Mechanism of the CO-sensing heme protein CooA: New insights from the truncated heme domain and UVRR spectroscopy

Available online at www.sciencedirect.comJOURNAL OF

www.elsevier.com/locate/jinorgbio

Journal of Inorganic Biochemistry 101 (2007) 1776–1785

InorganicBiochemistry

Mechanism of the CO-sensing heme protein CooA: New insightsfrom the truncated heme domain and UVRR spectroscopy

Mohammed Ibrahim a, Michael Kuchinskas b, Hwan Youn c, Robert L. Kerby c,Gary P. Roberts c, Thomas L. Poulos b, Thomas G. Spiro a,*

a Department of Chemistry, Princeton University, Princeton, NJ 08544, United Statesb Department of Molecular Biology and Biochemistry, University of California, Irvine, CA 92697-3900, United States

c Department of Bacteriology, University of Wisconsin, Madison, WI 53706, United States

Received 29 May 2007; received in revised form 5 July 2007; accepted 5 July 2007Available online 18 July 2007

In memory of Edward I. Stiefel

Abstract

The bacterial CO-sensing heme protein CooA activates expression of genes whose products perform CO-metabolism by binding itstarget DNA in response to CO binding. The required conformational change has been proposed to result from CO-induced displacementof the heme and of the adjacent C-helix, which connects the sensory and DNA-binding domains. Support for this proposal comes fromUV Resonance Raman (UVRR) spectroscopy, which reveals a more hydrophobic environment for the C-helix residue Trp110 when CObinds. In addition, we find a tyrosine UVRR response, which is attributable to weakening of a Tyr55-Glu83 H-bond that anchors theproximal side of the heme. Both Trp and Tyr responses are augmented in the heme domain when the DNA-binding domain has beenremoved, apparently reflecting loss of the inter-domain restraint. This augmentation is abolished by a Glu83Gln substitution, whichweakens the anchoring H-bond. The CO recombination rate following photolysis of the CO adduct is similar for truncated and full-length protein, though truncation does increase the rate of CO association in the absence of photolysis; together these data indicate thattruncation causes a faster dissociation of the endogenous Pro2 ligand. These findings are discussed in the light of structural evidence thatthe N-terminal tail, once released from the heme, selects the proper orientation of the DNA-binding domain, via docking interactions.� 2007 Elsevier Inc. All rights reserved.

Keywords: CooA; Carbon monoxide; Activation mechanism; UV resonance Raman spectroscopy

1. Introduction

Heme protein sensors have emerged as key biologicaltransducers, whose enzymatic or DNA-binding activity isregulated by the binding of diatomic molecules, CO, NOor O2 [1]. There is great interest in elucidating their mech-anisms of action. The transcription factor CooA is anexemplar of this class of proteins [2,3]. When CO bindsto its heme, CooA activates a suite of genes responsiblefor CO oxidation in CO-metabolizing bacteria.

0162-0134/$ - see front matter � 2007 Elsevier Inc. All rights reserved.

doi:10.1016/j.jinorgbio.2007.07.010

* Corresponding author.E-mail address: [email protected] (T.G. Spiro).

Structural, spectroscopic and functional analysis hasprovided a detailed picture of CooA from Rhodospirillum

rubrum (Rr), for which a mechanism of activation has beenproposed [4–7]. The protein is a homodimer, showingcooperativity in CO binding [8]. Each chain contains anN-terminal sensory domain and a C-terminal DNA-bind-ing domain, connected by a long C-helix [9]. The architec-ture is homologous to that of a cAMP-activatedtranscription factor, cAMP receptor protein (CRP) [10–12]. Fig. 1 compares the structure of CO-free (inactive)CooA with that of cAMP-bound (active) CRP, which sug-gests the required reorientation of the DNA-bindingdomains. A remarkable feature of the CooA structure is

mailto:[email protected]

Fig. 1. X-ray crystal structures of Rr-CooA (left, Protein Data Bank (PDB) code 1FT9) and CRP (PDB code 1G6N) showing the difference in orientationof the DNA binding domains, relative to the effector binding domains. In the CooA crystal, the C-helix connecting the effector and DNA binding domainsis fully extended in the B chain but is bent over in the A chain (though not as much as in CRP), probably due to crystal forces [9].

M. Ibrahim et al. / Journal of Inorganic Biochemistry 101 (2007) 1776–1785 1777

that each heme is ligated by the N-terminal Pro2 residue ofthe opposite chain. This endogenous ligand is displaced byCO [13].

Truncation of CooA at the C-helix interdomain connec-tor yields a sensory domain tr-CooA, whose structure isessentially the same as in the full-length protein [14]. How-ever, tr-CooA binds CO faster, and with higher affinity andcooperativity, reflecting a functional connection with theDNA-binding domain in the full-length protein.

Resonance Raman (RR) spectroscopy of key site-direc-ted variants has identified side-chains that interact with thebound CO, and has revealed a characteristic weakening ofthe bond to the proximal histidine ligand (Fe-His77), aneffect that is augmented by DNA binding [6,7]. This weak-ening is attributed to displacement of the heme into anadjacent cavity, accompanied by motion of the C-helix. Itis this motion that is proposed to permit the reorientationof the DNA-binding domains. A recent crystal structure ofthe CO-bound form of a variant locked into the activeform (from Carboxydothermus hydrogenoformans [Ch])fully supports this model [15]. In this structure, one ofthe subunits is without its heme, but the remaining subunitcontains CO-bound heme, and displays the heme and C-helix displacement predicted on the basis of the RR analy-sis (Fig. 2).

