Dynamics of Carbon Monoxide Binding to Cystathionine β-Synthase*

Cystathionine β-synthase (CBS) condenses homocysteine, a toxic metabolite, with serine in a pyridoxal phosphate-dependent reaction. It also contains a heme cofactor to which carbon monoxide (CO) or nitric oxide can bind, resulting in enzyme inhibition. To understand the mechanism of this regulation, we have investigated the equilibria and kinetics of CO binding to the highly active catalytic core of CBS, which is dimeric. CBS exhibits strong anticooperativity in CO binding with successive association constants of 0.24 and 0.02 μm-1. Stopped flow measurements reveal slow CO association (0.0166 s-1) limited by dissociation of the endogenous ligand, Cys-52. Rebinding of CO and of Cys-52 following CO photodissociation were independently monitored via time-resolved resonance Raman spectroscopy. The Cys-52 rebinding rate, 4000 s-1, is essentially unchanged between pH 7.6 and 10.5, indicating that the pKa of Cys-52 is shifted below pH 7.6. This effect is attributed to the nearby Arg-266 residue, which is proposed to form a salt bridge with the dissociated Cys-52, thereby inhibiting its protonation and slowing rebinding to the Fe. This salt bridge suggests a pathway for enzyme inactivation upon CO binding, because Arg-266 is located on a helix that connects the heme and pyridoxal phosphate cofactor domains.

Cystathionine ␤-synthase (CBS) 5 is a key enzyme in the regulation of homocysteine in humans (1,2). Mutations in CBS are the most common cause of hereditary homocystinuria, in which elevated plasma homocysteine levels are associated with a range of pathologies, including mental retardation and cardiovascular disease (3). CBS is a pyridoxal phosphate (PLP)-dependent enzyme that catalyzes the condensation of homocysteine with serine to give cystathionine. In addition to the PLP cofactor, human CBS contains a heme group. The heme is not required for enzyme activity and is absent in CBS produced by yeast and protozoa (4). A fragment of the human enzyme in which the 69-residue N-terminal region is deleted no longer binds heme but retains ϳ40% activity (4). However, the heme can play a regulatory role; the Fe(II) form of the enzyme is inhibited upon binding CO or nitric oxide (5,6), whereas activity is doubled when the Fe(II) is oxidized to Fe(III) (7). The redox state of the heme is pH dependent, with reoxidation of Fe(II)-CBS to Fe(III)-CBS being favored at low pH conditions (8). Although the physiological implications of this regulatory behavior in CBS are uncertain, this protein joins the growing list of heme sensor proteins in which enzymatic or DNA binding activity is controlled by the ligation or redox status of the heme.
In addition to the N-terminal heme-binding domain, CBS has a C-terminal inhibitory domain; the inhibition is lifted upon binding of another regulatory molecule, S-adenosylmethionine (AdoMet) (9,10). Truncation of the AdoMet-binding domain leaves a fully active catalytic core that is less prone to aggregation than the full-length enzyme and is useful for spectroscopic and mechanistic studies (11). Catalysis is inhibited by CO binding, as in the full-length protein. The truncated protein in which the last 143 residues of the C-terminal domain have been deleted (CBS-⌬143) is dimeric (whereas the full-length protein forms tetramers and higher oligomers). The crystal structure of the catalytic core has revealed the arrangement of the heme-and PLP-binding domains, which are ϳ20 Å apart ( Fig. 1) (12,13).
In the present study, we have investigated CO binding to truncated CBS with the aim of elucidating the mechanism of regulation by this gas. Our principal findings are that the dimer displays equilibrium anticooperativity in CO binding, that the rate of CO binding is controlled by pre-dissociation of the endogenous Cys-52 ligand, and that rebinding of Cys-52 is relatively slow and independent of pH, between 7.6 and 10.5, implying a perturbed pK a for the dissociated Cys-52. These results support a model in which interaction between the heme and PLP cofactors is mediated by a hydrogen bond between Cys-52 and the Arg-266 residue, which is connected via an intervening helix to the PLP binding site.

Enzyme Purification and CO Equilibrium Constant Measurements-
The truncated dimer, CBS-⌬C143, was purified as described previously (11). The binding constants for CO for the truncated enzyme were determined exactly as described for the full-length enzyme (5).
