A further clue to understanding the mobility of mitochondrial yeast cytochrome c

A 15N T investigation of the oxidized and reduced species

Authors


I. Bertini, CERM and Department of Chemistry, University of Florence, Via L. Sacconi, 6 – Sesto Fiorentino, Italy. Fax: + 39 0554574271, Tel.: + 39 0554574272, E-mail: bertini@cerm.unifi.it

Abstract

A new approach was developed to overproduce 15N-enriched yeast iso-1-cytochrome c in the periplasm of Escherichia coli in order to perform a study of the motions in the ms–µs time scale on the oxidized and reduced forms through rotating frame 15N relaxation rates and proton/deuterium exchange studies. It is confirmed that the reduced protein is rather rigid whereas the oxidized species is more flexible. The regions of the protein that display increased internal mobility upon oxidation are easily identified by the number of residues experiencing conformational equilibria and by their exchange rates. These data complement the information already available in the literature and provide a comprehensive picture of the mobility in the protein. In particular, oxidation mobilizes the loop containing Met80 and, through specific contacts, affects the mobility of helix 3 and possibly of helix 5, and of a section of protein connecting the heme propionates to helix 2. The relevance of internal motions to molecular recognition and to the early steps of the unfolding process of the oxidized species is also discussed. In agreement with the reported data, subnanosecond mobility is found to be less informative than the ms–µs with respect to redox dependent properties.

Abbreviations
R

rotating-frame relaxation rate

inline image

off-resonance rotating-frame relaxation rate

Structural differences between the native state of the oxidized and reduced mitochondrial cytochromes c have been revealed by analysis of their crystal and solution structures. These differences are small but significant and mainly involve the heme propionate-7 and surrounding residues [1–6]. Such conformational differences may play a role both in the electron transfer process and in molecular recognition. Oxidized and reduced cytochrome c also have different stability towards unfolding agents, the reduced cytochrome c being significantly more stable than the oxidized protein towards unfolding [7,8]. The difference in stability of the two redox forms has been tentatively related to different protein flexibilities. This hypothesis was based mainly on the different exchange behavior of the backbone amides in the two redox forms of the protein [3–6,9–13].

Heteronuclear NMR spin relaxation spectroscopy constitutes a powerful experimental approach for characterizing time-dependent conformational fluctuations in proteins and so represents a key to the understanding of different biophysical phenomena, including thermodynamic stability, folding, molecular recognition, and catalysis [14,15].

However, the isotopically labeled mitochondrial cytochrome c required for these studies has been difficult to obtain due to the problem of expression of the native holoprotein in the preferred host, E. coli. 15N-Labeled samples of this protein have previously been obtained by two different strategies [13,16,17]. E. coli normally produces c-type cytochromes in the periplasm and recently heterologous expression has yielded 10–15 mg·L−1 of such cytochromes [18]. In the present work to give improved yields we have used a novel approach by coexpressing the structural gene for yeast iso-1-cytochrome c (CYC1) in the E. coli periplasm together with the E. coli ccm genes. The ccm gene products are responsible for the expression of endogenous and exogenous holo c-type cytochromes [19,20]. The cytochrome c gene was fused to the sequence for the signal peptide of E. coli cytochrome b562 in a plasmid known to produce very high levels of protein expression in the periplasm [21]. This expression system allows us to obtain a high level of holo-cytochrome c with the correct heme attachment.

We have used the 15N-labeled protein produced in this way for a characterization of the mobility of the oxidized and reduced forms in the ms–µs time scale and to quickly monitor NH exchange through heteronuclear NMR. Such mobility is most conveniently measured through off-resonance T measurements. They allow the detection of NH groups that experience chemical exchange between two conformations, or conformational equilibria, in the ms–µs time scale [22,23]. Sometimes, also the value of the correlation time for the exchange is extracted. Mobility on the subnanosecond time scale can be obtained through 15N relaxation times, T1 and T2, and 15N-1H NOE. The T2 data sometimes can indicate the presence of chemical exchange. Recently, a paper appeared reporting a subnanosecond analysis based only on T1 and T2 measurements, besides the H2O/D2O NH exchange, of the sample at different pH values [13,24]. The present analysis has allowed us to reveal a network of NH groups experiencing conformational equilibria, its relation with H2O/D2O exchange of NH groups and its redox-dependent mobility.

