrating that the
transmembrane segment is the determinant of the metal specificity observed for P-type ATPases.
Figure 5.2: Peptides used as models for the study of Wilson ATPase.
In order to investigate the role of each of the 3 cysteines present in the TM6 of ATP7B, the 2K8p
peptide (Figure 5.2), and three mutants of 2K8p (Figure 5.2) were synthesized in which one
cysteine residue at a time is replaced by a serine residue. Due to the importance of the metal-
binding sites in the function of Wilson ATPase, the investigation of the nature of their interaction
CH3CO-Lys1-Gly2-Met3-Thr4-Cys5-Ala6-Ser7-Cys8-Val9-His10-Asn11-Lys12-CONH2 2K10p
CH3CO-Lys1-Leu2-Cys3-Ile4-Ala5-Cys6-Pro7-Cys8-Ser9-Lys10-CONH2 2K8p
CH3CO-Lys1-Leu2-Cys3-Ile4-Ala5-Ser6-Pro7-Cys8-Ser9-Lys10-CONH2 2s
CH3CO-Lys1-Leu2-Cys3-Ile4-Ala5-Cys6-Pro7-Ser8-Ser9-Lys10-CONH2 3s
CH3CO-Lys1-Leu2-Ser3-Ile4-Ala5-Cys6-Pro7-Cys8-Ser9-Lys10-CONH2 1s
88
with Cu(I) comprises an essential step in the elucidation of the mechanism of regulation of this
protein and it’s mechanism of action. Following this concept, the comparison of the interactions
of all five of the above mentioned peptides (Figure 5.2) with Cu(I) has been the focus of this
study.
5.3 Materials and Methods
5.3.1 Homology modelling
The 2K10p and 2K8p peptides were modeled by homology using SWISS-MODEL and SWISS-
PDB-viewer (Peitsch 1995; Peitsch 1996; Guex and Peitsch 1997) to generate pictorial
representations of the peptides as previously described by Fatemi and Sarkar (Fatemi and Sarkar
2002).
5.3.2 Preparation of peptides
In all peptide sequences two lysine residues were added at both termini to increase the solubility
in aqueous media. All peptides were prepared by solid phase peptide synthesis utilizing a
peptide synthesizer (Pioneer) according to the Fmoc-strategy. Fmoc-PAL-PEG-PS resin was
used as the solid substrate. The deprotection of the Fmoc-amino acids (NOVA) was performed
by 20% piperidine solution in DMF. HBTU was used as a coupling reagent. Acetic anhydride
was used for the acylation of the peptide. Cleavage from the resin and deprotection of the side
chains were performed in one step using TFA:H2O:phenol:TIPS ratio of 88:5:5:2. The peptide
was precipitated in cold ether and isolated by centrifugation. Crude peptides were purified by
reverse phase HPLC using a Waters 600 chromatograph with a Waters 481 absorption detector
utilizing a C4 RPHPLC column for the separations. Chromatography solvents were HPLC
grade acetonitrile and water, both containing 0.1% TFA. The purification was performed using a
2%/min linear gradient of water (0.1% TFA) and acetonitrile (0.1% TFA) with an initial
water:acetonitrile ratio of 99:1 and a final ratio of 20:80. The peptides were then lyophilized and
stored in sealed containers at 4 ˚C until use. The purity of the peptides was checked by ESIMS.
89
5.3.3 Assays
5.3.3.1 Bicinchoninic acid assay
The bicinchoninic acid (BCA) assay kit was purchased from Pierce. BCA is a sensitive, stable,
highly specific reagent for the Cu(I) ion. The BCA Protein Assay combines the well-known
biuret reaction in which Cu(II) is reduced by the protein in an alkaline medium to produce Cu(I)
with the unique ability of BCA to form a purple reaction product by the interaction of two
molecules of BCA with the one Cu(I) ion (Smith, Krohn et al. 1985; Wiechelman, Braun et al.
1988). The purple product is water soluble and exhibits a strong absorbance at 562 nm which
allows the spectrophotometric quantitation of proteins in aqueous solutions. Stock standard
solution of purified Bovine Serum Albumin (BSA) was used for the construction of the standard
curve.
5.3.3.2 Ellman assay
Ellman’s reagent (Ellman 1959) (5,5’-dithiobis-(2-nitrobenzoic acid) or DTNB) readily forms a
mixed disulfide with thiols, liberating the chromophore 5-mercapto-2-nitrobenzoic acid
(absorption maximum 412 nm, ε = 13,600 M-1 cm-1). Concentration is calculated using the
molar absorption coefficient of the yellow product.
5.3.3.3 Cu assay
BCA is also incorporated in this method except that ascorbic acid served as a reductant of Cu(II)
to Cu(I). Atomic absorption standard Cu(II) solution was used for the preparation of standard
solutions and the construction of the standard curve. Removal of interfering substances was
achieved by precipitation with 15% TCA (trichloroacetic acid) solution.
5.3.4 CD spectroscopy
Peptide concentration was determined by BCA protein assay. The concentration of copper was
determined utilizing the Cu(I)-assay (Brenner and Harris 1995). Ellman assay was utilized for
the determination of the concentration of DTT. DTT served as a reductant for both Cu(II) and
the peptide disulfides.
All samples were prepared in Tris–HCl (25 mM), pH 8.00 buffer. Peptide concentration was 100
µM. For the insertion of the peptide in micelles, sodium dodecyl sulfate (SDS) was added to the
90
sample to the final concentration of 20 mM. Increasing amount of DTT was added to samples
containing the peptide as well as samples containing peptide:Cu(II) 1:1. A series of blank
solutions containing all substances except the peptide was also prepared and measured. Each
measurement was repeated in triplicate. CD spectra were recorded on a Jasco J-720
spectropolarimeter. Cuvette path length of 0.1 cm was used for the 190–250 nm region while a 2
cm path length cuvette was used for the 250–400 region. Blank was subtracted from the CD
spectrum of each sample. Smoothing of the curves was performed using the Adjacent Averaging
method from Microcal Origin 6.0.
5.3.5 NMR spectroscopy
5 mM peptide solutions were prepared in D2O. The appropriate amount of peptide was weighted
on a balance of high accuracy. Equivalent amount of Cu(II) stock solution was added to the
peptide solution. The concentration of Cu(II) stock solution had been determined by titration
with Titriplex. DTT served as a reductant of both Cu(II) to Cu(I) and the peptide disulfide
bonds. For this reason DTT/peptide stoichiometry 2:1 was applied in all samples. The pH was
adjusted to 8.0 by using NaOD and DCl solutions. Sample preparation was performed inside a
glovebox in order to eliminate the presence of molecular oxygen and the NMR tube was tightly
stoppered. 1D 1H NMR and 2D TOCSY, NOESY and ROESY were acquired on a Bruker
Avance 400 MHz instrument.
5.4 Results
5.4.1 Homology modelling
2K10p is shown in Figure 5.3A as it is found in a homology model of WD6, the sixth Cu(I)
binding domain of ATP7B. Figure 5.3B shows TM6, the sixth transmembrane domain of
ATP7B modeled on M4, the fourth transmembrane domain of the Ca-ATPase. Val 304, Ile 307
and Glu 309 in M4, that contribute to site II high affinity Ca2+ ion binding site in SERCA1a
(Toyoshima, Nakasako et al. 2000), correspond to Cys 980, 983 and 985 in the unwound part of
TM6, forming the CXXCPC motif. Site II is formed almost directly on M4 by the main-chain
carbonyl oxygen atoms of Val 304, Ala 305 and Ile 307 and the side-chain oxygen atoms of Glu
309. This coordination geometry is made possible because the M4 helical structure is unwound
between Ile 307 and Gly 310 and interrupted by a kink at Pro 308. Similarly in the 2K8p
91
peptide, the Cu(I) ion binding site is not expected to adopt the standard alpha helical
conformation due to the presence of the same conserved proline residue, Pro 984. The residues
Cys 980, 983 and 985 in the unwound part of TM6 have been shown to be essential for copper
transport, and according to the homology model generated here, could provide a suitable
coordination geometry for Cu(I) ion binding.
Figure 5.3: Homology modelling of model peptides of the Wilson ATPase.
Homology model of 2K10p and WD6 of ATP7B (A). The sixth transmembrane helix of ATP7B, TM6 modeled on the fourth transmembrane helix of SERCA1a, M4 (B). The sequence motifs equivalent to 2K8p and the corresponding region on M4 are enlarged.
5.4.2 CD spectroscopy
Interaction of Cu(I) with peptides can be investigated by CD spectroscopy in two different ways.
By monitoring the changes of molar ellipticity per residue (deg cm2 dmol-1) in the region 190–
250 nm, information about the secondary structure of the peptide molecule can be obtained
92
(Wingfield and Pain 1996). In addition, the LMCT transitions present in the region 250–400 nm
offer valuable information about the nature of donor atoms ligating to Cu(I) (Pountney,
Schauwecker et al. 1994).
Due to their size and the absence of specific secondary structure and folding, peptide molecules
are far more exposed to solvent and vulnerable to oxidation compared to the whole native
proteins. The protection of sulfhydryl groups from oxidation becomes a necessary measure.
DTT, a common thiol protectant used in the isolation, purification and study of proteins, was
utilized for the reduction of peptide disulfide bond and for the in situ generation of Cu(I) from
Cu(II). The effect of DTT in the system is assessed by observing the changes of CD spectral
features in response to its successive addition, in the absence and presence of Cu(II). The ratio
of peptide to Cu(II) used was 1:1. Further addition of Cu(II) did not result in any spectral
variation for all the peptides studied (data not shown).
93
Figure 5.4: Changes in molar ellipticity at 198 nm of model peptides.
Change in molar ellipticity (vertical axis) of 2Kp and the mutants 1s, 2s, and 3s vs DTT/peptide stoichiometry (horizontal axis), [peptide, ■; peptide + Cu(II) 1:1, ●].
In Figure 5.4, the results from the far-UV region as a function of DTT/peptide stoichiometry can
be seen. The conformational changes taking place are studied by monitoring the molecular
ellipticity at 198 nm, the local minimum observed for all peptides, indicative of unordered
conformation (Johnson 1985; Fasman 1996). The values of molar ellipticity determined for
peptide and peptide + Cu(II) 1:1 are included in the graphs corresponding to zero DTT/peptide
ratio. All peptide + Cu(II) 1:1 samples exhibit lower magnitude of molar ellipticity compared to
the free peptide. This can be interpreted by considering either coordination of the peptide
thiolates to Cu(II) or reduction of Cu(II) to Cu(I) by the peptide thiolates resulting into disulfide
bridge formation with a consequent change in structure.
The presence of Cu(I) (Cu(II) +DTT) in solution is followed by major changes revealing the
difference in the nature of Cu(I) interactions with each peptide. The decrease in the molar
ellipticity magnitude observed in the presence of Cu(I) (when peptide and peptide + Cu(II) +
94
DTT are compared) probably originates from its complexation by the free sulfhydryl groups.
When the magnitude of molar ellipticity for the free peptide is compared with the one for peptide
+Cu(II) +DTT, ligation of Cu(I) has the most dramatic effect in the peptide conformation in the
case of 2K8p followed closely by 1s whereas the impact is appreciably lower for 3s and 2s
(Figure 5.4). Since, the CPC motif remains intact only in 2K8p and 1s, it can be speculated that
this effect originates from the interaction of CPC with Cu(I).
Further structural information about Cu(I)–peptide complexes comes from a detailed analysis of
mammalian Cu(I)–metallothioneins (Pountney, Schauwecker et al. 1994). Bands in the high
energy region (below 280 nm) originate predominantly from electric dipole allowed CysS–Cu(I)
LMCT transitions, the intensity of the band reflecting the number of thiolate ligands involved in
metal-binding. On the other hand, the unusually strong CD bands in the low-energy region are
often observed in multinuclear Cu(I) complexes which exhibit relatively short Cu···Cu distances
in which intramolecular d10–d10 interaction of adjacent Cu(I) ions leads to a number of excited
states with largely metal character. These weak low-energy bands have been assigned to
formally spin-forbidden 3d → 4s metal cluster-centered transitions. The occurrence of such
transitions indicates the presence of Cu(I) polyhedra in these complexes. The magnitude of these
transitions parallels the number of bound metal ions. All these features are well-demonstrated in
the CD spectra of mammalian metallothioneins (Pountney, Schauwecker et al. 1994) and the C-
terminal (Hasler, Faller et al. 1998) and N-terminal (Faller and Vasak 1997) domain of neuronal
growth inhibitory factor.
95
Figure 5.5: CD spectroscopy of 2K8p and mutant peptides.
Changes in the CD spectra of 2K8p and the mutants 1s, 2s, and 3s in the region 250-400 nm induced by the addition of Cu(II) and DTT (peptide, solid line; peptide + DTT, dotted line; peptide + Cu(II) 1:1, dashed line; peptide + Cu(II) + DTT, dashed and dotted line).
The appearance of metal-centered transitions is seen in the near-UV CD spectra of 2K8p and 1s
in the presence of Cu(I) (Figure 5.5). For comparison, CD spectra of peptide, peptide +DTT 1:1,
peptide + Cu(II) 1:1 and peptide +Cu(II)+DTT 1:1:1 are shown. Further addition of DTT does
not result in any change of the CD spectra (data not shown). In the 2K8p CD spectrum, in the
presence of Cu(I) (Cu(II) +DTT), well-resolved bands are present exhibiting positive and
negative maxima at (+)270 nm, (-)293 nm, (-)320 nm and (+)354 nm. The relatively intense
bands observed above 280 nm suggest copper cluster formation for the system 2K8p–Cu(I).
Mutation of C3 to S in 1s changes the shape of the CD spectrum dramatically. Distinct bands are
present exhibiting positive and negative maxima at (-)283 nm, (+)305 nm, (-)322 nm and (+)348
nm. The spectral features in the CT region are markedly red shifted in 1s, compared to 2K8p,
suggesting that the nature of the thiolate ligands differs. Moreover, the signs for the first two
96
bands have been reversed, indicating a dramatically different type of interaction between Cu(I)
and thiolates. Previous studies on metallothioneins (Vasak and Kagi 1983; Willner, Vasak et al.
1987) have attributed this spectral behavior to a Cu(I) coordination by predominantly bridging
thiolates, due to their relatively larger polarization when compared to that of terminal thiolates.
These data are suggestive of the participation of C3 in the ligation of Cu(I), its presence giving
rise to a different binding mode compared to the CPC motif by itself. It is most probable that the
CPC motif promotes Cu(I) ions cluster formation predominantly via coordination of bridging
thiolates.
The absence of any metal-dependent feature is pronounced in the CD spectrum of 2s. There is
no characteristic shape indicating the formation of a complex although there is a small but
nonspecific effect coming from the presence of Cu(I) in the solution. In the CD spectra of 3s,
bands which exhibit positive and negative maxima at (+)266 nm, (-)280 nm and (+)293 nm are
seen but no bands are present in the low-energy region. This fact indicates that the first two
cysteines may coordinate to Cu(I) and form a complex in a less asymmetric way compared to 1s
and even less compared to 2K8p. The absence of bands in the low-energy region implies that
except for playing a major role in the coordination of Cu(I), it may also act as a bridging ligand
in the CysS–Cu(I) cluster formation.
A more relevant approach to the study of the interaction of 2K8p with Cu(I) was attempted by
inserting it into a membrane. SDS, extensively used as a mimic of membrane environment in
studies of hydrophobic protein segments (Khan, Williams et al. 1995; Coles, Bicknell et al.
1998), was utilized as the membrane model. The hydrophobic environment of SDS micelles is
known to impose restrictions on the secondary structure of proteins (Li and Deber 1993). This is
also the case for 2K8p where a decrease in the magnitude of molar ellipticity at 198 nm is
observed in the presence of micelles. Furthermore, the addition of DTT as well as Cu(II) induces
only minor changes in the conformation of 2K8p in the presence of SDS micelles (data not
shown).
Upon insertion in the SDS micelles, 2K8p exhibits characteristic CT bands due to its interaction
with Cu(I) (data not shown). The magnitude and the maximum of molar ellipticity are almost
the same for 2K8p (270 nm, 656 deg cm2 dmol-1) and 2K8p in 20 mM SDS (272 nm, 658 deg
cm2 dmol-1). On the other hand, the magnitude of the CT band in the high energy region keeps
97
increasing upon the addition of DTT, reaching its maximum upon the addition of the third
equivalent whereas, in the absence of SDS, the spectrum remains invariable upon the addition of
DTT (data not shown). The study of 2K8p in the presence of micelles has demonstrated the
existence of a stable conformation, which is only slightly affected by changes of the external
environment.
Figure 5.6A shows the conformational changes observed for 2K10p upon the addition of DTT
and Cu(II). The magnitude of molar ellipticity for 2K10p at 198 nm is much lower compared to
2K8p. This is indicative of a more restricted conformation of 2K10p, as already predicted by the
homology model. The addition of DTT results in a major increase of molar ellipticity, probably
due to the reduction of the already formed disulfide bond. The possible complex formation via
coordination of the two free sulfhydryl groups (–Cys–X–X–Cys– motif) to Cu(I) is demonstrated
by the decrease in the magnitude of molar ellipticity (Figure 5.6A).
98
Figure 5.6: CD spectroscopy of 2K10p model peptide.
Changes in molar ellipticity at 198 nm (vertical axis) of 2K10p vs DTT/peptide stoichiometry (horizontal axis), [2K10p, ■; 2K10p + Cu(II) 1:1, ●] (A). Changes in the CD spectra of 2K10p in the region 250-400 nm as induced by the addition of Cu(II) and DTT(2K10p, solid line; 2K10p + DTT, dotted line; 2K10p + Cu(II) 1:1, dashed line; 2K10p + Cu(II) + DTT, dashed and dotted line) (B). Comparison of DTT/peptide 1:1 (DTT1) to 2:1 (DTT2) stoichiometry.
The changes induced by Cu(I) in the near-UV region are not so marked in the case of 2K10p. As
shown in Figure 5.6B, a CT band of medium intensity having two local maxima ((+)268 nm,
102.1 deg cm2 dmol-1 and (+)277 nm, 105.4 deg cm2 dmol-1) is observed in the presence of
Cu(II). This band is more intense than the one observed in the presence of Cu(I), which is
slightly red shifted upon addition of one equivalent of DTT ((+)273 nm, 64.5 deg cm2 dmol-1 and
99
(+)285 nm, 56.8 deg cm2 dmol-1) and separated into two well-resolved peaks upon the addition
of two equivalents of DTT ((+)267 nm, 71.4 deg cm2 dmol-1 and (+)287 nm, 46.3 deg cm2 dmol-
1). At the same time, a new peak appears exhibiting a maximum at (+)324 nm, 37.4 deg cm2
dmol-1. The shape of 2K10p CD spectra is not at all reminiscent of that observed for 2K8p,
which is what was expected from the homology models of the two peptides. It is clear that the
nature of the interaction between Cu(I) and the two peptides differs in a qualitative and also
quantitative manner.
Table 5-1 1H NMR chemical shifts of 2K8p, 1s, 2s, and 3s that differ when Cu(I) is present in solution
at pH 8.00 and T = 298 K.
5.4.3 1H NMR spectroscopy
The assignment of DTT protons was performed according to Krezel et al. (Krezel, Lesniak et al.
2001). The NMR spectrum of reduced DTT (DTTred) in D2O consists of one signal for a
CH(OH) proton at 3.6–3.7 ppm and two signals of the non-equivalent CH2S protons in the region
of 2.5–2.9 ppm. In the NMR spectrum of the oxidized form of DTT (DTTox) the signal for the
CH(OH) proton is at 3.6–3.7 ppm whereas the two signals of the CH2S protons shift to 2.8–3.0
ppm.
100
The assignment of cysteines was based on the successive replacement of one cysteine by one
serine residue in the original 2K8p sequence, which results in the disappearance of the
corresponding cross-peak in the 2D 1H NMR spectrum of the mutant peptide sequences 1s, 2s
and 3s. No NOE cross-peaks were observed for any of the samples measured. 2D ROESY did
not add any information to the TOCSY experiment.
The interaction of Cu(I) with peptides containing cysteine residues results in appreciable
broadening of the peaks. The greatest impact on the peaks linewidth by Cu(I) is observed for
2K8p whereas there is only a small influence on the peaks of 2s peptide. The fact that all signals
of the peptide protons are subject to broadening suggests that the entire molecule contributes to
the relaxation process. A similar effect is observed for the interaction of Cu(I) with DTT
(Krezel, Lesniak et al. 2001) and in the study of the metal thiolate clusters in the N-terminal of
neuronal growth inhibitory factor (Faller and Vasak 1997). Moreover, the addition of Cu(I) to
apo-CopZ, the copper chaperone of Enterococcus hirae, results in the loss of most NMR signals
near the metal binding site. The increased rotational correlation time of Cu(I)–CopZ compared
to apo-CopZ and the concomitant reduced overall mobility has been attributed to self-
aggregation of Cu(I)–CopZ (Wimmer, Herrmann et al. 1999). This phenomenon is even more
prominent when small oligomers are formed.
2D TOCSY experiment can be very informative with regard to the differences in the interaction
between the five different peptides. In Table 1, the differences can be seen in 1H NMR chemical
shifts of 2K8p, 1s, 2s and 3s when Cu(I) is present in solution. In the case of 2K8p, the great
number of combinations in oligomer formation results in extensive broadening of the cysteine
signals. This results in the disappearance of the corresponding cross-peaks. The participation of
all cysteines in the coordination of Cu(I) is therefore suggested. The signal of DTTred remains
unaffected suggesting strong chelation of Cu(I) by 2K8p. Furthermore, the peak of proline (P7)
Hδ has also disappeared probably due to broadening, while there is a shift of about 0.05 ppm for
the L2 Hβ and P7 Hγ protons. These data indicate the coordination of Cu(I) to the neighboring
residues of L2 and P7, thus being C3, C6 and C8.