Kubo et al. [16] have provided additional support forthe heme and C-helix displacement mechanism by showingthat UV resonance enhancement of Raman bands associ-ated with the sole tryptophan residue, Trp110, is aug-mented upon CO binding. This residue is on the C-helix,and its indole side chain is directed toward the heme. TheUVRR intensification is attributed to a more hydrophobicenvironment produced by the heme and C-helix displace-

ment. This interpretation is also supported by the Ch-CooA–CO structure, which shows the homologous C-helixresidue, Leu115, displaced toward the heme (Fig. 2).

In the present study we confirm the observation of Kuboet al., and show that the effect is further augmented in tr-CooA, suggesting relaxation of constraints imposed bythe connection to the DNA-binding domain in the full-length protein. We have also discovered a tyrosine UVRRresponse to CO binding, which is also augmented in tr-CooA. This signal is attributed to weakening of a strongH-bond between Tyr55 and Glu83, which connects twoloop segments in the b-sheet structure proximal to theheme (Fig. 2); the homologous H-bond is weakened inthe Ch-CooA–CO structure. This assignment was con-firmed by a Glu83Gln replacement in tr-CooA, whichdiminishes the Tyr UVRR response. It also diminishesthe Trp110 response, showing that the Tyr55-Glu83 H-bond is critical to the C-helix displacement.

Finally, we measured CO recombination to 5-coordinateCooA and tr-CooA following a photolysis pulse. The rateis hardly affected by truncation, showing that the increasedrate of CO association to Pro2-bound protein is due toacceleration of Pro2 dissociation in tr-CooA.

These results are discussed in conjunction with the crys-tal structures to further develop the mechanism of CooAactivation.

2. Methods

2.1. CooA expression and purification

The purification of full-length wild-type (WT) CooAand Leu120Ser variant was performed with our standard

Fig. 2. Overlay of CO-bound LL-ChCooA (orange) and CO-free RrCooA (blue), aligned via residues 29–99 in Ch- and the homologous residues 24–94 inRr-CooA (A chain). (RMS = 1.30 A for 71 Ca atoms). Shifts in the portions of the hemes, and of the opposite C-helix (carrying the UVRR indicator,Trp110), are indicated by pink arrows. In LL-ChCooA–CO, the Arg138-Glu59 (Arg143-Glu64) saltbridge is broken, the hinge is bent and the D-helix,which is connected to the DNA-binding domain, is reoriented. The heme displacement weakens the His77-Asn42 (His82-Asn47) H-bond. Also shown arethe Tyr55-Glu83 (Tyr60-Gln88) contacts. The box is a close-up view of the heme pocket.

1778 M. Ibrahim et al. / Journal of Inorganic Biochemistry 101 (2007) 1776–1785

method as described previously [17]. The construction andpurification of truncated CooA (tr-CooA) has beendescribed previously [14]. The truncated Glu83Gln CooAvariant (tr-Glu83Gln) was constructed via PCR amplifica-tion of tr-CooA with primers designed to incorporate thedesired nucleotide change, as described elsewhere [18].The tr-Glu83Gln CooA variant was purified using the sameprocedure for tr-CooA. In all cases, the final protein prep-aration was >95% pure based on SDS–PAGE. The hemecontent of CooA preparations was estimated using theextinction coefficient of WT Fe(II)–CO CooA, and proteinconcentration was measured using the BCA assay (Pierce).

2.2. Sample preparation

Purified CooA was diluted into an appropriate buffer(25 mM MOPS/0.5 M NaCl, pH 7.4 for tr-CooA and itsGlu83Gln variant; 25 mM MOPS/0.1 M NaCl, pH 7.4 forall others) to give heme concentrations of �15–20 lM forvisible RR, �20–30 lM for visible time-resolved resonanceRaman (TR3), and �50–90 lM for UVRR experiments.For UVRR experiments, samples also contain �0.1 MNaClO4 as an internal standard. Ferrous and ferrous–COCooA samples were prepared as previously described [7].

2.3. Resonance Raman (RR) spectroscopy

Visible and UV-RR spectra were obtained, respectively,with excitation wavelengths of 406.7 nm from a Kr+ laser

(Spectra Physics, 2080-RS), and 229 nm from an intracav-ity doubled-argon ion laser (Innova 300 FReD, CoherentRadiation, Palo Alto, CA) in a backscattering samplegeometry. Photodissociation of the bound CO and sampledegradation were minimized by using low laser power (�1and �0.5 mW at the sample for visible and UVRR experi-ments, respectively) and by spinning the sample. For thevisible RR experiments, the scattered light was collectedand focused onto a triple spectrograph (Spex 1877)equipped with a CCD detector (Roper Scientific, Model7375-0001) operating at �110 �C. Spectra were calibratedwith dimethyl formamide and dimethylsulfoxide-d6. Forthe UVRR experiments, a single spectrograph (Spex1269) equipped with a UV enhanced CCD detector (Prince-ton Instruments, Model LN/CCD-1340/400) was used tocollect the scattered light. UVRR spectra were calibratedusing acetone.

For the visible TR3 measurements, the second harmonicof a Q-switched Nd:YLF laser (Photonics Industries Inter-national, GM-30-527) was used to pump a Ti:sapphirelaser (Photonics International TU-UV), which gave a nar-rowed laser frequency output (<0.1 cm�1) tunable between810 and 920 nm. The Ti:S laser output (�25 ns at 1 kHz)was frequency doubled using a non-linear lithium triboratecrystal to produce a 438-nm probe. The second harmonicoutput of a Q-switched Nd:YLF (527 nm) was used to pro-duce pump pulses (�250 ns at 1 kHz). The optimum pumplaser power to achieve maximum photolysis was 600 mW.Photolysis due to the probe laser itself was minimized by


keeping the power at �1 mW. A DG535 delay generator(Stanford Research Systems, Inc.) controlled the time delaybetween the two laser pulses (600 ns–250 ls). The beamswere overlapped and then focused with a pair of cylindricallenses onto the sample. The sample solution was containedin a sealed spinning NMR tube with a small magnetic stir-ring bar and was effectively mixed using a second magnetplaced outside the tube. The NMR tube was saturated withCO at a concentration of 1 mM. The scattered light wascollected and focused onto double spectrograph (Spex14018) equipped with a liquid nitrogen cooled CCD detec-tor (Roper Scientific, Model 7375-0001). Twelve 10-sacquisition scans were averaged. Spectra were calibratedwith dimethyl formamide. The spectra were deconvolutedusing the GRAMS/AI version 7.00 software (ThermoGalactic) and analyzed as described previously [8].