Materials and methods

Plasmid construction

The vector pMSV1, that expresses protein from the pTrc promotor and confers ampicillin resistance, was constructed as follows. The C102T variant of S. cerevisiae iso-1-cytochrome c gene (kindly donated by A. G. Mauk, University of British Columbia, Vancouver, Canada) was amplified by PCR methods using primers that added BglII and XbaI restriction sites to the 5′ and 3′ ends of the gene, respectively. The resulting PCR product was cloned, using these restriction sites, into the pPB10 plasmid [25]. This fuses the cytochrome c gene to the sequence coding for signal peptide of E. coli cytochrome b562 and adds three N-terminal residues (ADL) from the mature cytochrome b562 to the mature yeast cytochrome c gene to ensure correct export and processing. Indeed it was demonstrated that fusion of the cytochrome c gene directly to a periplasmic target sequence did not lead to any detectable apoprotein or holoprotein expression, probably because of the presence of positively charged residues in the N-terminal sequence (there is a lysine at position 4) that may block export to the periplasm [21]. The expected sequence was confirmed by standard dideoxy sequencing methods.

Protein expression and isolation

pMSV1 was used to transform BL21(DE3)C41 E. coli competent cells already harbouring the pEC86 plasmid (kind gift of L. Thöny-Meyer, ETH Zürich, Switzerland), which expresses the ccm genes under the control of the tet promoter [26]. This plasmid contains genes ccmABCDEFGH cloned into pACY184 together with the chloramphenicol marker. Cells harbouring both plasmids were selected by their ability to grow on 2 × YT plates containing 100 µg·mL−1 ampicillin and 40 µg·mL−1 chloramphenicol.

Protein was expressed in 2-L flasks, containing 1.3 L of 2 × YT medium supplemented with antibiotics. The mediun was inoculated with a liquid culture in exponential phase and culture was grown with shaking, at 37 °C, 100 r.p.m. Induction was performed with isopropyl thio-β-d-galactoside, after overnight growth. Maximal protein expression was obtained 22–24 h after inoculation. Protein was isolated through complete disruption of the cells and purified following the procedure reported by Pollock et al. [16]. After the last chromatographic step fractions with an R value > 4.5 (R = A416/A280) were pooled together and concentrated in an Amicon ultrafiltration cell. The correct processing of unlabelled, mature protein was confirmed by ESI mass spectrometry (expected mass 12 962.8 Da, observed mass 12 962.2 ± 1.0 Da).

The 15N-labelled ADL C102T yeast iso-1-cytochrome c was isolated from cultures performed in M9 minimal medium containing 1.2 g·L−1 (15NH4)2SO4, supplemented with a solution of trace elements, a vitamin mix, the appropriate antibiotics, δ-aminolevulinic acid (0.1 mm) and 2-mercaptoethanesulfonic acid (1 mm). Conical flasks (250 mL) containing 150 mL of minimal medium were inoculated with 1.5 mL of 2 × YT culture in exponential phase (D600 ≈ 0.6) and induced in late exponential phase with isopropyl thio-β-d-galactoside. Cultures were grown with shaking at 37 °C, 100 r.p.m., for 24 h. Protein isolation and purification was performed as reported above.

NMR sample preparation

The NMR samples in D2O were prepared by dissolving the lyophilized oxidized and reduced protein in 100 mm phosphate buffer at pH 7 to give 2 mm solutions.

The oxidized protein tended to autoreduce during the NMR experiments. In order to keep the sample oxidized, a catalytic amount of laccase was added and O2 was passed about 1–2 cm above the solution surface during the experiments. This introduced some extra noise but did maintain the protein in its oxidation state over the time: O2 bubbling affects neither the chemical shift values nor the relaxation times.