The spectrum least affected by the presence of Cu(I) is that of 2s. There is no significant
broadening of the spectrum but an appreciable shift for the cysteine protons is observed (0.1 ppm
for Hα protons of C3 and C8 and ~0.05 ppm for Hβ protons of C3 and C8), coming from the
101
interaction with Cu(I). The formation of a macrochelate resulting from the simultaneous
coordination of C3 and C8 is not favored thermodynamically. For this reason, C3 and C8
cannot act as stable bridging ligands in oligomer formation. The coordination mode in this case
appears to be a simpler one. Probably, Cu(I) interacts independently with each cysteine. Taking
into account, that only in 2s there is a separate signal observed for the free and complexed forms
of DTTred, it can be suggested that DTTred also participates in the coordination as a monodentate
ligand resulting in a ternary complex with the peptide. This simple mode of binding can explain
the presence of separate signals instead of the broad lines observed for the other peptides.
In the case of 1s (0.07–0.12 ppm shift for P7 Hδ, Hβ protons) and 3s (~0.09 ppm shift for I4 Hα
and Hγ protons and 0.07–0.16 ppm for P7 Hδ and Hγ protons), disappearance of the cross-peaks
of cysteines is attributed to the signal broadening and suggests the formation of oligomers. The
signal of CH2 protons of DTTred is absent in the presence of Cu(I) for both peptides, suggesting
its possible participation in the coordination. 1s and 3s bind Cu(I) weaker in comparison to
2K8p, making it possible for DTT which has relatively strong Cu(I) chelating properties to
participate in the coordination. DTT has the capacity to form complexes with Cu(I)
independently or participate in the formation of mixed complexes.
For 2K10p, which is a part of WD6 and is known to bind Cu(I), the interaction is mainly
indicated by the broadening of signals observed in the 1D 1H NMR (data not shown). The
formation of oligomers can account for this broadening, which results in no significant shift of
the separate signals of the peptide. DTT seems to participate once more in the coordination since
the CH2 signal of its reduced form is not observed. 2K10p is probably not as strong a chelator as
2K8p. Given the reversibility of Cu(I) binding to WD6, whether Cu(I) remains bound to WD6
or is transferred to the transmembrane binding site for transport across the membrane, is likely
determined by the Cu(I) load of the hepatocyte.
In order to exclude the possibility that the broadening of the peaks in the presence of Cu(I) may
originate from other causes such as paramagnetism, additional experiments have been carried
out. Extensive studies of the interaction of Cu(I) with aminothiols (Bagiyan, Koroleva et al.
1978) have revealed that there is a strong tendency to form polynuclear complexes of
stoichiometry (CuL)n and (Cu2L3)n at Cu(I) concentrations above 10-4 M. The proportion of the
two forms and the degree of polymerization depend on the [Cu(I)]:[L] ratio, the absolute Cu(I)
102
concentration and the structure of L. Mononuclear species dominate, however, as [L] is raised
and [Cu(I)] is reduced. We decided to shift the equilibrium of polynuclear species formation by
modifying the [Cu(I)]:[L] ratio. 3s peptide has been utilized for these purposes and 1:5 and 1:2.5
ratios (excess of DTT has also been added to the samples) have been tested (Table 2). The
selection of 3s for this experiment was based on the significant signal broadening observed,
combined with the disappearance of DTTred signals. At a 1:5 ratio, the peaks sharpen and the
changes in the chemical shifts are more easily discriminated (Table 2). In the 2D TOCSY
spectrum, the cysteine cross-peaks are present, revealing a Hα shift of ~0.1 ppm. Furthermore,
proline Hδ cross-peaks are also present and characterized by a shift >0.1 ppm. No significant
proton shift is observed for the reduced DTT detected in the sample, suggesting that DTT is not
complexed with Cu(I). Increase the [Cu(I)]:[L] ratio to 1:2.5 does not result in a further shift for
cysteine and proline protons, whereas DTTred cross-peaks are absent. This suggests that
additional Cu(I) is complexed by DTT molecules in solution. Moreover, there were no spectral
changes observed between DTT/peptide stoichiometry 1:1 and 1:2.
Table 5-2 Comparison of 1H NMR spectra at different [Cu(I)]:[3s] ratios at 298 K.
Temperature is another determinant of the equilibrium stability constants. Decreasing the
temperature to 277 K, results in an expected increase in the broadening of the peaks in the 1D
and decreased cross-peak volumes in the 2D spectrum accompanied by an upfield shift of 0.2
ppm (observed for all protons). On the other hand, the increase of temperature at 323 K results
in the sharpening of all peaks and increased cross-peak volumes, and a downfield shift of all
peaks by 0.2 ppm. From the variable temperature experiments, it can be concluded that low
103
temperatures shift the equilibrium toward the formation of polynuclear species, and high
temperatures favor the formation of monomeric species.
5.5 Discussion
The difficulty in studying the entire ATP7B, due to its high molecular weight and insolubility in
aqueous media, necessitated the incorporation of CPC into a peptide model. Study of the CPC
motif, part of Wilson TM6, by means of CD and 1H NMR spectroscopy, has revealed its
significant interaction with Cu(I), implying a key role in copper transport.
The presence of the CPC motif in both 2K8p and 1s results in significant conformation changes
upon their interaction with Cu(I), as indicated by the intense CT bands in the CD spectra.
Moreover, the presence of C3 in 2K8p leads to the formation of different species compared to 1s,
with cysteine thiolates acting predominantly as bridging ligands in the case of 1s. DTT does not
participate in Cu(I) binding when C3 is present in the case of 2K8p. Cysteine C6 of the CPC
motif is important for the interaction with Cu(I), its absence in the sequence of 2s resulting in
weak association with Cu(I). In this case, DTT becomes an important ligand participating in
Cu(I) coordination. The absence of C8 in 3s results in smaller conformational changes upon
Cu(I) binding compared to 2K8p and significant participation of DTT in Cu(I) ligation.
2K10p, part of WD6, was used primarily as a reference in the study of peptide–Cu(I) interactions
since its binding mode is known. Clearly the position of cysteine residues and the context of a
peptide sequence result in different binding modes, which have to be examined thoroughly.
Although 2K10p and 3s may seem to have the same CXXC motif, the positioning of these
residues with respect to each other is very different in the 3D homology models of these
peptides. This conformational difference is reflected in the data, showing that the extent of
polynuclear species formation does not appear to be the same. P7 residue in 3s is probably
important for the final conformation of the peptide molecule affecting and probably promoting
the intrinsic capacity of cysteine thiolates to act as bridges.
Banci et al. (Banci, Bertini et al. 2003) investigated the effect of DTT on the coordination of
Cu(I) by the metal-binding motif MXCXXC present in CopZ, a copper chaperone in Bacillus
subtilis, using X-ray absorption and NMR techniques. Their results suggested that Cu(I) reaches
three-coordination in all cases where the third S ligand may either come from an exogenous
104
molecule like DTT or a second protein molecule to form a homodimeric complex. Furthermore,
exogenous thiols such as DTT or GSH might prevent intermolecular interactions whereas
homodimer formation appears to be favored in the presence of carboxylate salts and when copper
is reduced by ascorbate or provided as the acetonitrile Cu(I) complex. In our study, DTT appears
to be involved in Cu(I) coordination in all cases except 2K8p. The three cysteines present in the
2K8p molecule are adequate to saturate the Cu(I) coordination sphere but still cannot prevent the
intermolecular interactions. Furthermore, intermolecular interactions and DTT participation in
complex formation are minimized when the Cu(I):peptide ratio is decreased. This could be the
basis for the formation of the donor–acceptor adduct in intermolecular copper transport.
The significant shift noticed for proline protons as well as the fact that they have been subject to
significant broadening has to be evaluated. Moreover, significant shifts for aliphatic amino acids
such as leucine and isoleucine have also been observed. The events taking place upon Cu(I)
binding resulting in different conformational states can account for the above mentioned shifts.
Special characteristics have been attributed to the presence of a proline residue within a peptide
sequence: (i) Structural role including purely steric effects such as that arising from a kinked
helix as well as electronic effects that stem from increased local polarity, (ii) dynamic roles
which describe the participation of Pro in conformational changes and these include the cis–trans
isomerization of Xaa–Pro peptide bonds and Pro-mediated interconversions among all-trans
conformational states (Williams and Deber 1991). Proline is the only mammalian residue for
which the cis peptide bond is energetically accessible, converting the H-bonding patterns, thus
resulting in different conformational states. cis Xaa–Pro peptide bonds have been documented in
a number of proteins (Kordel, Forsen et al. 1990) as well as peptides (Torchia 1972). It has been
proposed that proline residues may be important in the context of membrane proteins. Brandl
and Deber (Brandl and Deber 1986) proposed that ligand-mediated cis–trans isomerization of
intramembranous Xaa–Pro bonds may regulate transport channel opening and closing.
The importance of P7 in the complexation of Cu(I) has been illustrated by the NMR data which
show a significant shift in proline protons. These data are suggestive of a dynamic role for
proline, introducing a stereochemically specific mode of Cu(I) binding to CPC. 2K8p exhibits
unique spectral behavior upon its complexation with Cu(I). This behavior is closely resembled
by 1s, which undergoes similar conformational changes as 2K8p. The picture is much different
105
when the cysteines of the CPC motif are not present. These observations provide some
preliminary indications concerning the importance of the CPC motif in Cu(I) binding.
Cys–X–Cys motif is present in metallothioneins where it is known to participate in polycopper
cluster formation (Bertini and Rosato 2008). Distinct metal–thiolate clusters formed by the
participation of many CXC motifs are observed at the N-terminal domain of neuronal growth
inhibitory factor (Faller and Vasak 1997). Cox17, a copper metallochaperone necessary for
delivery of copper ions to the mitochondrion, exists as an oligomeric polycopper species formed
by coordination of Cu(I) by the cysteinyl thiolates of a Cys–Cys–X– Cys motif (Heaton, George
et al. 2001). Therefore, CXC is a metal-binding motif in the vicinity of other metal-binding sites
that can easily form polynuclear clusters. In the case of CCS, the copper chaperone of
superoxide dismutase, the formation of a cysteine-bridged dinuclear copper cluster coordinated
by the two cysteines of the MT/HCXXC motif of domain I and the two cysteines of the CXC
motif of domain III, has been proposed (Rosenzweig 2001). The CXC motif of domain III has
been suggested to deliver Cu(I) to the CXXC motif of domain I. CXC and CXXC are two motifs
appearing in tandem in TM6. The cooperativity in Cu(I) binding between two different metal-
binding sites of the same protein may provide a basis for the Cu(I) transport pathway.
Our results suggest a Cu(I) binding geometry specific to the CPC motif. This Cu(I) interaction is
maintained even in the presence of SDS micelles, but is abolished when one of the two cysteines
of the CPC motif is mutated. The significance of proline, a highly conserved residue among P-
type ATPases, is indicated by the 1H NMR shifts upon Cu(I) binding and a possible role in Cu(I)
transport is proposed for the first time. Therefore, the CPC motif in the Wilson ATPase could
possibly act as an intermediate in the transport of Cu(I) from WD6 across the trans-Golgi
membrane. Such a hypothesis necessitates further investigation concerning interactions between
different donor–acceptor copper chaperone proteins. This would be a key step in the elucidation
of the mechanism of Cu(I) transport across membranes.
106
Chapter 6
NMR Characterization of Copper-Binding Domains 4-6 of ATP7B
This chapter is from Biochemistry, 49, Fatemi, N, Korzhnev, D, Velyvis, A, Sarkar, B, and
Forman-Kay, JD. NMR Characterization of Copper-binding domains 4-6 of ATP7B, 8468-8477,
2010, with permission from © ACS Publications 2010. This work was supported by funds from
the Canadian Institutes of Health Research to J.D.F.-K. N.F. was supported by the Canadian
Institutes of Health Research Strategic Training Program in the Structural Biology of Membrane
Proteins Linked to Disease. I wrote all sections and performed all experiments. Dmitry
Korzhnev assisted with model-free analysis. Apo and copper-loaded WCBD1-6 and the
corresponding TROSY spectra were produced by Algirdas Velyvis. We thank L. E. Kay and R.
Muhandiram for assistance with NMR experiments.
107
6 NMR Characterization of Copper-binding Domains 4-6 of ATP7B
6.1 Summary
The Wilson disease protein (ATP7B) is a copper-transporting member of the P-type ATPase
superfamily, which plays a central role in copper homeostasis and interacts with the copper
chaperone Atox1. The N-terminus of ATP7B is comprised of six copper-binding domains
(WCBDs), each capable of binding one copper atom in the +1 oxidation state. To better
understand the regulatory effect of copper binding to these domains, we have performed NMR
characterization of WCBD4-6 (domains 4-6 of ATP7B). 15N relaxation measurements on the apo
and Cu(I)-bound WCBD4-6 show that there is no dramatic change in the dynamic properties of
this three-domain construct; the linker between domains 4 and 5 remains flexible, domains 5 and
6 do not form a completely rigid dimer but rather have some flexibility with respect to each
other, and there is minimal change in the relative orientation of the domains in the two states. We
also show that, contrary to previous reports, the protein-protein interaction between Atox1 and
the copper-binding domains takes place even in the absence of copper. Comparison of apo and
Cu(I)-bound spectra of WCBD1-6 shows that binding of Cu(I) does not induce the formation of
a unit that tumbles as a single entity, consistent with our results for WCBD4-6. We propose that
copper transfer to and between the N-terminal domains of the Wilson Cu-ATPase occurs via
protein interactions that are facilitated by the flexibility of the linkers and the motional freedom
of the domains with respect to each other.
6.2 Introduction
Copper is an essential trace element which plays an important role in mammalian cellular
metabolism (Sarkar 1999). While trace amounts of copper are needed to sustain life, excess
copper is extremely toxic. Although various aspects of copper transport and metabolism have
been investigated, many specific details of intracellular copper transport remain elusive. Cloning
of the genes responsible for the two closely related major genetic disorders of copper metabolism
in humans, Menkes disease (ATP7A) (Chelly, Tumer et al. 1993; Mercer, Livingston et al. 1993;
Vulpe, Levinson et al. 1993) and Wilson disease (ATP7B) (Bull, Thomas et al. 1993; Petrukhin,
108
Fischer et al. 1993; Tanzi, Petrukhin et al. 1993), was a major breakthrough in the understanding
of intracellular copper transport. Both genes encode copper-transporting P-type ATPases that
play central roles in copper transport and homeostasis, using ATP to transport copper across the
membrane against an electrochemical concentration gradient (Axelsen and Palmgren 1998).
Both transporters are localized to the trans-Golgi network where they receive copper from the
copper delivery protein Atox1 (also known as HAH1) and transport it across the membrane for
incorporation into copper-dependent enzymes (Dierick, Adam et al. 1997; Schaefer, Roelofsen et
al. 1999; La Fontaine and Mercer 2007). They also remove excess copper from the cytosol by
vesicular trafficking to the plasma membrane in response to elevated intracellular copper levels
(Petris, Mercer et al. 1996; Schaefer, Hopkins et al. 1999; Roelofsen, Wolters et al. 2000).
ATP7A and ATP7B share certain core structural features with other members of the P-type
ATPase family: a transmembrane domain (TM), a nucleotide-binding domain (N), a
phosphorylation domain (P), and an actuator domain (A) (Axelsen and Palmgren 1998). A
crucial feature of the human copper-transporting ATPases is the presence of a large N-terminal
segment which binds 6 copper atoms in a cooperative manner (DiDonato, Narindrasorasak et al.
1997; Jensen, Bonander et al. 1999) to six copper-binding domains, each capable of binding on
copper atom in the +1 oxidation state (DiDonato, Narindrasorasak et al. 1997; Ralle, Cooper et
al. 1998). The six copper-binding domains have ferredoxin folds and are connected by linkers of
various lengths. ATP7B performs its role by interacting with many cellular components in order
to receive, transport, traffic, and excrete copper (Fatemi and Sarkar 2002). ATP7B is known to
interact with at least two proteins, the copper chaperone Atox1 (Hamza, Schaefer et al. 1999;
Larin, Mekios et al. 1999; Hamza, Prohaska et al. 2003) and COMMD1 (Tao, Liu et al. 2003),
which has been shown to bind copper in the Cu(II) state (Narindrasorasak, Kulkarni et al. 2007).
In both cases, the N-terminal segment of ATP7B has been identified as the site of interaction. It
has also been shown that this region interacts in a copper-dependent manner with the
cytoplasmic nucleotide-binding/phosphorylation domain of ATP7B, with copper decreasing the
interaction (Tsivkovskii, MacArthur et al. 2001). An intact and functional N-terminal region is
also required for the copper-induced phosphorylation of ATP7B (Vanderwerf, Cooper et al.
2001) which is in turn required for vesicular trafficking (Cater, La Fontaine et al. 2007).
Mutations found within Wilson copper-binding domains (WCBD) 1, 5 and 6 are known to give
rise to Wilson disease (Loudianos, Dessi et al. 1998), and show an impaired interaction with
109
Atox1 (Hamza, Schaefer et al. 1999), demonstrating their importance in ATP7B function.
Domains 5 and 6, the domains closest to the membrane spanning segment, have been shown to
play a critical role for copper transport as well as copper-dependent intracellular trafficking of
ATP7B (Cater, Forbes et al. 2004). However, the mechanism by which these domains receive
copper prior to transport across the membrane is not clear.
Investigations of the closely related Menkes protein ATP7A using a three-domain construct
(MNK4-6) (Banci, Bertini et al. 2005) and more recently on the entire N-terminal region
(MNK1-6) (Banci, Bertini et al. 2007) show that MNK1 and MNK4 form a stable adduct with
Cu(I) Atox1, and are therefore likely the preferential sites for this interaction. On the other hand
in ATP7B, domains 2 and 4 are thought to interact with Atox1 (van Dongen, Klomp et al. 2004;
Achila, Banci et al. 2006) showing that, although these two proteins are closely related, there are
subtle yet significant differences in their N-terminal copper-binding domains. In a study
involving the single-domain WCBD4 added to the two-domain WCBD5-6, it was shown that
Atox1 can transfer copper through the formation of a stable adduct to WCBD4 but not WCBD5-
6, and WCBD4 is able to transfer copper to WCBD5-6 (Achila, Banci et al. 2006). Further
emphasizing the distinction between the Wilson and Menkes N-terminal region, these results are
in contrast to the Menkes N-terminus, where domain 6 can bind metal independent of domain 4
(Banci, Bertini et al. 2007). Until the recent characterization of the entire N-terminal domain of
ATP7B (Banci, Bertini et al. 2009), which showed that all six domains can be metallated by
Cu(I) Atox1, most studies focused on single or two-domain constructs (Achila, Banci et al. 2006;
Banci, Bertini et al. 2008). Dynamic information is available only for the two domain constructs
of WCBD5-6 in the apo state (Achila, Banci et al. 2006) and WCBD3-4 in the apo and Cu(I)-
bound states (Banci, Bertini et al. 2008). There has been no dynamic characterization of the
entire WCBD1-6 nor of other constructs containing greater than two domains, with the exception
of MNK4-6, which is lacking data from domain 5 in the Cu(I)-bound state (Banci, Bertini et al.
2005).
In order to understand the mechanism of intracellular regulation and transport of copper, we
chose to analyze an intact segment of the ATP7B N-terminal copper-binding domains spanning
domains 4-6 (WCBD4-6). Interest in this particular construct stems from the inherent
importance of domains 5 and 6 to copper transport, and domain 4 as a target site for copper
delivery from Atox1, with domains 4 and 5 joined by an uncharacterized flexible linker. We
110
demonstrate the interaction of apo WCBD4-6 with apo Atox1 without the need for copper. We
present the results of 15N NMR relaxation studies of WCBD4-6 with coverage of three domains
and the linker region in both the apo and Cu(I)-bound states. We show that the Cu(I)-binding
domains of WCBD4-6 have independent mobility in the apo and the Cu(I)-bound state and that
copper binding does not induce a change in the relative orientation of the domains with respect to
one another. Comparison of NMR spectra of WCBD1-6 in the apo and Cu(I)-bound states
suggests that domains 1-6 also retain some degree of flexibility with respect to one another, even
in the Cu(I)-bound state. Thus, the flexibility of the N-terminal copper-binding domains with
respect to each other appears to be an essential feature of the Wilson ATPase.