3. Results

3.1. UVRR Trp and Tyr signals

Excitation at 229 nm selectively enhances Raman bandsarising from sidechain vibrations of tryptophan and tyro-sine residues in proteins [19]. The band intensities areresponsive to the sidechain environments, especially solva-tion and H-bonding, which shift the resonant electronictransitions, thereby altering the resonance enhancementfactors. The 229 nm-excited UVRR spectra of CooA andits CO adduct are similar to those of other proteins(Fig. 3). Kubo et al. [16] reported that Trp band intensitiesare increased when CO binds, suggesting increased hydro-phobicity of the Trp environment. This result supportedthe previously proposed heme/C-helix displacement model[4–6], inasmuch as the only Trp residue, Trp110, is on theC-helix, not far from the heme, and the displacements were

1000 1200 1400 1600 1800

(Tr)x5

(E83Q-Tr)x5

(L120S)x5

(WT)x5WT-CO

WT+CO

1770

W16

+W18

1616

Y8a

1601

Y8b

1555

W3

1361

W7

1340

W7

1176

Y9a

1009

W16

932

ClO

4-

874

W17

ex = 229 nm

Raman shift (cm-1)

Fig. 3. RrCooA UVRR spectra with and without CO, and the differencespectra (CooA–CO minus CooA) for the indicated variants (5 · amplifi-cation). Band frequencies and assignments are labeled.

expected to further protect the Trp110 indole sidechainfrom solvent. The Ch-CooA–CO structure [15] supportsthis expectation, since the sidechain of the residue at thehomologous position, Leu115, is closer to the hydrophobicCO-binding pocket than is Trp110 in Rr-CooA(CaÆÆÆFe = 9.9 and 11.5 A, respectively – Fig. 2).

We confirmed Kubo et al.’s finding (Fig. 3) Subtractingthe CooA UVRR spectrum from that of CooA–CO leavesresidual intensity for all the Trp bands (labeled Wn – seeRef. [20] for assignments). Table 1 lists intensity differencesof the most prominent bands, W16 and W3, measured viaband deconvolution of the parent spectra, and expressed asa fraction of the CO adduct intensity. The estimated uncer-tainty is 1%. The augmentation on binding CO differs fordifferent Trp bands, reflecting different responses of thevibrational mode displacements to environmental change.

These intensity increments are diminished in theLeu120Ser variant of CooA (Fig. 3 and Table 1). Thisbehavior is consistent with independent evidence fromheme-resonant RR spectra for lesser heme displacementin Leu120Ser [7] (see below). The Leu120 sidechain formspart of the hydrophobic cavity toward which the hemeslides upon CO binding (Fig. 2), and substitution by thepolar Ser has been suggested to lower the driving forcefor the heme displacement [7].

However, the increments are augmented in tr-CooA(Fig. 3, Table 1), indicating an increase in the extent ofC-helix displacement upon truncation. As discussed below,this structural effect correlates with increased CO affinityand binding rate.

We also detect small but reproducible difference bandsfor WT CooA associated with Tyr vibrations (Fig. 3), espe-cially Y9a (see Ref. [21] for assignments). These did notappear in Kubo et al.’s data, which had somewhat lowersignal/noise. The Tyr difference signals are more prominentwhen the UVRR spectrum of tr-CooA is subtracted fromthat of tr-CooA–CO (Fig. 3). Rr-CooA has four Tyr resi-dues per monomer, while the tr-CooA has two. However,only Tyr55 is buried from solvent and the environmentsof the surface Tyr residues are unlikely to be affected bythe conformational change associated with CO binding.Tyr55 forms an H-bond to Glu83 (OÆÆÆO distance = 3.4 A)

Table 1Percentage UVRR intensity differences induced by CO binding, forselected bands of the indicated Rr-CooA variants

Modes DI/ICooA–CO (%)a

WT L120S E83Q-TR TR

Tryptophan

W16 10 5 8 17W3 5 4 5 8

Tyrosine

Y9a 3 0 6 9

a Measured by subtraction of the CooA from the CooA–CO bands, afterdeconvolution from the parent spectra. The uncertainty is estimated to be1%.

440 480 520

FeC

CO

1985

1979

1980

1982

1982487

483

485

487

487WT+DNA

L120S

E83Q-Tr

Tr

WT

Raman shift (cm-1)

1950 1975 2000

ex = 407 nm

Fig. 4. Heme-resonant FeC and CO stretching RR bands for the indicatedRrCooA–CO variants.

1920 1940 1960 1980 2000460

480

500

520

CooA

Fe

C

O

N

N

H

O OH

Fe

C

O

N

NH

CH

CNH

O

L120F

Mb

E83Q-Tr

Mb (H64V/V68T)

Mb (-H64)

Mb (L29F)

WT+DNA

Mb (H64)

L120S

WT,Tr

FeC

[cm

-1]

CO [cm-1]

Fig. 5. mFe–C/mCO back-bonding plot showing data for CO adducts ofmyoglobin variants (h) [6] and of the CooA variants in this (d) andprevious (s, [7]) studies. The CooA variants deviate horizontally from theMb line, as expected for weakening of the Fe–His bond. The lower andupper inserts show the proximal histidine H-bond arrangements in Mband CooA, respectively.


in the CO-free Rr-CooA [9], and in tr-CooA [14], whichshould affect the UVRR intensity [22].