Sample Preparation-For Raman spectroscopy, aliquots of Fe(III)-CBS protein were transferred to NMR tubes and diluted using specified buffers to achieve a typical concentration of 25 M in heme. The tubes were sealed with rubber septa and parafilm and purged with high purity nitrogen gas for 2 h. Sodium dithionite was added to reduce the iron to the Fe(II) form under anaerobic conditions. To prepare the CO-bound form of CBS following the reduction step above, the samples were immediately purged with CO for 3-5 min. For the rapid mixing experiments, Fe(II)-CBS for the stopped flow experiments was prepared with the same procedure as for Raman experiments. The final concentration was 14 M before mixing. CO-saturated buffers for use in rapid mixing experiments were prepared by first degassing with alternate application * This work was supported by National Institutes of Health Grants GM33576 (to T. G. S.) and HL58984 (to R. B.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. □ S The on-line version of this article (available at http://www.jbc.org) contains supplemental Fig. S1 and supplemental of vacuum and N 2 flushing followed by 2 h of CO purging. Solutions saturated with CO are 1 mM in dissolved CO. Buffer with lower concentrations of CO was obtained by mixing the CO-saturated buffer with an appropriate volume of N 2 -saturated buffer. The buffers used were phosphate at pH 7.6, Tris at pH 8.6, and glycine-NaOH at pH 10.6. These latter two pH values were above those at which significant anaerobic reoxidation of Fe(II)-CBS has been observed; the reoxidation of Fe(II)-CBS is slow at pH 7.6 (8) and so did not interfere in the conversion to the CO adduct, which was stable. Stopped Flow Measurements of CO Association Rates-A solution of CBS (75 l, 14 M) in buffer (Tris, pH 8.6) was rapidly mixed with an equal volume of buffer solution (Tris, pH 8.6) containing dissolved CO using a stopped flow instrument (HI-TECH Scientific SF-61 DX2 Double Mixing Stopped Flow System) with a diode array detector. The concentration of CO was varied by dilution of CO-saturated buffer (1.0 mM). A total of 300 scans were acquired over a period of 4500 s. The spectral range observed was from 300 to 700 nm. All reactions were carried out at 24.5 Ϯ 1°C. Global analysis of the time-dependent changes in the spectra was carried out using MATLAB v.7.0.4 on the data from 380 to 650 nm.
Time-resolved Resonance Raman Measurements of Photorecombination-The set-up has been described in detail previously (14,15). The second harmonic of a Q-switched Nd:YLF laser (Photonics International GM-30-527) was used to pump a Ti:sapphire laser (Photonics International TU-UV) that gave a narrowed laser frequency output (Ͻ0.1 cm Ϫ1 ) tunable between 810 and 920 nm. The Ti:S laser output (ϳ25 ns at 1 kHz) was frequency doubled using a non-linear LBO crystal to obtain pump pulses at 419 nm. Probe laser pulses at 426 nm were generated by a second Ti:S laser with the same characteristics. The optimum pump laser power to achieve maximum photolysis (i.e. no further increase in the intensity of the 4 resonance Raman band of the five-coordinate heme at the earliest time) was 140 milliwatt. Photolysis from the probe laser was minimized by keeping its power at ϳ1.0 milliwatt. The time delay between pump and probe pulses (0.20 -300 s) was controlled by a DG535 delay generator.
The sample solution in an NMR tube was spun for efficient mixing of the sample. Spinning ensured that the volume sampled was replaced between probe pulses and that there was no permanent product formation.
The pump and probe beams were spatially overlapped and then focused onto the sample with a pair of cylindrical lenses. The scattered light was collected and focused onto a single spectrograph (SPEX 1269, 3600 grooves/mm) equipped with a gated intensified photo diode array detector. Ten scans, each of 30-s acquisition time, were averaged. Spectra were calibrated with N,N-dimethylformamide, acetonitrile, and methylcyclohexane. Spectral deconvolution was carried out with the GRAMS/AI v.7.00 software. Parameters for curve fitting included use of a mixed Gaussian/Lorentzian (1:9) band shape and the incorporation of fixed bandwidths: 9 cm Ϫ1 (1372 cm Ϫ1 ); 12 cm Ϫ1 (1354 cm Ϫ1 ); and 11 cm Ϫ1 (1361 cm Ϫ1 ). The spectra were divided by control spectra, obtained by reversing the order of the pump and probe pulses, to compensate for any slow changes in sample composition.

