- Email: [email protected]
1H NMR spectroscopy of cytochrome cd1 derivatives
ARCHIVES OF BIOCHEMISTRY AND BIOPHYSICS Vol. 240, No. 2, August 1, pp. 689-697, 1985
‘H NMR Spectroscopy
of Cytochrome
RUSSELL TIMKOVICH,2
MARGARET
.Department
Illinois
of Chemistry,
Received
December
S. CORK,
Institute
cd, Derivatives’ AND
of Chemistry,
26, 1984, and in revised
form
PRISCILLA Chicago, March
Illinois
V. TAYLOR 60616
25, 1985
Proton nuclear magnetic resonance spectra are reported for cytochrome cdl from Pseudwmxmas aeruginosa (ATCC 19429) in several forms including complexes of the ferricytochrome with cyanide, azide, and fluoride, a quasi-apo form in which the noncovalently associated heme dl has been removed but the covalently bound heme c is retained, and the reduced state of both native and the quasi-apo forms. Comparisons are made to the previously reported spectrum of ferricytochrome cdl. The following points are :made. (i) The spectra of the azide and fluoride complexes and the ferric quasi-apo form show perturbation of resonances assignable to the site of heme dr, and leave relatively unperturbed resonances assignable to the site of heme c. The heme dl associated resonances are at 46.0, 35.4, 23.3, 17.5, -2.9, and -16 ppm, and the heme c associated resonances are at 42.0, 33.7, 15.0, 13.9, -7.5, -14, and -33 ppm in native ferricytochrome cdl. (ii) The similarity of the hyperfine resonances of the ferric quasi-apo form to the heme c resonances of intact ferricytochrome cdl is evidence that removal of heme d1 leaves the heme c binding site relatively unaltered. (iii) Linewidths and relaxation times suggest that the relaxation times of the unpaired electron spins of the ferric hemes c and d1 are on the same order of magnitude. (iv) Although i,t is paramagnetic, ferrocytochrome cdl does not demonstrate an experimentally detectable hyperfine shifted spectrum under present conditions. Possible reasons for this are discussed. (v) The presence of a narrow resonance at -2.8 ppm in both ferrocytochrome cdl and the reduced state of the quasi-apo form suggests that methionine may be a ligand to heme c. o 19% Academicpress,I~~.
Cytochrome cd1 is a dissimilatory nitrite reductase present in facultative denitrifying bacteria. The native enzyme is a dimer composed of two identical subunits, each with a molecular weight of approximately 60,000 and containing one covalently bonded heme c and one noncovalently associated heme d1 (1). The heme c is believed to be an iron chelated form of i This work was supported by National Institutes of Health Grants GM23869 (R.T.) and GM2607102Sl (from the NIGMS Shared Instrumentation Program for support of the NMR facility). a To whom correspondence should be addressed, at the present address: Department of Chemistry, P. 0. Box H, University of Alabama, University, Ala. 35486. 689
protoporphyrin IX in which two polypeptide cysteinyl residues have added across the original vinylic substituents to form covalent thioether bonds. Although there is only optical spectroscopic evidence for this in terms of the absorption maxima and intensities of the pyridine hemochromogen (2), such evidence has been upheld in those cases of where additional structural studies have confirmed the porphyrin structure. Heme d1 has recently been shown to be the iron chelate of an unusual chlorin macrocycle, termed acrylochlorin (3). The paramagnetism of the ferric heme protein causes large hyperfine chemical shifts in the ‘H NMR spectrum that resolve some resonances from the usual 0- to lo-ppm range. The hyperfine 0003-9861185
$3.00
Copyright0 1985by AcademicPress,Inc. All rights of reproductionin anyform reserved.
690
TIMKOVICH,
CORK.
shifted resonances are either protons directly bonded to the paramagnetic hemes or protons from nearby amino acid side chains. The hyperfine shifted spectrum is thus an excellent probe of the active sites of the cytochrome. An initial investigation (4) reported the detection of hyperfine shifted resonances in ferricytochrome cdl and examined the effects of pH and temperature. In order to make full use of these hyperfine shifted resonances as probes of the heme sites, it would be necessary to have a complete assignment to specific protons. This is a formidable task for such a large macromolecule. A useful and feasible first step is to make a partial assignment of resonances in terms of whether they arise from the heme c or heme di site. This is the main objective of the present report. In addition, the limited data obtainable on the ‘H NMR spectrum of the reduced state of cytochrome cdl will be presented. MATERIALS
AND
METHODS
The purification of cytochrome cd, from Pseudo monos aeruginosa ATCC 19429 has been described (4). Several procedures have been reported for the extraction of the noncovalently bound heme di (2, 5, 6). The resulting reddish protein will be called quasiapo protein because it retains the covalently bound heme c. The procedures cited rely upon acidic acetone or low pH extraction of heme d, and produce a red precipitate that is the quasi-apo protein. In our hands these procedures all successfully extracted heme di, but we were unable to resolubilize the quasi-apo protein at concentrations in deuterium oxide buffers suitable for NMR spectroscopy. The acid treatment may have denatured at least some portion of the sample, rendering it insoluble and/or preventing the solubilization of other more nativelike material. A more soluble quasi-apo protein was prepared by mild aqueous acid extraction of heme d,. Cold native enzyme was titrated with glacial acetic acid to pH 3 with constant stirring. A reddish precipitate was collected by centrifugation and redissolved in 50 mM sodium carbonate/bicarbonate buffer at pH 9. This was dialyzed in the cold overnight against phosphate buffer and chromatographed on DEAE-cellulose under the same conditions for the corresponding step in the isolation of the native cytochrome (4). A reddish band eluted at 0.15 to 0.20 M in the KC1 gradient. Yields were variable and generally low. The quasi-apo protein so obtained was found to match a red pigment obtained at the same elution volume from DEAE-cellulose during
AND
TAYLOR
the normal isolation of the enzyme. Both had identical uv-visible spectra, molecular weights on gel filtration columns, and could be reconstituted in dilute solution with heme di extracts as described by Hill and Wharton (5). When the reconstituted protein was assayed with Pseudmmm ferrocytochrome ~551 as donor (5), it was found to have ‘70100% of the specific activity of native enzyme, which had a maximal velocity of 170 pmol donor oxidized min-’ pmol enzyme-‘. The range of recovered activity observed was somewhat broader than previously reported (5), but this may be due in part to ambiguity in assaying initial rates, since the reaction never has a true zero-order rate dependence on donor concentration (7). The latter pigment presumably has been produced by accidental loss of heme d, during the enzyme isolation. The visible spectra of both pigments matched a report for quasi-apo protein prepared by acidic acetone extraction [Fig. 1, Ref. (5)]. In order to obtain sufficient quantities for NMR studies, the various preparations of quasi-apo protein were combined. The NMR spectrum of this material agreed with the native enzyme in the general aromatic and aliphatic regions, although lack of resolution in these crowded spectra preclude conclusions concerning similarity of tertiary structure. Preparation of samples for NMR spectroscopy and conditions for obtaining spectra at 300 MHz have been described (4). Chemical shifts are reported in parts per million from sodium 4,4-dimethyl-4-silapentane sulfonate. Spin-lattice relaxation times were measured by an inversion-recovery pulse sequence. Recycle time was 160 ms or at least lOTi for the slowest hyperfine shifted resonance. For each of seven delay times, 350,000 transients were collected. Mixtures of ferricytochrome cd, and potassium cyanide or sodium azide were prepared by adding the ligands as aliquots from concentrated stock solutions in deuterium oxide, whose pH had been previously adjusted to match the protein solution. Sodium fluoride was added as a solid. In all three cases, the final pH was checked after additions with a glass micro-combination electrode and any fine adjustments in pH were made with stock solutions of NaO’H or *HCl. Values of pH represent the uncorrected glass electrode reading and will be designated pH*. NMR samples in the presence of exogenous ligands were examined by optical spectroscopy either in 1-mm-pathlength cells or after dilution with deuterium oxide in standard cells. Cyanide (8) and azide (9) caused optical changes that have been previously reported. The optical effects of even high concentrations of fluoride were subtle. In difference spectra with and without fluoride, weak features were detected at 643 and 470 nm, similar to a published report (10). The question of whether fluoride forms a true complex with cytochrome cd, has been equivocal, with one group reporting a complex
NMR
OF
CYTOCHROME
with heme di based upon optical and EPR results (10) and another group claiming no spectral effects for fluoride (8). From NMR data to be presented, it will be clear that high concentrations of fluoride do exert an effect on the NMR spectrum of ferricytochrome cdl. But, it is conceivable that it does so by acting as a selective denaturant of the heme di site, without forming a true stable complex. The germaine issue for the present study is only that fluoride does selectively perturb the NMR spectrum. For brevity, reference will be made to a fluoride complex. Samples of ferrocytochromes were prepared by the addition of a slight excess of solid sodium dithionite, or concentrated sodium ascorbate from a neutralized solution in deuterium oxide. Reduced samples were
691
cd, DERIVATIVES
sealed in glass 5-mm NMR tubes after alternate cycles of vacuum degasing and flushing with argon. Complete reduction was confirmed after NMR spectroscopy by breaking open the sample tubes, transfering to optical cuvettes with l-mm pathlengths, and examining the ferrous 01 bands (11) before and after the addition of still more reductant.