We hypothesized that the CO-induced conformationalchange might weaken this H-bond, inducing an increasein resonance enhancement, as documented for other pro-teins with H-bonded Tyr residues [22]. The smaller numberof surface Tyr residues on tr-CooA accounts in part for thegreater relative prominence of the putative Tyr55 differencesignature. However, the percentage change in the Y9aintensity is triple that of the full-length protein (Table 1),whereas a doubling would have been expected on the basisof the number of Tyr’s.

The Tyr55 hypothesis is strengthened by the Ch-CooA–CO structure, in which the OÆÆÆO distance of the homolo-gous H-bond pair, Tyr60ÆÆÆGln88 (Fig. 2), is lengthenedto 4.3 A. However, the weaker H-bond might result fromthe altered conformation, or from Gln being a weaker H-bond acceptor than Glu. To explore the issue further, wereplaced Glu83 in tr-CooA with Gln. The difference spec-trum for the tr-Glu83Gln variant (Fig. 3) shows twice theY9a percentage intensity as WT Rr-CooA (Table 1), theratio expected from truncation if the Tyr55ÆÆÆGln83 H-bondis weakened to the same extent as the Tyr55ÆÆÆGlu83 H-bond in full-length protein.

The Trp110 percentage difference intensity is the samefor tr-Glu83Gln as for WT Rr-CooA (Fig. 3 and Table1), indicating the same extent of C-helix displacement.However, as noted above, the increments for tr-CooA itselfare higher, and the Tyr 9a difference is higher than in tr-Glu83Gln. This means that the Tyr55ÆÆÆGlu83 H-bond isconnected to the C-helix displacement, and that theGlu83Gln substitution abolishes the truncation-inducedincrease in the C-helix displacement.

3.2. mFe–C and mC–O

Fe–C and C–O stretching RR bands were recorded at407 nm (Fig. 4), in resonance with the heme Soret band.These data have been reported previously for WT CooA–CO, with and without bound DNA, and for the L120S var-iant [7]. For tr-CooA–CO, the band positions are similar tothose of WT protein, while they are both 2 cm�1 lower forthe Glu83Gln variant.

The mFe–C/mC–O data for CooA variants are consistentwith a pattern of variable strength of the proximal Fe–His bond [6,7]. The points on the mFe–C/mC–O plot (Fig. 5)deviate from the standard backbonding correlation formyoglobin variants along a horizontal line, at a positionindicating a hydrophobic CO binding pocket. This is thebehavior expected if the Fe–His bond is weakened whenthe His H-bond to a protein acceptor is weakened. Theresulting diminution in the His anion character diminishesFe–CO backbonding, thereby strengthening the C–O bondand weakening the Fe–C bond, but this weakening is com-pensated by concomitant Fe–C strengthening due to less-ened r competition from the His ligand. ConsequentlymC–O increases with little change in mFe–C [23,24]. The crys-

tal structure of Ch-CooA–CO [15] shows that proximal HisH-bonding does indeed weaken upon CO binding, theNÆÆÆO distance to the Asn acceptor increasing to 4.7 A fromthe 2.7 A seen in the crystal structure of CO-free Rr-CooA[9].

However, the extent of Fe–His bond weakening is vari-able, increasing (higher mC–O) when DNA binds to WTCooA, but decreasing for the variants Leu120Ser andLeu120Phe [7]. This behavior was interpreted as reflectingvariations in the extent of heme/C-helix displacement. Thisdisplacement is augmented by DNA binding, which stabi-lizes the fully active conformation of CooA, and it is inhib-ited by the Leu120Ser or Leu120Phe substitutions, near the

Fig. 6. Time-resolved resonance Raman spectra (438-nm probe and 527-nm pump) showing the m4 bands obtained after photolysis of CooA–CO atthe indicated delay times.

Fig. 7. Changes in the percentages of CooA–CO, CooA, and CooA (5c)as a function of time, based on deconvolution of the Raman m4 bands. Theinset shows the band deconvolution into components (see Section 2).


heme. The mFe–C and mC–O positions are the same in tr-CooA–CO as in WT CooA–CO, indicating the same extentof heme displacement. In the tr-Glu83Gln variant, how-ever, mC–O is shifted down 2 cm�1, consistent with inhibi-tion of the heme displacement. This observation indicatesthat the Glu83Gln substitution in tr-CooA reduces boththe heme and the C-helix displacement.

3.3. CO Photolysis and rebinding, monitored via heme RR

bands

The rate of rebinding after CO photolysis was measuredfor WT CooA and tr-CooA. As in our earlier study of WTCooA [8], we used time-resolved heme-resonant RR spec-troscopy to separately monitor CO-bound, Pro2-boundand 5-coordinate (5-c) heme (the immediate photolysisproduct). Part of the motivation of this experiment wasto see if rebinding of the endogenous Pro2 ligand to the5-c heme could be monitored by lengthening the photolysispulse to 250 ns (from the earlier 25 ns); this was accom-plished by utilizing a Q-switched YLF laser, with sufficientpower at 527 nm to completely photolyze CooA–CO viaexcitation in the heme Q absorption bands. Because of effi-cient geminate recombination, the maximum fraction of 5-cphotoproduct at the start of time-resolved measurementswas 30% with a 25 ns photolysis pulse [8]. By lengtheningthe pulse we were able to increase this yield to 45%. Wealso shifted the probe wavelength to 438 nm (from429 nm) to better enhance the 5-c heme signal (Soret bandmaximum at �440 nm, vs. 425 and 422 nm for CooA andCooA–CO).

