6
Dynamic allostery governs cyclophilin AHIV capsid interplay Manman Lu a,b,1 , Guangjin Hou a,b,1 , Huilan Zhang a,b , Christopher L. Suiter a,b , Jinwoo Ahn b,c , In-Ja L. Byeon b,c , Juan R. Perilla d , Christopher J. Langmead e , Ivan Hung f , Peter L. Gorkov f , Zhehong Gan f , William Brey f , Christopher Aiken b,g , Peijun Zhang b,c , Klaus Schulten d , Angela M. Gronenborn b,c,2 , and Tatyana Polenova a,b,2 a Department of Chemistry and Biochemistry, University of Delaware, Newark, DE 19716; b Pittsburgh Center for HIV Protein Interactions, University of Pittsburgh School of Medicine, Pittsburgh, PA 15260; c Department of Structural Biology, University of Pittsburgh School of Medicine, Pittsburgh, PA 15260; d Center for Biophysics and Computational Biology and Beckman Institute for Advanced Science and Technology and Departments of Physics, Chemistry, and Biochemistry, University of Illinois at UrbanaChampaign, Urbana, IL 61801; e School of Computer Science, Computational Biology Department, Carnegie Mellon University, Pittsburgh, PA 15213; f National High Magnetic Field Laboratory, Florida State University, Tallahassee, FL 32310; and g Department of Pathology, Microbiology and Immunology, Vanderbilt University School of Medicine, Nashville, TN 37232 Edited by Gerhard Wagner, Harvard Medical School, Boston, MA, and approved October 21, 2015 (received for review August 25, 2015) Host factor protein Cyclophilin A (CypA) regulates HIV-1 viral infec- tivity through direct interactions with the viral capsid, by an unknown mechanism. CypA can either promote or inhibit viral infection, depending on host cell type and HIV-1 capsid (CA) protein sequence. We have examined the role of conformational dynamics on the nanosecond to millisecond timescale in HIV-1 CA assemblies in the escape from CypA dependence, by magic-angle spinning (MAS) NMR and molecular dynamics (MD). Through the analysis of back- bone 1 H- 15 N and 1 H- 13 C dipolar tensors and peak intensities from 3D MAS NMR spectra of wild-type and the A92E and G94D CypA escape mutants, we demonstrate that assembled CA is dynamic, particularly in loop regions. The CypA loop in assembled wild-type CA from two strains exhibits unprecedented mobility on the nanosecond to micro- second timescales, and the experimental NMR dipolar order parame- ters are in quantitative agreement with those calculated from MD trajectories. Remarkably, the CypA loop dynamics of wild-type CA HXB2 assembly is significantly attenuated upon CypA binding, and the dynamics profiles of the A92E and G94D CypA escape mutants closely resemble that of wild-type CA assembly in complex with CypA. These results suggest that CypA loop dynamics is a deter- mining factor in HIV-1s escape from CypA dependence. magic angle spinning NMR | HIV-1 capsid | CA protein assemblies | escape mutations | conformational dynamics C yclophilin A (CypA), a peptidyl-prolyl isomerase, is a host factor critical in the regulation of the HIV-1 infection, in- volving a direct interaction with the capsid (CA) protein (13). The mechanism by which CypA modulates the viral infectivity is complex and poorly understood, being dependent on the CA pro- tein primary sequence and the host cell type (46). For example, it is known that mutations in the CypA-binding loop of the CA protein dramatically reduce virus infectivity (7, 8). The A92E and G94D escape mutants bind CypA with similar affinity to wild-type CA, but exhibit only 10% of the activity of wild-type CA in the presence of CypA, and full infectivity can be restored if CypA is inhibited with cyclosporin A in the host cells (8), as shown schematically in SI Appendix, Fig. S1. Alas, the molecular mechanisms underlying CypA escape remain elusive, despite numerous virological, bio- chemical, and structuralbiological studies. The present study investigates the internal conformational dynamics of a CA protein assembly. Although static structures of HIV-1 proteins and complexes with host factors provide important clues into their assembly architecture and conformational details of the interactions, structures alone are insufficient for understanding molecular mechanisms. It is well known that biological functions can be dynamically regulated, at multiple levels of organization, from internal dynamics of individual protein molecules (9) to entire cells. This dynamic regulation certainly also applies to HIV-1 because numerous dynamic processes are associated with HIV-1 assembly, disassembly, release, and maturation (10, 11). For example, we previously demonstrated that internal conformational dynamics of the CA protein and its structural plasticity determine its ability to assemble into pleiomorphic conical capsids (12, 13) (Fig. 1). We also uncovered that, in the HIV-1 CA-SP1 maturation intermediate, dynamic disorder in the SP1 peptide plays an important role in the final step of virus maturation, permitting condensation of CA into the cores of infectious virions (14). In this study, we examined the residue-specific mobility of CA protein from HXB2 and NL4-3 sequence polymorphs (SI Ap- pendix, Fig. S2) in tubular assemblies on the nanoseconds to milliseconds timescales. In particular, we compared wild-type and A92E and G94D escape mutants of the NL4-3 strain as well as wild- type HXB2 CA alone and in complex with CypA. As discussed previously (14, 15), tubular assemblies recapitulate the hexameric lattice, the predominant symmetry arrangement of the conical HIV-1 capsid core, illustrated in Fig. 1A. Dipolar tensors and resonance intensities extracted from a series of 2D and 3D ho- monuclear and heteronuclear magic-angle spinning (MAS) NMR experiments revealed that certain regions in both HXB2 and NL4-3 wild-type CA are unusually dynamic on all timescales. These motions are significantly attenuated upon CypA binding. Most remarkably, the dynamic profiles of the A92E and G94D escape mutants closely resemble that of CA when bound by CypA. To Significance The mechanisms of how Cyclophilin A (CypA) regulates HIV-1 infectivity remain poorly understood. We examined the role of dynamics in capsid (CA) protein assemblies by magic-angle- spinning NMR. The assembled CA is highly dynamic. Dipolar tensors calculated from molecular dynamics trajectories are in quantitative agreement with the NMR results. Motions in the CypA loop are sequence-dependent and attenuated in the es- cape mutants A92E and G94D. Dynamics are similar in escape mutants and CA/CypA complex. These findings suggest that CA escapes from CypA dependence through dynamic allostery. Thus, a host factors function in HIV infectivity may not be primarily associated with a structural change of the capsid core, but with altering its dynamics, such as the reduction of motions for the CypA loop. Author contributions: G.H., K.S., A.M.G., and T.P. designed research; M.L., G.H., H.Z., C.L.S., I.-J.L.B., J.R.P., C.J.L., I.H., P.L.G., Z.G., W.B., P.Z., and T.P. performed research; J.A., I.-J.L.B., and C.A. contributed new reagents/analytic tools; M.L., G.H., H.Z., C.L.S., A.M.G., and T.P. ana- lyzed data; M.L., G.H., H.Z., J.R.P., C.J.L., I.H., A.M.G., and T.P. wrote the paper; and P.L.G. and W.B. designed and built the Low-E MAS NMR probe that was used in this research. The authors declare no conflict of interest. This article is a PNAS Direct Submission. 1 M.L. and G.H. contributed equally to this work. 2 To whom correspondence may be addressed. Email: [email protected] or amg100@ pitt.edu. This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. 1073/pnas.1516920112/-/DCSupplemental. www.pnas.org/cgi/doi/10.1073/pnas.1516920112 PNAS | November 24, 2015 | vol. 112 | no. 47 | 1461714622 BIOPHYSICS AND COMPUTATIONAL BIOLOGY Downloaded by guest on November 13, 2021