RESULTS
Measured spin-lattice relaxation times for the resolved hyperfine resonances of ferricytochrome cdl are reported in Table I. NMR spectra were obtained for com-
TABLE
I
SPIN-LATTIC:E RELAXATION TIMESAND CHEMICAL SHIFTSFOR VARIOUS FORMSOF FERRICYTOCHROMEcdl Native (ppm)
T, (ms)
46.0 42.0 35.4
10.7 10.1 11.4
33.7 32.5
10.0 10.8
23.3
15.0
17.5
8.2
Azide (mm)
Fluoride
51. 49.
57.
41.9 34.7 33.5 26-23
Quasi-apo (ppm)
Cyanide bpm)
42.0
41.3
42-40
33.4 32.5
33.5
(mm)
(?)
25.0 21.8
17 (?)
22-20
17. 16-14
15.0 13.9
-2.9
13.6
15. 13.5
16.3 14.8 14.0 12.7 -2.1
15.6 13.9
6.9 -3.3
(?)
-4.1 -7.5
(?)
-3.3
-3.8 -7.5 -14. -16. -24. -33.
4.5 3.7 3.7
-14.
-7.5 -10.9 -13.7
-24. -32.
-32.9
Note. Chemical shifts are reported in parts per million at 25°C and pH* 7.5. Because of the widths of the peaks, the precision of shift values was generally fO.l ppm for the narrowest and f0.4 ppm for the very broadest. Spin-lattice relaxation times, T,, are reported in milliseconds. Statistical deviations based upon the linearity of the usual log plot of residual magnetization versus delay time were on the order of 10%. For some weak resonances, relaxation times could not be reliably measured, and no value is quoted. The symbol (?) indicates especially weak resonances.
692
TIMKOVICH,
CORK,
plexes of ferricytochrome cc& with the exogenous ligands azide, fluoride, and cyanide. Typical data are presented in Figs. 1 and 2. The spectra of such complexes were inevitably noisier than for the native ferricytochrome. Based upon either visible or EPR spectroscopy, azide (9) and fluoride (10) have been reported to bind only to heme dl, while cyanide binds to both heme d1 and heme c in the ferric state (9, 10). The spectrum of the azide complex, Fig. lB, showed perturbation of some but
AND
TAYLOR
H B
I.” -10
-
I -20
’
”
1
I -30
’
1.1 PP”
FIG. 2. Upfield hyperfine NMR spectra of various forms of Pseudomonas ferricytochrome cdl. Conditions were as described for Fig. 1. (A) Native ferricytochrome cdl; (B) plus 100 mM sodium azide; (C) plus 500 mM sodium fluoride; (D) the quasi-apo ferric form; (E) native plus 100 mM potassium cyanide. The vertical scales are not equivalent.
FIG. 1. Downfield hyperfine NMR spectra of various forms of Pseudomonas ferricytochrome cdl. All spectra were recorded at 25°C. Protein concentrations were 1 mM, expressed as the subunit concentration. The solvent was 99.7% deuterium oxide buffered with sodium phosphate to a final pH* of 7.5 with additional reagents as listed. (A) Native ferricytochrome cdl; (B) plus 100 mM sodium azide; (C) plus 500 mM sodium fluoride; (D) the quasi-apo ferric form; (E) native ferricytochrome cd, plus 100 mM potassium cyanide. The vertical scales of the spectra were chosen in order to give a reasonable presentation of the main features in a concise manner, and are not the same for all spectra. The intensity of the peak at circa 42 ppm may be used as an approximate internal standard. The absolute integrated intensity of this resonance was the same within our experimental accuracy for (A)-(D) but for (E) the 42-ppm feature was fourfold weaker.
not all of the hyperfine shifted resonances found in the native ferricytochrome. The 46.0-ppm resonance was absent, but a new major resonance appeared at 49 ppm with some satellite peaks. The 35.4- and 1’7.5ppm resonances were absent. The 23.3ppm resonance was broader with new satellites. The upfield region (Fig. 2B) was especially noisy and was difficult to interpret. The -2.9-ppm resonance appeared to be affected and a new weak resonance appeared at -3.8 ppm. In the spectrum shown, the resonances at -7.5, -14, and -16 ppm appear to be absent, but in other spectra of the complex there was evidence of peaks at -7.5 and -14 ppm. Compared to these changes, the resonances at 42.0, 33.7, 32.5, and 15.0 ppm were much less perturbed. In the fluoride complex, the 46ppm resonance also disappeared and a new broad feature at 5’7 ppm appeared. The 35.4- and -16-ppm resonances disappeared, while the resonances at 23.3, 17.5, and -2.9 ppm decreased in intensity.
