Structure of an Electron Transfer Complex I. COVALENT CROSS-LINKING OF CYTOCHROME PEROXIDASE AND CYTOCHROME c*

Cytochrome c peroxidase and cytochrome c form a noncovalent electron transfer complex in the course of the peroxidase-catalyzed reduction of HzOz. The two hemoproteins were cross-linked in 40% yield to a co- valent 1:l complex with the aid of l-ethyl-3-(3-di-methylaminopropy1)carbodiimide. The covalent com- plex was found to be a valid model of the noncovalent electron transfer complex for the following reasons. 1) The covalent complex 5% residual peroxidase exogeneous ferrocytochrome c indicat- ing that the cross-linked cytochrome c covers the elec-tron-accepting of cytochrome c peroxidase. peroxidase activity site

Electron transfer reactions between hemoproteins are fundamental to many biochemical pathways such as respiratory, photosynthetic, and microsomal electron transfer chains. Recent progress on heme enzyme structure and function indicates that quite often the electron has to overcome large distances between hemes (1,2). Therefore, the question arises as to what extent functional groups of the protein or the polypeptide backbone are involved in the electron transfer process. For a satisfactory answer we need to know more about the intermolecular interface of electron transfer complexes.
There is no x-ray structure available of an electron transfer complex composed of two hemoproteins, but there are several indirect methods to elucidate the intermolecular interface. Closest to an actual crystal structure comes a fitting procedure whereby individual crystal structures are brought together, using a computer-aided graphics display system to optimize noncovalent interactions. In this way, hypothetical structures were proposed of complexes of cytochrome b5 and cytochrome c peroxidase with cytochrome c (1)(2)(3), of cytochrome b5 with methemoglobin (4), and of the nonphysiological complex formed between cytochrome c and flavodoxin (5, 6). These proposals are speculative and in need of experimental support. Some support has come from comparison of primary structures and from chemical modification of amino acid side chains of hemoproteins. The most abundant results pertain to mitochondrial cytochrome c (reviewed in Refs. 7 and 8). Synthetic model compounds may also help to define the contribution by the polypeptide chain to the electron transfer process (9, 10).
We have concentrated on the cytochrome c peroxidasecytochrome c complex for several reasons. First, both primary and three-dimensional structure of the participating hemoproteins are known in detail (7,(30)(31)(32)(33). Second, there is ample information about the kinetics of the peroxidase reaction to guide speculations on structure-function relationships (Refs. 34 and 35; see reviews in Refs. 2 and 36). Third, an attractive hypothetical model of the complex has been proposed. The model is based on the crystal structures of cytochrome c peroxidase and cytochrome c (1,2). Finally, the peroxidase reaction is akin to the reduction of dioxygen to water catalyzed by mitochondrial cytochrome c oxidase (37); thus, the cytochrome c peroxidase-cytochrome c complex is a valid model for the cytochrome oxidase-cytochrome c complex which is less well amenable to the sort of analysis presented here.
In this paper we report our results on cross-linking of the cytochrome c peroxidase-cytochrome c complex by the water-soluble carbodiimide EDC.' We had shown before the formation of a covalent 1:1 complex by treatment of cytochrome c peroxidase and cytochrome c with dithiobis(succinimidylpropionate) or with EDC (16,21). We now show that the EDC cross-linked complex has the features of a virtual electron transfer complex and that cross-linking occurs only to the amino-terminal region of cytochrome c peroxidase where we locate two sites of cross-linking to within a The abbreviations used are: EDC, l-ethyl-3-(3-dimethylamino-propy1)carbodiimide hydrochloride; CNBr fragments, peptides ohtained by cleavage with CNBr; HPLC, high performance liquid chromatography; SDS, sodium dodecyl sulfate; dansyl, 5-dimethylaminonaphthalene-1-sulfonyl. 5 184 few residues. Independently, the recognition site for cytochrome c on cytochrome c peroxidase has also been investigated by differential chemical modification of carboxyl groups (38). Results from the two experimental strategies are complementary and in very good agreement.