Dynamic allostery governs cyclophilin A HIV capsid interplay

  • Upload
    others

  • View
    4

  • Download
    0

Embed Size (px)

Citation preview

Page 1: Dynamic allostery governs cyclophilin A HIV capsid interplay

Dynamic allostery governs cyclophilin A–HIVcapsid interplayManman Lua,b,1, Guangjin Houa,b,1, Huilan Zhanga,b, Christopher L. Suitera,b, Jinwoo Ahnb,c, In-Ja L. Byeonb,c,Juan R. Perillad, Christopher J. Langmeade, Ivan Hungf, Peter L. Gor’kovf, Zhehong Ganf, William Breyf,Christopher Aikenb,g, Peijun Zhangb,c, Klaus Schultend, Angela M. Gronenbornb,c,2, and Tatyana Polenovaa,b,2

aDepartment of Chemistry and Biochemistry, University of Delaware, Newark, DE 19716; bPittsburgh Center for HIV Protein Interactions, University ofPittsburgh School of Medicine, Pittsburgh, PA 15260; cDepartment of Structural Biology, University of Pittsburgh School of Medicine, Pittsburgh, PA 15260;dCenter for Biophysics and Computational Biology and Beckman Institute for Advanced Science and Technology and Departments of Physics, Chemistry, andBiochemistry, University of Illinois at Urbana–Champaign, Urbana, IL 61801; eSchool of Computer Science, Computational Biology Department, CarnegieMellon University, Pittsburgh, PA 15213; fNational High Magnetic Field Laboratory, Florida State University, Tallahassee, FL 32310; and gDepartment ofPathology, Microbiology and Immunology, Vanderbilt University School of Medicine, Nashville, TN 37232

Edited by Gerhard Wagner, Harvard Medical School, Boston, MA, and approved October 21, 2015 (received for review August 25, 2015)

Host factor protein Cyclophilin A (CypA) regulates HIV-1 viral infec-tivity through direct interactionswith the viral capsid, by an unknownmechanism. CypA can either promote or inhibit viral infection,depending on host cell type and HIV-1 capsid (CA) protein sequence.We have examined the role of conformational dynamics on thenanosecond to millisecond timescale in HIV-1 CA assemblies in theescape from CypA dependence, by magic-angle spinning (MAS)NMR and molecular dynamics (MD). Through the analysis of back-bone 1H-15N and 1H-13C dipolar tensors and peak intensities from 3DMAS NMR spectra of wild-type and the A92E and G94D CypA escapemutants, we demonstrate that assembled CA is dynamic, particularlyin loop regions. The CypA loop in assembled wild-type CA from twostrains exhibits unprecedented mobility on the nanosecond to micro-second timescales, and the experimental NMR dipolar order parame-ters are in quantitative agreement with those calculated from MDtrajectories. Remarkably, the CypA loop dynamics of wild-type CAHXB2 assembly is significantly attenuated upon CypA binding, andthe dynamics profiles of the A92E and G94D CypA escape mutantsclosely resemble that of wild-type CA assembly in complex withCypA. These results suggest that CypA loop dynamics is a deter-mining factor in HIV-1’s escape from CypA dependence.

magic angle spinning NMR | HIV-1 capsid | CA protein assemblies |escape mutations | conformational dynamics

Cyclophilin A (CypA), a peptidyl-prolyl isomerase, is a hostfactor critical in the regulation of the HIV-1 infection, in-

volving a direct interaction with the capsid (CA) protein (1–3).The mechanism by which CypA modulates the viral infectivity iscomplex and poorly understood, being dependent on the CA pro-tein primary sequence and the host cell type (4–6). For example, it isknown that mutations in the CypA-binding loop of the CA proteindramatically reduce virus infectivity (7, 8). The A92E and G94Descape mutants bind CypA with similar affinity to wild-type CA, butexhibit only 10% of the activity of wild-type CA in the presenceof CypA, and full infectivity can be restored if CypA is inhibitedwith cyclosporin A in the host cells (8), as shown schematically in SIAppendix, Fig. S1. Alas, the molecular mechanisms underlyingCypA escape remain elusive, despite numerous virological, bio-chemical, and structural–biological studies.The present study investigates the internal conformational

dynamics of a CA protein assembly. Although static structures ofHIV-1 proteins and complexes with host factors provide importantclues into their assembly architecture and conformational details ofthe interactions, structures alone are insufficient for understandingmolecular mechanisms. It is well known that biological functions canbe dynamically regulated, at multiple levels of organization, frominternal dynamics of individual protein molecules (9) to entire cells.This dynamic regulation certainly also applies to HIV-1 becausenumerous dynamic processes are associated with HIV-1 assembly,disassembly, release, and maturation (10, 11). For example, we