NMR
OF
CYTOCHROME
Resonances at 42.0, 33.7, 32.5, 15.0, -7.5, and -14 ppm were less affected. The very broad resonance at -33 ppm has been difficult to detect in some native ferricytochrome spectra, but appeared in the fluoride complex. Cyanide caused complex, time-dependent changes in the hyperfine spectrum. In this context it must be stressed that observation of any ferricytochrome cdl spectrum required extensive signal averaging. A minimum of 6 h of accumulation of averaged transients was required before hyperfine peaks could be detected above baseline noise. Changes taking place during this time would not be apparent but would increase the noise of spectra. The spectrum displayed in Fig. 1E is representative of spectra obtained within 12 h of mixing cyanide and protein. Very broad peaks were detected around 26,22,15, and -6 ppm. The broad resonance observed around 42 ppm in Fig. 1E was absent in some spectra, and was replaced by a resonance of comparable intensity and line width at 60 ppm. After 12 h, the previously observed peaks disappeared and no hyperfine shifted resonances could be detected. The NMR spectrum of the ferric form of the quasi-apo protein is shown in Fig. 1D. Hyperflne shifted resonances at 41.4, 33.6, 14.67, -7.5, -13.69, and -32.9 ppm were similar to resonances observed in native ferricytochrome cdl, but narrower. Additional resonances were observed at 16.4,12.8, -2.1, -3.3, and -10.9 ppm. Their intensity is such that they may have been obscured by noise or by overlap with heme di associated resonances in the native ferricytochrome spectrum. Ferrocytochrome cdl is known to be a paramagnetic heme protein with high spin ferrous heme d1 based upon magnetic circular dichroism (12) and magnetic susceptibility (13) data. NMR spectroscopy failed to detect hyperfine shifted resonances in the reduced state. Independent batches of protein and the alternative reducing reagents dithionite and ascrobate were examined. The same samples sealed in the NMR tubes were examined by resonance Raman spectroscopy (Lewis, Cotton, and
693
cd, DERIVATIVES
Timkovich, unpublished) and were found to give the native resonance Raman spectrum of ferrocytochrome cdl (14). The only noteworthy feature was a clearly resolved resonance at -2.8 ppm. Its linewidth (80 Hz full width at half height) is reasonably sharp for a protein the size of cdl and it is unlikely to be a hyperfine shifted resonance. The reduced state of the quasi-apo protein also did not show any detectable hyperfine shifted resonances, but included the -2.8-ppm resonance. DISCUSSION
For the sake of brevity in the ensuing discussion, the designations “heme c resonances” or “heme di resonances” will include covalently bonded heme protons as well as polypeptide protons in the vicinity of the heme that might appear in the paramagnetically shifted spectrum. Spin-lattice relaxation times were measured for ferricytochrome cdl because it was believed that such data might contribute evidence toward the respective assignment of hyperfine resonances to either the heme c or heme d1 site. It has been reported for ferricytochrome cdl that ferric heme di has an electron spin relaxation time much longer than ferric heme c (15). The unpaired electron spin is certainly a major factor in the mechanism of proton resonance relaxation for hyperfine shifted resonances. For macromolecules in general, relaxation times may be influenced by magnetization transfer (spin diffusion), but this is not likely to be of importance for the hyperfine shifted resonances of cdl because electron-proton interactions are governed by a correlation time much shorter than proton-proton dipole interactions. It was anticipated that resonances arising from heme d, should have significantly different spin-lattice relaxation times then those arising from heme c. It was surprising that the observed data did not fall into two clear categories. The heme methyl groups in particular had relaxation times all on the same general time scale. Observed differences cover the same spread as has been observed for
694
TIMKOVICH,
CORK,
heme methyls in other cytochromes, such as ferricytochromes ~550 (16) and ~554 (17). The differences could thus be ascribed to differences in the exact distance from the paramagnetic centers and to different contributions of contact, pseudocontact, and proton dipole-dipole mechanisms to the total relaxation rate. Our data do not allow a calculation of electron spin correlation times, but suggest to a first approximation that they are of the same order of magnitude for heme di and heme c. The hyperfine spectrum of the ferric form of the quasi-apo protein is the best available evidence for assignment of resonances to the heme c region. Resonances at 41.3, 33.5, 14.8, 14.0, -7.5, -13.7, and -32.9 ppm from the quasi-apo protein are aligned in Table I with resonances at 42.0, 33.7, 15.0, 13.9, -7.5, -14, and -33 ppm in ferricytochrome cdl because they are believed to represent the same spin systems. They may be directly bonded heme protons or protons of nearby amino acid side chains. Other hyperfine shifted resonances in ferricytochrome cd, are attributed to the heme di region. The hyperfine chemical shift of a resonance is a sensitive probe of environment. The similarity of the spectra of the heme c associated resonances between native and quasi-apo forms is suggestive that the heme c environment has not been radically perturbed by the loss of heme di. Magnetic circular dichroism data have been interpreted as indicating that there are not strong heme-heme interactions (such as heme stacking) in cytochrome cdl (18), and the NMR data would support this conclusion. There are some differences in the resonances of the quasi-apo form. Most notable are that the partially resolved pair at 33.7 and 32.5 ppm in the native ferricytochrome may have collapsed to a single peak at 33.5 ppm in the quasiapo form, the resonance at 14.7 ppm may be more intense in the quasi-apo form, and the -32.9-ppm resonance is narrower. The reasons for these changes are unknown. They might arise from modest conformational changes at the binding site.
AND
TAYLOR
The hyperfine NMR spectra of the azide and fluoride complexes of ferricytochrome cdl on the whole support the relative assignments deduced from comparing native and quasi-apo forms. Since visible and EPR spectroscopy have indicated that heme d1 is the binding site for these exogenous ligands, it is reasonable that resonances associated with heme d1 should be most severely affected by complexation. Complexation could be expected to alter the g-tensor as well as the unpaired electron spin density around heme di, and therefore could perturb both contact and pseudocontact shifted resonances. The actual extent of perturbation on resonances could not be predicted a priori and at this time it would not be possible to analyze the perturbations in detail in terms of structure. By themselves, the spectra of the complexes are somewhat ambiguous to interpret, because they are noisier and because the differential effects on resonances have to be judged by degree. Nevertheless, the total impact of the data largely confirms the relative assignments. Cyanide is a strong field ligand that can lower the spin state and/or decrease the electron spin relaxation time and so sharpen hyperfine shifted resonances in some ferric heme proteins. It is well known to accomplish this in model heme complexes and the globins. Since cyanide has been shown by EPR and visible spectroscopy to bind to both hemes of cdl, it was reasonable to expect that many native hyperfine shifted resonances would be perturbed. Although both hemes in the native ferric state are already low spin (4, g-14), a change in the electron spin relaxation time and/or the g-tensor anisotropy was a reasonable expectation. It was surprising to find that most of the hyperfine spectrum has disappeared, and there is a slow process with a time scale of hours that eliminates all observable resonances. Since the resonances of the intermediate form are already broader than the native form, we interpret the disappearance of the hyperfine resonances as due to further broadening. The possible mechanism of this broadening will be discussed in subsequent paragraphs. EPR
NMR
OF
CYTOCHROME
has been used to probe the electronic environment for the native form and the cyanide complex of ferricytochrome cc&. While there is agreement concerning the native spectrum, the effects of cyanide complexation have been contentious. One report (19) indicated no change for signals arising from heme c while stating that the heme dl signals disappeared. A more recent report (10) from the same group refined their measurements and reported changes in heme c signals as well as newly assigned heme d1 signals. Another group (12) reported slight changes in g values for the heme c signals and more extensive changes for the heme d1 signals, but the heme d1 remained low spin. These measurements indicate that at least heme c and probably heme d1 as well do not undergo m,ajor changes in spin state or relaxation time as a result of cyanide binding, although the crystal field symmetry may change for both hemes. Therefore, it is unlikely that the line broadening observed in the NMR spectrum is the result of ;a change in gross electronic properties, such as conversion to highspin ferric iron. The linewidths observed for the quasiapo protein are obviously narrower than for the native enzyme and these in turn are narrower than for the complexes with exogenous ligands. Several mechanisms could account for this, but there are difficulties for all but one. Possibilities to consider include (a) dimer to monomer dissociatio:n, (b) restricted local motion in the native form, (c) pseudocontact interactions on heme c resonances by ferric heme dl, (d) chemical exchange broadening due to a mixture of oxidation states (17, 20), and (e) chemical shift heterogeneity. Point a is unlikely because the operative correlation time would probably be dominated by the electron relaxation time rather than the isotropic rotational tumbling time. At any rate, on gel-filtration columns, the quasi-apo form behaved with the same apparent molecular weight as the native form. Points b and c are unlikely because they would preferentially affect only some resonances, while the observation is that all are broadened in