RESULTS
Synthesis and Purification of Covalent Complex-We have shown before that cytochrome c is covalently cross-linked to cytochrome c peroxidase by treatment with the water-soluble carbodiimide EDC (21). This cross-linking reagent promotes the formation of amide bonds between side chain amino and carboxyl groups (39)(40)(41). For this reason it is known as "zerolength" cross-linker. The yield of complex formation depends on the type of buffer used, on the ratio of EDC to protein, and on the timepoint of addition of EDC. The complex was obtained in highest yield (about 40%) in 10 mM cacodylate HCI, pH 6.0, using 10 mM EDC and equimolar amounts of cytochrome c and cytochrome c peroxidase. When cytochrome c peroxidase was first treated with EDC and cytochrome c added afterwards, or when using phosphate buffer, the yield was much lower. No covalent complex was formed at 0.2 M ionic strength (data not shown).
Increasing the ratio of cytochrome c to cytochrome c peroxidase had a small effect on the yield of the 1:l complex which always constituted the main reaction product. Only when the concentration of cytochrome c was five times above that of cytochrome c peroxidase, some 2:l complex was being formed (Fig. 1, faint band in lane D with mobility between covalent complex and dimeric cytochrome c peroxidase). Ad- ditional reaction products at higher ratios of cytochrome c to cytochrome c peroxidase were dimers and trimers of cytochrome c and some dimeric cytochrome c peroxidase ( Fig.-l). All of these observations indicate that cross-linking of cytochrome c with cytochrome c peroxidase by EDC is a specific event, consequent to the high-affinity noncovalent interaction of the two hemoproteins (34, 42).
The covalent 1:l complex was purified by gel chromatography (Fig. 2). After two purification steps the complex was run as a single band of M, = 48,000 in SDS-polyacrylamide gel electrophoresis (not shown). Chromatography under nondenaturing conditions at pH 8.8, however, showed a somewhat fuzzy band and some smear (Fig. 3). This could indicate heterogeneity of the number and sites of cross-links.
Proof that the material with M, = 48,000 is a covalent 1:l complex of cytochrome c and cytochrome c peroxidase was given in our previous communication (21). The proof included careful molecular weight analysis by SDS-polyacrylamide gel electrophoresis and use of 3H-methylated cytochrome c in the cross-linking reaction. In addition, we have now performed spectroscopic analysis of the complex. The ratio of heme c to heme b was found to be 1.04, 1.12, and 1.16 in three different preparations, indicating that little if any of the noncovalently bound heme b was lost during purification. The visible absorption spectrum of the oxidized and dithionite-reduced complex very closely resembled the spectrum of a 1:l mixture of cytochrome c and cytochrome c peroxidase (Fig. 4). The amino acid composition of the complex was in very good agreement with that expected for a 1:l complex (Table 11).
Residual Peroxidase Activity and Reducibility by Ascorbate of Covalent Complex-Even though the composition of the covalent complex had been clearly established, it remained to be shown that the complex is a valid model of the noncovalent electron transfer complex. To this end, two different tests were performed measurement of the residual peroxidase activity of the covalent complex toward exogeneous ferrocytochrome c, and reduction by ascorbate of heme c of cytochrome c in the covalent complex. The residual peroxidase activity, measured at close to saturating concentrations of ferrocyto-  Fig. 2). Right lane, native cytochrome c peroxidase. Electrophoresis was performed a t pH 8.8 using a discontinuous Tris glycine-HC1 system (49) without SDS and mercaptoethanol (12% acrylamide in separation gel). The cathode is at the top. chrome c and H202, was around 5% at 10 mM ionic strength and was almost independent of ionic strength (Table I). Since in SDS-polyacrylamide gel electrophoresis the preparation was not visibly contaminated by free cytochrome c peroxidase this residual activity is probably intrinsic to the covalent complex?
Similarly, the rate of reduction of heme c by ascorbate was about 20 times slower in the covalent complex than in free cytochrome c, and. the rate was again independent of ionic strength. In contrast, the reducibility of free cytochrome c is Any contaminating cytochrome c peroxidase must have been EDC modified, a reaction which inhibits peroxidase activity (21).