The reduced protein was obtained by adding sodium dithionite under anaerobic conditions. All the samples were in phosphate buffer 100 mm at pH 7. The final protein concentration was about 1 mm.

NMR spectroscopy

All NMR experiments were carried out at 303 K on a Bruker Avance 600 NMR spectrometer operating at a proton Larmor frequency of 600.13 MHz. Exceptions to these conditions are indicated in the text.

A assignment of the 15N and 1H amide resonances was obtained from 15N NOESY-HSQC experiments [27]. The spectra were recorded with spectral windows of 10 000 Hz (1H) × 3048 Hz (15N) × 10 000 Hz (1H) for 1024 (1H) × 64 (15N) × 256 (1H) data points.

A series of 1H-15N HSQC spectra on the sample prepared dissolving the lyophilized proteins in D2O solution were collected as a function of time at 700 MHz with spectral windows of 11160 Hz (1H) × 3547 Hz (15N) for 1024 (1H) × 512 (15N) data points. Phase sensitivity improvement was applied using Echo/Antiecho gradient selection [28]. Each HSQC experiment was acquired every 56 min.

The 15N longitudinal relaxation rates, R1, were measured as described previously [29] by using delays in the pulse sequence varying from 20 to 2560 ms. The 15N transverse relaxation rates R2 were measured by using the CPMG sequence as described elsewhere [30]. The relaxation delays used varied from 7.7 to 269.5 ms. All experiments were recorded with a spectral width of 3048 Hz in the F1 (15N frequency) dimension and of 9328 Hz in the F2 (1H frequency) dimension. A total of 200 experiments in t1, each having 2048 real data points, were recorded. Each free induction decay consisted of eight scans. A recycle delay of 3 s was used for all the spectra. Quadrature detection in F1 was obtained by using the time-proportional phase incrementation method [31].

15N off-resonance rotating-frame relaxation rates ( inline image) were measured as a function of the effective magnetic field amplitude (ωeff) by using a pulse sequence reported previously [32,33]. 15N RF irradiation was applied with an amplitude ω1 and with an offset Δω with respect to the center of the amide nitrogen resonances, as reported previously [33–36]. The ωeff values used varied from 994 to 3500 Hz for the oxidized protein and from 1019 to 3479 Hz for the reduced protein. For each ω1 amplitude value, a series of 2D experiments was performed in which the relaxation delay was set at values in the range 10–300 ms. The acquisition parameters for inline image experiments were the same as those used for R1 and R2 measurements, except that a total of 16 scans and recycle delay of 2 s were used. For the oxidized protein, maps were acquired for a larger number of ωeff values than for the reduced form. This was due to the lower quality spectra, in terms of the signal-to-noise ratio. In fact, extra noise was introduced by the apparatus needed for O2 bubbling into the NMR tube (as described above).

All NMR data were processed using the uxnmr Bruker software. Only the downfield part of the spectra (in the 1H dimension), containing the HN-N connectivities (4.5–12 p.p.m.), was kept for data analysis. All 2D spectra were transformed with 2K × 512 points in the F2 and F1 dimensions, respectively. All spectra acquired with the same ω1 amplitude were processed by using the same processing parameters (phasing parameters, baseline correction, etc.). Subsequent integration of cross peaks for all spectra was performed by using the standard routine of the uxnmr program.

Determination of relaxation rates

R 1, R2, and ROFF relaxation rates were determined by fitting the cross peak volumes, measured as a function of the relaxation delay, to a single exponential decay by using the Levenberg–Marquardt algorithm [37,38]. Uncertainties were evaluated with a Monte Carlo approach [39].