cdl
DERIVATIVES
695
the native form. Point d is unlikely because kinetic measurements of electron transfer rates in the native enzyme indicate that exchange would be slow compared to the NMR time scale (21), and because spectra were insensitive to the excess external oxidant ferricyanide. The prefered interpretation is point e, chemical shift heterogeneity in the native versus the quasi-apo form. Each apparent peak in the native spectrum represents a grouping of peaks with narrower intrinsic linewidths, but slightly different chemical shifts. The observed linewidths in the quasi-apo form are interpreted as more representative of the intrinsic linewidth of hyperfine resonances in cytochrome cc&. Excessive broadening in the native form is attributed to chemical shift heterogeneity due to a range of heme dl-protein noncovalent interactions in our NMR samples. It should be stressed that we are not interpreting the native samples as being grossly denatured. Because of the strength of paramagnetic interactions, it would require only modest perturbations in conformations and/or electronic properties to produce a sizable spread of chemical shifts. To illustrate, consider AZcaligenes ~554 where ionization of a remote propionate group in the heme pocket causes chemical shifts in the range of 23 ppm for hyperfine shifted heme methyl resonances (1’7). The existence of a range of readily accessible conformations only slightly different could reflect a “loose” heme d1 crevice that would correlate with the lability of the heme dl-protein association. The line broadening and concommittent low signal-to-noise ratio in the spectra of complexes with exogenous ligands may result from the ligand further distorting the heme binding site and creating an even larger range of shifts. The chemical shift heterogeneity interpretation is supported by two sets of observations. In the first case, selective decoupling of the hyperfine shifted resonances of the native form burns a hole in the center of the resonances without saturating the entire apparent peak. Equivalent decoupler irradiation of the hyperfine shifted resonances of the quasi-apo
696
TIMKOVICH,
CORK,
form led to complete collapse of the peak. A lower decoupler power could not be found sufficient to just burn a hole in the quasi-apo peaks. This argues that in the native form, the apparent peaks are actually superimposed resonances in slow exchange on the NMR time scale with respect to their chemical shift differences. Of course, the limit of slow exchange is static, noninterconverting conformations. In the second case, supportive evidence comes from a comparison of spin-lattice ( T1) and spin-spin (Tz) relaxation times. T2 values may be estimated from the observed linewidths since the intrinsic linewidth for this size macromolecule is much greater than field inhomogeneities. For the native protein, the ratio of T,/T, is on the order of 10 to 20. For example, the resonance at 46 ppm in the native protein has an apparent Tz of 0.8 ms while Tl is 10.7 ms. At 23 ppm the respective values are 0.7 and 15 ms. For the quasiapo protein the ratio is on the order of 3. For example, for the 34-ppm resonance the values are 3.0 and 10 ms. A paramagnetic macromolecule the size of cytochrome cd1 is likely to be beyond the extreme narrowing limit for resonances at 300 MHz, so a Tl longer than T2 is not wholly surprising. However, the magnitude of the ratio in the native form spectrum is excessive when compared to most heme proteins. For other cytochromes studied by this lab, the ratio of Tl (determined by inversion-recovery experiments) to T2 (determined from line widths) has been on the order of 3-4. For example, the ratio for the heme methyls in Parucoca cytochrome ~550 (16) and Alcaligenes cytochrome ~554 (17) average to 3.3, which is comparable to the ratio in the quasi-apo form. We are not attempting to attach any deep significance to this number because the precise mechanistic contributions to Tl and T2 are not accurately known, but are simply using the ratio as a benchmark. The macromolecular rotational correlation time for cytochrome edi can be estimated from the gross molecular weight to be in the range of l-2 X 10d7 s, but if this were the determining correla-
AND
TAYLOR
tion time for magnetic relaxation, no detectable resonances would be observed. Therefore, some faster process must be responsible for the observed magnetic relaxation. The relaxation contribution from paramagnetic effects can be expected to be major. In the most general case [see Ref. (22) for a review of the theory] they could affect Tl and T2 unequally. The electron spin correlation time for heme dl has been reported to be 3.2 X lo-’ s based upon measurements of the solvent water proton relaxation (15). Such a correIation time would lead to large TJT, ratios for heme d1 resonances. However, the same report described the electron spin correlation time for heme c as being much shorter, which would predict noticeably different ratios for heme c resonances in the ‘H NMR hyperfine spectrum. The critical observation is that the observed linewidths and Tl’s are comparable for both heme sites in the native form. For what it is worth, in frozen solutions at low temperatures the ESR linewidths for heme di and heme c have appeared to be similar (lo), suggesting that in this physical state the electron spin relaxation times are similar for the two hemes. The inability to detect hyperfine shifted resonances in ferrocytochrome cd1 was an experimental situation that was accepted only after examining independent batches of protein reduced with both ascorbate and dithionite at a variety of temperatures. Magnetic circular dichroism data (12) and direct magnetic susceptibility measurements (13) have both concured that the ferrocytochrome is paramagnetic with high-spin heme di. It is possible that the electron spin relaxation time of ferrous heme d1 is so long that proton hyperfine shifted resonances would be too broad to detect. In this circumstance one might expect narrow EPR resonances for ferrous heme dr. It has not been possible to detect EPR signals in the ferrocytochrome at cryogenic temperatures [(19), but see also Ref. (23)]. It may be that the electronic relaxation time of the ferrous heme is strongly temperature dependent. Magnetic susceptibility measure-
NMR
OF
CYTOCHROME
ments near physiological temperatures have shown that the magnetic susceptibility of ferrocytochrome cdl changes in a manner opposite to expectations based upon Curie’s law. A possible hypothesis to explain the behavior was advanced that a spin state equilibrium existed that mixed a small portion of an unknown spin state with a dominant high spin, S = 2, state. Such a spin state fluctuation could account for very broad, undetectable hyperfine proton resonances for ferrocytochrome cdl. The single resolved resonance at -2.8 ppm in both ferrocytochrome cdl and the reduced quasi-apo protein may be an indication that methionine is a ligand to heme c. In low-molecular-weight, low-spin ferrocytochromes c a methionyl methyl resonance arising from ligated methionine is observed within 0.3 ppm of -3 ppm [see Ref. (17) a.nd references cited therein]. The unusual shift of this methyl is ascribable to the diamagnetic heme ring current, and is expected to be approximately constant for similar iron-methionine coordi:nation geometries.
cdl
7.
8. 9.
10. 11.
12.
13. 14.
15. 16. 17. 18.
REFERENCES 19. 1. KURONEN, T., SARASTE, M., AND ELLFOLK, N. (1975) Biochim Biophys. Acta 393, 48-54. 2. YAMANAKA, T., AND OKUNUKI, K. (1963) Biochim. Biophys. Acta 67, 407-416. 3. TIMKOVICH:, R., CORK, M. S., AND TA~OR, P. V. (1984) J’. BioL C&m. 259, 1577-1585. 4. TIMKOVICH, R., AND CORK, M. S. (1982) Biochemistry 21,5119-5123. 5. HILL, K. E:., AND WHARTON, D. C. (1978) J. BioL Chem 253,489-495. 6. WALSH, T. A., JOHNSON, M. K., BARBER, D.,
20. 21. 22.
23.
DERIVATIVES
697
THOMSON, A. J., AND GREENWOOD, C. (1980) J. Inorg. Bicdem. 14.15-31. ROBINSON, M. K., MARTINKUS, K., KENNELLY, P. J., AND TIMKOVICH, R. (1979) Biochemistry 18, 3921-3926. BARBER, D., PARR, S. R., AND GREENWOOD, C. (1978) Biochem. J. 175,239-249. WALSH, T. A., JOHNSON, M. K., THOMSON, A. J., BARBER, D., AND GREENWOOD, C. (1981) J. Inorg. B&hem. 14, l-14. MUHOBERAC, B. B., AND WHARTON, D. C. (1983) J. BioL Chem. 258, 3019-3027. SILVESTRINI, M. C., COLOSIMO, A., BRUNORI, M., WALSH, T. A., BARBER, D., AND GREENWOOD, C. (1979) Biochem J. 183,701-709. WALSH, T. A., JOHNSON, M. K., GREENWOOD, C., BARBER, D., SPRINGALL, J. P., AND THOMSON, A. J. (1979) Biochem. J. 177, 29-39. TIMKOVICH, R., AND CORK, M. S. (1983) B&him Biophys. Acta 742, 162-168. CHING, Y., ONDRIAS, M. R., ROUSSEAU, D. L., MUHOBERAC, B. B., AND WHARTON, D. C. (1982) FEBS I&t. 138, 239-244. BLATT, Y., BENKO, B., PECHT, I., AND VUK-PAVLOVIC, S. (1981) J BioL Chem. 256, 2297-2301. TIMKOVICH, R., CORK, M. S., AND TAYLOR, P. V. (1984) Biochemists 23, 3526-3533. TIMKOVICH, R., AND CORK, M. S. (1984) Biochemistry 23, 851-860. VICKERY, L. E., PALMER, G., AND WHARTON, D. C. (1978) B&&em Biophys. Res. Commun 80, 458-463. GUDAT, J. C., SINGH, J., AND WHARTON, D. C. (1973) Biochim. Biophys. Actu 292, 376-390. KELLER, R. M., WUTHRICH, K., AND PECHT, I. (1976) FEB.9 Id&t. 70, 180-184. SCHICHMAN, S. A., AND GRAY, H. B. (1981) J. Amer. Chem. Sot 103, 7794-7795. SWIFT, T. J. (1973) in NMR of Paramagnetic Molecules (LaMar, G. N., Horrocks, W. D., and Holm, R. H., eds.), pp. 53-83, Academic Press, New York. MUHOBERAC, B., AND WHARTON, D. C. (1980) J. BioL Chem. 255, 8437-3442.