previously demonstrated that internal conformational dynamics ofthe CA protein and its structural plasticity determine its ability toassemble into pleiomorphic conical capsids (12, 13) (Fig. 1). Wealso uncovered that, in the HIV-1 CA-SP1 maturation intermediate,dynamic disorder in the SP1 peptide plays an important role in thefinal step of virus maturation, permitting condensation of CA intothe cores of infectious virions (14).In this study, we examined the residue-specific mobility of CA

protein from HXB2 and NL4-3 sequence polymorphs (SI Ap-pendix, Fig. S2) in tubular assemblies on the nanoseconds tomilliseconds timescales. In particular, we compared wild-type andA92E and G94D escape mutants of the NL4-3 strain as well as wild-type HXB2 CA alone and in complex with CypA. As discussedpreviously (14, 15), tubular assemblies recapitulate the hexamericlattice, the predominant symmetry arrangement of the conicalHIV-1 capsid core, illustrated in Fig. 1A. Dipolar tensors andresonance intensities extracted from a series of 2D and 3D ho-monuclear and heteronuclear magic-angle spinning (MAS) NMRexperiments revealed that certain regions in both HXB2 and NL4-3wild-type CA are unusually dynamic on all timescales. Thesemotions are significantly attenuated upon CypA binding. Mostremarkably, the dynamic profiles of the A92E and G94D escapemutants closely resemble that of CA when bound by CypA. To

Significance

The mechanisms of how Cyclophilin A (CypA) regulates HIV-1infectivity remain poorly understood. We examined the role ofdynamics in capsid (CA) protein assemblies by magic-angle-spinning NMR. The assembled CA is highly dynamic. Dipolartensors calculated from molecular dynamics trajectories are inquantitative agreement with the NMR results. Motions in theCypA loop are sequence-dependent and attenuated in the es-cape mutants A92E and G94D. Dynamics are similar in escapemutants and CA/CypA complex. These findings suggest that CAescapes from CypA dependence through dynamic allostery.Thus, a host factor’s function in HIV infectivity may not beprimarily associated with a structural change of the capsidcore, but with altering its dynamics, such as the reduction ofmotions for the CypA loop.

Author contributions: G.H., K.S., A.M.G., and T.P. designed research; M.L., G.H., H.Z., C.L.S.,I.-J.L.B., J.R.P., C.J.L., I.H., P.L.G., Z.G.,W.B., P.Z., and T.P. performed research; J.A., I.-J.L.B., andC.A. contributed new reagents/analytic tools; M.L., G.H., H.Z., C.L.S., A.M.G., and T.P. ana-lyzed data; M.L., G.H., H.Z., J.R.P., C.J.L., I.H., A.M.G., and T.P. wrote the paper; and P.L.G. andW.B. designed and built the Low-E MAS NMR probe that was used in this research.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.1M.L. and G.H. contributed equally to this work.2To whom correspondence may be addressed. Email: [email protected] or [email protected].

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1516920112/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1516920112 PNAS | November 24, 2015 | vol. 112 | no. 47 | 14617–14622

BIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

Dow

nloa

ded

by g

uest

on

Nov

embe

r 13

, 202

1

Page 2: Dynamic allostery governs cyclophilin A HIV capsid interplay

gain further understanding of the sequence-dependent dynamicsprofiles of CA assemblies, we performed extensive moleculardynamics (MD) simulations. The motionally averaged dipolartensors extracted from the MD trajectories are in remarkablequantitative agreement with the NMR results. Together, ourresults suggest that changes in the sequence-dependent con-formational dynamics may be a key determinant in the escapemechanism of HIV-1 CA capsid mutants from CypA dependence.

ResultsCA/CypA Assemblies: CypA Binding and Sample Morphology. In thepresence of high concentration of NaCl (0.5–2.4 M), the CAprotein assembles into tubes (14). As shown in Fig. 1 B and C,the assembly is highly efficient for wild-type CA, as well as theA92E and G94D escape mutants, enabling their MAS NMRcharacterization. To examine whether CypA binding interferes

with the tubular assemblies, we conducted a cosedimentationexperiment, in which varying concentrations of CypA were addedto preassembled CA tubes. As shown in Fig. 1B, CypA copre-cipitates with CA tubes, and, with increasing CypA concentra-tions, more CypA becomes part of the CA/CypA complex.Furthermore, our studies indicate that at the CA/CypA ratiosbelow 2:1 CypA disrupts the tubes. Therefore, we used a 2:1 CA:CypA ratio for our MAS NMR studies. At this ratio, we observedefficient binding of CypA to tubular CA assemblies without anydisruption of the tubes, as illustrated in Fig. 1C.

MAS NMR Spectra of CA and CA/CypA Assemblies. The five CA as-semblies under investigation (wild-type and A92E and G94D es-cape mutants of the NL4-3 strain, as well as wild-type HXB2 CAalone and in complex with CypA) yielded extremely well-resolvedMAS NMR spectra, confirming previous observations (14). We

Fig. 1. (A, Left) All-atom MD-derived model of mature HIV-1 capsid constructed on the basis of cryo-electron tomography (cryo-ET) and solution NMR studies(13). The capsid comprises 216 hexamers (orange) and 12 pentamers (blue) [Protein Data Bank (PDB) ID 3J3Y]. Structural organization of a hexamer ofhexamers (HOH) building block is illustrated in the expansion. Color coded are individual hexameric units comprising the HOH building block. (A, Right) The3D structure of CA monomer [HXB2 sequence polymorph [PDB file 3NTE (42)]. (B) Cosedimentation assay of CA with CypA illustrating the efficiency ofcosedimentation for different CA/CypA molar ratios. S, supernatant; P, pellet. (C) Transmission electron microscopy (TEM) images of tubular assemblies of CAand CA/CypA. (C, Upper) CA NL4-3 (Left), CA NL4-3 A92E (Center), and CA NL4-3 G94D (Right). (C, Lower) HXB2 (Left) and CA HXB2/CypA (Right).(D) Expansions around the aliphatic region for 2D NCA and combined R2-driven (CORD) MAS NMR spectra for CA HXB2 (black) and CA HXB2/CypA (orange),illustrating the multiple chemical shift perturbations observed upon formation of the complex. These perturbations are mapped onto the structure of CAmonomer (A) and are confined to flexible loops and residue variation sites. The spectra are recorded at 20.0 T and the MAS frequency of 14 kHz. (E) Ex-pansions of glycine regions for 2D NCA MAS NMR spectra for (from left to right): HXB2, HXB2/CypA, NL4-3, NL4-3 A92E, and NL4-3 G94D. Dashed linesindicate the G89 cross-peaks associated with cis- and trans-P90.