Off-resonance rotating frame relaxation rates are given by:

inline image

[(1)34,36,40], where R1 is the longitudinal relaxation rate of the backbone 15N nucleus of the ith residue, RON,∞ is the on-resonance rotating frame relaxation rate for an infinitely large effective field amplitude (where all exchange contributions are dispersed to zero), K is a constant equal to papbδΩ2, where pa and pb are the populations of the two states a and b between which the exchange process occurs, δΩ is the difference in the chemical shift of the 15N in the two states, and τex is the time constant for the exchange process. The effective field amplitude for the ith spin is ωeff,i = (Δinline image + inline image)1/2 where ω1 is the amplitude of the applied RF irradiation, which is off-resonance with respect to the resonating frequency of the ith spin by an offset Δωi. The angle θi= arctg (ω1/Δωi) is the angle between the effective magnetic field and the static magnetic field axes. It should be noted that Eqn (1) is valid in the limit of fast exchange with respect to the chemical shift separation, i.e. δΩ·τex ≪ 1. As the angle θi is not equal for all spins, it is convenient to rearrange Eqn (1) as follows:

image(2)

The experimental inline image values were fitted as a function of the amplitude of the effective applied spin-lock according to Eqn (2), where K and τex were used as adjustable parameters.

Results and discussion

The protein produced by our expression system has the mass expected for C102T yeast iso-1-cytochrome c with the N-terminal extension of ADL and heme attached in the normal manner. The yield was as much as 20–25 mg·L−1 on rich media and 10 mg·L−1 on minimal medium. The optical spectra of both the oxidized and reduced proteins are identical to that of the native yeast iso-1-cytochrome c (data not shown). 1D 1H NMR spectra and 2D 1H-15N HSQC spectra recorded for both the oxidized and reduced forms of ADL C102T yeast iso-1-cytochrome c were used to monitor the heme environment and the protein folding by comparison with reported data for the native yeast iso-1-cytochrome c[3,6,13,41]. The hyperfine shifted signals from the heme moiety and surrounding residues in the oxidized protein are identical to the corresponding resonances in the native protein. The same applies to the upfield resolved resonances in the reduced protein (which belong to the Met80 axial ligand). 1H-15N HSQC spectra on both the oxidized and the reduced proteins indicate that the overall fold is also conserved. The 3D 1H-15N NOESY-HSQC experiments allowed assignment of the amide resonances of 94 and 97 residues in the reduced and oxidized proteins, respectively. 1H and 15N assignments for the backbone amide in the two oxidation states of the protein are not shown. They agree with previously reported assignments for the native protein obtained at pH 4.6 and 298 K [13,41]. We conclude that the protein obtained with our expression system has properties almost identical to yeast mitochondrial cytochrome c such that it can be used as a model for obtaining structural and dynamic information about this important cytochrome. One interesting point to note about the expression of this protein is the absolute requirement for the presence of 2-mercaptoethanesulfonic acid in cultures grown on minimal media whereas expression in rich media does not require this reagent. This observation is under further investigation but we note that this thiol compound has previously been shown to provide reductant for c-type cytochrome biosynthesis in E. coli[42].

Figure 1 shows the 1H-15N HSQC spectra of oxidized and reduced cytochrome c obtained dissolving the lyophilized protein in D2O solution. The residues whose amide proton resonances do not exchange with the bulk solvent within 30 min (violet) and within 14 h (brown) from the dissolution in D2O of the proteins are indicated in Fig. 2A (oxidized) and Fig. 2B (reduced). The comparison indicates that a larger number of amide protons exchange in the oxidized form. If exchangeability is related to solvent accessibility [9–12], then the NH groups in the oxidized form will be more solvent accessible. As shown in Fig. 2, the structure of cytochrome c is characterized by the presence of five α helices (3–13, 50–55, 61–69, 71–74, 88–101, denoted H1–H5) connected by loops [2]. The two axial iron ligands belong to the first and the last loops. Most of the less accessible residues in both the oxidation states are concentrated in the C-terminal helix (H5). Other residues that do not exchange or exchange slowly are present in helices located above the heme plane, on the side of the axial ligand Met80. These are H3 and H4 for the reduced protein, and H4 for the oxidized protein. In addition, some of the residues that do not exchange in either oxidation state are located in loop 1. Residues 53 and 55 of helix H2 appear to be slightly less accessible to the solvent in the reduced protein. In fact, the entire region surrounding the heme propionate-7 experiences different exchange behavior as a function of the redox state. Val57 and Trp59 exchange more slowly in the reduced protein, although on different time scales. The cross peak of Val57 of the reduced protein is still detected in the HSQC spectrum recorded 30 min after dissolution, whereas it was not detectable in the corresponding spectrum of the oxidized protein: the cross peak of Trp59 in the reduced form is still detectable after 14 h, whereas it is not in the oxidized protein. Trp59 backbone carbonyl and amide are hydrogen bonded to Gly37 NH and Arg38 CO, respectively. The latter two residues also exchange more slowly in the reduced protein, as their resonances are still observed after 30 min.