6.3 Materials and Methods
6.3.1 Cloning, purification and NMR sample preparation
6.3.1.1 WCBD4-6
The DNA sequence encoding WCBD4-6 (amino acids 357-632 of ATP7B) was amplified via
PCR and subcloned into a pET SUMO vector in E. coli BL21 (DE3) cells. Cells were grown at
37 ºC in M9 D2O minimal media, supplemented with 10 mg/L biotin, 10 mg/L thiamin, 0.3%
glucose (13C-glucose was used in samples prepared for backbone assignment experiments), and
0.1% 15NH4Cl. Expression of the fusion protein was induced at A600 of 0.6-0.8 by the addition of
0.1 mM IPTG for 18 h at 18 ºC. Cells were harvested at 4420 x g for 15 minutes, at 4 ºC. The
supernatant was poured off and the bacterial pellets were either stored at -20 ºC or lysed
immediately. All buffers reported herein were prepared in such a way as to minimize cysteine
oxidation: DTT-less solution was made, degassed under vacuum, purged with argon (for at least
30 minutes at RT) and then supplemented with DTT. The cell pellet was resuspended in binding
buffer A (20 mM Tris-HCl, 500 mM NaCl, 10 mM imidazole, 2 mM DTT, pH 8, and 1 mini-
tablet EDTA-free Complete protease inhibitor cocktail per 100 mL buffer A), lysed by sonication
on ice for 5 x 1 minute (pulsed), and centrifuged for 45 minutes at 118,000 x g, at 4 ºC. The
residual pellets were resuspended in the same buffer, sonicated, and centrifuged once more. The
pooled supernatants were filtered through a 0.8 µm filter to remove cell debris and applied to a
Ni2+ affinity column equilibrated with binding buffer A. After a 1 h incubation at 4 °C while
mixing, the column was washed extensively with the same buffer. The fusion protein was eluted
with binding buffer A containing 500 mM imidazole. The eluate was allowed to digest with
111
SUMO protease (Ulp) in the cold overnight. The digest was concentrated and applied to a
Superdex-75 Hi-load gel-filtration column equilibrated with NMR sample buffer B (20 mM
NaK-phosphate, 130 mM NaCl, 5mM DTT, pH 6.0). Fractions corresponding to each protein
were collected and concentrated in NMR sample buffer B.
6.3.1.2 Atox1
The pGEX-6P-2 plasmid containing Atox1 was transformed into E. coli BL21 (DE3) cells, and
the protein was expressed and purified as previously described protocol (Narindrasorasak, Zhang
et al. 2004) with a few changes. The fusion protein was digested on-column according to
standard protocol with Pre-Scission protease. The flow-through containing Atox1 was
concentrated and, as a final purification step, it was applied to a Superdex-75 Hi-load XK16/60
column (120 mL bed volume) equilibrated with NMR sample buffer B. Fractions corresponding
to Atox1 were collected and concentrated in NMR sample buffer B.
6.3.1.3 WCBD1-6
The DNA sequence encoding WCBD1-6 (amino acids 1-632 of ATP7B) was amplified via PCR
and subcloned into a pGEX-6P-2 vector in E. coli BL21 (DE3) cells. 1 L of cell culture was
grown at 37 °C according to standard protocols. Protein expression was induced with 1 mM
IPTG at room temperature for 16 h. The cells were harvested and the cell pellet was resuspended
in buffer C (100 mM NaH2PO4, 200 mM NaCl, 10 mM DTT, pH 7.0) and either stored at -20 °C
or lysed immediately to extract protein.
Cells were thawed and incubated at room temperature with 1 mg/mL lysozyme for 15 minutes. 1
mM PMSF, 1 mM EDTA and 1 mini-tablet of EDTA-free Complete protease inhibitor cocktail
were added. Cells were lysed by sonication on ice, lysate centrifuged 30 minutes at 50,000 x g
and filtered through a 0.8 µm filter to remove cell debris. 20 mL of glutathione-sepharose resin
was equilibrated in buffer C and mixed with cleared lysate for 30 minutes at 4 °C. The
suspension was drained, washed with 160 mL of buffer C supplemented with 10 g/L Triton X-
100, then washed with 140 mL of buffer C. Target protein was eluted in 40 mL of buffer C
containing 10 mM of reduced glutathione. 30 units of PreScission™ protease were added to
eluted protein and the mixture was dialyzed against 2 L of buffer D (100 mM NaH2PO4, 200 mM
NaCl, 5 mM DTT, pH 7.0) for one day at 4 °C. As the flow-through fraction from the
112
glutathione-sepharose resin contained an appreciable amount of target protein, the glutathione
affinity step was repeated on the flow-through fraction. 15 units of PreScission™ protease were
added to protein obtained from repeat purification and likewise dialyzed against 2 L of buffer D
for one day.
PreScission™ protease cleavage was continued for the second day in the freshly made 2 L of
buffer D, after which dialysates were passed through to glutathione-sepharose resin equilibrated
in buffer C. The flow-through fraction containing the WCBD1-6 fragment was then dialyzed
against 2 L of buffer E (50 mM NaH2PO4, 1 mM EDTA, 5 mM DTT, pH 7) overnight, loaded
onto a 20 mL bed volume Q-sepharose column and eluted with a 400 mL gradient of 0 to 1 M
NaCl in buffer E.
Fractions containing WCBD1-6 were pooled, 10% w/w glycerol and additional DTT to 20 mM
were added. Into this solution solid urea was dissolved to 5.2 M solution, incubated 30 minutes
on ice and dialyzed twice against 2 L of refolding buffer (buffer E with 10% v/v glycerol).
Refolded WCBD1-6 was further purified on Superdex200 HiLoad XK16/60 column (120 mL
bed volume) in 100 mM NaH2PO4, 200 mM NaCl, 1 mM DTT and 20 mM BME. Relevant
fractions were pooled and dialyzed against 1.1x concentration of NMR buffer F (50 mM
NaH2PO4, 150 mM arginine, 1 mM EDTA, 10 mM DTT, pH 6.5).
6.3.2 Copper loading of protein samples
Protein purification according to the above protocols typically yielded apo Atox1 containing no
copper, and ‘near-apo’ WCBD4-6, which can contain as much as 8% copper, where 100%
copper content would be equivalent to 3 copper atoms per protein molecule. Treatment with
copper-specific chelator BCS (bathocuproine disulfonic acid) followed by extensive dialysis did
not affect the copper content of WCBD4-6. Protein concentrations were confirmed with amino
acid analysis and copper content of every Atox1 and WCBD4-6 sample was determined using
the Cu(I)-assay described by Brenner and Harris (Brenner and Harris 1995). EPR spectroscopy
of the samples confirmed the absence of significant (<100 nM) concentrations of paramagnetic
Cu(II) from the samples (supplemental data) (Peisach and Blumberg 1974; Narindrasorasak,
Kulkarni et al. 2007). WCBD1-6 as purified from E. coli cells, according to ICP-AES analysis,
contains 0.35 copper atoms per protein molecule, or 0.35/6=0.058 copper per domain, if all
copper atoms were distributed evenly across all 6 domains. It has been previously reported that
113
all the copper-binding domains have comparable affinities for binding copper (Wernimont,
Yatsunyk et al. 2004) and that most are exchangeable except for one or two (Walker, Tsivkovskii
et al. 2002); for example domain 2 retains copper in excess BCS (Walker, Huster et al. 2004).
We therefore refer to Atox1, WCBD4-6 and WCBD1-6, as purified from E. coli cells minus the
metal removal step described in previous protocols (DiDonato, Hsu et al. 2000), as the apo
protein.
Copper-loading of WCBD4-6 was achieved according to established protocols (DiDonato, Hsu
et al. 2000) using a 10-fold molar excess CuSO4 with DTT acting as a reductant. Precipitation
was removed by centrifugation, and unbound copper was removed by dialysis in buffer A and
gel-filtration in NMR sample buffer B. WCBD4-6 NMR samples were 450 - 500 µM in 0.5 mL
of NMR sample buffer B containing 0.5 mM DSS, 0.05% azide, 10 mM benzamidine and 10%
(v/v) D2O, and were sealed in an argon-blanketed NMR tube. Copper-loaded (1:1 molar ratio)
Atox1 was prepared as was described for WCBD4-6.
WCBD1-6 was copper-loaded by adding 200 mM CuCl2 in the presence of 20 mM DTT to
achieve 5-fold excess of copper over total binding sites. Precipitation was removed by
centrifugation and the sample was dialyzed overnight into 1.1x NMR sample buffer F.
6.3.3 NMR spectroscopy
WCBD4-6 NMR spectra were acquired at 35 ºC on 4-channel Varian Unity Inova spectrometers
operating at 500 MHz and 600 MHz 1H frequencies. Apo and Cu(I)-bound WCBD1-6 NMR
spectra were acquired at 25 ºC on 4-channel Varian Unity Inova spectrometers operating at 500
MHz. HSQC spectra of apo WCBD4-6 and apo WCBD1-6 in NMR sample buffer B were also
acquired at 45 ºC at 500 MHz for comparative purposes. Data were processed using NMRPipe
(Delaglio, Grzesiek et al. 1995) and analyzed using NMRView (Johnson and Blevins 1994).
6.3.3.1 Assignment experiments
1HN, 15N, 13Cα, 13Cβ and 13CO resonance assignments were performed on 500 µM apo WCBD4-
6 using three-dimensional HNCO, HN(CA)CO, HNCA, HN(CO)CA, HNCACB,
HN(CO)CACB, and (H)CC(CO)NH-TOCSY experiments (Cavanagh, Fairbrother et al. 2007).
114
6.3.3.2 NMR Cu-titrations of WCBD4-6
The Cu(I)-loaded state was assigned by titrating small amounts of copper and following the
chemical shifts of cross-peaks. HSQC spectra were recorded at 500 MHz at 0, 200, 600, 1000,
1500 and 2000 µM CuSO4. The NMR sample tube was sealed and argon-blanketed after each
successive addition. To prevent dilution during the course of the titration the NMR sample was
concentrated as required to maintain a sample volume of approximately 0.5 mL and protein
concentration of 500 µM.
6.3.3.3 NMR titration of WCBD4-6 with Atox1
Titrations were carried out at 35 ºC using a 500 MHz spectrometer and 1H-15N TROSY spectra
were recorded at each point. The apo WCBD4-6 sample was at a concentration of 450 µM in a
450 µL volume. Apo Atox1 and Cu(I) Atox1 were added directly into the NMR tube containing
the apo WCBD4-6 sample. Spectra were recorded after each addition at 0, 5, 10, 20, 50, 100,
275, 510, 700, 850, 1500, 1700, and 2600 µM for apo Atox1, and at 150, 225, 450, 900, 1350,
and 3500 µM for Cu(I) Atox1. NMR samples were concentrated at several steps during the
course of the titrations to maintain signal intensity and sample volumes of approximately 0.5 mL.
To ensure that the interaction was fully saturated, titrations were performed to the point where no
additional spectral changes were observed, indicating that titration endpoints for the apo Atox1
and Cu Atox1 titration had been reached. For the apo Atox1 titration, there were no changes
beyond 1500 µM, and no changes beyond 1350 µM for the Cu(I) Atox1 titration.
6.3.3.4 NMR relaxation
15N T1, T1ρ and heteronuclear 1H-15N NOE measurements were performed as described
elsewhere (Farrow, Muhandiram et al. 1994; Korzhnev, Skrynnikov et al. 2002). Spectra of the
apo and Cu(I)-loaded states were recorded at 35˚C at 600 MHz for T1, T1ρ, and heteronuclear
NOE, and at 500 MHz for T1 and T1ρ. Longitudinal T1 and rotating frame T1ρ relaxation times
and their uncertainties were obtained from exponential fits of peak intensities in a series of 2D 1H-15N correlation spectra using the DFIT module of the DASHA program (Orekhov, Nolde et
al. 1995). The resonances for about 30% of residues were excluded from the analysis due to
spectral overlap. Transverse relaxation rates (R2) were calculated from R1 and R1ρ using the
equation R1ρ = R1cos2θ + R2sin2θ, where θ = arctan(ωSL/∆ω), R1=1/T1, and R1ρ =1/T1ρ, ∆ω is the
resonance offset from the spin-lock carrier, and ωSL is the spin-lock field strength (1779.4 Hz and
115
1811.6 Hz at 500 MHz and 1953.1 Hz and 1968.5 Hz at 600 MHz for the apo and Cu(I)-loaded
samples, respectively). 1H-15N steady state NOE values were determined from the ratio of the
intensities of the respective cross-peaks in 1H-15N correlation spectra obtained with and without
proton saturation. Errors for the NOE values were determined from signal to noise ratios in
spectra. To account for possible systematic errors in the experimental data minimal errors of 5%
and 0.05 were assumed for R1 and R2 rates and NOE values, respectively.
The overall rotational correlation times (τc) and the rotational diffusion tensors of the individual
domains of both the apo and Cu(I)-bound forms of WCBD4-6 were determined from 15N R1 and
R2 values using the program DASHA (Orekhov, Nolde et al. 1995), which uses the model-free
approach of Lipari and Szabo (Lipari and Szabo 1982). For each domain, R1 and R2 data at 600
MHz from residues in rigid parts of the domains (those with NOE>0.6) were globally fit to
isotropic and asymmetric anisotropic models of overall rotation (Orekhov, Nolde et al. 1995).
The former model includes one parameter of molecular overall rotational: τc, while the later
includes six parameters that describe anisotropic rotational diffusion: τc = 1/(2(Dx+Dy+Dz)),
Dx/Dz, Dy/Dz, where Dx, Dy and Dz are eigenvalues of the rotational diffusion tensor, and Euler
angles α, β, and γ that orient the tensor within the PDB frame. The selection of the appropriate
form of overall rotation using an F-test has shown that the anisotropic model is always required
to properly fit the data. The structures of apo domains-3-4 (PDB id: 2rop) and apo domains-5-6
(PDB id: 2ew9) were used for the fit of the rotational diffusion models for domain 4, and
domains 5 and 6, respectively.
6.3.3.5 Model-free analysis
Model-free analysis was performed on apo WCBD4-6 only due to a small inconsistency between
500 MHz and 600 MHz data collected on the Cu(I)-bound protein. The inconsistency may be
due to paramagnetic relaxation contributions from oxidation of Cu(I) to Cu(II). Bertini et al.
(Bertini, Luchinat et al. 2001) have documented the paramagnetic contribution of Cu(II) to 15N
R2. For a protein with a 10 ns rotation correlation time, the paramagnetic contribution from
Cu(II) at 5Å distance is 3000-20000/s to 1H R2, and ~30-200/s to 15N R2 (~100 times less), in a 1
mM protein sample containing as little as 100 nm Cu(II). Therefore in a 0.5 mM WCBD4-6
sample, as low as 50 nM Cu(II) may result in a 0.1/s contribution to 15N R2 values. Although
precautions were taken to keep the oxidation of copper to a minimum during sample preparation,
116
it could not be totally avoided once the NMR sample was placed in the magnet and experiments
begun.
Following the characterization of molecular overall rotation, the parameters of molecular
rotational diffusion tensors of individual domains of apo WCBD4-6 were fixed and used in
model-free analysis. The residues belonging to the flexible linkers were excluded from the
analysis. 15N R1, R2 and 1H-15N NOE data at two magnetic fields (up to five experimental data
per residue, see above) were fit using the following models of spectral density function including
different number of parameters of internal motions (shown in brackets): {S2}, { S2, Rex}, { S2, τe},
{ S2, τe, Rex}, and {Ss2, Sf
2, τs}, where S2 and Ss2, Sf
2 (S2 = Ss2, Sf
2) are generalized order
parameters, τe, τs are correlation times for the internal motions, and Rex is the exchange
contribution to transverse relaxation (at 500 MHz spectrometer) due to conformational exchange
on µs-ms time-scales. The order parameters reflect angular amplitudes of the internal motions of
the amide bond vectors, while the correlation times reflect the time-scale of the internal motions.
In the original Lipari-Szabo approach order parameters S2 and correlation times τe are used to
characterize fast picosecond time-scale dynamics (Lipari and Szabo 1982). The theory was
extended to account for internal motions on two time-scales: picosecond with the correlation
time τf and order parameter Sf2, and sub-nanosecond with correlation time τs and order parameter
Ss2 (Clore, Szabo et al. 1990). The model selection was performed based on the values of the χ2
target function obtained in fits of relaxation data; more complex models were selected if they led
to significant improvements in the fits (F-test confidence level of 0.25). Despite this relatively
loose criterion for acceptance of more complex models assuming dynamics on micro-millisecond
or sub-nanosecond time-scales, the data for most of the residues indicate restricted mobility on
the picosecond time-scale consistent with the simplest spectral density models (see below).
6.4 Results
6.4.1 Protein structure characterization
Triple resonance experiments were used to assign 88% of the non-proline 1HN, 15N, 13CO, 13Cα
and 13Cβ chemical shifts for WCBD4-6, residues T357-A632 of ATP7B. Assignments for
resonances (around the CXXC motif, in particular) were challenging due to degeneracy of the
sequence in the three domains leading to overlap of signals. Secondary chemical shifts of the
backbone 13C nuclei are in agreement with those expected from previously published secondary
117
structures of WCBD domains (Achila, Banci et al. 2006). Figure 6.1 shows the secondary
structure propensity (SSP) values that combine different chemical shifts into a single residue-
specific score of α-helical or β-strand population by calculating the weighted experimental
chemical shift difference from random coil relative to that expected for stable helix or strand
(Marsh, Singh et al. 2006). As expected, each domain has the βαββαβ fold seen in all heavy
metal-associated domains including the previously published structures of WCBDs (Walker,
Huster et al. 2004; Achila, Banci et al. 2006; Banci, Bertini et al. 2008). The various secondary
structure elements within each domain are numbered as follows: β1-L1-α1-L2-β2-L3-β3-L4-α2-
L5-β4. Residues 357-428 comprise domain 4, 429-485 the flexible linker, 486-556 domain 5
and 565-632 domains 6. The residues connecting β-strands 2 and 3 also show some α-helical
character but are part of a turn.
Figure 6.1: Secondary structure of apo WCBD4-6.
Cα and Cβ and Co backbone chemical shifts were used in the SSP program (Marsh, Singh et al. 2006) to generate a secondary structure propensity plot. In folded proteins, SSP values of +1 are expected for α-helical structure and -1 for β-strand structure. Domains 4, 5, and 6 are shown to have the βαββαβ fold characteristic of heavy metal-associated domains.
6.4.2 Effect of Cu(I) binding on WCBD4-6
HSQC spectra were recorded at each point during the titration of apo WCBD4-6 with CuSO4.
This series of spectra made it possible to assign 79% of the non-proline backbone resonances of
the Cu(I)-bound spectra and to detect chemical shifts in response to copper binding on a per
residue basis. Again, assignments around the CXXC motif were more difficult. The 1H and 15N
118
chemical shift differences between the apo and Cu(I)-bound protein are shown in Figure 6.2 as
the weighted average chemical shift differences, ∆δ = {[(∆H)2 + (∆N/5)2]/2} 1/2, where ∆H and
∆N are chemical shift differences for 1H and 15N, respectively. Copper-loading of WCBD4-6
results in small chemical shift changes that are localized to the residues surrounding the copper
binding sites, the CXXC motif, while the rest of the residues remain largely unaffected (Figure
6.2). These copper-dependent chemical shifts are observed for analogous residues across the 3
domains. Shifts observed outside the immediate region of the CXXC motifs may be suggestive
of transient interactions between the domains, consistent with the reported transfer of copper
from domain 4 to a separate construct of domains 5-6 (Achila, Banci et al. 2006). While a
quantitative comparison is not possible due to differences in temperature and buffer conditions,
the chemical shifts observed for domain 4 in the context of WCBD4-6 change similarly upon
copper binding to those observed for domain 4 in a domain 3-4 construct (Banci, Bertini et al.
2008). Likewise, chemical shifts observed for domain 5 and domain 6 in WCBD4-6 change
similarly upon copper binding to those seen for domain 5-6 in isolation (Achila, Banci et al.
2006). These observations suggest that when alone these domains bind copper in a similar
fashion to when present together in the context in which they occur naturally.
119
Figure 6.2: Cu(I) binding to WCBD4-6.
Chemical shift differences between apo WCBD4-6 and Cu(I) WCBD4-6 are shown as a function of residue as well as mapped on to the surfaces of domain 4 (PDB id: 2rop) and domains 5-6 (PDB id: 2ew9). The surface having the greatest chemical shift differences for each of domains 4, 5 and 6 is displayed. Data for domains 5 and 6 are shown in different orientations of the surface of domains 5-6, with the orientation chosen for the domain 5 data completely obscuring domain 6. The domain graphics were created using the molecular graphics program PyMOL (DeLano 2008), using the scripts color_b.py and data2bfactor.py (provided by Dr. Robert L. Campbell). Surfaces of domains 4, 5, and 6 are shown in red-white, green-white, and blue-white gradients, respectively. The gradient is from dark to light with darkest shades showing a chemical shift of 0.06 ppm or greater and white for residues experiencing a chemical shift of 0.02 ppm or less. Unassigned residues are colored gray. Cysteine residues from the CXXC motif are shown in yellow: Cys 370 and Cys 373 in domain 4, Cys 499 and Cys 502 in domain 5, and Cys 575 and Cys 578 in domain 6.
120
6.4.3 The flexible linker maintains its flexibility in both apo and Cu(I)-bound states
1H-15N steady-state NOEs of WCBD4-6 shown in Figure 6.3A indicate the presence of three
stable domains and a long highly flexible linker E429-V485 which shows low NOE values, sharp
resonances and chemical shifts characteristic of disordered protein regions (e.g. 1HN chemical
shifts around 8 ppm). This flexible linker joins domain 4 to domains 5-6 which are associated
much more closely and are separated by a very short (but also somewhat flexible, according to
NOE values) linker of only a few residues. The backbone chemical shifts measured for the
linker region between the domain 4 and domains 5-6 also indicate the lack of secondary structure
in this region (Figure 6.1). Low NOE values suggest that the linker retains its flexibility both in
the apo and Cu(I)-bound states (Figure 6.3A), and becomes even more mobile in the Cu(I)-bound
state, based on the somewhat lower NOE values obtained for residues in the linker region in the
Cu(I)-bound form of a protein (Figure 6.3B). As shown in Figure 6.3C, the resonances
originating from the linker also show a significant increase in intensity due to a narrower
linewidth when WCBD4-6 is Cu(I)-bound, in contrast to the ordered domains, which clearly
demonstrates the enhanced flexibility in the linker region. Note that this effect is observable for
WCBD4-6 samples which are loaded directly with copper or by copper titration.