These changes improved the signal strength substan-tially, allowing more precise monitoring of the time course.The time-resolved RR spectra are shown in Fig. 6 and thetime course of the species population is shown in Fig. 7. Asbefore [8], the strong porphyrin m4 RR band envelope wasdeconvoluted into contributions from heme-CO, heme-Pro2 and 5-c heme, and the band intensities were convertedto populations after correction for differential resonanceenhancement (equal for CooA and CooA–CO and 1.5-foldhigher for 5-c heme). The time-course is well-described bysingle exponential curves for loss of 5-c heme and forma-tion of heme-CO, both with 70 ls time constants, and pro-ceeding to 90% completion by the last time point, 250 ls.Since the solution was saturated with CO (1000 lM), thecorresponding recombination rate constant isk0CO ¼ 14 lM�1 s�1.

The heme-Pro2 population was undetectable over the250 ls time course. This negative observation is consistentwith our previous analysis [8] of pulse-probe absorptiondata, at longer times and lower CO concentration, whichindicated that Pro2 recombination occurs on the millisec-ond time scale (kPro � 1000 s�1).

The previous short-pulse RR (and absorption) datayielded biexponential decays [8], interpreted as two rebind-ing processes with k0CO ¼ 67–32 and 2 lM�1 s�1 (Aono andcoworkers [25] reported three exponential phases, with

k0CO ¼ 35, 6.8 and 1.2 lM�1 s�1). With the current, highersignal/noise data, we see no evidence for a second, slower


process, and find a single rebinding phase with a rate con-stant intermediate between the previously reported fast andslow phases. However, it is possible that our time-resolu-tion was insufficient to distinguish two phases, since,because of the longer photolysis pulse, our first time pointwas 600 ns.

The data for tr-CooA–CO (Figs. 6 and 7) are very sim-ilar, except that the fitted time constant is slightly shorter,58 ls. However, the difference is within the uncertainty ofthe measurements (±10 ls for each run). Thus, there isno significant difference between rebinding rates for full-length and truncated proteins.

4. Discussion

4.1. Heme and C-helix displacement

On the basis of mutational studies combined with RRspectroscopy we have proposed a heme/C-helix displace-ment model for the mechanism of CooA activation. In thismodel, displacement of the endogenous Pro2 ligand by COinduces an upward movement (in the orientation of Figs. 1and 2) of the heme into an adjacent hydrophobic cavity.This movement induces a concerted displacement of theC-helix toward the opposite heme, forming the CO-bindingpocket. The C-helix displacement rearranges contacts atthe ‘hinge’ between the C and D helices, allowing theDNA-binding domains to swing into the proper orienta-tion for binding. The concerted motions required for bothsubunits (the initially bound Pro2 ligand also belongs tothe opposite chain) accounts for the cooperativity in CObinding that is observed experimentally [8].

The recent crystal structure of a variant of Ch-CooA–CO that is poised in the active form [15] strongly supportsthis model. Rr- and Ch-CooA are highly similar, andsuperposition (Fig. 2) of the Rr-CooA (blue) and Ch-CooA–CO (orange) structures, reveals exactly the displace-ments anticipated by the activation model. On the basis ofmutational evidence regarding the proximity of Leu116(Rr-CooA), the upward motion of the heme was predicted[6] to be �2 A. The heme displacement seen in Fig. 2(arrow) is 2.05 A, and the homologous Ch-CooA residue,Leu121, is indeed part of the CO-binding pocket. Displace-ment of the opposite C-helix toward the heme was posited,in order to bring Rr-CooA residues Gly117 and Ile113 intocontact with the bound CO. Such a displacement is evidentin Fig. 2, and the homologous Ch-CooA residues, Gly122and Val118 are again part of the binding pocket. As aresult of the C-helix displacement Leu120 moves towardthe heme, and its replacement by the bulky Phe or the polarSer would be expected to inhibit the heme displacement, asinferred previously from the mFe–C/mC–O data [7], and nowfrom the Trp110 UVRR data for Leu120Ser.

The Ch-CooA–CO structure also reveals a weakened H-bond between the proximal His residue and a nearby Asnsidechain, again providing support for our analysis of thespectroscopic data, with respect to mFe–C/mC–O variation.

The horizontal mFe–C/mC–O deviations from the standardbackbonding line for CooA variants (Fig. 5) is evidencefor weakening of the Fe–His bond, as discussed above.Our hypothesis was that heme displacement weakens thisbond, either by mechanical tension, or by weakening theproximal His H-bonding. However, recent Density Func-tional Theory (DFT) calculations [24] indicate thatmechanical tension should produce vertical displacement(selective mFe–C increase) from the backbonding line, whileH-bond weakening does indeed produce horizontal dis-placement (selective mC–O increase). The structural compar-ison in Fig. 2 confirms this interpretation. The NÆÆÆOdistance is 2.7 A for the His77-Asn42 contact in CO-freeRr-CooA, but 4.7 A for the homologous His82-Asn47 con-tact in Ch-CooA–CO. (There had been some doubt aboutthe H-bond interpretation, because replacing Asn42 withthe non-H-bonding residue Ala, failed to weaken the Fe-His bond, as measured by the mFe–His stretching frequencyin a variant with significant 5-coordinate heme population[7]. However, it was recognized that this negative resultmight be explained by a water molecule entering the prox-imal region in the Asn42Ala variant, and forming a H-bond with His77, in place of Asn42.)