‘H NMR Spectroscopy
of Cytochrome
RUSSELL TIMKOVICH,2
MARGARET
.Department
Illinois
of Chemistry,
Received
December
S. CORK,
Institute
cd, Derivatives’ AND
of Chemistry,
26, 1984, and in revised
form
PRISCILLA Chicago, March
Illinois
V. TAYLOR 60616
25, 1985
Proton nuclear magnetic resonance spectra are reported for cytochrome cdl from Pseudwmxmas aeruginosa (ATCC 19429) in several forms including complexes of the ferricytochrome with cyanide, azide, and fluoride, a quasi-apo form in which the noncovalently associated heme dl has been removed but the covalently bound heme c is retained, and the reduced state of both native and the quasi-apo forms. Comparisons are made to the previously reported spectrum of ferricytochrome cdl. The following points are :made. (i) The spectra of the azide and fluoride complexes and the ferric quasi-apo form show perturbation of resonances assignable to the site of heme dr, and leave relatively unperturbed resonances assignable to the site of heme c. The heme dl associated resonances are at 46.0, 35.4, 23.3, 17.5, -2.9, and -16 ppm, and the heme c associated resonances are at 42.0, 33.7, 15.0, 13.9, -7.5, -14, and -33 ppm in native ferricytochrome cdl. (ii) The similarity of the hyperfine resonances of the ferric quasi-apo form to the heme c resonances of intact ferricytochrome cdl is evidence that removal of heme d1 leaves the heme c binding site relatively unaltered. (iii) Linewidths and relaxation times suggest that the relaxation times of the unpaired electron spins of the ferric hemes c and d1 are on the same order of magnitude. (iv) Although i,t is paramagnetic, ferrocytochrome cdl does not demonstrate an experimentally detectable hyperfine shifted spectrum under present conditions. Possible reasons for this are discussed. (v) The presence of a narrow resonance at -2.8 ppm in both ferrocytochrome cdl and the reduced state of the quasi-apo form suggests that methionine may be a ligand to heme c. o 19% Academicpress,I~~.
Cytochrome cd1 is a dissimilatory nitrite reductase present in facultative denitrifying bacteria. The native enzyme is a dimer composed of two identical subunits, each with a molecular weight of approximately 60,000 and containing one covalently bonded heme c and one noncovalently associated heme d1 (1). The heme c is believed to be an iron chelated form of i This work was supported by National Institutes of Health Grants GM23869 (R.T.) and GM2607102Sl (from the NIGMS Shared Instrumentation Program for support of the NMR facility). a To whom correspondence should be addressed, at the present address: Department of Chemistry, P. 0. Box H, University of Alabama, University, Ala. 35486. 689
protoporphyrin IX in which two polypeptide cysteinyl residues have added across the original vinylic substituents to form covalent thioether bonds. Although there is only optical spectroscopic evidence for this in terms of the absorption maxima and intensities of the pyridine hemochromogen (2), such evidence has been upheld in those cases of where additional structural studies have confirmed the porphyrin structure. Heme d1 has recently been shown to be the iron chelate of an unusual chlorin macrocycle, termed acrylochlorin (3). The paramagnetism of the ferric heme protein causes large hyperfine chemical shifts in the ‘H NMR spectrum that resolve some resonances from the usual 0- to lo-ppm range. The hyperfine 0003-9861185
$3.00
Copyright0 1985by AcademicPress,Inc. All rights of reproductionin anyform reserved.
690
TIMKOVICH,
CORK.
shifted resonances are either protons directly bonded to the paramagnetic hemes or protons from nearby amino acid side chains. The hyperfine shifted spectrum is thus an excellent probe of the active sites of the cytochrome. An initial investigation (4) reported the detection of hyperfine shifted resonances in ferricytochrome cdl and examined the effects of pH and temperature. In order to make full use of these hyperfine shifted resonances as probes of the heme sites, it would be necessary to have a complete assignment to specific protons. This is a formidable task for such a large macromolecule. A useful and feasible first step is to make a partial assignment of resonances in terms of whether they arise from the heme c or heme di site. This is the main objective of the present report. In addition, the limited data obtainable on the ‘H NMR spectrum of the reduced state of cytochrome cdl will be presented. MATERIALS
AND
METHODS
The purification of cytochrome cd, from Pseudo monos aeruginosa ATCC 19429 has been described (4). Several procedures have been reported for the extraction of the noncovalently bound heme di (2, 5, 6). The resulting reddish protein will be called quasiapo protein because it retains the covalently bound heme c. The procedures cited rely upon acidic acetone or low pH extraction of heme d, and produce a red precipitate that is the quasi-apo protein. In our hands these procedures all successfully extracted heme di, but we were unable to resolubilize the quasi-apo protein at concentrations in deuterium oxide buffers suitable for NMR spectroscopy. The acid treatment may have denatured at least some portion of the sample, rendering it insoluble and/or preventing the solubilization of other more nativelike material. A more soluble quasi-apo protein was prepared by mild aqueous acid extraction of heme d,. Cold native enzyme was titrated with glacial acetic acid to pH 3 with constant stirring. A reddish precipitate was collected by centrifugation and redissolved in 50 mM sodium carbonate/bicarbonate buffer at pH 9. This was dialyzed in the cold overnight against phosphate buffer and chromatographed on DEAE-cellulose under the same conditions for the corresponding step in the isolation of the native cytochrome (4). A reddish band eluted at 0.15 to 0.20 M in the KC1 gradient. Yields were variable and generally low. The quasi-apo protein so obtained was found to match a red pigment obtained at the same elution volume from DEAE-cellulose during
AND
TAYLOR
the normal isolation of the enzyme. Both had identical uv-visible spectra, molecular weights on gel filtration columns, and could be reconstituted in dilute solution with heme di extracts as described by Hill and Wharton (5). When the reconstituted protein was assayed with Pseudmmm ferrocytochrome ~551 as donor (5), it was found to have ‘70100% of the specific activity of native enzyme, which had a maximal velocity of 170 pmol donor oxidized min-’ pmol enzyme-‘. The range of recovered activity observed was somewhat broader than previously reported (5), but this may be due in part to ambiguity in assaying initial rates, since the reaction never has a true zero-order rate dependence on donor concentration (7). The latter pigment presumably has been produced by accidental loss of heme d, during the enzyme isolation. The visible spectra of both pigments matched a report for quasi-apo protein prepared by acidic acetone extraction [Fig. 1, Ref. (5)]. In order to obtain sufficient quantities for NMR studies, the various preparations of quasi-apo protein were combined. The NMR spectrum of this material agreed with the native enzyme in the general aromatic and aliphatic regions, although lack of resolution in these crowded spectra preclude conclusions concerning similarity of tertiary structure. Preparation of samples for NMR spectroscopy and conditions for obtaining spectra at 300 MHz have been described (4). Chemical shifts are reported in parts per million from sodium 4,4-dimethyl-4-silapentane sulfonate. Spin-lattice relaxation times were measured by an inversion-recovery pulse sequence. Recycle time was 160 ms or at least lOTi for the slowest hyperfine shifted resonance. For each of seven delay times, 350,000 transients were collected. Mixtures of ferricytochrome cd, and potassium cyanide or sodium azide were prepared by adding the ligands as aliquots from concentrated stock solutions in deuterium oxide, whose pH had been previously adjusted to match the protein solution. Sodium fluoride was added as a solid. In all three cases, the final pH was checked after additions with a glass micro-combination electrode and any fine adjustments in pH were made with stock solutions of NaO’H or *HCl. Values of pH represent the uncorrected glass electrode reading and will be designated pH*. NMR samples in the presence of exogenous ligands were examined by optical spectroscopy either in 1-mm-pathlength cells or after dilution with deuterium oxide in standard cells. Cyanide (8) and azide (9) caused optical changes that have been previously reported. The optical effects of even high concentrations of fluoride were subtle. In difference spectra with and without fluoride, weak features were detected at 643 and 470 nm, similar to a published report (10). The question of whether fluoride forms a true complex with cytochrome cd, has been equivocal, with one group reporting a complex
NMR
OF
CYTOCHROME
with heme di based upon optical and EPR results (10) and another group claiming no spectral effects for fluoride (8). From NMR data to be presented, it will be clear that high concentrations of fluoride do exert an effect on the NMR spectrum of ferricytochrome cdl. But, it is conceivable that it does so by acting as a selective denaturant of the heme di site, without forming a true stable complex. The germaine issue for the present study is only that fluoride does selectively perturb the NMR spectrum. For brevity, reference will be made to a fluoride complex. Samples of ferrocytochromes were prepared by the addition of a slight excess of solid sodium dithionite, or concentrated sodium ascorbate from a neutralized solution in deuterium oxide. Reduced samples were
691
cd, DERIVATIVES
sealed in glass 5-mm NMR tubes after alternate cycles of vacuum degasing and flushing with argon. Complete reduction was confirmed after NMR spectroscopy by breaking open the sample tubes, transfering to optical cuvettes with l-mm pathlengths, and examining the ferrous 01 bands (11) before and after the addition of still more reductant.