121
Figure 6.3: The effect of Cu(I) binding to WCBD4-6 on the flexible linker.
1H-15N steady-state NOEs of apo WCBD4-6 and Cu(I) WCBD4-6 (A). An expansion of the linker regions shows lower NOE values in the Cu(I)-bound (black) than in the apo (grey) forms (B). The ratio of the peak intensities of the Cu(I)-bound to Cu(I)-free states of WCBD4-6 (C) shows that the flexible linker increases in flexibility in the Cu(I)-bound state. The dotted line (C) shows the peak intensity ratio = 1.
122
6.4.4 Copper(I) binding does not change the relative orientation of domains
The 15N R1 and R2 relaxation rates and 1H-15N NOE values are effective reporters of overall
rotation diffusion of the molecule and intermolecular motions of the internuclear N-H vectors,
routinely used for characterization of the conformational flexibility of different protein regions.
For relatively rigid residues, the measured relaxation data primarily reflect the overall tumbling
of the molecule in solution. To further analyze the effects of Cu(I) binding to WCBD4-6, 15N-
spin relaxation data for apo WCBD4-6 and Cu(I) WCBD4-6 were measured. 15N R1 and R2 rates
and 1H-15N NOE values for the apo and Cu(I)-loaded WCBD4-6, plotted in Figure 6.4A, 6.4B,
and 6.3A, respectively, were used for characterization of fully asymmetric anisotropic rotational
diffusion of individual domains of WCBD4-6 (Figure 6.5) and for the analysis of local
intermolecular dynamics of the domains (Figure 6.6) (as described in Materials and Methods).
Figure 6.4C shows correlation times (τc) obtained from 15N R2/R1 ratios for individual residues of
apo and Cu(I)-bound WCBD4-6. Figure 6.5 shows the τc and the parameters of the rotational
diffusion tensor obtained from fitting R1 and R2 (at 600 MHz) using the fully anisotropic model.
Domain 4 has the greatest motional freedom and a τc of ~ 6.3 ± 0.03 ns (N = 43) (Figure 6.5),
where N is the number of residues used in calculations of the rotational diffusion tensor. The τc
value of domain 4, while reflecting significant independent tumbling of this domain, is likely
increased due to restriction of motion by the long flexible linker connecting it to domain 5; the
effect of flexible linkers on the relaxation behavior of connected domains has recently been
investigated by Wright et al. (Bae, Dyson et al. 2009). Although the closely associated domains
5 and 6 have correlated motion due to the short linker joining them, they have flexibility with
respect to one another based on differences in their correlation times, 9.7 ± 0.08 ns (N = 32) for
domain 5 and 8.2 ± 0.07 ns (N = 32) for domain 6 (see Figure 6.5). This is in contrast to a
previous report that the isolated domain 5-6 construct tumbles as a rigid dumbbell (Achila, Banci
et al. 2006) but in agreement with molecular dynamic simulations suggesting the N-terminal
metal-binding domains of related ATPases have motional freedom to reorient with respect to one
another depending on the length of their linkers (Sharma and Rosato 2009). Each of the 3
domains and the linker region display distinct rotation correlation times in the apo state (Figure
6.4C), suggesting that the domains reorient at different rates, are flexible with respect to each
other, and do not tumble as a single large protein. The R1, R2 and τc for the domains do not show
large changes between the apo and Cu(I)-loaded WCBD4-6 (Figure 6.4C), indicating that the
123
overall dynamics of WCBD4-6 in the apo and Cu(I)-bound states remain the same in the two
states.
Figure 6.4: Dynamic properties of the WCBD4-6 residues in the apo and Cu(I)-bound
states.
R1 and R2 for WCBD4-6 residues in the apo (A) and Cu(I)-bound (B) states, in s-1. Local effective correlation time (τc) calculated from 15N R2/R1 ratios for individual amide groups in the apo and Cu(I)-bound state (C). Data were collected at 500 MHz (black) and 600 MHz (grey).
124
Rotational diffusion tensors of the apo and Cu(I)-bound domains 4, 5 and 6 are shown in Figure
6.5. Tensors are nearly axially symmetric (Dx/Dz ≈ Dy/Dz), and the parameters of rotational
anisotropy (Dx/Dz and Dy/Dz) are similar in both the apo and the Cu(I)-bound states. The
directions of principal axes of the rotational diffusion tensors are similar in both the apo and the
Cu(I)-bound states; the directions of Z axes between apo and Cu(I)-bound forms differ by no
more than 23 degrees for domain 4, 29 degrees for domain 5 and 28 degrees for domain 6.
Together with the similar (although not identical) overall rotational correlation times for
individual domains in apo and Cu(I)-bound states, the data show that the relative orientations and
mobilities of the domains in WCBD4-6 do not change significantly upon binding copper.
Figure 6.5: Rotational diffusion tensors of the apo and Cu(I)-bound domains 4, 5 and 6
in the apo and Cu(I)-bound states.
Directions of principle axes of the diffusion tensors in the apo (black) and Cu(I)-bound (red) states mapped on to the structures of domains 4, 5 and 6. Rotational correlation times and parameters of rotational anisotropy in both states are provided below each domain.
15N relaxation data were analyzed by model-free analysis assuming a fully asymmetric rotational
diffusion tensor. The model-free parameters S2, τe, and Rex for apo WCBD4-6 are shown in
Figure 6.6 A-C. Most residues from folded domains have order parameters S2 > 0.6, indicating
that the internal motions in the domains are restrained on the picosecond to nanosecond time-
scales. Slightly lower order parameters of S2 ≈ 0.5 where observed for 4 residues in domain 4 -
I408 and E423 in loops L4 and L5, respectively, I390 and E412 in β2 and α2, respectively.
125
Motions on pico- to nanosecond time-scales with correlation times (τe or τs) in the range 0.07-2.4
ns were detected for 17 residues in domain 4, 20 residues in domain 5, and 9 residues in domain
6. Conformational exchange terms (Rex) of up to 2.6/s pointing to micro- to millisecond time-
scale dynamics, were detected for 8 residues in domain 4, 11 residues in domain 5, and 6
residues in domain 6. The first residue of the defined boundaries for domain 5, A486, has a low
order parameter (S2 of 0.2) and higher internal correlation time (τs = 1.3 ns), which demonstrates
that it is part of the long flexible linker connecting domains 4 and 5. Note that model-free
parameters for the Cu(I)-bound state are not reported due to inconsistencies in the data (see
Methods).
Figure 6.6: Model-free parameters for the local motion of the backbone amide N-H of
domains 4, 5 and 6 in the apo state.
S2, the squared generalized order parameter (A), τe, internal motion correlation time (in nanoseconds) (B), and Rex, exchange rate (in reciprocal seconds) (C) are plotted as a function of the residue number. Values are not shown for residues whose relaxation could not be measured accurately (due to signal overlap) or for the flexible linker region. The S2 (A) and τe (B) data shown for residues fit using the {Ss
2, Sf2, τs} model are S2 =
Ss2Sf
2 and τs, respectively.
126
6.4.5 Protein-protein interactions with Atox1
HSQC spectra can be used to map protein-protein interaction surfaces at residue-level resolution,
by monitoring the perturbation in the positions or intensities of the resonances upon titration of a
binding partner. Apo Atox1 and Cu(I) Atox1 were titrated into apo WCBD4-6 with final spectra
representing fully saturated complexes (see Methods). The final titration points contain an excess
of Atox1 to WCBD4-6, 5.8:1 and 7.8:1 for apo Atox1 and Cu(I) Atox1, respectively. The
titration with apo Atox1 into apo WCBD4-6 shows that an interaction is detectable in the
absence of Cu(I), or in the presence of the very low levels of Cu(I) (see Methods). Chemical shift
differences between apo WCBD4-6 and each saturated complex (final titration point) are shown
in Figure 6.7 are weighted average chemical shift differences calculated in a similar manner as
those shown in Figure 6.2 for Cu(I) binding. Significant chemical shift changes are observed for
domain 4 and to a lesser extent for domain 5 (Figure 6.7A). The linker region and domain 6
show no chemical shift changes. This interaction pattern is consistent with previously published
results showing a complex in fast exchange between Atox1 and WCBD4 in the presence of Cu(I)
(Achila, Banci et al. 2006). This is also in agreement with recent work on titration of domains 1-
6 with Cu(I) Atox1 (Banci, Bertini et al. 2009) showing a differential interaction of domain 4
versus domains 5 and 6, although that study did not show binding in the absence of copper. Our
chemical shift changes for WCBD4-6 interaction with apo Atox1 (Figure 6.7A) are only
somewhat smaller than those measured for Cu(I) Atox1 (Figure 6.7B), suggesting that the
interaction between Atox1 and WCBD4-6 is driven primarily by protein-protein interactions and
that the presence of Cu(I), while not required, does lead to enhanced binding.
When WCBD4-6 binds Cu(I) Atox1 (Figure 6.7B), presumably enabling Cu(I) transport, the
residues affected are very different from residues involved in the binding of free copper (Figure
6.2). These data suggest that the targeting of copper by Atox1 to domain 4 is driven by favorable
protein-protein interactions; this is further supported by the interaction of Atox1 even in the apo
state with domain 4 (Figure 6.7A). Our findings are consistent with the data for MNK4-6 and
MNK1-6 showing that domain 4 is the domain that receives copper from Atox1 (Banci, Bertini
et al. 2005; Banci, Bertini et al. 2007).
127
Figure 6.7: Interaction of apo WCBD4-6 with apo and Cu(I) Atox1.
Chemical shift differences between apo WCBD4-6 and apo WCBD4-6 in a fully saturated complex with apo Atox1 (molar ratio 1:5.8) (A). Chemical shift differences between apo WCBD4-6 and apo WCBD4-6 in a fully saturated complex with Cu(I) Atox1 (molar ratio 1:7.8) (B). Differences between initial and endpoints of the titration are shown as a function of residue as well as mapped on to the surfaces of the structures of domain 4 and domains 5-6 as for Figure 6.2, with the orientation chosen for the domain 5 data obscuring domain 6 in both (A) and (B) and the orientation for domain 6 data obscuring domain 5 in (A).
128
Figure 6.8: WCBD1-6 spectra in apo and Cu(I)-bound states.
HSQC spectra (45 ºC) of apo WCBD4-6 (blue) and apo WCBD1-6 (black) show overlap in a significant number of peaks (A). TROSY spectra (25 ºC) of WCBD1-6 in the apo (black) and Cu(I)-bound states (red), is typical of smaller molecules with high mobility (B).
129
6.4.6 Copper(I) binding to WCBD1-6 does not appear to induce a significant change in the relative mobility of the domains
Comparison of spectra of apo WCBD1-6 to those of apo WCBD4-6 demonstrates overlap of
many of the resonances (Figure 6.8A), indicative of limited interactions between domains 1-3
and 4-6. The spectra for domains 1-6 in the apo and Cu(I)-bound states are overlaid in Figure
6.8B, with the large degree of overlap suggestive of similar overall behavior in the absence and
presence of copper. Visual inspection of WCBD1-6 spectra in both states reveals sharp peaks
having lineshapes typical of smaller proteins with reasonably rapid overall tumbling. The
assignment of NMR resonances of N-terminal copper-binding domains 1-6 of ATP7B, a 72 kDa
protein, by Banci et al (Banci, Bertini et al. 2007; Banci, Bertini et al. 2009) without the need for
specialized methods for “large” protein NMR (Tugarinov, Hwang et al. 2004; Tugarinov,
Kanelis et al. 2006) also argues for significant motional flexibility. These observations are
inconsistent with WCBD1-6 tumbling as a single unit in either apo or copper-bound states.
Correlated motions of domains 5 and 6, consistent with the results for WCBD4-6, are expected
and it is possible that motions of additional domain are also correlated. Overall, however, the
results are supportive of a model in which the binding of copper to WCBD1-6 leaves the
domains largely flexible with respect to one another.
6.5 Discussion
Many proteins contain multiple binding domains connected by linkers. These multiple domains
can participate in inter-domain motions that play a key role in molecular recognition and
regulation. The WCBD N-terminal region of ATP7B contains six copper-binding domains
joined through linkers of various lengths and flexibility. These domains are involved in many
aspects of copper metabolism. WCBD receives copper through interaction with the copper
chaperone Atox1 (Hamza, Schaefer et al. 1999; Larin, Mekios et al. 1999; Hamza, Prohaska et
al. 2003) and binds copper in a cooperative manner (DiDonato, Narindrasorasak et al. 1997;
DiDonato, Hsu et al. 2000). There have been many studies alluding to a conformational change
or transition in the N-terminal region domains of both the Wilson (DiDonato, Narindrasorasak et
al. 1997; DiDonato, Hsu et al. 2000) and the Menkes ATPase (DiDonato, Narindrasorasak et al.
1997; Ralle, Cooper et al. 1998; Cobine, George et al. 2000; Ralle, Lutsenko et al. 2004) upon
copper binding. In order to more specifically characterize the conformational transitions that
have been suggested to take place upon binding copper and to detect interactions with protein
130
partners that can ultimately regulate copper transport by ATP7B, we have employed NMR
spectroscopy to probe the structural and dynamic properties of WCBD4-6, metal binding
domains 4 to 6 of the N-terminal metal-binding domain of ATP7B, as well as to describe
WCBD1-6, the full N-terminal domain.
Our results reveal that the three domains in WCBD4-6 have independent mobility in solution.
Domain 4 tumbles independently and is connected by a long flexible linker to domains 5-6
which are joined only by a very short linker and tumble with some flexibility with respect to
each other (Figure 6.3A and 6.3B). Analysis of the correlation times for WCBD4-6 (Figure 6.4
and Figure 6.5) suggests that domains 4, 5 and 6 tumble as separate entities in the apo state and
Cu(I)-bound state, with distinct τc values of 6.3 ± 0.03 ns, 9.7 ± 0.08 ns and 8.2 ± 0.07 ns,
respectively (Figure 6.5). The closely associated domains 5 and 6 have different correlation
times (9.7 ± 0.08 ns and 8.2 ± 0.07 ns, respectively; see Figure 6.5), and therefore have
flexibility with respect to one another which is in contrast to that reported previously for the
isolated domain 5-6 construct (Achila, Banci et al. 2006). The correlation times for domains 4
and 5 are larger than those previously reported for isolated domain 6 and 5-6 (4.5 ns and 9.1 ns,
respectively, at 25 ºC (Achila, Banci et al. 2006)), with the long flexible linker connecting
domain 4 to domain 5 possibly leading to an increase in the correlation times (Bae, Dyson et al.
2009). This linker effect may also contribute to the reported increases in R2 values for residues
of domain 5 of the ATP7A domains 4-6 (Banci, Bertini et al. 2005).
Our relaxation analysis of WCBD4-6 shows no evidence in support of significant copper-
dependent conformational change due to the rearrangement or reorientation of domains 4, 5 and
6 relative to each other (Figure 6.5). The fairly sharp resonances of WCBD1-6 in the apo and
Cu(I)-bound states (Figure 6.8B) point to domains that do not tumble as a single unit but instead
are somewhat flexible with respect to each other, as in WCBD4-6. Importantly, copper binding
to WCBD1-6 does not induce the formation of a larger uniformly tumbling globular unit as there
is no drastic change in the overall dynamic character of the N-terminal domains upon binding
Cu(I). Secondary and tertiary conformational changes reported by DiDonato et al. (DiDonato,
Narindrasorasak et al. 1997; DiDonato, Hsu et al. 2000) in WCBD1-6 by the binding of copper
are probably due to the effect of copper-binding on residues within each domain as detected by
the chemical shifts changes in our Cu(I)-binding experiments on domains 4-6. In addition we
131
find that, even in the absence of copper, Atox1 interacts preferentially with domain 4, which has
greater motional freedom than domains 5 and 6.
Based on these data and the work of others on the metal-binding domains of ATP7B and ATP7A
(Banci, Bertini et al. 2007; Banci, Bertini et al. 2009; Sharma and Rosato 2009; Rodriguez-
Granillo, Crespo et al. 2010), the flexibility between the domains in the apo and Cu(I)-bound
state appears to be a key feature of the N-terminal region of multi-domain Cu-ATPases and
copper binding. Domain-domain flexibility in this region provides motional freedom to the
domains that, by increasing their accessibility, could facilitate protein-protein interactions with
the copper chaperone Atox1 in order to receive copper, to subsequently transfer copper to the
other N-terminal copper-binding domains, and finally to the transmembrane site for transport
across the membrane. The idea of a transition from disorder and flexibility to order upon binding
has long been a central paradigm in protein structure and dynamics; in fact, the stabilization of
tertiary structure by metal binding is a key feature of metalloregulatory proteins (O'Halloran
1993). Nevertheless, there are examples of retention of dynamics or even increases in motion
upon binding (Forman-Kay 1999; Mittag, Kay et al. 2010). This appears to be another such
example, with retained flexibility and relative conformation freedom between the domains,
creating an ensemble of states that may facilitate the multi-step process of copper transport by
copper-transporting ATPases.
132
Chapter 7
Summary and Future Directions
133
7 Summary and Future directions
7.1 Summary
The Wilson disease copper transporting ATPase plays a central role in maintaining copper
homeostasis. This transmembrane protein has a large cytoplasmic N-terminus consisting of six
tandem copper-binding domains that bind one copper each. A variety of approaches have been
used in this thesis in order to characterize the structural consequences of copper-binding to the
N-terminal domains, their role in the transport of copper by the ATPase, and their interactions
with the copper chaperone Atox1.
Homology modelling of the ATP7B transporter as a whole has provided a 3D structural model to
aid in the understanding of the roles and possible interactions of the cytoplasmic domains in
transport of copper by P-type Cu-ATPases. This model is also the basis for investigation of
Cu(I) binding to the CPC motif of the transmembrane site as a possible intermediate step in
copper transport from the cytoplasmic copper-binding domains across the membrane, and the
important role played by the conserved proline in a highly specific geometry for CPC-Cu(I)
binding.
Finally the N-terminal domains have been examined at the molecular level to provide residue
specific information on copper binding to the domains, their interaction with the copper
chaperone Atox1 as well as characterization of the overall and internal dynamics of the domains
and the flexible linker in the apo and copper-bound states. It was hypothesized that the
secondary and tertiary structural transitions induced in CD spectra of WCBD upon binding
copper were due to copper-induced changes in the structural conformation of WCBD, such as
changes in domain-domain interaction or orientation, that would allow for the cooperative
binding of copper as observed by the 65Zn(II) blotting experiments . Our relaxation analyses of
WCBD4-6 show only subtle changes in the diffusion tensors of the free and copper-bound state,
and show no evidence in support of copper-dependent conformational change due to domain-
domain interactions or reorientation of domains 4, 5 and 6 relative to each other. The flexible
linkers and the mobility of the domains may allow for the formation of transiently populated
tertiary contacts, which may be enough to facilitate cooperativity.
134
7.2 Future directions
The exact role of the N-terminal copper-binding domains WCBD1-6 in the transport of copper is
still not fully understood, with many question that remain to be answered particularly with
regards to cooperative binding of copper and conformational changes induced by copper binding
to the domains. This body of work has attempted to address some of these questions by
characterizing WCBD4-6, and found no evidence in support of copper-dependent conformational
change due to the rearrangement or reorientation of domains relative to each other. Although
WCBD4-6 appears to be a good model of WCBD1-6, definitive extension of these conclusions
to WCBD1-6 can be made only by analyzing WCBD1-6 itself. Factors such as spectral
crowding, the presence of long flexible linkers, and high sequence similarity among the copper
domains have made NMR analysis of WCBD4-6 very difficult. NMR characterization of
WCBD1-6 will be even more challenging, but will be greatly aided by the ground work laid by
the WCBD4-6 studies.
NMR analysis of WCBD1-6 will help in characterizing both the cooperative binding of copper,
and the copper-induced conformational changes detected by CD spectroscopy. NMR
characterization of WCBD1-6 will also aid to identify the molecular and mechanistic link
between copper-dependent kinase-mediated phosphorylation and trafficking. The kinase-
mediated phosphorylation and the intracellular localization of ATP7B (Vanderwerf, Cooper et al.
2001) and ATP7A (Voskoboinik, Fernando et al. 2003) are regulated by copper, suggesting a
regulatory link between phosphorylation and trafficking. In ATP7B, this phosphorylation is on
serine residues and is distinct from the acyl-phosphate site at the invariant aspartate in the
DKTGT motif (Vanderwerf, Cooper et al. 2001) and ATP7A (Voskoboinik, Fernando et al.
2003). Mass spectrometry and limited proteolysis of the N-terminal domains has identified the
linker between domains 3 and 4 as the target of a kinase-mediated copper-dependent
phosphorylation (Bartee, Ralle et al. 2009), suggesting that binding of copper results in structural
rearrangements in the N-terminus of ATP7B, and the exposure of phosphorylation sites within
the domain 3-4 linker region.
Assignment of WCBD1-6 will require a similar set of 3D assignment experiments, relaxation,
and RDC. Separate resonance assignment of the apo and the copper-bound states are necessary
for the interpretation of Cu(I) titration and relaxation experiments, and on the intermediates that
135
are formed when the protein is partially copper-bound. Resonance assignment of WCBD1-6 in
the phosphorylated and non-phosphorylated states will identify the phosphorylated residues and
the conformational changes that facilitate access and phosphorylation by the kinase. These
experiments will not only help to differentiate the role of domains in WCBD1-6 and WCBD4-6
but they will also help to explain why the dynamical characterization of WCBD4-6 presented in
this body of work shows no evidence of domain association or reorientation while other studies
suggest conformational change. DiDonato et al. observed cooperative binding of copper and
copper-indueced conformational change in WCBD1-6. Bartee et al. reported on copper-induced
conformational changes that facilitate kinase-mediated phosphorylation on the domain 3-4
linker. Residual dipolar coupling experiments will be useful for the measurement of domain-
domain reorientations.