Measurement of Relative Raman Cross-sections-Relative
Raman cross-sections were determined in two ways. In one experiment, samples of Fe(II)-CBS and of its CO adduct were made at the same concentration, and spectra were recorded under identical conditions. The laser power at the sample was 0.2 milliwatt. Measured intensities of the 4 band were used to obtain the relative crosssections. To obtain cross-section of CBS(5c) relative to CBS-CO, the laser intensity was increased stepwise to cause different amounts of photolysis. The intensities of 4 were then used to calculate the relative cross-section using a linear least squares procedure under the constraint that total heme concentration is constant. In another experiment, CBS was reduced under nitrogen atmosphere and the spectrum of Fe(II)-CBS was recorded. Following this, the sample solution was kept under 1 atm CO for 2 h. CO binding was monitored by recording the absorption spectrum periodically. When CO binding was complete, the Raman spectrum of CBS-CO was recorded. In both experiments, intensities were determined by deconvolution of the spectrum in the 4 region into three bands at 1354, 1361, and 1372 cm Ϫ1 as described above for the TR3 measurements.

RESULTS
Equilibrium Constants for CO Binding-Equilibrium dissociation constants were measured by titration of Fe(II)-CBS with CO (not shown). Two dissociation constants were obtained: 3.9(2.0) M and 50 (8) M, which are similar to those reported for wild-type human CBS, 1.5(0.1) M and 68 (14) M (5). Thus, CO binding to one subunit reduces the CO binding affinity of the second subunit.
CO Association Rates- Fig. 2 shows the spectra at various times after mixing CO with Fe(II)-CBS in the stopped flow apparatus. The heme  Soret band shifts from 449 nm in Fe(II)-CBS to 419 nm in the CO adduct. However, a side reaction was discovered, which we attribute to a small amount of O 2 in the Fe(III)-CBS solution prior to dithionite addition. When the O 2 was allowed to increase, complete bleaching of the 449 nm Fe(II)-CBS absorption band was observed, and yet subsequent addition of CO led to complete recovery of the 419 nm Fe(II)-CBS-CO absorption band (supplemental Fig. S1). Presumably, radical species resulting from the reaction of O 2 with dithionite attack the Fe(II)-CBS heme, but the nature of the product, which reverts back to the normal CO adduct, is currently a puzzle. The extent of bleaching could be reduced, but not eliminated, by a freeze-pump-thaw cycle on the Fe(III)-CBS solution.
In the stopped flow experiments, the extent of bleaching was minor, but somewhat variable, and produced an extraneous rapid initial increase in the 419-nm absorbance due to CO reaction with the bleach product. To eliminate the influence of this artifact on the course of the CO binding to authentic Fe(II)-CBS, we applied global analysis to extract the spectral component that changed over time because of the binding of CO to Fe(II)-CBS (Fig. 3a). The kinetics of the change in this spectral component (Fig. 3b) could be fit to two successive exponentials with the fast phase having a much larger amplitude, on average four times greater, than the slow phase (supplemental Table S1).
The successive pseudo-first-order rate constants, k obs (Table 1), differ by an order of magnitude. The fast phase was [CO] dependent and increased non-linearly with [CO], showing saturation behavior (Fig. 4). This is the behavior expected when the rate of CO association is limited by the prior dissociation of an endogenous ligand (Cys-52 in the case of CBS) (18). The kinetic scheme, illustrated in Fig. 5, leads to Equation 1.
when CO dissociation is much slower than binding, kϪ CO Ͻ Ͻ kЈ CO [CO], and when Cys-52 dissociation is much slower than rebinding of either ligand, kϪ cys Ͻ Ͻ k cys ϩ kЈ CO [CO] (16). The CO saturation limit gives the Cys-52 dissociation rate constant, kϪ cys , whereas the curvature is determined by the ligand rebinding rate ratio, k cys /kЈ CO . The fitting of Equation 1 to the data for k obs1 in Fig. 4 gives kϪ cys ϭ 0.0166 (8) s Ϫ1 (standard error in the last significant figure given in parentheses) and k cys /kЈ CO ϭ 109 (18) M at pH 8.6. An independent estimate of the latter is available from flash photolyis (see next paragraph). The slow phase of the CO binding was [CO] independent and had a much smaller amplitude than the fast phase. We attribute it to a subpopulation of Fe(II)-CBS molecules that are unable to bind CO, perhaps because the Cys-52 ligand is not primed to dissociate (see "Discussion") and requires a conformation change to do so. This conformation change is slow, occurring in hours (k obs2 ϳ0.0011 s Ϫ1 , Table 1).