14618 | www.pnas.org/cgi/doi/10.1073/pnas.1516920112 Lu et al.

Dow

nloa

ded

by g

uest

on

Nov

embe

r 13

, 202

1

Page 3: Dynamic allostery governs cyclophilin A HIV capsid interplay

speculate that the high spectral resolution in these assemblies isattained owing to their high conformational homogeneity andwell-defined structure, similar to that observed in other supra-molecular protein assemblies (16, 17), and making MAS NMR anattractive method for their structural and dynamics analysis withatomic resolution. Indeed, the quality of the data is such that dif-ferences in chemical shifts upon CypA binding and for CypA mu-tants are readily observed. As illustrated in Fig. 1D and SI Appendix,Fig. S3, assemblies of CA with bound CypA or the A92E and G94Descape mutants exhibit multiple chemical shift changes, respectively.These changes map onto the residues in the CypA loop, which areaffected by the CypA binding or by the mutations in the loop resi-dues. It should be noted that the four residues that differ in theprimary sequence in the HXB2 and NL4-3 strains of HIV-1 ex-hibit signal changes as well (SI Appendix, Fig. S3), and so do theirneighbors. It is important to note that the chemical shift differ-ences are modest and do not exceed 2 ppm, indicating that thetertiary structure is not perturbed significantly. Indeed, secondarychemical shift analysis of the CypA loop residues reveals that thestructure is similar in all five samples.We also ascertained that the P90 residue in the CypA loop of CA

does not undergo cis–trans isomerization to any detectable extent inthe absence of CypA, as evidenced by a single N–Cα cross-peak forG89 associated with the trans-P90 in all samples investigated. Thisfinding is in contrast to solution NMR results on the CA N-terminaldomain, where the cis conformer for P90 was readily detected by usand others as a minor conformer (populated at ∼14% at roomtemperature) from two sets of resonances in G89 (SI Appendix, Fig.S7; refs. 18 and 19). The cis–trans isomerization is catalyzed byCypA and Nup358Cyp (18, 19). The uncatalyzed reaction is slow(exchange rates of <0.1 s−1), whereas in the presence of CypA andNUP358Cyp, isomerization rates of 6.6 and 12.1 s−1 were observed,respectively. In contrast, in the tubular assembly, P90 isomerization isabsent. This finding is important because, as discussed below, theCypA loop undergoes motions on the nanoseconds to millisecondstimescales, and it is necessary to establish that these motions are nota result of cis–trans proline isomerization. Interestingly, in CA/CypAassemblies, the N–Cα cross-peak(s) of G89 is (are) missing (Fig. 1E),suggesting the presence of motions on the microseconds to millisec-onds timescales that interfere with the polarization transfers inNCACX experiments. It is not likely that these motions may be as-sociated with a catalyzed cis–trans proline exchange process becausethe timescale for the latter is too slow to interfere with the NCACXexperiment.

Dynamics in CA/CypA Assemblies and CypA Escape Mutants onNanosecond to Millisecond Timescales: 1H-15N and 1H-13C DipolarInteraction Parameters and Spectral Intensities. The one-bond1H-15NH and 1H-13Cα dipolar coupling constants in the absence ofany motion are 11.34 (20) and 22.7 (21) kHz, respectively. Ifmotions that are faster than the magnitude of the rigid-limitdipolar coupling constant (microseconds to nanoseconds) arepresent, the dipolar tensor is dynamically averaged, resulting in areduced effective dipolar coupling (22). The dipolar couplingconstants and the asymmetry parameters of the motionally averageddipolar tensors are very sensitive probes of motional amplitudes andsymmetries in the nanosecond to microsecond timescales, as we andothers have demonstrated (22–24). We used 3D RN-symmetrybased experiments (25) for the measurement of 1H-15NH and 1H-13Cα

dipolar lineshapes of individual resonances of CA assemblies. Thedipolar interaction parameters were extracted by numerical simula-tions of the dipolar lineshapes (Materials and Methods).The 1H-15N and 1H-13C dipolar order parameters (S) are plotted

vs. residue number in Fig. 2 and SI Appendix, Fig. S4, respectively.The corresponding dipolar interaction parameters are compiled in SIAppendix, Table S1. As illustrated in Fig. 3 and SI Appendix, Figs. S5and S6, the 1H-15N and 1H-13C dipolar lineshapes for the fivesamples are remarkably diverse, suggesting that a large number ofresidues undergo motions on the nanoseconds to microsecondstimescales, as evidenced by their dynamically averaged lineshapesand greatly reduced dipolar order parameters. At the same time,

for many resonances, the lineshapes and the corresponding dipolarinteraction parameters are characteristic of rigid-limit behavior,with the majority of the rigid residues belonging to alpha helices,which are the main secondary structure element in the CA protein.For wild-type CA from both the HXB2 and NL4-3 strains,