Figure 1.

1H 15N HSQC spectra region of the oxidized (A) and reduced (B) forms of yeast iso-1-cytochrome c in D2O at the first point in the exchange experiment, i.e. 30 min after dissolution in D2O. The spectra are recorded at 700 MHz and 303 K. The unlabeled cross peaks belong to the extra residues present in this protein with respect to the wild-type. Their assignment is outside the scope of the present manuscript.

Figure 2.

Schematic representation of the structure of cytochrome c, color-labeled according to the exchange behavior of the backbone amide protons for the oxidized (A) and reduced (B) forms. Violet residues are still present 30 min after dissolving the lyophilized protein in D2O. The brown ones are still present after 14 h.

Recently, a hydrogen exchange study was reported for the same protein under different experimental conditions (pH 4.6, 298 K and 500 MHz) [24]. Apparently the increase of pH and temperature affects the exchange properties of the protein and in particular those of the oxidized form. In any case, the oxidized form always possesses an increased solvent accessibility. The same authors also performed R1 and R2 measurements and tried to extract information on mobility at the picosecond level and on the exchange at the millisecond level [13].

Our R1 and R2 values (data not shown), which are necessary for R analysis, compare well with those reported if allowance is made for the different experimental conditions [13]. The off-resonance rotating frame relaxation rates, inline image, of the backbone amide nitrogens of oxidized and reduced cytochrome c were measured for those 15N resonances that do not show overlap and are not too low in intensity and therefore can be accurately integrated. inline image measurements could be performed reliably on 82 15N nuclei in the oxidized protein and 80 in the reduced form. R1 relaxation rates for the amide nitrogens that do not show signal overlap were measured and these values used to calculate inline image by using Eqn (2). Eighteen 15N for the oxidized protein and six 15N for the reduced form displayed a dependence of inline image with ωeff. Consequently, these residues do experience an exchange process in the ms–µs time scale. The approximate correlation times for the exchange processes, τex, have been obtained by fitting the data to Eqn (2). An example of inline image dependence on ωeff and of its fitting to Eqn (2) is shown in Fig. 3. As discussed previously [43], the paramagnetic contribution to the nitrogen relaxation rates is negligible because of the low gyromagnetic ratio of the 15N nucleus. This is particularly true for a low-molecular mass and S = 1/2 low-spin heme iron protein as cytochrome c. The values estimated for τex range between 38 ± 12 µs (Leu94) and 167 ± 61 µs (Asn63) in the oxidized protein (Fig. 4A) and between 49 ± 16 µs (Glu44) and 98 ± 20 µs (Tyr97) in the reduced form (Fig. 4B). From the fitting to Eqn (2), an estimate of the δΩ values can also been obtained; assuming that the two states in the conformational equilibrium are equally populated, δΩ2 results equal to K/0.25. The obtained values never exceeded 0.36 p.p.m. The Rex values can be determined by the product Kτex[22]. A direct comparison between these figures and the Rex given in the previous subnanosecond analysis [13] shows some similarities in the residues found to experience a conformational equilibria, but only a direct measurement of the exchange provide the number of amino acids experiencing mobility and reliable values of the rate constant.

Figure 3.