RESULTS
Measured spin-lattice relaxation times for the resolved hyperfine resonances of ferricytochrome cdl are reported in Table I. NMR spectra were obtained for com-
TABLE
I
SPIN-LATTIC:E RELAXATION TIMESAND CHEMICAL SHIFTSFOR VARIOUS FORMSOF FERRICYTOCHROMEcdl Native (ppm)
T, (ms)
46.0 42.0 35.4
10.7 10.1 11.4
33.7 32.5
10.0 10.8
23.3
15.0
17.5
8.2
Azide (mm)
Fluoride
51. 49.
57.
41.9 34.7 33.5 26-23
Quasi-apo (ppm)
Cyanide bpm)
42.0
41.3
42-40
33.4 32.5
33.5
(mm)
(?)
25.0 21.8
17 (?)
22-20
17. 16-14
15.0 13.9
-2.9
13.6
15. 13.5
16.3 14.8 14.0 12.7 -2.1
15.6 13.9
6.9 -3.3
(?)
-4.1 -7.5
(?)
-3.3
-3.8 -7.5 -14. -16. -24. -33.
4.5 3.7 3.7
-14.
-7.5 -10.9 -13.7
-24. -32.
-32.9
Note. Chemical shifts are reported in parts per million at 25°C and pH* 7.5. Because of the widths of the peaks, the precision of shift values was generally fO.l ppm for the narrowest and f0.4 ppm for the very broadest. Spin-lattice relaxation times, T,, are reported in milliseconds. Statistical deviations based upon the linearity of the usual log plot of residual magnetization versus delay time were on the order of 10%. For some weak resonances, relaxation times could not be reliably measured, and no value is quoted. The symbol (?) indicates especially weak resonances.
692
TIMKOVICH,
CORK,
plexes of ferricytochrome cc& with the exogenous ligands azide, fluoride, and cyanide. Typical data are presented in Figs. 1 and 2. The spectra of such complexes were inevitably noisier than for the native ferricytochrome. Based upon either visible or EPR spectroscopy, azide (9) and fluoride (10) have been reported to bind only to heme dl, while cyanide binds to both heme d1 and heme c in the ferric state (9, 10). The spectrum of the azide complex, Fig. lB, showed perturbation of some but
AND
TAYLOR
H B
I.” -10
-
I -20
’
”
1
I -30
’
1.1 PP”
FIG. 2. Upfield hyperfine NMR spectra of various forms of Pseudomonas ferricytochrome cdl. Conditions were as described for Fig. 1. (A) Native ferricytochrome cdl; (B) plus 100 mM sodium azide; (C) plus 500 mM sodium fluoride; (D) the quasi-apo ferric form; (E) native plus 100 mM potassium cyanide. The vertical scales are not equivalent.
FIG. 1. Downfield hyperfine NMR spectra of various forms of Pseudomonas ferricytochrome cdl. All spectra were recorded at 25°C. Protein concentrations were 1 mM, expressed as the subunit concentration. The solvent was 99.7% deuterium oxide buffered with sodium phosphate to a final pH* of 7.5 with additional reagents as listed. (A) Native ferricytochrome cdl; (B) plus 100 mM sodium azide; (C) plus 500 mM sodium fluoride; (D) the quasi-apo ferric form; (E) native ferricytochrome cd, plus 100 mM potassium cyanide. The vertical scales of the spectra were chosen in order to give a reasonable presentation of the main features in a concise manner, and are not the same for all spectra. The intensity of the peak at circa 42 ppm may be used as an approximate internal standard. The absolute integrated intensity of this resonance was the same within our experimental accuracy for (A)-(D) but for (E) the 42-ppm feature was fourfold weaker.
not all of the hyperfine shifted resonances found in the native ferricytochrome. The 46.0-ppm resonance was absent, but a new major resonance appeared at 49 ppm with some satellite peaks. The 35.4- and 1’7.5ppm resonances were absent. The 23.3ppm resonance was broader with new satellites. The upfield region (Fig. 2B) was especially noisy and was difficult to interpret. The -2.9-ppm resonance appeared to be affected and a new weak resonance appeared at -3.8 ppm. In the spectrum shown, the resonances at -7.5, -14, and -16 ppm appear to be absent, but in other spectra of the complex there was evidence of peaks at -7.5 and -14 ppm. Compared to these changes, the resonances at 42.0, 33.7, 32.5, and 15.0 ppm were much less perturbed. In the fluoride complex, the 46ppm resonance also disappeared and a new broad feature at 5’7 ppm appeared. The 35.4- and -16-ppm resonances disappeared, while the resonances at 23.3, 17.5, and -2.9 ppm decreased in intensity.