mobile residues are found in the CypA-binding loop as well as inseveral short loops that connect helices 1 and 2; 2 and 3; 3 and4; 9 and 10; and 10 and 11, as well as between the N-terminal beta-hairpin and helix 1. Our results are in agreement with prior so-lution NMR studies on unassembled N-terminal domain (NTD)and full-length CA NL4-3, which identified the same mobile andrigid regions of the protein on nanosecond to microsecond time-scales, on the basis of R1, R2, and heteronuclear NOEs (26, 27). Ithad also been reported that loops in assembled CA exhibit acertain degree of mobility (28). However, completely unexpectedwas our finding that the dipolar order parameters are reducedtremendously for resonances of the CypA-binding loop residues,particularly those of G89 and A92. Resonances of these two res-idues, which are located in the middle of the loop, exhibit essen-tially isotropic lineshapes, indicating that the H–N bond vectorssample a very large conformational space. The fact that we areable to detect these loop residue resonances in the 3D spectra,despite their high flexibility, is very unusual. Indeed, it is fortuitousthat the associated motions fall into a regime where they are fastenough not to interfere with cross-polarization (CP), decoupling,and MAS. Other resonances of residues in the CypA loop are alsodynamic on the microsecond to nanosecond timescales, and theirlineshapes are characteristic of rhombic dipolar tensors (e.g., H87and A88). Interestingly, several residues that reside in the hingesof the different loops exhibit two conformations—a dynamic one(with reduced S) and a static one (with rigid-limit S)—despite asingle chemical shift (Figs. 2 and 3). Overall, the above findingsreflect an unprecedented degree of motion for the loops in wild-type CA protein assemblies from both HXB2 and NL4-3 strains.

Fig. 2. The 1H-15N S in tubular assemblies of HXB2 CA (A) and NL4-3 A92ECA (B) plotted vs. the residue number. Experimental MAS NMR 1H-15N S forassigned residues are shown as markers. Red markers for HXB2 associatedwith several residues represent a minor conformer. 1H-15N S computed fromMD trajectories are shown as lines. The computed S of NL4-3 A92E wereextracted from the following MD trajectories: averaged over six CA moleculescomprising different hexamers in one HOH building block (solid lines color-codeddifferently) and averaged over all molecules in a HOH building block (dashedyellow line). The MD trajectories for the HOH building block are extracted fromall-atom MD-derived model of mature HIV-1 capsid (PDB ID 3J3Y) (13).

Lu et al. PNAS | November 24, 2015 | vol. 112 | no. 47 | 14619

BIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

Dow

nloa

ded

by g

uest

on

Nov

embe

r 13

, 202

1

Page 4: Dynamic allostery governs cyclophilin A HIV capsid interplay

In contrast, in the complex assemblies of CA/CypA, the mo-tions of residues in all loop regions are significantly reducedcompared with the CA assemblies in the absence of CypA (Figs.

2 and 3). The CypA loop residue A92 no longer exhibits isotropiclineshapes, and the dipolar order parameter is 0.44 (SI Appendix,Table S1). For G89, no amide resonances were detected, sug-gesting that motions occurring in the intermediate regime withrespect to the CP and/or decoupling and/or MAS are present,rendering the corresponding peak in the 3D experiments invisible.This finding is not surprising per se, given that CypA binds directlyto the CypA loop (2) and may attenuate motion. Alternatively, theG89 peak disappearance may be associated with the conformationalexchange associated with cis–trans isomerization of the P90 peptidebond, but this possibility is not very likely because the isomerizationtimescale is too slow to interfere with the MAS NMR experiments.Also unexpected was that motions of other loop regions in the CAprotein were attenuated (SI Appendix, Table S1).Surprisingly, the dynamics profiles of assemblies of the two

CypA escape mutants, A92E and G94D, resemble that of theCA/CypA rather than wild-type CA assemblies. The dipolarlineshapes associated with the CypA-binding loop residues inboth mutants reveal that motions in this loop are reduced con-siderably, compared with wild-type CA assemblies, as illustratedin Fig. 3 and SI Appendix, Table S1. Remarkably, the dipolarorder parameters of the CypA-binding loop residues in the G94Dmutant closely match those in the CA/CypA assembly. Interestingly,earlier-solution NMR studies on an unassembled NTD of theG94D mutant CA indicated no changes in dynamics or conforma-tion with respect to the wild-type CA (26), in contrast with ourresults. Therefore, it appears that in this mutant the modulation ofCypA loop conformation and dynamics requires the presence of theC-terminal domain (CTD) and/or assembly. For the A92E mutant,the S values are higher than those in G94D and CA/CypA, indi-cating that motions are further restricted. The lineshapes of G89 andA92/E92 are clearly anisotropic with S of 0.21(G89)/0.32(A92) and0.43(G89)/0.55(E92) for the G94D and A92E mutants, respectively.The above findings are also illustrated in Fig. 4, where we

mapped the N–H dipolar order parameters for all five assembliesonto the 3D structure of an isolated CA protein. Overall, theNMR dipolar parameters indicate (i) a remarkably high degreeof motions on nanosecond–microsecond timescales for residuesin the CypA-binding loop of wild-type CA from both HXB2 andNL4-3 strains; (ii) a significant decrease in the loop motions inCA/CypA assemblies and the two CypA escape mutants, A92Eand G94D; and (iii) similar dynamic profiles for CA/CypA,A92E, and G94D assemblies.We next turned our attention to dynamics on the microsecond–

millisecond timescales. As reported previously by us, suchmotions are seen in the hinge region of CA (residues 144–148)and are directly connected to the ability of the CA protein toassemble into pleiomorphic conical capsids (12). Dynamics inthis time regime can be detected qualitatively from the peak inten-sities in the heteronuclear correlation spectra because such motionswould interfere with CP, decoupling, or MAS, causing reductions inpeak intensity or complete peak disappearance.Fig. 4 displays normalized peak intensities plotted vs. residue

number for all five assemblies measured from the NCA spectraand their mapping onto the 3D structure of a single CA molecule.Missing or weakened signals are noted for many residues, indicatingthat assembled CA is mobile on this timescale. Consistent with ourfindings for nanosecond to microsecond dynamics, the loop regionsof wild-type CA exhibit significant motions in the microsecond tomillisecond regime. In the CA/CypA assemblies, the motions aresomewhat attenuated, but, interestingly, CypA-binding loop stillremains flexible in this time window. This observation is also madefor the CypA-binding loop in the G94D mutant, whereas in theA92E mutant these loop motions are somewhat slower.More surprisingly, the microsecond–millisecond timescale

motions are not restricted to only loops, but occur throughoutthe entire protein, including the helices of both the CTD and theNTD. Among the five assemblies, motions of the wild-type CAassembly are the fastest on this slow timescale, which, interestingly,are not perturbed significantly upon CypA binding. These motionsare likely important for the relative repositioning of the NTD and