Off-resonance rotating-frame relaxation ratesinline imageof the amide nitrogen of Lys55 as a function of the effective magnetic field amplitude, τeff, for the oxidized yeast iso-1-cytochrome c. The solid curve represents the fit to a Lorentzian-type function (Eqn 2).

Figure 4.

Exchange correlation times (τex) estimated for the backbone amide nitrogens for residues exhibiting conformational equilibria in both the oxidized (A) and reduced (B) yeast iso-1-cytochrome c.

Figure 4 indicates that residues experiencing conformational equilibria are clustered in selected regions of the protein. In the reduced form, these residues are Glu21 and Glu44, which belong to loop 1, Tyr74 which belongs to the H4α helix, and Tyr97, Leu98, and Lys100 which belong to the C-terminal helix (H5). In the oxidized form, many more residues show conformational equilibria. They are concentrated mainly in the helical regions of the protein. They are residue 10 (H1 helix), residues 32 and 44 (loop 1), residues 51 and 53–55 (H2 helix), residues 63 and 66 (H3 helix), residues 75, 78 and 79 (H4 helix), residues 85 (loop 4), residues 91, 94, 97, 98 and 101 (H5 helix).

Before going into a detailed analysis of the above data, note that exchange of bulk solvent and the conformational equilibria detected for backbone NH could represent two effects of the same phenomenon. Exchangeable amide protons that are involved in hydrogen bonds can exchange with bulk solvent if they become transiently exposed to the solvent upon H-bond breaking [11]. This description is equivalent to assuming that a certain NH has two different conformations (one where it forms an H-bond and it is not solvent accessible, and one where it is solvent accessible). If the chemical exchange between these two conformations occurs on the ms–µs time scale, it could also be detected with T measurements. Therefore, we will discuss our experimental data and try to establish a relationship between the two sets of experimental observations.

In both oxidation states, the amino acids of the C-terminal helix experience conformational exchange processes, with a larger number involved in the oxidized form. It is interesting to note that, as discussed above, the NH backbone residues of the same helix show low solvent accessibility. This may indicate that the exchange processes on the ms–µs time scale involve equilibria between conformations that do not involve variations in exposure of the amide nitrogens to the solvent.

The Glu21 amide in the reduced form has a τex of 73.3 µs whereas in the oxidized form its cross peak is not detectable, possibly as a result of fast exchange with the bulk solvent. The Glu44 amide experiences conformational equilibria in both oxidation states, although the τex value is larger for the oxidized protein. Consistently, the NH exchanges with D2O.

The Asn63 and Glu66 amides experience conformational equilibria only in the oxidized form. The former is connected to Tyr67 through a H-bond between its backbone carbonyl and the backbone NH of residue 67, the latter precedes Tyr67. The aromatic ring of Tyr67 is known to experience conformational equilibria in the case of the oxidized cytochrome c, but not in the reduced one [3,6,44]. (This is equivalent to setting an upper limit for the flipping time in the oxidized protein of the order of 1–2 ms.) No direct information could be gained on the mobility properties of the backbone amide of Tyr67 in the oxidized protein, because it exchanges fast with the bulk solvent and the signal intensity is too weak to be analyzed as a function of ωeff. In the reduced form, the same resonance does not show any exchange process on the ms–µs time scale. Therefore, the conformational equilibrium detected by NMR on the oxidized protein involving the aromatic ring of Tyr67 could affect the backbone of the same residue, of the preceding one, and of the H-bonded Asn63. These data are complemented by the observation provided by X-ray crystallography that higher thermal factors for the backbone of residues 65–72 and for the side chain of Tyr67 are obtained for the oxidized cytochrome c[1], and also by the fact that in 1H-15N HSQC experiments obtained by dissolving the sample in D2O, the NH of Glu66 is more protected in the reduced than in the oxidized cytochrome c(Fig. 2).