NMR
OF
CYTOCHROME
Resonances at 42.0, 33.7, 32.5, 15.0, -7.5, and -14 ppm were less affected. The very broad resonance at -33 ppm has been difficult to detect in some native ferricytochrome spectra, but appeared in the fluoride complex. Cyanide caused complex, time-dependent changes in the hyperfine spectrum. In this context it must be stressed that observation of any ferricytochrome cdl spectrum required extensive signal averaging. A minimum of 6 h of accumulation of averaged transients was required before hyperfine peaks could be detected above baseline noise. Changes taking place during this time would not be apparent but would increase the noise of spectra. The spectrum displayed in Fig. 1E is representative of spectra obtained within 12 h of mixing cyanide and protein. Very broad peaks were detected around 26,22,15, and -6 ppm. The broad resonance observed around 42 ppm in Fig. 1E was absent in some spectra, and was replaced by a resonance of comparable intensity and line width at 60 ppm. After 12 h, the previously observed peaks disappeared and no hyperfine shifted resonances could be detected. The NMR spectrum of the ferric form of the quasi-apo protein is shown in Fig. 1D. Hyperflne shifted resonances at 41.4, 33.6, 14.67, -7.5, -13.69, and -32.9 ppm were similar to resonances observed in native ferricytochrome cdl, but narrower. Additional resonances were observed at 16.4,12.8, -2.1, -3.3, and -10.9 ppm. Their intensity is such that they may have been obscured by noise or by overlap with heme di associated resonances in the native ferricytochrome spectrum. Ferrocytochrome cdl is known to be a paramagnetic heme protein with high spin ferrous heme d1 based upon magnetic circular dichroism (12) and magnetic susceptibility (13) data. NMR spectroscopy failed to detect hyperfine shifted resonances in the reduced state. Independent batches of protein and the alternative reducing reagents dithionite and ascrobate were examined. The same samples sealed in the NMR tubes were examined by resonance Raman spectroscopy (Lewis, Cotton, and
693
cd, DERIVATIVES
Timkovich, unpublished) and were found to give the native resonance Raman spectrum of ferrocytochrome cdl (14). The only noteworthy feature was a clearly resolved resonance at -2.8 ppm. Its linewidth (80 Hz full width at half height) is reasonably sharp for a protein the size of cdl and it is unlikely to be a hyperfine shifted resonance. The reduced state of the quasi-apo protein also did not show any detectable hyperfine shifted resonances, but included the -2.8-ppm resonance. DISCUSSION
For the sake of brevity in the ensuing discussion, the designations “heme c resonances” or “heme di resonances” will include covalently bonded heme protons as well as polypeptide protons in the vicinity of the heme that might appear in the paramagnetically shifted spectrum. Spin-lattice relaxation times were measured for ferricytochrome cdl because it was believed that such data might contribute evidence toward the respective assignment of hyperfine resonances to either the heme c or heme d1 site. It has been reported for ferricytochrome cdl that ferric heme di has an electron spin relaxation time much longer than ferric heme c (15). The unpaired electron spin is certainly a major factor in the mechanism of proton resonance relaxation for hyperfine shifted resonances. For macromolecules in general, relaxation times may be influenced by magnetization transfer (spin diffusion), but this is not likely to be of importance for the hyperfine shifted resonances of cdl because electron-proton interactions are governed by a correlation time much shorter than proton-proton dipole interactions. It was anticipated that resonances arising from heme d, should have significantly different spin-lattice relaxation times then those arising from heme c. It was surprising that the observed data did not fall into two clear categories. The heme methyl groups in particular had relaxation times all on the same general time scale. Observed differences cover the same spread as has been observed for
694
TIMKOVICH,
CORK,
heme methyls in other cytochromes, such as ferricytochromes ~550 (16) and ~554 (17). The differences could thus be ascribed to differences in the exact distance from the paramagnetic centers and to different contributions of contact, pseudocontact, and proton dipole-dipole mechanisms to the total relaxation rate. Our data do not allow a calculation of electron spin correlation times, but suggest to a first approximation that they are of the same order of magnitude for heme di and heme c. The hyperfine spectrum of the ferric form of the quasi-apo protein is the best available evidence for assignment of resonances to the heme c region. Resonances at 41.3, 33.5, 14.8, 14.0, -7.5, -13.7, and -32.9 ppm from the quasi-apo protein are aligned in Table I with resonances at 42.0, 33.7, 15.0, 13.9, -7.5, -14, and -33 ppm in ferricytochrome cdl because they are believed to represent the same spin systems. They may be directly bonded heme protons or protons of nearby amino acid side chains. Other hyperfine shifted resonances in ferricytochrome cd, are attributed to the heme di region. The hyperfine chemical shift of a resonance is a sensitive probe of environment. The similarity of the spectra of the heme c associated resonances between native and quasi-apo forms is suggestive that the heme c environment has not been radically perturbed by the loss of heme di. Magnetic circular dichroism data have been interpreted as indicating that there are not strong heme-heme interactions (such as heme stacking) in cytochrome cdl (18), and the NMR data would support this conclusion. There are some differences in the resonances of the quasi-apo form. Most notable are that the partially resolved pair at 33.7 and 32.5 ppm in the native ferricytochrome may have collapsed to a single peak at 33.5 ppm in the quasiapo form, the resonance at 14.7 ppm may be more intense in the quasi-apo form, and the -32.9-ppm resonance is narrower. The reasons for these changes are unknown. They might arise from modest conformational changes at the binding site.
AND
TAYLOR
The hyperfine NMR spectra of the azide and fluoride complexes of ferricytochrome cdl on the whole support the relative assignments deduced from comparing native and quasi-apo forms. Since visible and EPR spectroscopy have indicated that heme d1 is the binding site for these exogenous ligands, it is reasonable that resonances associated with heme d1 should be most severely affected by complexation. Complexation could be expected to alter the g-tensor as well as the unpaired electron spin density around heme di, and therefore could perturb both contact and pseudocontact shifted resonances. The actual extent of perturbation on resonances could not be predicted a priori and at this time it would not be possible to analyze the perturbations in detail in terms of structure. By themselves, the spectra of the complexes are somewhat ambiguous to interpret, because they are noisier and because the differential effects on resonances have to be judged by degree. Nevertheless, the total impact of the data largely confirms the relative assignments. Cyanide is a strong field ligand that can lower the spin state and/or decrease the electron spin relaxation time and so sharpen hyperfine shifted resonances in some ferric heme proteins. It is well known to accomplish this in model heme complexes and the globins. Since cyanide has been shown by EPR and visible spectroscopy to bind to both hemes of cdl, it was reasonable to expect that many native hyperfine shifted resonances would be perturbed. Although both hemes in the native ferric state are already low spin (4, g-14), a change in the electron spin relaxation time and/or the g-tensor anisotropy was a reasonable expectation. It was surprising to find that most of the hyperfine spectrum has disappeared, and there is a slow process with a time scale of hours that eliminates all observable resonances. Since the resonances of the intermediate form are already broader than the native form, we interpret the disappearance of the hyperfine resonances as due to further broadening. The possible mechanism of this broadening will be discussed in subsequent paragraphs. EPR
NMR
OF
CYTOCHROME
has been used to probe the electronic environment for the native form and the cyanide complex of ferricytochrome cc&. While there is agreement concerning the native spectrum, the effects of cyanide complexation have been contentious. One report (19) indicated no change for signals arising from heme c while stating that the heme dl signals disappeared. A more recent report (10) from the same group refined their measurements and reported changes in heme c signals as well as newly assigned heme d1 signals. Another group (12) reported slight changes in g values for the heme c signals and more extensive changes for the heme d1 signals, but the heme d1 remained low spin. These measurements indicate that at least heme c and probably heme d1 as well do not undergo m,ajor changes in spin state or relaxation time as a result of cyanide binding, although the crystal field symmetry may change for both hemes. Therefore, it is unlikely that the line broadening observed in the NMR spectrum is the result of ;a change in gross electronic properties, such as conversion to highspin ferric iron. The linewidths observed for the quasiapo protein are obviously narrower than for the native enzyme and these in turn are narrower than for the complexes with exogenous ligands. Several mechanisms could account for this, but there are difficulties for all but one. Possibilities to consider include (a) dimer to monomer dissociatio:n, (b) restricted local motion in the native form, (c) pseudocontact interactions on heme c resonances by ferric heme dl, (d) chemical exchange broadening due to a mixture of oxidation states (17, 20), and (e) chemical shift heterogeneity. Point a is unlikely because the operative correlation time would probably be dominated by the electron relaxation time rather than the isotropic rotational tumbling time. At any rate, on gel-filtration columns, the quasi-apo form behaved with the same apparent molecular weight as the native form. Points b and c are unlikely because they would preferentially affect only some resonances, while the observation is that all are broadened in