Characterization of the protein interactions involving the WCBD are of considerable importance
for understanding Cu-transport and regulation of the protein. Two potentially important
interacting protein partners are COMMD1 (Tao, Liu et al. 2003) which exerts its regulatory role
in copper homeostasis through the regulation of ATP7B stability (de Bie, van de Sluis et al.
2007), and DCTN4 (Lim, Cater et al. 2006) which is involved in vesicular trafficking and may
be a key regulatory step in the copper-induced trafficking of ATP7B to vesicles traveling to the
cell membrane and the subsequent removal of excess copper into bile. NMR titration
experiments with potential binding partners such as COMMD1 and DCTN4, the copper
chaperone Atox1, and the other cytoplasmic domains of ATP7B with WCBD1-6 will also yield
information on the interactions involving the N-terminal domains. Relaxation experiments will
further characterize the effect of these interactions on the dynamics of the copper-binding
domains.
7.3 Conclusion
Since the cloning of the Wilson disease genes in 1993, a great deal of progress has been made in
the characterization of the copper-transporter associated with this disease. The activity of the
Cu-ATPases is tightly regulated through complex interactions between modulatory proteins and
in response to signaling pathways. Copper transport is modulated through conformational
changes in inter-domain interactions. The focus of this thesis has been on the detailed structural
136
and dynamical characterization of the Wilson disease ATPase and its N-terminal copper-binding
domains (WCBD).
Although the WCBD can bind a variety of transition metals including copper with varying
affinities, previous CD and 65Zn(II) blotting studies had suggested that the binding of copper to
WCBD may lead to unique structural consequences. CD spectroscopic analyses of copper-
induced conformational changes in WCBD indicated that both secondary and tertiary structural
changes take place upon copper binding. Competition 65Zn(II) blotting analyses showed a sharp,
reproducible transition only with copper as the competitor, suggesting that copper binding has
some degree of cooperativity. To better characterize the structural consequences of copper-
binding to the N-terminal domains, model-free analysis of the NMR relaxational properties of
WCBD4-6 was performed to test the hypothesis that copper binding to the WCBD induced
structural conformation changes, such as changes in domain-domain interactions or orientation,
which would allow for the cooperative binding of copper.
Cooperativity requires communication between the subunits of the binding protein. Once one
subunit binds its substrate, it becomes progressively easier for the other subunits to bind the
substrate. Each increment of ligand concentration results in a little more binding than the
previous increment did. The net effect is a sigmoidal binding curve, and the upward bend of the
curve means that the system is responding in a collective or cooperative fashion.
Communication between the subunits allows the whole protein to behave collectively in a way
that non-cooperative, truly independent subunits would not.
There are two major models for cooperative ligand binding, the symmetrical MWC model
described by Monod, Wyman, and Changeux (Monod, Wyman et al. 1965), and the sequential
KNF model put forth by Koshland, Nemethy, and Filmer (Koshland, Nemethy et al. 1966). In
both of these models enzyme subunits exist in one of two conformations, tense (T) or relaxed
(R), which bind substrate more readily than the tense state. The two models differ most in their
assumptions about subunit interaction and the preexistence of both states. In the MWC
symmetry-style model, if a ligand binds to one subunit, all subunits undergo the R↔T transition
simultaneously; there are no "hybrid" protein molecules with some subunits in R state and some
in T state. The subunits are connected in such a way that a conformational change in one subunit
is necessarily conferred to all other subunits, thus all subunits must exist in the same
137
conformation. The model further holds that, in the absence of any ligand, the equilibrium
favours one of the conformational states, T or R. The equilibrium can be shifted to the R or T
state through the binding of one ligand.
The KNF sequential-style model does allow for "hybrid" 4o structures, with some subunits in R
state and some in T state. Ligand binding can induce a change in conformation of an individual
subunit in a multimeric protein without all the rest of the subunits switching conformations. The
subunits are not connected in such a way that a conformational change in one induces a similar
change in the others, thus, all subunits need not exist in the same conformation. Ligand binding
is via an induced fit protocol. Ligand binding occurs via an induced fit, converting a subunit
from the tense to relaxed state. Conformational changes are not propagated to all subunits.
Instead, ligand-binding at one subunit only slightly alters the structure of adjacent subunits,
increasing ligand affinity in adjacent subuntis.
The existing data does not differentiate between these models. Cooperative behaviors of
different proteins are best described by different models, with some being best modeled by a
blend of the two models; it is possible that none of these models are applicable. While it is clear
from the 65Zn(II) blotting studies that copper-binding is cooperative, the mechanism behind this
cooperativity is not clear from the structural studies. The current findings are not in support of
copper-dependent conformational change due to domain-domain interactions or reorientation of
domains relative to each other. The results of the model-free analysis show only slight changes
in the rotational diffusion tensors of the domains upon binding copper. What is clear is that there
is no cooperative structural change in WCBD4-6.
Currently, there is no structural model that explains how copper-binding is cooperative and how
the binding of one copper at one domain is communicated to the remaining copper-binding
domains in WCBD. Is it possible to have cooperative copper-binding without communication
between the subdomains? The flexible linkers and the mobility of the domains do not exclude
the possibility of communication between the domains through the formation of transiently
populated tertiary contacts, which may be enough to facilitate cooperativity in the full length
WCBD. And, is it possible to have cooperativity without conformational change? More
significant changes in the rotational diffusion tensor indicative of the types of conformational
change seen with cooperative binding may require the presence of more than just domains 4, 5
138
and 6. The subtle changes detected for domains 4, 5 and 6 may translate to more significant
domain-domain interactions or reorientation when all six domains are present. More information
is needed on the nature of the intermediates that are formed when the cooperative protein is
partially saturated, to provide a structural model that shows how copper-binding is cooperative in
the full length WCBD. This work has laid a foundation for the study of the Wilson N-terminal
copper-binding domains as well as other biologically relevant multi-domain metal-binding
proteins. Advances in the understanding of these physiologically important metal binding
proteins will provide a clearer picture of their crucial importance in metal transport and their
related disorders.
139
Bibliography
Abragam, A. (1961). The Principles of Nuclear Magnetism. Oxford, Clarendon Press.
Achila, D., L. Banci, I. Bertini, J. Bunce, S. Ciofi-Baffoni and D. L. Huffman (2006). "Structure of human Wilson protein domains 5 and 6 and their interplay with domain 4 and the copper chaperone HAH1 in copper uptake." Proc Natl Acad Sci U S A 103(15): 5729-34.
Adams, M. D., S. E. Celniker, R. A. Holt, C. A. Evans, J. D. Gocayne, P. G. Amanatides and e. al. (2000). "The genome sequence of Drosophila melanogaster." Science 287(5461): 2185-95.
Arnesano, F., L. Banci, I. Bertini, S. Ciofi-Baffoni, E. Molteni, D. L. Huffman and T. V. O'Halloran (2002). "Metallochaperones and metal-transporting ATPases: a comparative analysis of sequences and structures." Genome Res 12(2): 255-71.
Axelsen, K. B. and M. G. Palmgren (1998). "Evolution of substrate specificities in the P-type ATPase superfamily." J Mol Evol 46(1): 84-101.
Bae, S. H., H. J. Dyson and P. E. Wright (2009). "Prediction of the rotational tumbling time for proteins with disordered segments." J Am Chem Soc 131(19): 6814-21.
Bagiyan, G. A., I. K. Koroleva and N. V. Soroka (1978). Russ J Inorg Chem 23: 1337-1341.
Baker, D. and A. Sali (2001). "Protein structure prediction and structural genomics." Science 294(5540): 93-6.
Banci, L., I. Bertini, F. Cantini, C. T. Chasapis, N. Hadjiliadis and A. Rosato (2005). "A NMR study of the interaction of a three-domain construct of ATP7A with copper(I) and copper(I)-HAH1: the interplay of domains." J Biol Chem 280(46): 38259-63.
Banci, L., I. Bertini, F. Cantini, N. Della-Malva, M. Migliardi and A. Rosato (2007). "The different intermolecular interactions of the soluble copper-binding domains of the menkes protein, ATP7A." J Biol Chem 282(32): 23140-23146.
Banci, L., I. Bertini, F. Cantini, C. Massagni, M. Migliardi and A. Rosato (2009). "An NMR study of the interaction of the N-terminal cytoplasmic tail of the Wilson disease protein with copper(I)-HAH1." J Biol Chem 284(14): 9354-60.
Banci, L., I. Bertini, F. Cantini, A. C. Rosenzweig and L. A. Yatsunyk (2008). "Metal binding domains 3 and 4 of the Wilson disease protein: solution structure and interaction with the copper(I) chaperone HAH1." Biochemistry 47(28): 7423-9.
Banci, L., I. Bertini, S. Ciofi-Baffoni, M. D'Onofrio, L. Gonnelli, F. C. Marhuenda-Egea and F. J. Ruiz-Duenas (2002). "Solution structure of the N-terminal domain of a potential copper-translocating P-type ATPase from Bacillus subtilis in the apo and Cu(I) loaded states." J Mol Biol 317(3): 415-29.
Banci, L., I. Bertini, S. Ciofi-Baffoni, D. L. Huffman and T. V. O'Halloran (2001). "Solution structure of the yeast copper transporter domain Ccc2a in the apo and Cu(I)-loaded states." J Biol Chem 276(11): 8415-26.
140
Banci, L., I. Bertini, R. Del Conte, S. Mangani and W. Meyer-Klaucke (2003). "X-ray absorption and NMR spectroscopic studies of CopZ, a copper chaperone in Bacillus subtilis: the coordination properties of the copper ion." Biochemistry 42(8): 2467-74.
Bartee, M. Y., M. Ralle and S. Lutsenko (2009). "The loop connecting metal-binding domains 3 and 4 of ATP7B is a target of a kinase-mediated phosphorylation." Biochemistry 48(24): 5573-81.
Beard, S. J., R. Hashim, J. Membrillo-Hernandez, M. N. Hughes and R. K. Poole (1997). "Zinc(II) tolerance in Escherichia coli K-12: evidence that the zntA gene (o732) encodes a cation transport ATPase." Mol Microbiol 25(5): 883-91.
Bertini, I., C. Luchinat and G. Parigi (2001). Solution NMR of Paramagnetic Molecules Current Methods in Inorganic Chemistry. Amsterdam, Elsevier Science. 2: 84 (Table 3.1).
Bertini, I. and A. Rosato (2008). "Menkes disease." Cell Mol Life Sci 65(1): 89-91.
Bissig, K. D., H. Wunderli-Ye, P. W. Duda and M. Solioz (2001). "Structure-function analysis of purified Enterococcus hirae CopB copper ATPase: effect of Menkes/Wilson disease mutation homologues." Biochem J 357(Pt 1): 217-23.
Booth, V., D. W. Keizer, M. B. Kamphuis, I. Clark-Lewis and B. D. Sykes (2002). "The CXCR3 binding chemokine IP-10/CXCL10: structure and receptor interactions." Biochemistry 41(33): 10418-25.
Brandl, C. J. and C. M. Deber (1986). "Hypothesis about the function of membrane-buried proline residues in transport proteins." Proc Natl Acad Sci U S A 83(4): 917-21.
Brenner, A. J. and E. D. Harris (1995). "A quantitative test for copper using bicinchoninic acid (vol 226, pg 80, 1995)." Analytical Biochemistry 230(2): 360.
Brewer, G. J., G. M. Hill, A. S. Prasad, Z. T. Cossack and P. Rabbani (1983). "Oral zinc therapy for Wilson's disease." Ann Intern Med 99(3): 314-9.
Bull, P. C., G. R. Thomas, J. M. Rommens, J. R. Forbes and D. W. Cox (1993). "The Wilson disease gene is a putative copper transporting P-type ATPase similar to the Menkes gene." Nat Genet 5(4): 327-37.
Bunce, J., D. Achila, E. Hetrick, L. Lesley and D. L. Huffman (2006). "Copper transfer studies between the N-terminal copper binding domains one and four of human Wilson protein." Biochim Biophys Acta 1760(6): 907-12.
Cater, M. A., J. Forbes, S. La Fontaine, D. Cox and J. F. Mercer (2004). "Intracellular trafficking of the human Wilson protein: the role of the six N-terminal metal-binding sites." Biochem J 380(Pt 3): 805-13.
Cater, M. A., S. La Fontaine and J. F. Mercer (2007). "Copper binding to the N-terminal metal-binding sites or the CPC motif is not essential for copper-induced trafficking of the human Wilson protein (ATP7B)." Biochem J 401(1): 143-53.
Cater, M. A., S. La Fontaine, K. Shield, Y. Deal and J. F. Mercer (2006). "ATP7B mediates vesicular sequestration of copper: insight into biliary copper excretion." Gastroenterology 130(2): 493-506.
Cavanagh, J., W. J. Fairbrother, A. G. Palmer, III, M. Rance and N. J. Skelton (2007). Protein NMR spectroscopy: Principles and practice. San Diego, Academic Press: 613-655.
141
Chelly, J., Z. Tumer, T. Tonnesen, A. Petterson, Y. Ishikawa-Brush, N. Tommerup, N. Horn and A. P. Monaco (1993). "Isolation of a candidate gene for Menkes disease that encodes a potential heavy metal binding protein." Nat Genet 3(1): 14-9.
Christodoulou, J., D. M. Danks, B. Sarkar, K. E. Baerlocher, R. Casey, N. Horn, Z. Tumer and J. T. Clarke (1998). "Early treatment of Menkes disease with parenteral copper-histidine: long-term follow-up of four treated patients." Am J Med Genet 76(2): 154-64.
Clore, G. M., A. Szabo, A. Bax, L. E. Kay, P. C. Driscoll and A. M. Gronenborn (1990). "Deviations from the simple two-parameter model-free approach to the interpretation of nitrogen-15 nuclear magnetic relaxation of proteins." J Am Chem Soc 112(12): 4989-4991
Cobbold, C., S. Ponnambalam, M. J. Francis and A. P. Monaco (2002). "Novel membrane traffic steps regulate the exocytosis of the Menkes disease ATPase." Hum Mol Genet 11(23): 2855-66.
Cobine, P. A., G. N. George, D. J. Winzor, M. D. Harrison, S. Mogahaddas and C. T. Dameron (2000). "Stoichiometry of complex formation between Copper(I) and the N-terminal domain of the Menkes protein." Biochemistry 39(23): 6857-63.
Coles, M., W. Bicknell, A. A. Watson, D. P. Fairlie and D. J. Craik (1998). "Solution structure of amyloid beta-peptide(1-40) in a water-micelle environment. Is the membrane-spanning domain where we think it is?" Biochemistry 37(31): 11064-77.
Cousins, R. J. (1985). "Absorption, transport, and hepatic metabolism of copper and zinc: special reference to metallothionein and ceruloplasmin." Physiol Rev 65(2): 238-309.
Cox, D. W. (1999). "Disorders of copper transport." Br Med Bull 55(3): 544-55.
Danks, D. M. (1995). The Metabolic Basis of Inherited Disease. C. Scriver, A. Beaudet, W. Sly and D. Valle. New York, McGraw-Hill: 2211-2235.
de Bie, P., B. van de Sluis, E. Burstein, P. V. van de Berghe, P. Muller, R. Berger, J. D. Gitlin, C. Wijmenga and L. W. Klomp (2007). "Distinct Wilson's disease mutations in ATP7B are associated with enhanced binding to COMMD1 and reduced stability of ATP7B." Gastroenterology 133(4): 1316-26.
De Feo, C. J., S. G. Aller, G. S. Siluvai, N. J. Blackburn and V. M. Unger (2009). "Three-dimensional structure of the human copper transporter hCTR1." Proc Natl Acad Sci U S A 106(11): 4237-42.
Delaglio, F., S. Grzesiek, G. W. Vuister, G. Zhu, J. Pfeifer and A. Bax (1995). "NMRPipe: a multidimensional spectral processing system based on UNIX pipes." J Biomol NMR 6(3): 277-93.
DeLano, W. L. (2008). The PyMOL Molecular Graphics System. Palo Alto, CA, USA., DeLano Scientific LLC.
DiDonato, M., H. F. Hsu, S. Narindrasorasak, L. Que, Jr. and B. Sarkar (2000). "Copper-induced conformational changes in the N-terminal domain of the Wilson disease copper-transporting ATPase." Biochemistry 39(7): 1890-6.
142
DiDonato, M., S. Narindrasorasak, J. R. Forbes, D. W. Cox and B. Sarkar (1997). "Expression, purification, and metal binding properties of the N- terminal domain from the wilson disease putative copper-transporting ATPase (ATP7B)." J Biol Chem 272(52): 33279-82.
DiDonato, M. and B. Sarkar (1997). "Copper transport and its alterations in Menkes and Wilson diseases." Biochim Biophys Acta 1360(1): 3-16.
DiDonato, M., J. Zhang, L. Que, Jr. and B. Sarkar (2002). "Zinc binding to the NH2-terminal domain of the Wilson disease copper- transporting ATPase: implications for in vivo metal ion-mediated regulation of ATPase activity." J Biol Chem 277(16): 13409-14.
Dierick, H. A., A. N. Adam, J. F. Escara-Wilke and T. W. Glover (1997). "Immunocytochemical localization of the Menkes copper transport protein (ATP7A) to the trans-Golgi network." Hum Mol Genet 6(3): 409-16.
Dierick, H. A., L. Ambrosini, J. Spencer, T. W. Glover and J. F. Mercer (1995). "Molecular structure of the Menkes disease gene (ATP7A)." Genomics 28(3): 462-9.
Dijkstra, M., G. J. van den Berg, H. Wolters, G. In't Veld, M. J. Slooff, H. S. Heymans, F. Kuipers and R. J. Vonk (1996). "Adenosine triphosphate-dependent copper transport in human liver." J Hepatol 25(1): 37-42.
Ellman, G. L. (1959). "Tissue sulfhydryl groups." Arch Biochem Biophys 82(1): 70-7.
Faller, P. and M. Vasak (1997). "Distinct metal-thiolate clusters in the N-terminal domain of neuronal growth inhibitory factor." Biochemistry 36(43): 13341-13348.
Fan, B., G. Grass, C. Rensing and B. P. Rosen (2001). "Escherichia coli CopA N-Terminal Cys(X)(2)Cys Motifs Are Not Required for Copper Resistance or Transport." Biochem Biophys Res Commun 286(2): 414-8.
Farrow, N. A., R. Muhandiram, A. U. Singer, S. M. Pascal, C. M. Kay, G. Gish, S. E. Shoelson, T. Pawson, J. D. Forman-Kay and L. E. Kay (1994). "Backbone dynamics of a free and phosphopeptide-complexed Src homology 2 domain studied by 15N NMR relaxation." Biochemistry 33(19): 5984-6003.
Fasman, G. D. (1996). Circular Dichroism and the Conformational Analysis of Biomolecules. New York, Plenum Press.
Fatemi, N., D. M. Korzhnev, A. Velyvis, B. Sarkar and J. D. Forman-Kay (2010). "NMR characterization of copper-binding domains 4-6 of ATP7B." Biochemistry 49(39): 8468-77.
Fatemi, N. and B. Sarkar (2002). "Insights into the mechanism of copper transport by the Wilson and Menkes disease copper-transporting ATPases." Inorg Chim Acta 339: 179-187.
Fatemi, N. and B. Sarkar (2002). "Structural and functional insights of Wilson disease copper-transporting ATPase." J Bioenerg Biomembr 34(5): 339-49.
Field, L. S., E. Luk and V. C. Culotta (2002). "Copper chaperones: personal escorts for metal ions." J Bioenerg Biomembr 34(5): 373-9.
Fiser, A., M. Feig, C. L. Brooks, 3rd and A. Sali (2002). "Evolution and physics in comparative protein structure modeling." Acc Chem Res 35(6): 413-21.
Forbes, J. R. and D. W. Cox (1998). "Functional characterization of missense mutations in ATP7B: Wilson disease mutation or normal variant?" Am J Hum Genet 63(6): 1663-74.
143
Forbes, J. R. and D. W. Cox (2000). "Copper-dependent trafficking of Wilson disease mutant ATP7B proteins." Hum Mol Genet 9(13): 1927-35.
Forbes, J. R., G. Hsi and D. W. Cox (1999). "Role of the copper-binding domain in the copper transport function of ATP7B, the P-type ATPase defective in Wilson disease." J Biol Chem 274(18): 12408-13.
Forman-Kay, J. D. (1999). "The 'dynamics' in the thermodynamics of binding." Nat Struct Biol 6(12): 1086-7.
Fu, D., T. J. Beeler and T. M. Dunn (1995). "Sequence, mapping and disruption of CCC2, a gene that cross-complements the Ca(2+)-sensitive phenotype of csg1 mutants and encodes a P-type ATPase belonging to the Cu(2+)-ATPase subfamily." Yeast 11(3): 283-92.
Ge, Z., K. Hiratsuka and D. E. Taylor (1995). "Nucleotide sequence and mutational analysis indicate that two Helicobacter pylori genes encode a P-type ATPase and a cation-binding protein associated with copper transport." Mol Microbiol 15(1): 97-106.
Gitschier, J., B. Moffat, D. Reilly, W. I. Wood and W. J. Fairbrother (1998). "Solution structure of the fourth metal-binding domain from the Menkes copper-transporting ATPase." Nat Struct Biol 5(1): 47-54.