4.2. Truncation of the heme domain

The CO-induced UVRR intensity gain of Trp modes,discovered by Kubo et al. [16], is an independent monitorof C-helix displacement, which brings the Trp110 sidechaininto the hydrophobic CO binding pocket (Fig. 2). It istherefore of considerable interest that the extent of theTrp intensity change is higher for tr-CooA than it is forthe full-length protein (Table 1), even though mFe–C andmC–O are the same. We infer that the CO-induced heme dis-placement is unaltered when CooA is truncated, but thatthe C-helix displacement is increased. Evidently, C-helixdisplacement is restrained in the full-length protein, andthis restraint is released upon truncation. Truncationoccurs at residue 131, just before the ‘hinge’ for bendingbetween the C and D helices [14]. Thus, restraint in thefull-length protein is likely associated with this bendingmotion, which is required for proper alignment of theDNA-binding domain. As pointed out previously [14],the energy required by this conformational change is evi-dent in the diminished CO affinity (2-fold) and cooperativ-ity (n � 1.5 vs. �2) of the full-length, relative to thetruncated protein. The augmented C-helix displacementafter truncation can be seen as the mechanical manifesta-tion of the energy release.

Truncation also increases the rate of CO associationwith CooA (10-fold), as measured in stopped-flow experi-ments [14]. In the current work, we find essentially nochange in the rate of CO recombination after photolysisof CooA–CO. Thus, the CO binding pocket appears tobe unaffected by the truncation. This result supports theinference [14] that acceleration of CO association resultsfrom faster dissociation of the endogenous Pro2 ligand in


the truncated heme domain. The Rr-CooA crystal struc-ture [9] reveals that the ligating N-terminus, Pro2, is innon-bonded contacts with C-helix residues Ile113, Leu116and Gly117 (which are also part of the CO binding pocket,as noted above.) The loss of C-helix constraints upontruncation may loosen these contacts, facilitating Pro2dissociation.

4.3. C-helix displacement and domain reorientation

A number of contacts between different regions of CooAhave been identified that may mediate domain reorienta-tion [9]. One of these is a salt-bridge between the hinge-region residue Arg138 and Glu59, located at the tip ofthe b4/5 hairpin loop, which extends from the hemedomain (Fig. 2). One strand of this loop extends to theheme ligand, His77, providing a mechanical connectionbetween Glu59 and the heme. The displacement of theheme in the Ch-CooA–CO structure is accompanied bymovement of the b4/5 hairpin away from the C-helix(Fig. 2), and the homologous Arg143-Glu64 salt-bridge isbroken, allowing the hinge to bend. This salt-bridges wouldcontribute to the energy penalty for the CO-induced con-formation change, and is eliminated in tr-CooA. Highlight-ing the importance of the b4/5 hairpin connection ismutational evidence linking the C-helix residue 128 andb4/5 hairpin residue 61 in constitutively active variants ofthe homologous CRP protein [26].

The present results suggest that another contributingcontact may be the Tyr55-Glu83 H-bond (Fig. 2), whoseweakening accounts for the CO-induced intensity loss inTyr UVRR bands, in both full-length and truncated CooA.Replacement of the H-bond acceptor, Glu83 with Gln hasa dramatic effect on the spectral characteristics of tr-CooA.The augmented C-helix displacement is abolished, as seenby the return of the Trp110 UVRR difference intensity tothat of WT-CooA. But in addition, the extent of heme dis-placement is reduced (lower mC–O) to about the same extentas in the Leu120Ser variant of the full-length protein. Weinterpret this to mean that the CO-induced displacementof the heme and of the C-helix is diminished when theTyr55ÆÆÆGlu83 H-bond is weakened by the Gln substitution.Tyr55 is on the second strand of the b4/5 hairpin, and theH-bond to Glu83 connects it to another b loop, down-stream from the His77 ligand (Fig. 2). Thus, weakening thisH-bond may attenuate the connection between the hemeand b4/5 hairpin, and diminish the displacement whenCO binds. Although weakened, the substitute Tyr55-ÆÆÆGln83 H-bond is nevertheless further weakened as aresult of this partial displacement, as evident from theY8a UVRR enhancement, which is the same as in full-length WT CooA.

We speculate that the relaxed ligand specificity of Ch-CooA compared to that of Rr-CooA, e.g. the ability ofthe former but not the latter to bind imidazole [27], mayresult from substitution of Gln for Glu at the position 88(homologous to position 83 in Rr-CooA) weakening the

proximal constraint to heme motion. It is interesting thatin the crystal structure of imidazole-bound Ch-CooA,recently reported by Komori et al. [27], the heme is foundto be rotated, rather than displaced, unlike that in Ch-CooA–CO.

4.4. Role of the N-terminus in activation

Bending at the hinge region between helices C and D is anecessary but not sufficient condition for the proper align-ment of the DNA-binding domains in CooA. Thus, the Asubunit in the CO-free Rr-CooA structure has a bent hingeregion (Fig. 1), in contrast to the B subunit. But the DNA-binding domain of neither subunit is oriented properly forbinding to DNA (Fig. 1). The hinge region bending in theA subunit was attributed to the influence of crystal contacts[9]. Likewise the hinge region is bent, but the DNA-bindingdomains are improperly aligned in the crystal structure ofCh-CooA-Im [27]. Thus, it is apparent that, once the hingebends, a variety of conformations are available to theDNA-binding domain.

Borjigin et al. [15] have proposed that the N-terminusmight play a role in activation. In the adduct-free protein,the N-terminus binds the heme of the opposite chain, butdisplacement by CO was shown by RR spectroscopy toexpel the N-terminus from the region of the binding pocket[6]. In Ch-CooA–CO, the N-terminal tail is sandwichedbetween the DNA-binding domain and the heme domain,held in place by several tertiary contacts [15]. These con-tacts are proposed to lock the DNA-binding domain inthe active conformation. Such a role is supported by thesuggestion of Komori et al. [27] that in the imidazoleadduct of Ch-CooA, an H-bond from the backbone car-bonyl of Met5, near the N-terminus, to the Ne atom ofthe bound imidazole, restrains the conformation changerequired for activity. This H-bond would in fact preventdocking of the N-terminal tail between the heme andDNA-binding domains.