cdl
DERIVATIVES
695
the native form. Point d is unlikely because kinetic measurements of electron transfer rates in the native enzyme indicate that exchange would be slow compared to the NMR time scale (21), and because spectra were insensitive to the excess external oxidant ferricyanide. The prefered interpretation is point e, chemical shift heterogeneity in the native versus the quasi-apo form. Each apparent peak in the native spectrum represents a grouping of peaks with narrower intrinsic linewidths, but slightly different chemical shifts. The observed linewidths in the quasi-apo form are interpreted as more representative of the intrinsic linewidth of hyperfine resonances in cytochrome cc&. Excessive broadening in the native form is attributed to chemical shift heterogeneity due to a range of heme dl-protein noncovalent interactions in our NMR samples. It should be stressed that we are not interpreting the native samples as being grossly denatured. Because of the strength of paramagnetic interactions, it would require only modest perturbations in conformations and/or electronic properties to produce a sizable spread of chemical shifts. To illustrate, consider AZcaligenes ~554 where ionization of a remote propionate group in the heme pocket causes chemical shifts in the range of 23 ppm for hyperfine shifted heme methyl resonances (1’7). The existence of a range of readily accessible conformations only slightly different could reflect a “loose” heme d1 crevice that would correlate with the lability of the heme dl-protein association. The line broadening and concommittent low signal-to-noise ratio in the spectra of complexes with exogenous ligands may result from the ligand further distorting the heme binding site and creating an even larger range of shifts. The chemical shift heterogeneity interpretation is supported by two sets of observations. In the first case, selective decoupling of the hyperfine shifted resonances of the native form burns a hole in the center of the resonances without saturating the entire apparent peak. Equivalent decoupler irradiation of the hyperfine shifted resonances of the quasi-apo
696
TIMKOVICH,
CORK,
form led to complete collapse of the peak. A lower decoupler power could not be found sufficient to just burn a hole in the quasi-apo peaks. This argues that in the native form, the apparent peaks are actually superimposed resonances in slow exchange on the NMR time scale with respect to their chemical shift differences. Of course, the limit of slow exchange is static, noninterconverting conformations. In the second case, supportive evidence comes from a comparison of spin-lattice ( T1) and spin-spin (Tz) relaxation times. T2 values may be estimated from the observed linewidths since the intrinsic linewidth for this size macromolecule is much greater than field inhomogeneities. For the native protein, the ratio of T,/T, is on the order of 10 to 20. For example, the resonance at 46 ppm in the native protein has an apparent Tz of 0.8 ms while Tl is 10.7 ms. At 23 ppm the respective values are 0.7 and 15 ms. For the quasiapo protein the ratio is on the order of 3. For example, for the 34-ppm resonance the values are 3.0 and 10 ms. A paramagnetic macromolecule the size of cytochrome cd1 is likely to be beyond the extreme narrowing limit for resonances at 300 MHz, so a Tl longer than T2 is not wholly surprising. However, the magnitude of the ratio in the native form spectrum is excessive when compared to most heme proteins. For other cytochromes studied by this lab, the ratio of Tl (determined by inversion-recovery experiments) to T2 (determined from line widths) has been on the order of 3-4. For example, the ratio for the heme methyls in Parucoca cytochrome ~550 (16) and Alcaligenes cytochrome ~554 (17) average to 3.3, which is comparable to the ratio in the quasi-apo form. We are not attempting to attach any deep significance to this number because the precise mechanistic contributions to Tl and T2 are not accurately known, but are simply using the ratio as a benchmark. The macromolecular rotational correlation time for cytochrome edi can be estimated from the gross molecular weight to be in the range of l-2 X 10d7 s, but if this were the determining correla-
AND
TAYLOR
tion time for magnetic relaxation, no detectable resonances would be observed. Therefore, some faster process must be responsible for the observed magnetic relaxation. The relaxation contribution from paramagnetic effects can be expected to be major. In the most general case [see Ref. (22) for a review of the theory] they could affect Tl and T2 unequally. The electron spin correlation time for heme dl has been reported to be 3.2 X lo-’ s based upon measurements of the solvent water proton relaxation (15). Such a correIation time would lead to large TJT, ratios for heme d1 resonances. However, the same report described the electron spin correlation time for heme c as being much shorter, which would predict noticeably different ratios for heme c resonances in the ‘H NMR hyperfine spectrum. The critical observation is that the observed linewidths and Tl’s are comparable for both heme sites in the native form. For what it is worth, in frozen solutions at low temperatures the ESR linewidths for heme di and heme c have appeared to be similar (lo), suggesting that in this physical state the electron spin relaxation times are similar for the two hemes. The inability to detect hyperfine shifted resonances in ferrocytochrome cd1 was an experimental situation that was accepted only after examining independent batches of protein reduced with both ascorbate and dithionite at a variety of temperatures. Magnetic circular dichroism data (12) and direct magnetic susceptibility measurements (13) have both concured that the ferrocytochrome is paramagnetic with high-spin heme di. It is possible that the electron spin relaxation time of ferrous heme d1 is so long that proton hyperfine shifted resonances would be too broad to detect. In this circumstance one might expect narrow EPR resonances for ferrous heme dr. It has not been possible to detect EPR signals in the ferrocytochrome at cryogenic temperatures [(19), but see also Ref. (23)]. It may be that the electronic relaxation time of the ferrous heme is strongly temperature dependent. Magnetic susceptibility measure-
NMR
OF
CYTOCHROME
ments near physiological temperatures have shown that the magnetic susceptibility of ferrocytochrome cdl changes in a manner opposite to expectations based upon Curie’s law. A possible hypothesis to explain the behavior was advanced that a spin state equilibrium existed that mixed a small portion of an unknown spin state with a dominant high spin, S = 2, state. Such a spin state fluctuation could account for very broad, undetectable hyperfine proton resonances for ferrocytochrome cdl. The single resolved resonance at -2.8 ppm in both ferrocytochrome cdl and the reduced quasi-apo protein may be an indication that methionine is a ligand to heme c. In low-molecular-weight, low-spin ferrocytochromes c a methionyl methyl resonance arising from ligated methionine is observed within 0.3 ppm of -3 ppm [see Ref. (17) a.nd references cited therein]. The unusual shift of this methyl is ascribable to the diamagnetic heme ring current, and is expected to be approximately constant for similar iron-methionine coordi:nation geometries.
cdl
7.
8. 9.
10. 11.
12.
13. 14.
15. 16. 17. 18.
REFERENCES 19. 1. KURONEN, T., SARASTE, M., AND ELLFOLK, N. (1975) Biochim Biophys. Acta 393, 48-54. 2. YAMANAKA, T., AND OKUNUKI, K. (1963) Biochim. Biophys. Acta 67, 407-416. 3. TIMKOVICH:, R., CORK, M. S., AND TA~OR, P. V. (1984) J’. BioL C&m. 259, 1577-1585. 4. TIMKOVICH, R., AND CORK, M. S. (1982) Biochemistry 21,5119-5123. 5. HILL, K. E:., AND WHARTON, D. C. (1978) J. BioL Chem 253,489-495. 6. WALSH, T. A., JOHNSON, M. K., BARBER, D.,
20. 21. 22.
23.
DERIVATIVES
697
THOMSON, A. J., AND GREENWOOD, C. (1980) J. Inorg. Bicdem. 14.15-31. ROBINSON, M. K., MARTINKUS, K., KENNELLY, P. J., AND TIMKOVICH, R. (1979) Biochemistry 18, 3921-3926. BARBER, D., PARR, S. R., AND GREENWOOD, C. (1978) Biochem. J. 175,239-249. WALSH, T. A., JOHNSON, M. K., THOMSON, A. J., BARBER, D., AND GREENWOOD, C. (1981) J. Inorg. B&hem. 14, l-14. MUHOBERAC, B. B., AND WHARTON, D. C. (1983) J. BioL Chem. 258, 3019-3027. SILVESTRINI, M. C., COLOSIMO, A., BRUNORI, M., WALSH, T. A., BARBER, D., AND GREENWOOD, C. (1979) Biochem J. 183,701-709. WALSH, T. A., JOHNSON, M. K., GREENWOOD, C., BARBER, D., SPRINGALL, J. P., AND THOMSON, A. J. (1979) Biochem. J. 177, 29-39. TIMKOVICH, R., AND CORK, M. S. (1983) B&him Biophys. Acta 742, 162-168. CHING, Y., ONDRIAS, M. R., ROUSSEAU, D. L., MUHOBERAC, B. B., AND WHARTON, D. C. (1982) FEBS I&t. 138, 239-244. BLATT, Y., BENKO, B., PECHT, I., AND VUK-PAVLOVIC, S. (1981) J BioL Chem. 256, 2297-2301. TIMKOVICH, R., CORK, M. S., AND TAYLOR, P. V. (1984) Biochemists 23, 3526-3533. TIMKOVICH, R., AND CORK, M. S. (1984) Biochemistry 23, 851-860. VICKERY, L. E., PALMER, G., AND WHARTON, D. C. (1978) B&&em Biophys. Res. Commun 80, 458-463. GUDAT, J. C., SINGH, J., AND WHARTON, D. C. (1973) Biochim. Biophys. Actu 292, 376-390. KELLER, R. M., WUTHRICH, K., AND PECHT, I. (1976) FEB.9 Id&t. 70, 180-184. SCHICHMAN, S. A., AND GRAY, H. B. (1981) J. Amer. Chem. Sot 103, 7794-7795. SWIFT, T. J. (1973) in NMR of Paramagnetic Molecules (LaMar, G. N., Horrocks, W. D., and Holm, R. H., eds.), pp. 53-83, Academic Press, New York. MUHOBERAC, B., AND WHARTON, D. C. (1980) J. BioL Chem. 255, 8437-3442.