Goodyer, I. D., E. E. Jones, A. P. Monaco and M. J. Francis (1999). "Characterization of the Menkes protein copper-binding domains and their role in copper-induced protein relocalization." Hum Mol Genet 8(8): 1473-8.
Guex, N. and M. C. Peitsch (1997). "SWISS-MODEL and the Swiss-PdbViewer: An environment for comparative protein modeling." Electrophoresis 18: 2714-2723.
Halliwell, B. (1994). "Free radicals and antioxidants: a personal view." Nutr Rev 52(8 Pt 1): 253-65.
Hamza, I., J. Prohaska and J. D. Gitlin (2003). "Essential role for Atox1 in the copper-mediated intracellular trafficking of the Menkes ATPase." Proc Natl Acad Sci U S A 100(3): 1215-20.
Hamza, I., M. Schaefer, L. W. Klomp and J. D. Gitlin (1999). "Interaction of the copper chaperone HAH1 with the Wilson disease protein is essential for copper homeostasis." Proc Natl Acad Sci U S A 96(23): 13363-8.
Harrison, M. D., C. E. Jones, M. Solioz and C. T. Dameron (2000). "Intracellular copper routing: the role of copper chaperones." Trends Biochem Sci 25(1): 29-32.
Hasler, D. W., P. Faller and M. Vasak (1998). "Metal-thiolate clusters in the C-terminal domain of human neuronal growth inhibitory factor (GIF)." Biochemistry 37(42): 14966-73.
Heaton, D. N., G. N. George, G. Garrison and D. R. Winge (2001). "The mitochondrial copper metallochaperone Cox17 exists as an oligomeric, polycopper complex." Biochemistry 40(3): 743-51.
Hisano, T., Y. Hata, T. Fujii, J. Q. Liu, T. Kurihara, N. Esaki and K. Soda (1996). "Crystal structure of L-2-haloacid dehalogenase from Pseudomonas sp. YL. An alpha/beta hydrolase structure that is different from the alpha/beta hydrolase fold." J Biol Chem 271(34): 20322-30.
Hoogenraad, T. U. (1998). "Zinc treatment of Wilson's disease." J Lab Clin Med 132(4): 240-1.
144
Hou, Z. J., S. Narindrasorasak, B. Bhushan, B. Sarkar and B. Mitra (2001). "Functional analysis of chimeric proteins of the Wilson Cu(I)-ATPase (ATP7B) and ZntA, a Pb(II)/Zn(II)/Cd(II)-ATPase from Escherichia coli." J Biol Chem 276(44): 40858-63.
Huffman, D. L. and T. V. O'Halloran (2001). "Function, Structure, and Mechanism of Intracellular Copper Trafficking Proteins." Annu Rev Biochem 70: 677-701.
Hung, I. H., R. L. Casareno, G. Labesse, F. S. Mathews and J. D. Gitlin (1998). "HAH1 is a copper-binding protein with distinct amino acid residues mediating copper homeostasis and antioxidant defense." J Biol Chem 273(3): 1749-54.
Hung, I. H., M. Suzuki, Y. Yamaguchi, D. S. Yuan, R. D. Klausner and J. D. Gitlin (1997). "Biochemical characterization of the Wilson disease protein and functional expression in the yeast Saccharomyces cerevisiae." J Biol Chem 272(34): 21461-6.
Huster, D. and S. Lutsenko (2003). "The distinct roles of the N-terminal copper-binding sites in regulation of catalytic activity of the Wilson's disease protein." J Biol Chem 278(34): 32212-8.
Jensen, P. Y., N. Bonander, L. B. Moller and O. Farver (1999). "Cooperative binding of copper(I) to the metal binding domains in Menkes disease protein." Biochim Biophys Acta 1434(1): 103-13.
Johnson, B. A. and R. A. Blevins (1994). "NMRView: A computer program for the visualization and analysis of NMR data." J Biomol NMR 4: 603-14.
Johnson, W. C., Jr. (1985). "Circular dichroism and its empirical application to biopolymers." Methods Biochem Anal 31: 61-163.
Jones, C. E., N. L. Daly, P. A. Cobine, D. J. Craik and C. T. Dameron (2003). "Structure and metal binding studies of the second copper binding domain of the Menkes ATPase." J Struct Biol 143(3): 209-18.
Jones, D. T., W. R. Taylor and J. M. Thornton (1992). "A new approach to protein fold recognition." Nature 358(6381): 86-9.
Kanelis, V., J. D. Forman-Kay and L. E. Kay (2001). "Multidimensional NMR methods for protein structure determination." IUBMB Life 52(6): 291-302.
Khan, A. R., K. A. Williams, J. M. Boggs and C. M. Deber (1995). "Accessibility and dynamics of Cys residues in Bacteriophage IKe and M13 major coat protein mutants." Biochemistry 34(38): 12388-97.
Kodama, H. (1993). "Recent developments in Menkes disease." J Inherit Metab Dis 16(4): 791-9.
Kordel, J., S. Forsen, T. Drakenberg and W. J. Chazin (1990). "The rate and structural consequences of proline cis-trans isomerization in calbindin D9k: NMR studies of the minor (cis-Pro43) isoform and the Pro43Gly mutant." Biochemistry 29(18): 4400-9.
Korzhnev, D. M., N. R. Skrynnikov, O. Millet, D. A. Torchia and L. E. Kay (2002). "An NMR experiment for the accurate measurement of heteronuclear spin-lock relaxation rates." J Am Chem Soc 124(36): 10743-53.
Koshland, D. E., Jr., G. Nemethy and D. Filmer (1966). "Comparison of experimental binding data and theoretical models in proteins containing subunits." Biochemistry 5(1): 365-85.
145
Krezel, A., W. Lesniak, M. Jezowska-Bojczuk, P. Mlynarz, J. Brasun, H. Kozlowski and W. Bal (2001). "Coordination of heavy metals by dithiothreitol, a commonly used thiol group protectant." J Inorg Biochem 84(1-2): 77-88.
La Fontaine, S. and J. F. Mercer (2007). "Trafficking of the copper-ATPases, ATP7A and ATP7B: Role in copper homeostasis." Arch Biochem Biophys 463(2): 149-67.
Larin, D., C. Mekios, K. Das, B. Ross, A. S. Yang and T. C. Gilliam (1999). "Characterization of the interaction between the Wilson and Menkes disease proteins and the cytoplasmic copper chaperone, HAH1p." J Biol Chem 274(40): 28497-504.
Lee, J., M. M. Pena, Y. Nose and D. J. Thiele (2002). "Biochemical characterization of the human copper transporter Ctr1." J Biol Chem 277(6): 4380-7.
Li, S. C. and C. M. Deber (1993). "Peptide environment specifies conformation. Helicity of hydrophobic segments compared in aqueous, organic, and membrane environments." J Biol Chem 268(31): 22975-8.
Lim, C. M., M. A. Cater, J. F. Mercer and S. La Fontaine (2006). "Copper-dependent interaction of dynactin subunit p62 with the N terminus of ATP7B but not ATP7A." J Biol Chem 281(20): 14006-14.
Lim, C. M., M. A. Cater, J. F. Mercer and S. La Fontaine (2006). "Copper-dependent interaction of glutaredoxin with the N termini of the copper-ATPases (ATP7A and ATP7B) defective in Menkes and Wilson diseases." Biochem Biophys Res Commun 348(2): 428-36.
Linder, M. C. (1991). Biochemistry of Copper. E. Frieden. New York, Plenum: 89 and 119.
Lipari, G. and A. Szabo (1982). "Model-free approach to the interpretation of nuclear magnetic resonance relaxation in macromolecules." J Am Chem Soc 104(17): 4546-4559.
Lipsky, M. A. and J. L. Gollan (1987). "Treatment of Wilson's disease: in D-penicillamine we trust--what about zinc?" Hepatology 7(3): 593-5.
Lockhart, P. J. and J. F. Mercer (2000). "Identification of the copper chaperone SAH in Ovis aries: expression analysis and in vitro interaction of SAH with ATP7B." Biochim Biophys Acta 1490(1-2): 11-20.
Loudianos, G., V. Dessi, M. Lovicu, A. Angius, A. Nurchi, G. C. Sturniolo, M. Marcellini, L. Zancan, P. Bragetti, N. Akar, R. Yagci, A. Vegnente, A. Cao and M. Pirastu (1998). "Further delineation of the molecular pathology of Wilson disease in the Mediterranean population." Hum Mutat 12(2): 89-94.
Loudianos, G. and J. D. Gitlin (2000). "Wilson's disease." Semin Liver Dis 20(3): 353-64.
Lowe, J., A. Vieyra, P. Catty, F. Guillain, E. Mintz and M. Cuillel (2004). "A mutational study in the transmembrane domain of Ccc2p, the yeast Cu(I)-ATPase, shows different roles for each Cys-Pro-Cys cysteine." J Biol Chem 279(25): 25986-94.
Lutsenko, S., N. L. Barnes, M. Y. Bartee and O. Y. Dmitriev (2007). "Function and regulation of human copper-transporting ATPases." Physiol Rev 87(3): 1011-46.
Lutsenko, S., R. G. Efremov, R. Tsivkovskii and J. M. Walker (2002). "Human copper-transporting ATPase ATP7B (the Wilson's disease protein): biochemical properties and regulation." J Bioenerg Biomembr 34(5): 351-62.
146
Lutsenko, S., K. Petrukhin, M. J. Cooper, C. T. Gilliam and J. H. Kaplan (1997). "N-terminal domains of human copper-transporting adenosine triphosphatases (the Wilson's and Menkes disease proteins) bind copper selectively in vivo and in vitro with stoichiometry of one copper per metal-binding repeat." J Biol Chem 272(30): 18939-44.
Mandal, A. K., Y. Yang, T. M. Kertesz and J. M. Arguello (2004). "Identification of the transmembrane metal binding site in Cu+-transporting PIB-type ATPases." J Biol Chem 279(52): 54802-7.
Mandel, A. M., M. Akke and A. G. Palmer, 3rd (1995). "Backbone dynamics of Escherichia coli ribonuclease HI: correlations with structure and function in an active enzyme." J Mol Biol 246(1): 144-63.
Marsh, J. A., V. K. Singh, Z. Jia and J. D. Forman-Kay (2006). "Sensitivity of secondary structure propensities to sequence differences between alpha- and gamma-synuclein: implications for fibrillation." Protein Sci 15(12): 2795-804.
Marti-Renom, M. A., A. C. Stuart, A. Fiser, R. Sanchez, F. Melo and A. Sali (2000). "COMPARATIVE PROTEIN STRUCTURE MODELING OF GENES AND GENOMES
doi:10.1146/annurev.biophys.29.1.291." Annual Review of Biophysics and Biomolecular Structure 29(1): 291-325.
Mense, M., L. A. Dunbar, R. Blostein and M. J. Caplan (2000). "Residues of the fourth transmembrane segments of the Na,K-ATPase and the gastric H,K-ATPase contribute to cation selectivity." J Biol Chem 275(3): 1749-56.
Mercer, J. F., J. Livingston, B. Hall, J. A. Paynter, C. Begy, S. Chandrasekharappa, P. Lockhart, A. Grimes, M. Bhave, D. Siemieniak and et al. (1993). "Isolation of a partial candidate gene for Menkes disease by positional cloning." Nat Genet 3(1): 20-5.
Mitra, B. and R. Sharma (2001). "The cysteine-rich amino-terminal domain of ZntA, a Pb(II)/Zn(II)/Cd(II)- translocating ATPase from Escherichia coli, is not essential for its function." Biochemistry 40(25): 7694-9.
Mittag, T., L. E. Kay and J. D. Forman-Kay (2010). "Protein dynamics and conformational disorder in molecular recognition." J Mol Recog 23(2): 105-116.
Moller, L. B., C. Petersen, C. Lund and N. Horn (2000). "Characterization of the hCTR1 gene: genomic organization, functional expression, and identification of a highly homologous processed gene." Gene 257(1): 13-22.
Monod, J., J. Wyman and J. P. Changeux (1965). "On the Nature of Allosteric Transitions: A Plausible Model." J Mol Biol 12: 88-118.
Morth, J. P., B. P. Pedersen, M. S. Toustrup-Jensen, T. L. Sorensen, J. Petersen, J. P. Andersen, B. Vilsen and P. Nissen (2007). "Crystal structure of the sodium-potassium pump." Nature 450(7172): 1043-9.
Myari, A., N. Hadjiliadis, N. Fatemi and B. Sarkar (2004). "Copper(I) interaction with model peptides of WD6 and TM6 domains of Wilson ATPase: regulatory and mechanistic implications." J Inorg Biochem 98(9): 1483-94.
147
Nagano, K., K. Nakamura, K. I. Urakami, K. Umeyama, H. Uchiyama, K. Koiwai, S. Hattori, T. Yamamoto, I. Matsuda and F. Endo (1998). "Intracellular distribution of the Wilson's disease gene product (ATPase7B) after in vitro and in vivo exogenous expression in hepatocytes from the LEC rat, an animal model of Wilson's disease." Hepatology 27(3): 799-807.
Narindrasorasak, S., P. Kulkarni, P. Deschamps, Y. M. She and B. Sarkar (2007). "Characterization and Copper Binding Properties of Human COMMD1 (MURR1)." Biochemistry 46(11): 3116-28.
Narindrasorasak, S., P. Yao and B. Sarkar (2003). "Protein disulfide isomerase, a multifunctional protein chaperone, shows copper-binding activity." Biochem Biophys Res Commun 311(2): 405-14.
Narindrasorasak, S., X. Zhang, E. A. Roberts and B. Sarkar (2004). "Comparative Analysis of Metal Binding Characteristics of Copper Chaperone Proteins, Atx1 and ATOX1." Bioinorg Chem Appl. 2(1-2): 105-123.
Nucifora, G., L. Chu, T. K. Misra and S. Silver (1989). "Cadmium resistance from Staphylococcus aureus plasmid pI258 cadA gene results from a cadmium-efflux ATPase." Proc Natl Acad Sci U S A 86(10): 3544-8.
O'Halloran, T. V. (1993). "Transition metals in control of gene expression." Science 261(5122): 715-25.
O'Halloran, T. V. and V. C. Culotta (2000). "Metallochaperones, an intracellular shuttle service for metal ions." J Biol Chem 275(33): 25057-60.
Odermatt, A., H. Suter, R. Krapf and M. Solioz (1993). "Primary structure of two P-type ATPases involved in copper homeostasis in Enterococcus hirae." J Biol Chem 268(17): 12775-9.
Orekhov, V. Y., D. E. Nolde, A. P. Golovanov, D. M. Korzhnev and A. S. Arseniev (1995). "Processing of heteronuclear NMR relaxation data with the new software DASHA." Appl Magn Reson 9: 581-588.
Pedersen, B. P., M. J. Buch-Pedersen, J. P. Morth, M. G. Palmgren and P. Nissen (2007). "Crystal structure of the plasma membrane proton pump." Nature 450(7172): 1111-4.
Peisach, J. and W. E. Blumberg (1974). "Structural implications derived from the analysis of electron paramagnetic resonance spectra of natural and artificial copper proteins." Arch Biochem Biophys 165(2): 691-708.
Peitsch, M. C. (1995). "Protein modeling by E-mail." Bio/Technology 13: 658-660.
Peitsch, M. C. (1996). "ProMod and Swiss-Model: Internet-based tools for automated comparative protein modelling." Biochem Soc Trans 24(1): 274-9.
Petris, M. J. (2004). "The SLC31 (Ctr) copper transporter family." Pflugers Arch 447(5): 752-5.
Petris, M. J., J. F. Mercer, J. G. Culvenor, P. Lockhart, P. A. Gleeson and J. Camakaris (1996). "Ligand-regulated transport of the Menkes copper P-type ATPase efflux pump from the Golgi apparatus to the plasma membrane: a novel mechanism of regulated trafficking." Embo J 15(22): 6084-95.
148
Petris, M. J., I. Voskoboinik, M. Cater, K. Smith, B. E. Kim, R. M. Llanos, D. Strausak, J. Camakaris and J. F. Mercer (2002). "Copper-regulated Trafficking of the Menkes Disease Copper ATPase Is Associated with Formation of a Phosphorylated Catalytic Intermediate." J Biol Chem 277(48): 46736-42.
Petrukhin, K., S. G. Fischer, M. Pirastu, R. E. Tanzi, I. Chernov, M. Devoto, L. M. Brzustowicz, E. Cayanis, E. Vitale, J. J. Russo and et al. (1993). "Mapping, cloning and genetic characterization of the region containing the Wilson disease gene." Nat Genet 5(4): 338-43.
Petrukhin, K., S. Lutsenko, I. Chernov, B. M. Ross, J. H. Kaplan and T. C. Gilliam (1994). "Characterization of the Wilson disease gene encoding a P-type copper transporting ATPase: genomic organization, alternative splicing, and structure/function predictions." Hum Mol Genet 3(9): 1647-56.
Phung, L. T., G. Ajlani and R. Haselkorn (1994). "P-type ATPase from the cyanobacterium Synechococcus 7942 related to the human Menkes and Wilson disease gene products." Proc Natl Acad Sci U S A 91(20): 9651-4.
Pountney, D. L., I. Schauwecker, J. Zarn and M. Vasak (1994). "Formation of mammalian Cu8-metallothionein in vitro: evidence for the existence of two Cu(I)4-thiolate clusters." Biochemistry 33(32): 9699-705.
Powlowski, J. and L. Sahlman (1999). "Reactivity of the two essential cysteine residues of the periplasmic mercuric ion-binding protein, MerP." J Biol Chem 274(47): 33320-6.
Pufahl, R. A., C. P. Singer, K. L. Peariso, S. J. Lin, P. J. Schmidt, C. J. Fahrni, V. C. Culotta, J. E. Penner-Hahn and T. V. O'Halloran (1997). "Metal ion chaperone function of the soluble Cu(I) receptor Atx1." Science 278(5339): 853-6.
Rae, T. D., P. J. Schmidt, R. A. Pufahl, V. C. Culotta and T. V. O'Halloran (1999). "Undetectable intracellular free copper: the requirement of a copper chaperone for superoxide dismutase." Science 284(5415): 805-8.
Ralle, M., M. J. Cooper, S. Lutsenko and N. J. Blackburn (1998). "The Menkes disease protein binds copper via novel 2-coordinate Cu(I)-cysteinates in the N-terminal domain." J Am Chem Soc 120(51): 13525-13526.
Ralle, M., S. Lutsenko and N. J. Blackburn (2003). "X-ray absorption spectroscopy of the copper chaperone HAH1 reveals a linear two-coordinate Cu(I) center capable of adduct formation with exogenous thiols and phosphines." J Biol Chem 278(25): 23163-70.
Ralle, M., S. Lutsenko and N. J. Blackburn (2004). "Copper transfer to the N-terminal domain of the Wilson disease protein (ATP7B): X-ray absorption spectroscopy of reconstituted and chaperone-loaded metal binding domains and their interaction with exogenous ligands." J Inorg Biochem 98(5): 765-74.
Rensing, C., B. Fan, R. Sharma, B. Mitra and B. P. Rosen (2000). "CopA: An Escherichia coli Cu(I)-translocating P-type ATPase." Proc Natl Acad Sci U S A 97(2): 652-6.
Rensing, C., B. Mitra and B. P. Rosen (1997). "The zntA gene of Escherichia coli encodes a Zn(II)-translocating P-type ATPase." Proc Natl Acad Sci U S A 94(26): 14326-31.
Rensing, C., Y. Sun, B. Mitra and B. P. Rosen (1998). "Pb(II)-translocating P-type ATPases." J Biol Chem 273(49): 32614-7.
149
Rice, W. J., H. S. Young, D. W. Martin, J. R. Sachs and D. L. Stokes (2001). "Structure of Na+,K+-ATPase at 11-A resolution: comparison with Ca2+- ATPase in E1 and E2 states." Biophys J 80(5): 2187-97.
Roberts, E. A. and D. W. Cox (1998). "Wilson disease." Baillieres Clin Gastroenterol 12(2): 237-56.
Roberts, E. A. and B. Sarkar (2008). "Liver as a key organ in the supply, storage, and excretion of copper." Am J Clin Nutr 88(3): 851S-4S.
Roberts, E. A. and M. L. Schilsky (2003). "A practice guideline on Wilson disease." Hepatology 37(6): 1475-92.
Rodriguez-Granillo, A., A. Crespo and P. Wittung-Stafshede (2010). "Interdomain interactions modulate collective dynamics of the metal-binding domains in the Wilson disease protein." J Phys Chem B 114(5): 1836-48.
Roelofsen, H., H. Wolters, M. J. Van Luyn, N. Miura, F. Kuipers and R. J. Vonk (2000). "Copper-induced apical trafficking of ATP7B in polarized hepatoma cells provides a mechanism for biliary copper excretion." Gastroenterology 119(3): 782-93.
Rosenzweig, A. C. (2001). "Copper delivery by metallochaperone proteins." Acc Chem Res 34(2): 119-28.
Rosenzweig, A. C., D. L. Huffman, M. Y. Hou, A. K. Wernimont, R. A. Pufahl and T. V. O'Halloran (1999). "Crystal structure of the Atx1 metallochaperone protein at 1.02 A resolution." Structure Fold Des 7(6): 605-17.
Rost, B. (1999). "Twilight zone of protein sequence alignments." Protein Eng 12(2): 85-94.
Sali, A. and J. Kuriyan (1999). "Challenges at the frontiers of structural biology." Trends Cell Biol 9(12): M20-4.