Ligand Rebinding after Flash Photolysis-Resonance Raman spectroscopy was used to assess rebinding rates for CO and Cys-52 after flash photolysis of the CO adduct, at equilibrium with 1 mM CO. The 4 heme resonance Raman band has distinguishable positions for the native protein (bound Cys-52), 1361 cm Ϫ1 , the CO adduct, 1372 cm Ϫ1 , and the five-coordinate (5c) heme intermediate, 1354 cm Ϫ1 . Fig. 6 shows the time evolution of the 4 band envelope at different delay times between pump (419 nm) and probe (426 nm) lasers. Deconvolution of this envelope into the three components is illustrated for the 200-ns delay spectrum.    The deconvoluted band areas were converted to concentrations after determining the relative cross-sections. Comparison of pure solutions of CBS and the CO adduct revealed the 4 cross-sections to be the same at 426 nm. The cross-section of the five-coordinate heme photoproduct was found to be 1.7-fold larger, via a least squares fit of the intensities under the constraint of constant heme concentration. Fig. 7 shows the time evolution of the relative fractions of CBS-CO, CBS(5c), and Fe(II)-CBS. The intensities were corrected for probe laser photolysis by subtraction of the probe-only spectrum and normalization with respect to the probe-only intensity of the CBS-CO band. Nearly 70% of the photolyzed CO recombines geminately, so that the initial fraction of CBS(5c) is ϳ30%. This intermediate decays and is replaced by CBS-CO and by Fe(II)-CBS, but the former grows faster than the latter. The data are well described by single exponentials, with time constants of 45 (15) and 55 (12) s for CBS(5c) and CBS-CO for pH 8.6 solutions (Fig. 7b, curves); similar results are obtained at pH 7.6 and 10.5. The formation rate for the slowly evolving Fe(II)-CBS is best determined from the rate difference between CBS(5c) decay and CBS-CO formation as shown in Equation 2. k 5c,obs ϭ kЈ CO ͓CO͔ ϩ k cys (Eq. 2) The derived Fe(II)-CBS rate constants are the same within experimental error for all these pH values ( Table 2).
There is no evidence for two kinetic phases in the photo-induced rebinding data. Because geminate recombination accounts for 70% of the heme, we cannot exclude the possibility that the kinetic traces are monitoring only one of the two hemes in the CBS dimer However, the k cys /kЈ CO ratio obtained from the rebinding data, 220 (170) M at pH 8.6, is within experimental error of the value determined from the stopped flow CO binding experiment. Thus, the same elementary rates are monitored in the two experiments, one starting with Fe(II)-CBS and the other with Fe(II)-CBS-CO, suggesting equal rates for the two hemes. Table 3 lists the derived values of the rate constants for the CBS-CO system.

DISCUSSION
Ligation Dynamics-CBS belongs to the class of heme proteins in which exogenous ligand binding requires displacement of an endogenous ligand, unlike hemoglobin or myoglobin in which exogenous ligands bind directly to five-coordinate heme. Recently investigated proteins of this class include rice hemoglobin, rHb1 (20), in which CO replaces an endogenous histidine, and the gene activator protein CooA (17) from the bacterium Rhodospirillum rubrum, in which CO replaces the N-terminal amine group, which belongs to a proline residue.
As might be expected, the exogenous ligand affinity is diminished by the binding affinity of the endogenous ligand. Thus, the equilibrium constant for CO association is much lower for these proteins than it is   The recombination rate of the endogenous ligand is faster for CBS than for rHb1, 4000 versus 500 s Ϫ1 , again as might be expected for an anionic ligand attacking the positively charged Fe(II). For CooA, two rates are deduced from the requirement for cooperative CO binding, 4000 and 84 s Ϫ1 , which bracket the rHb1 value; the faster rate is the same as the value for CBS.
The equilibrium constant for Cys-52 association in CBS is calculated from the forward and reverse rate constants to be 2.4 ϫ 10 5 . This is much larger than the Pro association constants derived for CooA, 20,000 and 1,200, and nearly a million times larger than the His association constant in rHb1, 0.27.
On the other hand, the rate of CO rebinding to the five-coordinate photoproduct for CBS, 18 M Ϫ1 s Ϫ1 , is between the rates for rHb1 (6 M Ϫ1 s Ϫ1 ) and CooA (32 M Ϫ1 s Ϫ1 ). Thus, the dynamical differences among these three six-coordinate heme proteins lie in the nature of the endogenous ligand being replaced.