Fig. 3. Summary of nanosecond to microsecond timescales dynamics for CypAloop residues observed in tubular assemblies of HIV-1 CA protein. Experi-mental 1H-15N S plotted vs. the residue number (A) and corresponding di-polar lineshapes in B (from top to bottom): HXB2, HXB2/CypA, NL4-3, NL4-3 A92,and NL4-3 G94D. The experimental and simulated lineshapes are shown as solidblack and dashed blue lines, respectively. The simulated dipolar lineshapes andthe corresponding S for the minor R100 and S102 conformers are shown withdashed red lines and red markers, respectively. Note the large variations in Samong the five CA assemblies and that CypA loop in the wild-type CA has in-creased mobility vis-à-vis the escape mutants and CA/CypA assemblies.

14620 | www.pnas.org/cgi/doi/10.1073/pnas.1516920112 Lu et al.

Dow

nloa

ded

by g

uest

on

Nov

embe

r 13

, 202

1

Page 5: Dynamic allostery governs cyclophilin A HIV capsid interplay

CTD around the hinge region, associated with the varied curvaturein tubular and conical assemblies.

MD Simulations and Calculations of NMR Dipolar Parameters. Thesurprising experimental observation of the unusually high extentof dynamics of the CypA-binding loop residues in CA assemblies,which is significantly affected by a single amino acid change,prompted us to scrutinize these results. To that end, we pursuedall-atom MD simulations. With the currently available compu-tational resources, we were able to calculate MD trajectories fora single molecule of wild-type CA (HXB2 strain) and for the conicalA92E CA mutant in the capsid assembly (NL4-3 strain) to 100 ns(13). These simulations were sufficiently long to recapitulate theexperimental NMR dipolar order parameters, as discussed be-low, but not long enough to evaluate the motions that occur onmicrosecond–millisecond timescales.As depicted in Fig. 2, the 1H-15N dipolar order parameters

computed from the MD trajectories for both wild-type CA andthe A92E mutant are in remarkable agreement with the exper-imental NMR values. Indeed, it is striking that the calculationsnot only correctly predict the flexible regions, such as CypA-binding loop, but that the calculated S values are also in goodquantitative agreement with the experiment. Specifically, the MDsimulations capture the experimentally observed differences in theCypA loop dynamics well, substantiating that the loop in wild-typeCA assemblies is significantly more flexible than that in the A92Emutant assemblies.We also performed calculations of dipolar order parameters

from multiple MD trajectories for different substructures of capsidassemblies. To that end, we extracted a hexamer of hexamers(HOH) unit from the all atom model of a conical capsid (13). Asillustrated in Fig. 2B, the dipolar order parameters calculated foreach hexamer in the HOH building block agree well with eachother and with the experimental results, indicating the validity of ourapproach. To our knowledge, this work establishes, for the first time,quantitative agreement between MD simulation-derived and experi-mental order parameters for such a large assembly.

An important advantage of using the MD simulations is thatthe trajectories of the CypA-binding loop in the two CA sequencepolymorphs can be visualized as angular probability distributionsand in 3D scatter plots that depict the bond vector orientations inspace. These are plotted in SI Appendix, Fig. S8, for several rep-resentative rigid CA residues of CA, as well as for mobile residuesin the CypA-binding loop. Clearly, the remarkable mobility of theCypA-binding loop in wild-type CA assemblies is associated withthese loop residues sampling a large portion of the available 3Dspace. Interestingly, the corresponding angular probability dis-tributions do not exhibit a canonical Gaussian shape and mayeven possess two local maxima, such as seen for residue G94.

DiscussionThe effect of CypA on the HIV-1 infection is complex and poorlyunderstood. In humans, the HIV-1 infectivity is enhanced by theinteraction of CypA with the capsid (29, 30). On the contrary, inother primates, CypA interferes with the viral infection by en-hancing the activity of the restriction factors TRIM5α andTRIM-Cyp (31, 32). Moreover, depending on the cell type, thecapsid can be either stabilized or destabilized by CypA (5, 6).Escape mutations in the CypA loop result in the loss of 90% ofthe HIV-1 infectivity in the presence of CypA; the activity isrestored upon CypA inhibition (8).Changes in the overall CA structures due to mutations are

unlikely the explanation for the escape from CypA dependence:Given the relatively modest chemical shift differences, most ofwhich are limited to CypA loop residues, we are confident thatthe 3D structures of the wild-type CA and the two CypA escapemutants are very similar. Furthermore, the binding affinities ofCypA to the monomeric CypA escape mutants do not differmuch from those to the wild-type CA (8). We therefore suggestthat the mechanism of capsid’s escape from CypA dependenceinvolves a change in dynamics.The NMR and MD simulation data reveal an unequivocal

relationship between the dynamic profile of CA and the mutants.Wild-type CA of the HXB2 and NL4-3 strains is remarkably

Fig. 4. Summary of nanosecond to millisecond time-scales dynamics observed in tubular assemblies of HIV-1CA protein. (A) Peak intensities observed in the 2DNCACX MAS NMR correlation spectra plotted vs. resi-due number (for nonoverlapping peaks with resonanceassignments) for tubular assemblies of CA NL4-3 (red),CA NL4-3 A92E (light blue), CA NL4-3 G94D (green), CAHXB2 (black), and CA HXB2/CypA (yellow). The sec-ondary structure is indicated at the bottom: H, helix;S, sheet; L, loop. Note the attenuated peak intensi-ties or missing peaks in CypA loop and throughoutthe protein, which indicate the presence of motionson microsecond to millisecond timescales. (B) 1H-15NS (Upper) and normalized peak intensities (Lower)mapped onto the CA structure for (from left toright): CA HXB2, CA HXB2/CypA, CA NL4-3, CA NL4-3A92E, and CA NL4-3 G94D.