Sambongi, Y., T. Wakabayashi, T. Yoshimizu, H. Omote, T. Oka and M. Futai (1997). "Caenorhabditis elegans cDNA for a Menkes/Wilson disease gene homologue and its function in a yeast CCC2 gene deletion mutant." J Biochem (Tokyo) 121(6): 1169-75.
Sambrook, J., E. F. Fritsch and T. Maniatis (1989). Molecular Cloning: A Laboratory Manual. New York, Cold Spring Harbor Laboratory Press.
Sanchez, R. and A. Sali (1997). "Advances in comparative protein-structure modelling." Curr Opin Struct Biol 7(2): 206-14.
Sanchez, R. and A. Sali (1998). "Large-scale protein structure modeling of the Saccharomyces cerevisiae genome." Proc Natl Acad Sci U S A 95(23): 13597-602.
Saqi, M. A., R. B. Russell and M. J. Sternberg (1998). "Misleading local sequence alignments: implications for comparative protein modelling." Protein Eng 11(8): 627-30.
Sarkar, B. (1999). "Treatment of Wilson and Menkes Diseases." Chem Rev 99(9): 2535-2544.
Sarkar, B. (2000). "Copper transport and its defect in Wilson disease: characterization of the copper-binding domain of Wilson disease ATPase." J Inorg Biochem 79(1-4): 187-91.
Sarkar, B., K. Lingertat-Walsh and J. T. Clarke (1993). "Copper-histidine therapy for Menkes disease." J Pediatr 123(5): 828-30.
150
Sass-Kortsak, A. (1975). "Wilson's disease. A treatable liver disease in children." Pediatr Clin North Am 22(4): 963-84.
Sattler, M., J. Schleucher and C. Griesinger (1999). Heteronuclear multidimensional NMR experiments for the structure determination of proteins in solution employing pulsed field gradients 93-158.
Scarborough, G. A. (2000). "Crystallization, structure and dynamics of the proton-translocating P- type ATPase." J Exp Biol 203 Pt 1: 147-54.
Schaefer, M. and J. D. Gitlin (1999). "Genetic disorders of membrane transport. IV. Wilson's disease and Menkes disease." Am J Physiol 276(2 Pt 1): G311-4.
Schaefer, M., R. G. Hopkins, M. L. Failla and J. D. Gitlin (1999). "Hepatocyte-specific localization and copper-dependent trafficking of the Wilson's disease protein in the liver." Am J Physiol 276(3 Pt 1): G639-46.
Schaefer, M., H. Roelofsen, H. Wolters, W. J. Hofmann, M. Muller, F. Kuipers, W. Stremmel and R. J. Vonk (1999). "Localization of the Wilson's disease protein in human liver." Gastroenterology 117(6): 1380-5.
Scheinberg, I. H., M. E. Jaffe and I. Sternlieb (1987). "The use of trientine in preventing the effects of interrupting penicillamine therapy in Wilson's disease." N Engl J Med 317(4): 209-13.
Scheinberg, I. H. and I. Sternlieb (1984). Wilson's Disease. Philadelphia, Saunders.
Schilsky, M. L. (1996). "Wilson disease: genetic basis of copper toxicity and natural history." Semin Liver Dis 16(1): 83-95.
Seth, R., S. Yang, S. Choi, M. Sabean and E. A. Roberts (2004). "In vitro assessment of copper-induced toxicity in the human hepatoma line, Hep G2." Toxicol In Vitro 18(4): 501-9.
Sharma, R., C. Rensing, B. P. Rosen and B. Mitra (2000). "The ATP hydrolytic activity of purified ZntA, a Pb(II)/Cd(II)/Zn(II)- translocating ATPase from Escherichia coli." J Biol Chem 275(6): 3873-8.
Sharma, S. and A. Rosato (2009). "Role of the N-terminal tail of metal-transporting P(1B)-type ATPases from genome-wide analysis and molecular dynamics simulations." J Chem Inf Model 49(1): 76-83.
She, Y. M., S. Narindrasorasak, S. Yang, N. Spitale, E. A. Roberts and B. Sarkar (2003). "Identification of metal-binding proteins in human hepatoma lines by immobilized metal affinity chromatography and mass spectrometry." Mol Cell Proteomics 2(12): 1306-18.
Sherlock, S. and J. Dooley (1993). Wilson's disease. Diseases of the liver and biliary systems. Oxford, Blackwell Sci: 400-407.
Singleton, C. and N. E. Le Brun (2007). "Atx1-like chaperones and their cognate P-type ATPases: copper-binding and transfer." Biometals 20(3-4): 275-89.
Sippl, M. J. (1990). "Calculation of conformational ensembles from potentials of mean force. An approach to the knowledge-based prediction of local structures in globular proteins." J Mol Biol 213(4): 859-83.
151
Sippl, M. J. (1993). "Boltzmann's principle, knowledge-based mean fields and protein folding. An approach to the computational determination of protein structures." J Comput Aided Mol Des 7(4): 473-501.
Smith, P. K., R. I. Krohn, G. T. Hermanson, A. K. Mallia, F. H. Gartner, M. D. Provenzano, E. K. Fujimoto, N. M. Goeke, B. J. Olson and D. C. Klenk (1985). "Measurement of protein using bicinchoninic acid." Anal Biochem 150(1): 76-85.
Smith, S. D., Y. M. She, E. A. Roberts and B. Sarkar (2004). "Using immobilized metal affinity chromatography, two-dimensional electrophoresis and mass spectrometry to identify hepatocellular proteins with copper-binding ability." J Proteome Res 3(4): 834-40.
Solioz, M. and C. Vulpe (1996). "CPx-type ATPases: a class of P-type ATPases that pump heavy metals." Trends Biochem Sci 21(7): 237-41.
Steele, R. A. and S. J. Opella (1997). "Structures of the reduced and mercury-bound forms of MerP, the periplasmic protein from the bacterial mercury detoxification system." Biochemistry 36(23): 6885-95.
Sternlieb, I. (1990). "Perspectives on Wilson's disease." Hepatology 12(5): 1234-9.
Stokes, D. L. and N. M. Green (2000). "Modeling a dehalogenase fold into the 8-A density map for Ca(2+)-ATPase defines a new domain structure." Biophys J 78(4): 1765-76.
Strausak, D., S. La Fontaine, J. Hill, S. D. Firth, P. J. Lockhart and J. F. Mercer (1999). "The role of GMXCXXC metal binding sites in the copper-induced redistribution of the Menkes protein." J Biol Chem 274(16): 11170-7.
Sweadner, K. J. and C. Donnet (2001). "Structural similarities of Na,K-ATPase and SERCA, the Ca(2+)-ATPase of the sarcoplasmic reticulum." Biochem J 356(Pt 3): 685-704.
Tabata, K., S. Kashiwagi, H. Mori, C. Ueguchi and T. Mizuno (1997). "Cloning of a cDNA encoding a putative metal-transporting P-type ATPase from Arabidopsis thaliana." Biochim Biophys Acta 1326(1): 1-6.
Tanzi, R. E., K. Petrukhin, I. Chernov, J. L. Pellequer, W. Wasco, B. Ross, D. M. Romano, E. Parano, L. Pavone, L. M. Brzustowicz and et al. (1993). "The Wilson disease gene is a copper transporting ATPase with homology to the Menkes disease gene." Nat Genet 5(4): 344-50.
Tao, T. Y., F. Liu, L. Klomp, C. Wijmenga and J. D. Gitlin (2003). "The copper toxicosis gene product Murr1 directly interacts with the Wilson disease protein." J Biol Chem 278(43): 41593-6.
Terada, K., N. Aiba, X. L. Yang, M. Iida, M. Nakai, N. Miura and T. Sugiyama (1999). "Biliary excretion of copper in LEC rat after introduction of copper transporting P-type ATPase, ATP7B." FEBS Lett 448(1): 53-6.
Terada, K., Y. Kawarada, N. Miura, O. Yasui, K. Koyama and T. Sugiyama (1995). "Copper incorporation into ceruloplasmin in rat livers." Biochim Biophys Acta 1270(1): 58-62.
Terada, K., M. L. Schilsky, N. Miura and T. Sugiyama (1998). "ATP7B (WND) protein." Int J Biochem Cell Biol 30(10): 1063-7.
Terada, K. and T. Sugiyama (1999). "The Long-Evans Cinnamon rat: an animal model for Wilson's disease." Pediatr Int 41(4): 414-8.
152
Torchia, D. A. (1972). "Evidence for cis peptide bonds in copolypeptides of glycine and proline." Biochemistry 11(8): 1462-8.
Toyoshima, C., M. Nakasako, H. Nomura and H. Ogawa (2000). "Crystal structure of the calcium pump of sarcoplasmic reticulum at 2.6 A resolution." Nature 405(6787): 647-55.
Tsay, M. J., N. Fatemi, S. Narindrasorasak, J. R. Forbes and B. Sarkar (2004). "Identification of the "missing domain" of the rat copper-transporting ATPase, ATP7B: insight into the structural and metal binding characteristics of its N-terminal copper-binding domain." Biochim Biophys Acta 1688(1): 78-85.
Tsivkovskii, R., B. C. MacArthur and S. Lutsenko (2001). "The Lys1010-Lys1325 fragment of the Wilson's disease protein binds nucleotides and interacts with the N-terminal domain of this protein in a copper-dependent manner." J Biol Chem 276(3): 2234-42.
Tugarinov, V., P. M. Hwang and L. E. Kay (2004). "Nuclear magnetic resonance spectroscopy of high-molecular-weight proteins." Annu Rev Biochem 73: 107-46.
Tugarinov, V., V. Kanelis and L. E. Kay (2006). "Isotope labeling strategies for the study of high-molecular-weight proteins by solution NMR spectroscopy." Nat Protoc 1(2): 749-54.
Tumer, Z., N. Horn, T. Tonnesen, J. Christodoulou, J. T. Clarke and B. Sarkar (1996). "Early copper-histidine treatment for Menkes disease." Nat Genet 12(1): 11-3.
Tümer, Z., B. Vural, T. Tonnesen, J. Chelly, A. P. Monaco and N. Horn (1995). "Characterization of the exon structure of the Menkes disease gene using vectorette PCR." Genomics 26(3): 437-42.
Turski, M. L. and D. J. Thiele (2009). "New roles for copper metabolism in cell proliferation, signaling, and disease." J Biol Chem 284(2): 717-21.
Valverde, R. H., I. Morin, J. Lowe, E. Mintz, M. Cuillel and A. Vieyra (2008). "Cyclic AMP-dependent protein kinase controls energy interconversion during the catalytic cycle of the yeast copper-ATPase." FEBS Lett.
van De Sluis, B., J. Rothuizen, P. L. Pearson, B. A. van Oost and C. Wijmenga (2002). "Identification of a new copper metabolism gene by positional cloning in a purebred dog population." Hum Mol Genet 11(2): 165-73.
van Dongen, E. M., L. W. Klomp and M. Merkx (2004). "Copper-dependent protein-protein interactions studied by yeast two-hybrid analysis." Biochem Biophys Res Commun 323(3): 789-95.
Vanderwerf, S. M., M. J. Cooper, I. V. Stetsenko and S. Lutsenko (2001). "Copper specifically regulates intracellular phosphorylation of the Wilson's disease protein, a human copper-transporting ATPase." J Biol Chem 276(39): 36289-94.
Vanderwerf, S. M. and S. Lutsenko (2002). "The Wilson's disease protein expressed in Sf9 cells is phosphorylated." Biochem Soc Trans 30(4): 739-41.
Vasak, M. and J. H. R. Kagi (1983). Metal Ions in Biological Systems. H. Sigel. New York, Marcel Dekker, Inc. 15: 213-273.
153
Veglia, G., F. Porcelli, T. DeSilva, A. Prantner and S. J. Opella (2000). "The Structure of the Metal-Binding Motif GMTCAAC Is Similar in an 18-Residue Linear Peptide and the Mercury Binding Protein MerP." J. Am. Chem. Soc. 122(10): 2389-2390.
Veldhuis, N. A., A. P. Gaeth, R. B. Pearson, K. Gabriel and J. Camakaris (2009). "The multi-layered regulation of copper translocating P-type ATPases." Biometals.
Voskoboinik, I., R. Fernando, N. Veldhuis, K. M. Hannan, N. Marmy-Conus, R. B. Pearson and J. Camakaris (2003). "Protein kinase-dependent phosphorylation of the Menkes copper P-type ATPase." Biochem Biophys Res Commun 303(1): 337-42.
Voskoboinik, I., M. Greenough, S. La Fontaine, J. F. Mercer and J. Camakaris (2001). "Functional studies on the Wilson copper P-type ATPase and toxic milk mouse mutant." Biochem Biophys Res Commun 281(4): 966-70.
Voskoboinik, I., D. Strausak, M. Greenough, H. Brooks, M. Petris, S. Smith, J. F. Mercer and J. Camakaris (1999). "Functional analysis of the N-terminal CXXC metal-binding motifs in the human menkes copper-transporting P-type ATPase expressed in cultured mammalian cells." J Biol Chem 274(31): 22008-12.
Vulpe, C., B. Levinson, S. Whitney, S. Packman and J. Gitschier (1993). "Isolation of a candidate gene for Menkes disease and evidence that it encodes a copper-transporting ATPase." Nat Genet 3(1): 7-13.
Walker, J. M., D. Huster, M. Ralle, C. T. Morgan, N. J. Blackburn and S. Lutsenko (2004). "The N-terminal metal-binding site 2 of the Wilson's Disease Protein plays a key role in the transfer of copper from Atox1." J Biol Chem 279(15): 15376-84.
Walker, J. M., R. Tsivkovskii and S. Lutsenko (2002). "Metallochaperone Atox1 transfers copper to the NH2-terminal domain of the Wilson's disease protein and regulates its catalytic activity." J Biol Chem 277(31): 27953-9.
Walshe, J. M. (1956). "Penicillamine, a new oral therapy for Wilson disease." Am. J. Med. 21: 487-495.
Walshe, J. M. (1982). "Treatment of Wilson's disease with trientine (triethylene tetramine) dihydrochloride." Lancet 1(8273): 643-7.
Walshe, J. M. (1999). "Penicillamine: the treatment of first choice for patients with Wilson's disease." Mov Disord 14(4): 545-50.
Wernimont, A. K., D. L. Huffman, A. L. Lamb, T. V. O'Halloran and A. C. Rosenzweig (2000). "Structural basis for copper transfer by the metallochaperone for the Menkes/Wilson disease proteins." Nat Struct Biol 7(9): 766-71.
Wernimont, A. K., L. A. Yatsunyk and A. C. Rosenzweig (2004). "Binding of copper(I) by the Wilson disease protein and its copper chaperone." J Biol Chem 279(13): 12269-76.
Wiechelman, K. J., R. D. Braun and J. D. Fitzpatrick (1988). "Investigation of the bicinchoninic acid protein assay: identification of the groups responsible for color formation." Anal Biochem 175(1): 231-7.
Williams, K. A. and C. M. Deber (1991). "Proline residues in transmembrane helices: structural or dynamic role?" Biochemistry 30(37): 8919-23.
154
Willner, H., M. Vasak and J. H. Kagi (1987). "Cadmium-thiolate clusters in metallothionein: spectrophotometric and spectropolarimetric features." Biochemistry 26(19): 6287-92.
Wilson, S. A. K. (1912). "Progressive lenticular degeneration: A familial nervous disease associated with cirrhosis of the liver." Brain 34: 295-508.
Wimmer, R., T. Herrmann, M. Solioz and K. Wuthrich (1999). "NMR structure and metal interactions of the CopZ copper chaperone." J Biol Chem 274(32): 22597-603.
Wingfield, P. T. and R. H. Pain (1996). Determining the CD Spectrum of a Protein. Current Protocols in Protein Science. J. E. Coligan, B. M. Dunn, H. L. Ploegh, D. W. Speicher and P. T. Wingfield. New York, John Wiley & Sons, Inc. 1: 7.6.1-7.6.23.
Woessner, D. E. (1962). "Nuclear spin relaxation in ellipsoids undergoing rotational Brownian motion." J Chem Phys 37: 647-654.
Wu, J., J. R. Forbes, H. S. Chen and D. W. Cox (1994). "The LEC rat has a deletion in the copper transporting ATPase gene homologous to the Wilson disease gene." Nat Genet 7(4): 541-5.
Yamaguchi, Y., M. E. Heiny and J. D. Gitlin (1993). "Isolation and characterization of a human liver cDNA as a candidate gene for Wilson disease." Biochem Biophys Res Commun 197(1): 271-7.
Yamaguchi, Y., M. E. Heiny, M. Suzuki and J. D. Gitlin (1996). "Biochemical characterization and intracellular localization of the Menkes disease protein." Proc Natl Acad Sci U S A 93(24): 14030-5.
Yatsunyk, L. A. and A. C. Rosenzweig (2007). "Cu(I) binding and transfer by the N terminus of the Wilson disease protein." J Biol Chem 282(12): 8622-31.
Yoshimizu, T., H. Omote, T. Wakabayashi, Y. Sambongi and M. Futai (1998). "Essential Cys-Pro-Cys motif of Caenorhabditis elegans copper transport ATPase." Biosci Biotechnol Biochem 62(6): 1258-60.
Zhou, B. and J. Gitschier (1997). "hCTR1: a human gene for copper uptake identified by complementation in yeast." Proc Natl Acad Sci U S A 94(14): 7481-6.
155
Appendices
A.1 The relaxation equations
The typical set of heteronuclear NMR relaxation data consists of T1, T2 and heteronuclear NOEs
at one or several spectrometer frequencies. The theoretical values of these rates can be
calculated from (Abragam 1961):
R1 = d[J(ωH − ωN) + 3J(ωN) + 6J(ωH + ωN)] + cJ(ωN) ,
R2 = d/2[4J(0) + J(ωH − ωN) + 3J(ωN) + 6J(ωH) + 6J(ωH + ωN)]
+ c/6[4J(0) + 3J(ωN)] and
NOE = 1 + (γH / γN)(σNOE / R1) ,
with
d = (µ0/4π)2 (γHγNħ)2 / (4rNH
6) ,
c = (∆σNγNB0)2/3 and
σNOE = d[6J(ωH + ωN) − J(ωH − ωN)] ,
where J(ω) is the power spectral density function, µ0 is the permeability of free space, γH and γN
are the gyromagnetic ratios of the 1H and 15N nuclei respectively, ħ is Plank’s constant divided
by 2π, rNH is the 15N−1H bond length, ∆σN is the chemical shift anisotropy of the 15N nucleus, and
B0 is the static magnetic field strength.
A.2 Brownian rotational diffusion
Local fields are modulated by relative motions of nuclei in a molecular reference frame as well
as by overall rotational Brownian motion. The interpretation of nuclear spin relaxation due to
dipole-dipole interaction between two nuclei in a molecules undergoing rotational Brownian
diffusion was developed by Woessner (Woessner 1962), and described by the following generic
NMR correlation C(t) and spectral density J(ω) functions
5
C(t) = Σ Cn·e−(t/τ
n) and
n=1
J(ω) = FT[C(t)] ,
156
where the summation index n ranges over the number of exponential terms within the correlation
function. For isotropic rotational diffusion and approximately for anisotropic rotational diffusion,
the total correlation function is factored as (Lipari and Szabo 1982)
C(t) = CI(t) CO(t) .
The correlation functions for overall motion and internal motion, are given by
CO(t) = e(−t/τR) and
CI(t) = S2 + (1− S2)e(−t/τe) .
This equation is generic in that it can describe the diffusion of a sphere (isotropic diffusion), a
spheroid (anisotropic symmetric diffusion), or an ellipsoid (anisotropic asymmetric diffusion).
A.3 Models of overall rotation The rotation diffusion tensor is specified by eigen values (D1, D2 and D3) and eigen vectors (Dx,
Dy and Dz) of the diffusion tensor. The Euler angles α, β, and γ orient the tensor within the PDB
frame. These angles are defined using the z-y-z axis rotation notation where α is the initial
rotation angle around the z-axis, β is the rotation angle around the y-axis, and γ is the final
rotation around the z-axis again. Within the PDB frame, the NH bond vector is described using θ
and φ where θ is the polar angle and φ is the azimuthal angle.
The three unit vectors Dx, Dy and Dz defining the diffusion frame can be expressed using the
Euler angles α, β, and γ as:
Dx = [−sin α sin γ + cos α cos β cos γ]
[−sin α cos γ − cos α cos β sin γ]
[cos α sin β],
Dy = [cos α sin γ + sinα cos β cos γ]
[cos α cos γ − sinα cos β sin γ]
[sin α sin β] and
Dz = [−sin β cos γ]
[sin β sin γ]
157
[cos β] .
A.3.1 Isotropic diffusion − diffusion as a sphere
In the case of isotropic or spherical diffusion,
D1 = D2 = D3 = D and
τR = 1/[2(D1 + D2 + D3)] = 1/(6D) ,
where τR is the overall rotation correlation time. The most simple, isotropic, model of overall
rotation describes the rotational Brownian diffusion of a molecule with a spherical shape is
CO(t) = e(−t6D) = e(−t/τR) .
If the overall rotation and internal motions of the molecule are independent, and the overall
molecular rotation is isotropic, then the total correlation function is a product of internal and
overall correlation functions,
C(t) = [S2 + (1− S2)e(−t/τe)] e(−t/τR).
This relationship, although not entirely valid for anisotropic rotation, has been shown to be a
good approximation of anisotropic rotation. When a rigid anisotropic molecule is undergoing
rotational Brownian diffusion CI(t)=1 and C(t)=CO(t).