In contrast to CooA, the equilibrium binding of CO is anti-cooperative for CBS. However, there is no evidence for kinetic inequivalence between the two binding sites of the dimer, either in CO association to Fe(II)-CBS or in CO rebinding to photolysed Fe(II)-CBS-CO. The inequivalent equilibrium constants must therefore stem from inequivalent CO dissociation rates, which have not been measured in this study. The CO dissociation rate must diminish for the second heme when the first heme loses CO, perhaps due to an allosteric change that limits egress of CO from the heme pocket.
Mechanism of Enzyme Regulation-The heme group in CBS is 20 Å distant from the PLP active site, yet enzyme activity is diminished 1.7fold when Fe(III) heme is reduced to Fe(II) (7) and is abolished altogether by CO binding to the heme (5). An allosteric mechanism is presumed to be responsible for this regulatory behavior. Helix 8 separates the heme and PLP sites (Fig. 8) and is a likely candidate for an allosteric element. It has been noted that this helix supports Thr-257 and Thr-260 on the PLP side and Arg-266 on the heme side (4). The two threonine residues donate hydrogen bonds to the PLP phosphate group, whereas Arg-266 donates an H-bond to the Cys-52 ligand. Thus the Arg-266-Thr-257/260 segment may act as a lever between the heme and the PLP site, modulating the activity of the latter via mechanical tension that is mediated by the hydrogen bonds at either end.
An important finding in this connection is the absence of any significant pH effect on the photodissociation dynamics. Essentially the same rebinding rates are observed when the solution pH ranges from 7.6 to 10.5 (Table 2). It might have been expected that the Cys-52 thiolate would be subject to protonation after it is displaced by CO. The thiolate pK a is ϳ8.8 in aqueous solution (21). If the displaced Cys-52 were protonated at pH 7.6, its recombination rate should have slowed substantially relative to pH 10.5, but this did not happen. We conclude that Cys-52 remains unprotonated and that its pK a is shifted below 7.6 by interactions in the binding pocket. The obvious interaction is the hydrogen bond with Arg-266, which can convert to a salt bridge when the Cys-52 thiolate anion is displaced from the positive Fe(II). This interaction can explain the otherwise surprising fact that it is the anionic Cys-52 ligand that is displaced by CO, rather than the neutral His-65 ligand on the other side of the heme. The Cys-52 thiolate is primed for displacement by the availability of a salt bridge interaction. We speculate that the slow conformation change, represented by the low amplitude [CO]-independent phase of the stopped flow kinetics (above), arises from a subpopulation of Fe(II)-CBS molecules in which Arg-266 is misoriented for H-bonding to Cys-52, which is therefore resistant to displacement by CO.
The salt bridge could account for enzyme inhibition if its formation reorients the Arg-266 side chain and, with it, the Arg-266-Thr-257/260 lever. The resulting movement of Thr-257 and Thr-260, which are anchored to the PLP phosphate group, may misalign the pyridoxal end of the PLP, interfering with an optimal geometry for catalysis. It is possible that a more subtle shift of the Arg-266-Thr-257/260 lever could also account for the 1.7-fold reduction in activity between Fe(III) and Fe(II) forms of the heme. The Arg-266-Cys-52 H-bond is expected to weaken significantly when Fe(II) is oxidized to Fe(III), because of the reduced negative charge. This weakening could shift the Arg-266-Thr-257/260 lever slightly, inducing a suboptimal alignment of the PLP. Thus the proposed allosteric mechanism could account for heme regulation of PLP activity by both redox and ligand displacement processes.
Physiological Implications-Does CO binding regulate CBS activity in vivo? This is quite possible because the reported range of physiologically relevant CO concentrations, 3-30 M (19), is bracketed by the two dissociation constants of human CBS, 1.5 and 68 M (5) of full-length  human CBS, implying that the reduced enzyme is normally half-saturated with CO. It has also been suggested that heme redox state is a regulator of CBS (7) because the activity doubles when the reduced form is oxidized to Fe(III)-CBS. The two effects are not mutually exclusive, because the reduction potential of Fe(III)-CBS would be raised by CO binding to Fe(II)-CBS. The physiologically relevant active and inactive states might be Fe(III)-CBS on the one hand and Fe(II)-CBS-CO on the other. The relatively slow rate of CO binding complicates the issue. At 10 M CO, the half-time for CO binding is expected to be 500 s from the present data on the truncated protein, and it may be affected by interactions with other proteins or with effector molecules in vivo. However, a slow binding rate might act as a filter to avoid unproductive responses to transient changes in local CO levels, as has been suggested for CooA (18).