Lu et al. PNAS | November 24, 2015 | vol. 112 | no. 47 | 14621

BIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

Dow

nloa

ded

by g

uest

on

Nov

embe

r 13

, 202

1

Page 6: Dynamic allostery governs cyclophilin A HIV capsid interplay

flexible in the assembled state. In particular, the extent of mo-tional averaging seen for CypA-binding loop residues appears tobe unprecedented and, to our knowledge, has not been seen inany other proteins that have been investigated by MAS NMR.Binding of CypA quenches the motions in the CypA loop, as wellas of many other residues in the protein.To our surprise, the two CypA escape mutants exhibited dy-

namic signatures very similar to that of wild-type CA in complexwith CypA. Specifically, the CypA-binding loop dynamics in theA92E and G94D mutants resembles that of the CA/CypA complex.Remarkably, MD simulations of wild-type CA and the A92E mutantnear quantitatively recapitulate the experimentally derived NMRdipolar tensor parameters, suggesting that CypA escape is intimatelycoupled to dynamics.We further propose that dynamic allostery is the main mech-

anism for regulation of viral infectivity, providing a more subtleway for fine-tuning activity than would result from an overallchange in the structure of the protein. In fact, dynamic allosteryas a regulatory mechanism is commonly used in many biologicalprocesses (33), including, but not limited to, receptor activationin signaling (34), enzymatic catalysis (35, 36) including viral matu-ration involving proteolysis (37), intracellular transport (38–40),and others.

Concluding RemarksUsing MAS NMR experiments and MD simulations, we investigatedconformational dynamics in HIV-1 CA protein assemblies. Wediscovered that CA in the assembled state is highly mobile on the

nanosecond to millisecond timescales. A surprising finding wasthat the motional signatures of residues in the CypA-binding loop ofCypA escape mutants are similar to those in CA/CypA assemblies,suggesting a previously unidentified mechanism for CypA escape bydynamic allostery. New therapeutic intervention strategies may beenvisioned through modulation of CypA-binding loop dynamics bysmall-molecule interactors. Finally, our MAS NMR and MD ap-proach is also applicable for the analysis of CA assemblies with otherhost factors.

Materials and MethodsCA (HXB2, NL4-3, NL4-3/A92E, and NL4-3/G94D variants) and CypA wereexpressed and purified by using the same protocol as reported (14, 41).Tubular assemblies of CA protein were prepared and their morphologycharacterized as described (14) and detailed in SI Appendix, SI Text. Detailsof MAS NMR experiments, MD simulations, and spectral and dynamicsanalysis are described in SI Appendix, SI Text.

ACKNOWLEDGMENTS. This work was supported by the National Institutesof Health (NIH) National Institute of General Medical Sciences Grant P50GM082251; the National Science Foundation (NSF) Grant CHE0959496 (forthe acquisition of the 850 MHz NMR spectrometer at the University of Del-aware); and NIH Grants P30GM103519 and P30GM110758 (for the support ofcore instrumentation infrastructure at the University of Delaware). Workperformed at the National High Magnetic Field Laboratory was supportedby NSF Grant DMR-1157490 and the State of Florida. K.S. and J.R.P. weresupported by NIH Grants 9P41GM104601 and R01GM067887 and NSF GrantPHY1430124. MD simulations on the assembled CA A92E were performed onthe Blue Waters Supercomputers, supported by NSF Grants OCI-0725070 andACI-1238993 under Petascale Computational Resource Grant ACI-1440026.

1. Colgan J, Yuan HE, Franke EK, Luban J (1996) Binding of the human immunodefi-ciency virus type 1 Gag polyprotein to cyclophilin A is mediated by the central regionof capsid and requires Gag dimerization. J Virol 70(7):4299–4310.

2. Gamble TR, et al. (1996) Crystal structure of human cyclophilin A bound to the amino-terminal domain of HIV-1 capsid. Cell 87(7):1285–1294.

3. Luban J, Bossolt KL, Franke EK, Kalpana GV, Goff SP (1993) Human immunodeficiencyvirus type 1 Gag protein binds to cyclophilins A and B. Cell 73(6):1067–1078.

4. Hatziioannou T, Perez-Caballero D, Cowan S, Bieniasz PD (2005) Cyclophilin interac-tions with incoming human immunodeficiency virus type 1 capsids with opposingeffects on infectivity in human cells. J Virol 79(1):176–183.

5. Li Y, Kar AK, Sodroski J (2009) Target cell type-dependent modulation of human immu-nodeficiency virus type 1 capsid disassembly by cyclophilin A. J Virol 83(21):10951–10962.

6. Shah VB, et al. (2013) The host proteins transportin SR2/TNPO3 and cyclophilin A exertopposing effects on HIV-1 uncoating. J Virol 87(1):422–432.

7. Veillette M, et al. (2013) The V86M mutation in HIV-1 capsid confers resistance toTRIM5α by abrogation of cyclophilin A-dependent restriction and enhancement ofviral nuclear import. Retrovirology 10:25.

8. Ylinen LM, et al. (2009) Cyclophilin A levels dictate infection efficiency of human immu-nodeficiency virus type 1 capsid escape mutants A92E and G94D. J Virol 83(4):2044–2047.

9. Tzeng SR, Kalodimos CG (2012) Protein activity regulation by conformational entropy.Nature 488(7410):236–240.

10. Freed EO (2015) HIV-1 assembly, release and maturation. Nat Rev Microbiol 13(8):484–496.

11. Müller B, Krijnse-Locker J (2014) Imaging of HIV assembly and release. Methods MolBiol 1087:167–184.

12. Byeon IJ, et al. (2012) Motions on the millisecond time scale and multiple confor-mations of HIV-1 capsid protein: Implications for structural polymorphism of CA as-semblies. J Am Chem Soc 134(14):6455–6466.