The picture that emerges from these considerations isthat CO binding to CooA releases the bound N-terminus,and induces concerted heme and C-helix displacements.These are impelled by desolvation forces that drive theheme further into the protein interior, providing a hydro-phobic binding pocket for CO. The displacements releaseconstraints at the b4/5 hairpin allowing the hinge to bend.However, the protein is then in a ‘flexible’ state, with arange of orientations available to the N-terminal tail andto the DNA-binding domain. It has been reported thatelectron density maps on crystals of CooA variants, oftengive weak density for the DNA-binding domain, reflectingdisorder [15], and a small angle X-ray scattering study hasindicated surprisingly little shape variation between CooAand CooA–CO [28]. However, docking of the N-terminaltail between the heme- and DNA-binding domains canlock the latter in the proper orientation for DNA-binding.

It is consistent with this ‘flexible’ view of CooA thatDNA binding to CooA–CO augments the heme/C-helix

Fig. 8. View of LL-ChCooA–CO (chain A) [15], with Pro and Tyrmodeled as substitutions for Leu7 and Leu127, showing the proposed H-bond docking of the N-terminal tail in the active conformation of (A/R)YLLRL variants of RrCooA, in which Pro2 and Tyr122 are theposition 7 and 127 homologs.


displacement, as revealed by the mFe–C/mC–O data [6], and byan increase in the Trp110 UVRR intensity change [16]. Inthe absence of DNA, CooA–CO appears to be in a confor-mational equilibrium between active and inactive forms.The new crystallographic data suggests that CO bindingfrees the N-terminal tail, which induces activity by dockingbetween the domains in the active form. There appears tobe a dynamic equilibrium between docked and undockedmolecules, which is drawn toward the docked conforma-tion by DNA binding.

Supporting this view is the finding that substitutionswhich leave Rr-CooA constitutively active (without CObinding), also induce maximal heme/C-helix displacementin the CO adduct, with or without bound DNA [7]. Thesesubstitutions tilt the energetics toward the fully activeform. They involve replacement of a C-helix segment justbefore the hinge, residues 121–126 (TSCMRT), with A(or R) YLLRL. As discussed previously [4,7], these substi-tutions improve the ‘‘leucine zipper’’ contact between thetwo C-helices; in particular the two Cys123 residues areat the critical ‘d’ position of the heptad repeat, and arereplaced by Leu. In addition both Cys123 and Met124form part of the hydrophobic cavity into which the hemeis proposed to slide [5], and their replacement by Leuincreases the computed cavity size and makes it morehydrophobic [7]. The new Ch-CooA–CO structure, whichis also of a variant with substitutions on the C-helix, sug-gests that these substitutions also improve the docking ofthe N-terminus in the active form. Borjigin et al. [15] noteda close contact between Leu7, near the N-terminus, and anintroduced Leu127 residue (Asn127 in the wild-type pro-tein – the introduction of Leu at this position and at theadjacent position render Ch-CooA constitutively active[15]). The residue at the homologous position to Leu7 inCh-CooA is the N-terminal Pro2 of Rr-CooA, while theresidue at the homologous position to Leu127 of Rr-CooA,Thr122, is replaced by Tyr in A/RYLLRL of Rr-CooA.When modeled into the Ch-CooA–CO structure (Fig. 8),the Pro2 carbonyl and Tyr122 side-chain can be broughtinto proximity (2.15 A OÆÆÆO distance). The resulting H-bond could favor docking of the N-terminus, at theexpense of heme binding; the A/RYLLRL variants areknown to have significant populations of 5-coordinateheme. The wild-type residue at position 122 is Ser, whichis a weaker H-bond donor than Tyr. It may neverthelessbe strong enough to favor the docked conformation, oncethe N-terminus is released by CO binding.

However, other contacts are certainly involved [15] andthe docking interactions are flexible enough to allow for thedeletion of two residues near the Rr-CooA N-terminus, orthe addition of an extra one, with retention of some, albeitreduced, activity [29].

In summary, the CooA activation mechanism appears toinvolve release of the heme-bound N-terminus and its sub-sequent docking between the heme and the DNA-bindingdomains (see Fig. 5. Scheme in Ref. [15]), orienting the lat-ter for DNA-binding. In between these steps, a wide range

of domain orientations are available to the protein. Thisflexibility is facilitated by breaking of contacts betweenthe b4/5 loop on the heme domain, and the ‘hinge’ regionbetween the C and D helices. These contacts break whenthe heme and the C-helices are displaced as a result ofhydrophobic forces, after the N-terminus dissociates fromthe heme. Dissociation of the N-terminus accompaniesCO binding, but can also be induced by residue substitu-tions that promote heme/C-helix displacement and/ordocking of the N-terminus.

5. Conclusions

CO promotes CooA activation by displacing the N-ter-minal ligand and inducing a concerted heme and C-helixdisplacement. The heme displacement is monitored by themFe–C/mC–O frequencies, which reflect weakening of thebond to the proximal His ligand, as a result of tension ona His-Asn H-bond when the heme is displaced. It alsoresults in weakening of a Tyr55-Glu83 H-bond, on theproximal side of the heme, which can be detected bychanges in the tyrosine UVRR intensity. The C-helix dis-placement is monitored by the UVRR intensity of theTrp110 residue, which is drawn toward the hydrophobicCO-binding pocket.


The C-helix displacement breaks restraining contactsbetween the C-helix and the heme domain b4/5 loop,allowing the hinge to bend. Loss of these restraints upontruncation of the heme domain results in a 10-fold increasein the rate of Pro2 dissociation from the heme, reflectingloosened non-bonded contacts with surrounding C-helixresidues. In addition, the CO affinity is doubled and theC-helix displacement increases.