More difficult to rationalize is the behavior of Phe10 and Leu32. From the analysis of the structure, the side chains of these two residues and of Tyr97, belonging to the C-term helix and experiencing conformational equilibria, seem to form a patch of hydrophobic interactions in the interior of the protein. It could be that the conformational exchange process of Tyr97 somehow influences the dynamic properties of the NH backbone of the Phe10 and Leu32 residues. However, no experimental observation exists that allowed us to propose a mechanism that could induce such effects.

The most noticeable effects involve the H2 helix and the loop containing Met80. In the H2 helix, τex could be determined in the oxidized protein for four out of six residues forming the helix, while in the reduced form none of these amino acids experiences conformational equilibria. In the loop containing Met80, the amides of Ile75, Thr78, Lys79, Leu85 show conformational equilibria in the oxidized form while in the reduced form only Tyr74 experience a conformational equilibrium. From Fig. 5, where the structure elements containing the residues experiencing exchange processes only in the oxidized protein are highlighted, it is evident that redox-dependent exchange processes are mainly confined to the part of the protein facing the propionates (H2 helix) and above the heme on the site of the axial ligand Met80. It has been found that in both the solution and crystal structures, iron oxidation is accompanied by a change in the conformation of propionate-7 and a re-orientation of the side chain of Asn52 [3–6,45], resulting in the formation of a hydrogen bond between the two moieties. This is particularly evident from the different NOE patterns observed in the reduced and oxidized proteins and from the different exchange behavior of the terminal NH2 of Asn52 [44]. In addition, X-ray data have shown higher thermal factors for the backbone of residues 47–59 upon oxidation [1].

Figure 5.

Yeast iso-1-cytochrome c. Protein regions experiencing conformational equilibria only in the oxidized form are shown in red. The heme and its axial ligands are represented by orange sticks.

Asn52 is located within NOE distances from residues 75 and 78 in the reduced form of the protein. These dipolar interactions are removed as an effect of the change in conformation of Asn52 side chain upon oxidation [3,6]. It is therefore tempting to propose that such movement has also the effect of rendering the Met80 loop more solvent accessible to explain the observed conformational equilibrium in the oxidized protein. Indeed, the present data, as well as previously reported measurements of the exchange properties of the backbone amides [3,6,24], suggest a redox-dependent behavior of the backbone of this loop.

The subnanosecond analysis of the mobility [13] completes the picture of the dynamic properties of cytochrome c. Motions on a time scale at least one order of magnitude slower than the overall protein correlation time, have been postulated for two residues in the reduced protein and 27 residues in the oxidized form, consistent with the existence of extensive conformational equilibria in the iron(III) form of the protein demonstrated here by direct measurements of Rex. However, for the oxidized protein, the residues for which a Rex was proposed do not exactly match those here found to experience conformational equilibria. A general increase in the order parameters for the reduced form relative to the oxidized form has been also reported. A residue by residue analysis of the difference in the order parameter (S2) values between the two redox forms shows that the largest differences are detected for the 72–78 loop. Also the H5 helix contains a stretch of residues having significantly larger order parameter values in the reduced protein. The same holds for two of residues of the H2 helix. If allowance is made for the two time scales, the whole picture is consistent, although the subnanosecond analysis does not show any change in the H3 helix.

Conclusions

The results reported here could provide the key to explaining the decreased stability of oxidized cytochrome c, with respect to the reduced form, towards unfolding conditions such as high pH, high temperature, and denaturants such as guanidinium chloride [7,8,46]. Indeed, in all these cases the oxidized protein unfolds giving an intermediate where the bond with Met80 is broken and another axial ligand binds iron [46–48]. Up to now, no structure has been available for such an intermediate, but several protein amino acids have been proposed as possible ligands (His33, Lys73) [49,50]. The present findings favor Lys73 binding, as they show conformational equilibria in the 75–79 region. Such result is in agreement with the hypothesis that the change in the sixth ligand is the result of a conformational rearrangement probably due to the transcis isomerization of either Pro71 or Pro76 [51]. The possible release of the structure of this loop and the conformational equilibria of the residues forming it in the oxidized protein demonstrated here could constitute the structural/dynamic basis for the decreased stability of ferricytochrome c vs. ferrocytochrome c. On the basis of the residues involved in conformational exchange processes, it is also tempting to propose that, if a proline residues is involved in the process, it has to be Pro76, because there are some residues experiencing conformational equilibria in the oxidized cytochrome clustered around this proline, i.e. residues 75, 78, and 79.