13. Zhao G, et al. (2013) Mature HIV-1 capsid structure by cryo-electron microscopy andall-atom molecular dynamics. Nature 497(7451):643–646.

14. Han Y, et al. (2013) Magic angle spinning NMR reveals sequence-dependent structuralplasticity, dynamics, and the spacer peptide 1 conformation in HIV-1 capsid proteinassemblies. J Am Chem Soc 135(47):17793–17803.

15. Byeon IJ, et al. (2009) Structural convergence between cryo-EM and NMR revealsintersubunit interactions critical for HIV-1 capsid function. Cell 139(4):780–790.

16. Demers JP, et al. (2014) High-resolution structure of the Shigella type-III secretionneedle by solid-state NMR and cryo-electron microscopy. Nat Commun 5:4976.

17. Bertini I, et al. (2011) Solid-state NMR of proteins sedimented by ultracentrifugation.Proc Natl Acad Sci USA 108(26):10396–10399.

18. Bichel K, et al. (2013) HIV-1 capsid undergoes coupled binding and isomerization bythe nuclear pore protein NUP358. Retrovirology 10:81.

19. Bosco DA, Eisenmesser EZ, Pochapsky S, Sundquist WI, Kern D (2002) Catalysis of cis/trans isomerization in native HIV-1 capsid by human cyclophilin A. Proc Natl Acad SciUSA 99(8):5247–5252.

20. Yao L, Vögeli B, Ying J, Bax A (2008) NMR determination of amide N-H equilibriumbond length from concerted dipolar coupling measurements. J Am Chem Soc 130(49):16518–16520.

21. Al-Karaghouli AR, Koetzle TF (1975) Neutron diffraction study of l-phenylalaninehydrochloride. Acta Crystallogr B 31(10):2461–2465.

22. Hong M, Su Y (2011) Structure and dynamics of cationic membrane peptides andproteins: Insights from solid-state NMR. Protein Sci 20(4):641–655.

23. Yang J, Tasayco ML, Polenova T (2009) Dynamics of reassembled thioredoxin studiedby magic angle spinning NMR: Snapshots from different time scales. J Am Chem Soc131(38):13690–13702.

24. Meier BH, Graf F, Ernst RR (1982) Structure and dynamics of intramolecular hydrogenbonds in carboxylic acid dimers: A solid state NMR study. J Chem Phys 76(2):767–774.

25. Hou G, Byeon IJ, Ahn J, Gronenborn AM, Polenova T (2011) 1H-13C/1H-15N hetero-nuclear dipolar recoupling by R-symmetry sequences under fast magic angle spinning fordynamics analysis of biological and organic solids. J Am Chem Soc 133(46):18646–18655.

26. Campos-Olivas R, Summers MF (1999) Backbone dynamics of the N-terminal domainof the HIV-1 capsid protein and comparison with the G94D mutant conferring cy-closporin resistance/dependence. Biochemistry 38(32):10262–10271.

27. Deshmukh L, et al. (2013) Structure and dynamics of full-length HIV-1 capsid proteinin solution. J Am Chem Soc 135(43):16133–16147.

28. Bayro MJ, Chen B, Yau WM, Tycko R (2014) Site-specific structural variations accom-panying tubular assembly of the HIV-1 capsid protein. J Mol Biol 426(5):1109–1127.

29. Ikeda Y, Ylinen LM, Kahar-Bador M, Towers GJ (2004) Influence of gag on humanimmunodeficiency virus type 1 species-specific tropism. J Virol 78(21):11816–11822.

30. Song C, Aiken C (2007) Analysis of human cell heterokaryons demonstrates thattarget cell restriction of cyclosporine-resistant human immunodeficiency virus type 1mutants is genetically dominant. J Virol 81(21):11946–11956.

31. Berthoux L, Sebastian S, Sokolskaja E, Luban J (2005) Cyclophilin A is required forTRIM5α-mediated resistance to HIV-1 in Old World monkey cells. Proc Natl Acad SciUSA 102(41):14849–14853.

32. Towers GJ, et al. (2003) Cyclophilin A modulates the sensitivity of HIV-1 to host re-striction factors. Nat Med 9(9):1138–1143.

33. Ma B, Tsai C-J, Haliloglu T, Nussinov R (2011) Dynamic allostery: Linkers are not merelyflexible. Structure 19(7):907–917.

34. Khoury E, Clément S, Laporte SA (2014) Allosteric and biased g protein-coupled receptorsignaling regulation: potentials for new therapeutics. Front Endocrinol (Lausanne) 5:68.

35. Williams JC, McDermott AE (1995) Dynamics of the flexible loop of triosephosphateisomerase: the loop motion is not ligand gated. Biochemistry 34(26):8309–8319.

36. Goodey NM, Benkovic SJ (2008) Allosteric regulation and catalysis emerge via acommon route. Nat Chem Biol 4(8):474–482.

37. Domitrovic T, et al. (2013) Virus assembly and maturation: Auto-regulation throughallosteric molecular switches. J Mol Biol 425(9):1488–1496.

38. Moughamian AJ, Osborn GE, Lazarus JE, Maday S, Holzbaur EL (2013) Ordered re-cruitment of dynactin to the microtubule plus-end is required for efficient initiationof retrograde axonal transport. J Neurosci 33(32):13190–13203.

39. Cao L, et al. (2014) The structure of apo-kinesin bound to tubulin links the nucleotidecycle to movement. Nat Commun 5:5364.

40. Gigant B, et al. (2013) Structure of a kinesin-tubulin complex and implications forkinesin motility. Nat Struct Mol Biol 20(8):1001–1007.

41. Han Y, et al. (2010) Solid-state NMR studies of HIV-1 capsid protein assemblies. J AmChem Soc 132(6):1976–1987.

42. Du S, et al. (2011) Structure of the HIV-1 full-length capsid protein in a conforma-tionally trapped unassembled state induced by small-molecule binding. J Mol Biol406(3):371–386.

14622 | www.pnas.org/cgi/doi/10.1073/pnas.1516920112 Lu et al.

Dow

nloa

ded

by g

uest

on

Nov

embe

r 13

, 202

1