Bending the hinge region between the C- and D-helicesin the full-length protein allows a range of conformationsfor the DNA-binding domain. The conformation thatallows DNA to bind can be stabilized by docking of theN-terminal tail between the DNA-binding and hemedomains. By displacing the heme and C-helix, CO bindingleaves CooA in an equilibrium between ‘flexible’ and‘docked’ conformations. The equilibrium is shifted toward‘docked’ by DNA binding itself, and also by substitutionsthat stabilize the docking and displacement contacts.

Acknowledgments

This work was supported by NIH Grant GM 33576 (toT.G.S), NIH Grant GM53228 (to G.P.R.) and GM42614(to T.L.P.). We thank Mary Conrad for technicalassistance.

References

[1] M.A. Gilles-Gonzalez, G. Gonzalez, J. Inorg. Biochem. 99 (2005) 1–22.

[2] G.P. Roberts, R.L. Kerby, H. Youn, M. Conrad, J. Inorg. Biochem.99 (2005) 280–292.

[3] S. Aono, Acc. Chem. Res. 36 (2003) 825–831.[4] R.L. Kerby, H. Youn, M.V. Thorsteinsson, G.P. Roberts, J. Mol.

Biol. 325 (2003) 809–823.[5] H. Youn, R.L. Kerby, G.P. Roberts, J. Biol. Chem. 278 (2003) 2333–

2340.[6] C.M. Coyle, M. Puranik, H. Youn, S.B. Nielsen, R.D. Williams, R.L.

Kerby, G.P. Roberts, T.G. Spiro, J. Biol. Chem. 278 (2003) 35384–35393.

[7] M. Ibrahim, R.L. Kerby, M. Puranik, I.H. Wasbotten, H. Youn, G.P.Roberts, T.G. Spiro, J. Biol. Chem. 281 (2006) 29165–29173.

[8] M. Puranik, S.B. Nielsen, H. Youn, A.N. Hvitved, J.L. Bourassa,M.A. Case, C. Tengroth, G. Balakrishnan, M.V. Thorsteinsson, J.T.Groves, G.L. McLendon, G.P. Roberts, J.S. Olson, T.G. Spiro, J.Biol. Chem. 279 (2004) 21096–21108.

[9] W.N. Lanzilotta, D.J. Schuller, M.V. Thorsteinsson, R.L. Kerby,G.P. Roberts, T.L. Poulos, Nat. Struct. Biol. 7 (2000) 876–880.

[10] I.T. Weber, T.A. Steitz, J. Mol. Biol. 198 (1987) 311–326.[11] J.M. Passner, S.C. Schultz, T.A. Steitz, J. Mol. Biol. 304 (2000) 847–

859.[12] G. Parkinson, C. Wilson, A. Gunasekera, Y.W. Ebright, R.E.

Ebright, H.M. Berman, J. Mol. Biol. 260 (1996) 395–408.[13] K. Yamamoto, H. Ishikawa, S. Takahashi, K. Ishimori, I. Mori-

shima, H. Nakajima, S. Aono, J. Biol. Chem. 276 (2001) 11473–11476.

[14] M. Kuchinskas, H.Y. Li, M. Conrad, G. Roberts, T.L. Poulos,Biochemistry 45 (2006) 7148–7153.

[15] M. Borjigin, H.Y. Li, N.D. Lanz, R.L. Kerby, G.P. Roberts, T.L.Poulos, Acta. Crystallogr, Sect. D: Biol. Crystallogr. 63 (2007) 282–287.

[16] M. Kubo, S. Inagaki, S. Yoshioka, T. Uchida, Y. Mizutani, S. Aono,T. Kitagawa, J. Biol. Chem. 281 (2006) 11271–11278.

[17] D. Shelver, M.V. Thorsteinsson, R.L. Kerby, S.-Y. Chung, G.P.Roberts, M.F. Reynolds, R.B. Parks, J.N. Burstyn, Biochemistry 38(1999) 2669–2678.

[18] L.W. Chiang, I. Kovari, M.M. Howe, PCR methods and applications2 (1993) 210–217.

[19] J.C. Austin, K.R. Rodgers, T.G. Spiro, in: J.F. Riordan, B.L. Vallee(Eds.), Metallobiochemistry, Part C, Methods in Enzymology, vol.226, Academic Press Inc, San Diego, 1993, pp. 374–396.

[20] H. Takeuchi, I. Harada, Spectrochim. Acta Part. A 42 (1986) 1069–1078.

[21] H. Takeuchi, N. Watanabe, I. Harada, Spectrochim. Acta. Part A 44(1988) 749–761.

[22] P.G. Hildebrandt, R.A. Copeland, T.G. Spiro, J. Otlewski, M.Laskowski, F.G. Prendergast, Biochemistry 27 (1988) 5426–5433.

[23] S. Franzen, J. Am. Chem. Soc. 124 (2002) 13271–13281.[24] C. Xu, T.G. Spiro, J. Am. Chem. Soc., submitted for publication.[25] T. Uchida, H. Ishikawa, K. Ishimori, I. Morishima, H. Nakajima, S.

Aono, Y. Mizutani, T. Kitagawa, Biochemistry 39 (2000) 12747–12752.

[26] H. Youn, R.L. Kerby, M. Conrad, G.P. Roberts, J. Biol. Chem. 281(2006) 1119–1127.

[27] H. Komori, S. Inagaki, S. Yoshioka, S. Aono, Y. Higuchi, J. Mol.Biol. 367 (2007) 864–871.

[28] S. Akiyama, T. Fujisawa, K. Ishimori, I. Morishima, S. Aon, J. Mol.Biol. 341 (2004) 651–668.

[29] R.W. Clark, H. Youn, R.B. Parks, M.M. Cherney, G.P. Roberts, J.N.Burstyn, Biochemistry 43 (2004) 14149–14160.

Documents

Mechanism of the CO-sensing heme protein CooA: New insights from the truncated heme domain and UVRR spectroscopy