A.3.2 Anisotropic axially symmetric diffusion − diffusion as a spheroid
The correlation function of the overall rotation of spheroidal, or axially symmetric molecules is
described by
CO(t) = A·e(−t/τa) + B·e(−t/τb) + C·e(−t/τc) ,
where
τa = (4D║+2D┴)−1 , τb = (D║+5D┴)
−1 , τc = (6D┴)−1 ,
A = (3/4)sin4θ , B = 3sin2θcos2θ , C = (3cos2θ − 1)/2 ,
θ is the angle between NH bond vector and the axis of symmetry of the molecule, and D║ and D┴
are the eigen values of the diffusion tensor for an axially symmetric molecule:
158
D║=D1 ,
D┴=D2=D3 and
τR = 1/[2(D1 + D2 + D3)] = 1/[2(D║ + 2D┴)] .
A.3.3 Anisotropic fully asymmetric diffusion − diffusion as an ellipsoid
When all three eigen values of the diffusion tensor are different, the molecule diffuses as an
ellipsoid. This diffusion is also known as fully anisotropic, asymmetric, or rhombic. The
rotational Brownian diffusion of a fully anisotropic molecule or an ellipsoid, is represented by
the five-exponential correlation function and the corresponding spectral density function
(Woessner 1962):
5
C(t) = Σ Cn·e(−t/τn) and
n=1
5
J(ω) = Σ Cn·[ τn/(1+ω2
τn2)] .
n=1
The spectral density function coefficients are:
C1 = 6m2n2 ,
C2 = 6l2n2 ,
C3 = 6l2m2 ,
C4 = d − e and
C5 = d + e .
The five correlation times τn are
τ1-1 = 4D1 + D2 + D3 ,
τ2-1 = D1 + 4D2 + D3 ,
τ3-1 = D1 + D2 + 4D3 ,
τ4-1 = 6{D + (D2 − L2) ½} ,
τ5-1 = 6{D − (D2 − L2) ½} ,
and
τR = 1/[2(D1 + D2 + D3)] ,
159
where l, m and n are the direction cosines of a unit vector parallel to the NH bond vector, and
fixed with respect to the diffusion ellipsoid principal axis, and
d = ½{3(l4 + m4 + n4) − 1} ,
e = 1/6{ε1(3l4 + 6m2n2 − 1) + ε2(3m4 + 6l2n2 − 1) + ε3(3n4 + 6l2m2 − 1)} ,
εi = (Di − D)/{(D2 − L2) ½} ,
D = 1/3(D1 + D2 + D3) ,
L2 = 1/3(D1D2 + D1D3 + D2D3) and
i = 1, 2, 3 .
A.4 Models of internal motions
The relaxation equations are themselves dependent on the calculation of the values of the
spectral density function describing the probability of finding motions at a given angular
frequency,
J(ω) = 2τR/(1+ω2τR
2) .
When all internal motions are in the extreme narrowing limit, where their characteristic
correlation times are much shorter than the overall rotation correlation time, the corresponding
Lipari-Szabo autocorrelation function (Lipari and Szabo 1982) is:
CI(t) = S2 + (1− S2)e(−t/τe) ,
where S2 is the square of the Lipari and Szabo generalized order parameter and τe is the effective
correlation time. The order parameter reflects the amplitude of the motion and the correlation
time in an indication of the time scale of that motion. The theory was extended by (Clore, Szabo
et al. 1990) to include the modelling of two independent internal motions, fast, described by the
correlation time τf and order parameter Sf2, and intermediate-slow, with correlation time τs and
order parameter Ss2, using the equation,
CI(t) = Sf2Ss
2 + (1− Sf2)e(−t/τf) + (Sf
2− Sf2Ss
2) (−t/τs) ,
160
where Sf2 and τf are the amplitude and timescale of the faster of the two motions whereas Ss
2 and
τs are those of the slower motion. Sf2 and Ss
2 are related by the formula S2 = Sf2Ss
2 (Clore, Szabo
et al. 1990). In DASHA, these models are:
model 1 = { S2 } ,
model 2 = { S2, Rex } ,
model 3 = { S2, τe } ,
model 4 = { S2, τe, Rex } and
model 5 = { Ss2, Sf
2, τs } .
A.5 Statistical analysis, evaluation of model-free parameters and model selection
Experimental relaxation rates and NOEs (Vexp) are fitted to the theoretical values (Vth(ζ)) by
minimization of the loss function, chi-squared,
N χ2(ζ) = Σ [(Vi
th(ζ) − Viexp)2/(∆Vi
exp)2] , i=1
where ∆Vexp is the uncertainty of the experimental value, the index i is the summation index
ranging over all the experimentally collected relaxation data of all residues used in the analysis;
N is the number of experimental values, and ζ is the model parameter vector which is minimized.
The significance of the chi-squared equation is that the function returns a single value which is
then minimized by the optimization algorithm to find the model-free parameter values of the
given model. The Levenberg-Marquardt algorithm minimizes the total loss function for the
specified set of residues, consequently optimizing both local residue specific parameters of
internal dynamics (S, Ss, τe etc.) and parameters of the overall molecular tumbling.
161
A.6 1H, 15N, and 13C resonance assignments Table A.6-1 1H, 15N, and 13C resonance assignments of human apoWCBD4-6 at 35 ºC,
in 20 mM NaK-phosphate, 130 mM NaCl, 5mM DTT, pH 6.0.
# Residue N HN CA CB C
357 T - - 61.3 44.0 173.9
358 C 120.7 8.01 57.9 29.2 173.7
359 S 120.8 8.54 57.0 39.9 171.2
360 T 116.6 8.37 60.7 45.1 174.5
361 T 122.8 9.11 59.6 43.9 169.5
362 L 130.5 7.98 53.0 44.3 174.6
363 I 126.1 9.09 58.6 41.0 174.9
364 A 132.0 9.24 51.1 44.2 176.6
365 I 123.3 8.37 60.0 40.2 174.8
366 A 132.0 9.11 50.5 45.9 177.6
367 G 107.9 8.42 - - -
368 M - - - - -
369 T - - - - -
370 C - - - - -
371 A - - - - -
372 S - - - - -
162
373 C - - - - -
374 V - - 42.1 30.7 177.2
375 H 116.3 7.68 57.6 28.2 168.7
376 S 117.3 8.11 61.3 36.9 176.4
377 I 123.3 7.92 41.6 37.9 177.7
378 E 117.8 8.40 59.6 28.8 169.7
379 G 106.7 8.21 46.5 - 175.2
380 M 120.4 7.58 57.6 32.0 169.2
381 I 119.4 8.32 61.0 34.6 177.6
382 S 111.5 7.80 60.8 37.4 174.4
383 Q 116.8 6.88 55.1 28.9 176.3
384 L 122.0 7.43 54.3 40.5 177.4
385 E 125.8 8.47 57.3 28.2 176.1
386 G 110.5 8.48 44.6 - 174.4
387 V 121.6 7.60 63.6 31.1 175.4
388 Q 128.6 9.09 56.0 29.2 175.8
389 Q 117.0 7.76 55.2 32.2 172.8
390 I 121.6 8.66 58.0 39.6 172.2
391 S 121.0 8.23 55.7 39.2 173.7
392 V 130.0 9.70 60.6 32.0 174.2
163
393 S 121.5 8.77 54.5 37.7 175.3
394 L 131.7 8.81 57.3 41.1 168.6
395 A 120.2 8.29 54.4 43.7 169.7
396 E 112.9 7.61 55.6 29.9 177.0
397 G 110.3 7.89 46.2 - 174.0
398 T 107.4 7.37 58.4 47.5 173.0
399 A 122.4 9.53 49.8 47.1 175.3
400 T 120.5 8.73 61.8 43.5 174.5
401 V 128.0 9.03 60.2 33.4 173.8
402 L 130.9 8.95 53.6 42.6 176.3
403 Y 124.3 9.23 - - -
404 N - - - - -
405 P - - 62.8 31.8 177.1
406 A 120.7 7.87 52.7 43.8 177.6
407 V 116.6 7.80 62.1 34.6 173.9
408 I 121.9 8.21 59.7 41.6 170.2
409 S 117.2 7.44 - - -
410 P - - 41.3 31.3 177.4
411 E 115.7 8.18 58.3 27.8 168.7
412 E 120.1 7.48 58.7 29.5 169.9
164
413 L 121.6 7.45 57.3 41.4 177.0
414 R 119.7 8.30 59.2 28.6 177.9
415 A 120.9 8.21 54.4 43.1 169.4
416 A 121.0 7.58 54.3 42.8 170.2
417 I 119.6 7.46 40.9 37.5 168.4
418 E 122.6 8.29 58.4 28.5 171.4
419 D 121.3 8.31 56.2 40.2 177.5
420 M 116.9 7.50 56.4 33.2 175.8
421 G 105.9 7.98 44.2 - 173.1
422 F 119.1 6.81 56.2 41.0 174.0
423 E 120.7 7.92 55.4 29.8 175.3
424 A 128.1 8.87 50.0 48.2 176.1
425 S 115.3 8.64 56.5 40.0 172.7
426 V 126.6 9.01 63.6 31.2 176.2
427 V 130.0 8.87 62.7 31.8 176.0
428 S 117.3 7.99 57.4 38.8 173.9
429 E 124.4 8.58 56.3 30.3 175.7
430 S 116.8 8.40 57.7 38.2 174.1
431 C 121.9 8.38 58.0 28.0 174.5
432 S 118.9 8.37 58.1 38.1 174.5
165
433 T 115.9 8.02 61.3 43.7 173.9
434 N 122.2 8.23 - - -
435 P - - - - -
436 L - - - - -
437 G - - - - -
438 N - - - - -
439 H - - - - -
440 S - - 57.9 38.1 174.1
441 A 126.4 8.26 52.2 44.3 178.0
442 G 108.2 8.18 44.9 - 173.8
443 N 119.2 8.14 52.7 38.6 175.2
444 S 116.6 8.19 58.2 38.0 174.3
445 M 122.5 8.24 55.1 32.2 175.9
446 V 121.5 7.91 61.7 32.1 175.7
447 Q 124.7 8.34 55.1 29.0 175.8
448 T 116.3 8.17 61.2 43.9 174.6
449 T 116.0 8.13 61.3 43.8 174.2
450 D 123.0 8.22 53.9 40.8 176.4
451 G 109.6 8.16 44.7 - 173.9
452 T 117.4 7.97 - - -
166
453 P - - 62.9 31.4 177.1
454 T 115.0 8.13 61.7 43.8 174.6
455 S 118.2 8.12 57.8 38.1 174.3
456 L 124.4 8.11 54.9 41.4 177.0
457 Q 120.9 8.09 55.4 28.8 175.6
458 E 122.5 8.18 56.0 29.7 176.0
459 V 121.5 7.98 61.4 32.2 175.2
460 A 129.2 8.18 - - -
461 P - - - - -
462 H - - - - -
463 T - - 61.6 43.9 174.8
464 G 111.7 8.28 44.7 - 173.5
465 R 121.0 8.02 55.3 30.2 175.9
466 L 125.4 8.20 - - -
467 P - - 62.5 31.3 176.4
468 A 124.3 8.21 52.0 44.2 177.3
469 N 117.6 8.19 - - -
470 H - - - - -
471 A - - - - -
472 P - - 62.8 31.3 176.5
167
473 D 120.4 8.25 53.8 40.6 176.3
474 I 121.0 7.84 61.0 38.0 176.2
475 L 125.1 8.04 54.6 41.2 176.9
476 A 124.8 7.92 52.0 44.1 177.3
477 K 120.5 7.99 55.4 32.5 176.2
478 S 118.7 8.16 - - -
479 P - - 63.0 31.3 176.9
480 Q 120.3 8.30 55.5 28.8 176.0
481 S 117.3 8.18 57.9 38.1 174.6
482 T 116.5 8.04 61.4 43.8 174.1
483 R 124.0 8.12 55.4 30.2 175.3
484 A 126.8 8.24 51.7 44.3 177.0
485 V 121.4 8.03 61.4 32.4 175.0
486 A 130.9 8.35 - - -
487 P - - 62.2 32.0 176.1
488 Q 119.8 8.08 52.8 31.5 173.1
489 K 118.7 8.32 54.2 35.0 176.2
490 C 120.1 9.25 55.2 30.8 170.7
491 F 121.5 8.93 55.8 41.6 174.6
492 L 121.3 8.93 52.5 44.5 175.8
168
493 Q 122.8 8.98 54.5 29.9 175.3
494 I 126.8 8.47 59.7 40.2 175.4
495 K 128.9 9.01 54.9 34.1 176.3
496 G 108.9 8.44 - - -
497 M - - - - -
498 T - - - - -
499 C - - - - -
500 A - - - - -
501 S - - - - -
502 C - - - - -
503 V - - 42.1 30.7 177.3
504 S 113.4 7.80 60.6 36.9 176.3
505 N 119.7 7.75 - - -
506 I - - 41.2 37.7 177.2
507 E 118.1 8.49 60.2 29.2 168.6
508 R 116.4 8.39 58.4 44.4 169.1
509 N 115.3 7.55 56.3 39.1 177.3
510 L 120.4 8.41 56.9 41.3 178.0
511 Q 114.3 7.61 57.3 27.9 176.0
512 K 114.4 6.73 55.6 32.0 176.6
169
513 E 121.6 7.29 54.5 28.1 176.2
514 A 128.7 8.33 53.0 42.7 177.4
515 G 108.2 8.39 44.8 - 173.1
516 V 120.7 7.45 63.3 30.1 175.2
517 L 128.7 8.91 55.5 41.6 177.6
518 S 112.1 7.67 57.8 39.0 172.2
519 V 122.6 8.46 60.5 34.1 172.5
520 L 129.3 8.22 53.7 43.6 175.6
521 V 127.9 9.26 61.3 32.0 173.4
522 A 132.2 8.74 49.8 44.8 176.3
523 L 126.2 8.47 58.4 41.2 168.5
524 M 116.7 8.67 57.3 30.3 177.5
525 A 119.0 7.33 51.5 44.8 177.8
526 G 108.6 7.77 46.3 - 173.4
527 K 116.8 7.23 53.9 36.9 174.0
528 A 122.8 9.05 49.5 48.4 174.6
529 E 123.0 8.84 54.3 31.5 175.2
530 I 129.5 8.91 58.8 40.1 174.6
531 K 128.8 8.44 54.3 33.4 175.6
532 Y 124.6 9.28 - - -
170
533 D - - - - -
534 P - - 62.8 31.8 178.0
535 E 118.4 8.25 57.0 28.9 176.7
536 V 118.3 7.83 62.5 34.5 174.2
537 I 122.1 8.35 59.6 42.3 170.6
538 Q 119.4 7.56 - - -
539 P - - 42.1 32.0 177.2
540 L 116.1 8.53 57.7 40.7 169.5
541 E 118.4 7.24 58.1 28.9 169.0
542 I 121.6 7.27 41.1 36.6 177.3
543 A 120.2 7.96 54.9 42.0 169.1
544 Q 118.0 7.69 57.8 27.4 177.3
545 F 120.2 7.78 60.1 38.2 168.5
546 I 118.9 7.58 40.9 37.5 168.5
547 Q 122.6 8.29 57.8 27.1 171.3
548 D 122.4 8.40 56.3 39.5 177.8
549 L 119.3 7.46 54.9 41.4 176.8
550 G 105.2 7.81 44.0 - 173.1
551 F 119.6 7.08 56.0 40.5 173.8
552 E 120.6 7.96 55.8 30.0 175.2
171
553 A 126.4 8.08 50.1 48.3 175.2
554 A 122.7 8.41 50.1 47.5 176.0
555 V 123.4 8.79 62.9 31.4 176.2
556 M 128.4 8.80 54.7 33.2 175.5
557 E 123.6 8.47 56.2 29.9 175.8
558 D 120.9 8.34 53.5 40.5 175.3
559 Y 121.0 7.83 57.4 38.4 175.3
560 A 126.6 8.04 52.1 44.2 177.6
561 G 107.9 7.61 - - -
562 S - - - - -
563 D - - 53.6 40.6 176.4
564 G 108.5 8.05 44.5 172.9
565 N 119.4 7.53 52.4 41.4 173.9
566 I 119.1 8.99 59.3 42.3 172.1
567 E 123.8 8.24 53.8 31.7 175.3
568 L 123.0 9.06 52.4 45.2 176.5
569 T 119.0 8.95 61.6 43.7 174.4
570 I 128.3 8.60 58.9 38.9 175.8
571 T 120.9 8.97 60.8 44.7 174.4
572 G 110.0 8.56 - - -
172
573 M - - - - -
574 T - - - - -
575 C - - - - -
576 A - - - - -
577 S - - - - -
578 C - - - - -
579 V - - 41.9 30.8 177.3
580 H 117.0 7.69 58.1 28.3 177.8
581 N 119.7 8.11 55.7 37.5 177.2
582 I 119.5 7.79 41.3 38.0 177.2
583 E 118.0 8.44 60.5 29.1 169.2
584 S 114.2 8.27 60.8 37.2 175.8
585 K 120.7 7.38 57.6 31.3 169.6
586 L 117.5 8.10 57.7 40.8 168.5
587 T 108.9 8.11 41.0 - 175.6
588 R 118.7 7.05 55.5 29.4 175.9
589 T 120.5 7.56 62.9 43.4 173.9
590 N 128.2 8.78 54.8 37.6 174.5
591 G 112.9 8.42 44.9 - 172.9
592 I 120.3 7.64 61.0 35.6 176.4
173
593 T 120.3 8.85 61.6 43.3 174.9
594 Y 123.1 7.68 57.7 41.4 172.5
595 A 127.5 7.54 50.8 47.9 174.2
596 S 115.7 8.39 55.7 39.5 173.5
597 V 130.0 9.48 60.6 32.5 173.0
598 A 130.2 8.44 49.7 45.2 176.3
599 L 125.9 8.52 57.2 41.3 169.3
600 A 117.6 8.62 54.6 43.8 169.2
601 T 131.2 6.69 59.7 43.7 174.7
602 S 115.4 7.64 59.1 36.2 172.4
603 K 117.5 7.40 53.7 35.4 175.7
604 A 123.9 9.14 49.0 48.1 174.6
605 L 125.6 8.75 53.7 43.2 175.4
606 V 124.3 8.72 60.2 34.6 174.4
607 K 126.3 9.01 54.1 33.9 175.9
608 F 120.4 8.47 54.8 41.4 171.9
609 D 120.4 8.82 - - -
610 P - - 62.7 31.8 178.0
611 E 117.4 8.42 57.1 28.8 176.9
612 I 118.6 7.99 60.9 39.3 174.6
174
613 I 122.4 8.31 59.7 42.1 170.7
614 G 108.8 8.03 - - -
615 P - - 41.4 31.6 177.5
616 R 114.9 8.59 58.7 28.5 168.9
617 D 118.5 7.22 56.2 40.8 178.1
618 I 121.1 7.27 41.2 37.2 176.8
619 I 119.3 7.76 41.8 37.6 177.4
620 K 119.5 7.36 58.9 31.5 168.5
621 I 120.2 7.54 40.5 37.3 177.7
622 I 119.4 7.31 40.8 37.3 177.6
623 E 120.7 8.22 57.9 28.4 171.5
624 E 121.5 8.26 58.5 28.7 168.6
625 I 113.7 7.42 61.7 37.2 175.8
626 G 105.6 7.39 44.5 - 173.2
627 F 119.2 7.25 55.6 40.6 173.1
628 H 116.8 7.86 54.6 29.6 173.1
629 A 126.5 8.73 49.7 49.0 174.8
630 S 113.2 8.40 56.1 40.1 173.0
631 L 124.6 8.59 55.7 41.2 176.2
632 A 133.0 8.20 - - -
175
Copyright Acknowledgements
Contents of chapter 1 are from a review written for Environmental Health Perspectives, 110,
Fatemi, N., & Sarkar, B. Molecular mechanism of copper transport in Wilson disease, 695-698,
2002, with permission from The National Institute of Environmental Health Sciences (NIEHS).
Chapter 3 is adapted from publications in Inorganica Chimica Acta, 339, Fatemi, N., & Sarkar,
B. Insights into the mechanism of copper transport by the Wilson and Menkes disease copper-
transporting ATPase, 179-187, 2002, with permission from © Elsevier 2002, and Journal of
Bioenergetics and Biomembranes, 34, Fatemi, N., & Sarkar, B. Structural and Functional
Insights of Wilson Disease Copper-Transporting ATPase, 339-349, 2002, with permission from
© Springer Verlag 2002.
Chapter 4 is from Biochimica et Biophysica Acta, 1688, Tsay, M.J., Fatemi, N.,
Narindrasorasak, S., Forbes, J.R., & Sarkar, B. Identification of the “missing domain” of the rat
copper-transporting ATPase, ATP7B: insight into the structural and metal binding characteristics
of its N-terminal copper-binding domain, 78-85, 2004, with permission from © Elsevier 2004.
Chapter 5 is from Journal of Inorganic Biochemistry, 98, Myari, A., Hadjiliadis, N., Fatemi, N.,
& Sarkar, B. Copper(I) interaction with model peptides of WD6 and TM6 domains of Wilson
ATPase: regulatory and mechanistic implications, 1483-1494, 2004, with permission from ©
Elsevier 2004.
Chapter 6 is from Biochemistry, 49, Fatemi, N, Korzhnev, D, Velyvis, A, Sarkar, B, and
Forman-Kay, JD. NMR Characterization of Copper-binding domains 4-6 of ATP7B, 8468-8477,
2010, with permission from © ACS Publications 2010.