The crystal structure of the complex between cytochrome c and its biological partner cytochrome c peroxidase shows that complex formation is guided by van der Waals interactions, most of which involve residues present in the Met80 loop (i.e. Ala81, Phe82, Gly83, Lys86) [52]. Many authors have proposed that flexibility can represent an important effect on molecular recognition between biological partner molecules. The present findings are in agreement with this proposal, as many residues in the Met80 loop undergo conformational equilibria in the oxidized protein which can be interpreted as an indication of increased flexibility in this redox state. Thus, the increased dynamics of this loop could represent a favorable entropic contribution to protein release after the electron transfer process has occurred.

Cytochrome c has a small but measurable reorganization energy during the electron transfer [7]. Structural analysis in the two redox forms aimed at determining the structural differences, which could be related to the required reorganization energy, have shown that most of the differences between the reduced and oxidized proteins involve the distal site (Tyr67, Met80, and a water molecule involved in H-bonds with these two residues) and heme propionate-7 with surrounding residues. The data presented here reveal a unifying picture that allows us to interpret all the observed differences in terms of changes in the Met80 loop due to a weaker bond between the sulfur of the distal ligand and the oxidized iron [53]. Such differences propagates to the H2 helix (residues 51, 53, 54, 55) through Asn52, which points toward residues 75 and 78 and has van der Waals interactions with Ile75. As a consequence, propionate-7 changes its conformation thus affecting the H-bond network involving its terminal carboxylate, Trp59, Gly37, and Arg38. At the same time the changes in the H-bond network involving Met80, the Tyr67 ring and the distal water molecule influence the H3 helix. A further contribution to the destabilization of some structural elements upon oxidation can arise from the interaction between positive NH dipoles oriented towards the metal ion and the iron(III). Considering that the porphyrin ring has a charge −2, the reduced cytochrome c possesses a total 0 charge at the metal center, whereas the oxidized protein possesses a total +1 charge at the metal center. This change in positive charge of a positively charged protein would increase internal repulsion and render the protein more solvent accessible through the negative H2O dipoles. In particular, NH dipoles oriented with their positive pole towards the iron would experience larger repulsions. Analysis of the structure reveals that this is the case for the backbone amide of Ala81, belonging to the Met80 loop. Moreover, both the positive N-terminus of the H2 α helix and the positive part of the NH dipole of Thr49, preceding this helix, roughly point towards the iron, and could therefore further contribute to the increased mobility of this structural element in the oxidized protein.

A possible explanation for the increased mobility of the H5 helix (residues 91, 94, 97, 98, 101) could be drawn from the analysis of the structural details. The N-terminus of H5 helix, indeed, has van der Waals contacts with H3 helix. These contacts could represent a possible way by which conformational equilibria in H3 helix (residues 63, 66, 67) induce conformational equilibria in the H5 helix.

The above picture is also consistent with the description of the folding process of cytochrome c in terms of cooperative structural units [11,46], characterized by different free energy values for the unfolding. In this description, the 70–85 loop corresponds to the first unfolding stretch, and is followed by the 36–61 region (that contains the H2 helix). The 20–35 loop and the helix H3 unfold later and more or less at the same time. The last cooperative unit contains the two H1 and H5 helices.

Acknowledgements

This work was supported by the European Community (TMR-LSF Contract ERBFMGECT950033), by Italian CNR (Progetto Finalizzato Biotecnologie 99.0509.PF49) and by MURST COFIN99.

Supplementary material

The following material is available from http://www.ejbiochem.com

Table S1. Amide proton and nitrogen resonance assignments for oxidized and reduced yeast iso-1-cytochrome c (ADL-C102T) at 303 K and pH 7.0.

Ancillary