TABLE I Residual peroxidase activity and reducibility by ascorbate of covalent complex
The rates shown are turnover numbers for the peroxidase activity and first-order rate constants for the reduction of heme c by ascorbate.

Ionic
Peroxidase Reduction by strength activity" ascorbateb  (21), and on the basis of amino acid composition (Table 11). Fragments C1, C2, and C3 were obtained from the covalent complex only and not from free cytochrome c and cytochrome c peroxidase. These fragments must, therefore, contain the covalent cross-links. ionic strength dependent (Table I). The reduction by ascorbate was a simple first-order reaction under all conditions. The reduced heme c was not autooxidizable in the covalent complex.
These data are taken to indicate that in the covalent complex, access to the electron-accepting site of cytochrome c peroxidase is sterically hindered; hence cytochrome c is cross-linked at, or very close to, its proper recognition site on cytochrome c peroxidase (see "Discussion").
Locating Sites of Cross-linking-The covalent complex was cleaved at methionine residues with cyanogen bromide, and the CNBr fragments were separated by gel permeation chromatography (Fig. 5 ) . Three extra CNBr fragments (Cl, C2, C3) were obtained which contained the sites of cross-linking. On SDS-polyacrylamide gel electrophoresis two fragments, C1 and C2, stained with benzidine and, therefore, contained covalently linked heme c (not shown). From the amino acid composition the following assignments were made: fragment C1 is cytochrome c peroxidase 1-163 cross-linked to the whole molecule of cytochrome c; fragment C2 is cytochrome c peroxidase 1-119 with two cross-links to cytochrome c 1-65 and 81-104, respectively; fragment C3 is cytochrome c peroxidase 1-119 cross-linked to cytochrome c 81-104 (Table 11). PEPTIDE X1 CNBr fragments C1 to C3 were digested with trypsin. Tryptic peptides were purified by gel chromatography followed by HPLC. Representative examples of chromatograms are shown in Figs. 6 to 10. Peptides from HPLC were analyzed for amino acid composition and N-terminal residue. Peptides covering 64% of the sequence 1-163 of cytochrome c peroxidase and 65% of the sequence of cytochrome c were recovered from the tryptic digests. Conspicuously missing were peptides of the sequence 22-48 and 73-90 of cytochrome c peroxidase and of the sequence 9-22 and 81-88 of cytochrome c. Hence it was expected that cross-linking occurred somewhere in these sequences. Indeed, two peptides, called X1 and X2, were found which had 2 different N-terminal residues and whose amino acid composition corresponded to the sequence 30-48 of cytochrome c peroxidase and 9-22 and 81-87 of cytochrome c, respectively (Table 111, Fig. 11). Peptide X1 had covalently bound heme c. In accord with the proposed structure of CNBr fragments C1 to C3, peptides X1 and X2 were recovered from the tryptic digest of both fragment C1 and C2, whereas from fragment C3 only peptide X2 was obtained.

~I-F-v-Q-K-c-A-Q-c-H-T-v-E-K~~ r H E H E 7
It was not possible to determine the exact site of crosslinking in the two peptides because, after the many purification steps, the amount of material left over was not sufficient for stepwise Edman degradation.
Most of the peptides which were missing to account for the composition of CNBr fragments C1 to C3 were short and hydrophilic (Fig. 12) and were, therefore, poorly resolved on the reversed-phase column of the HPLC. But there were two exceptions, peptides 1-12 and 76-90 of cytochrome c peroxidase. We must assume that cross-linking to these peptides did occur but escaped our analysis? Actually, cross-linked peptides might have been hidden in a well resolved peak that could not be assigned unequivocally (peak U of Fig. 10). Several small peaks from HPLC could not be identified for lack of material (Figs. 6-10 and data not shown).

Is the Covalent Complex a Valid Model of the Electron
Transfer Complex?-The cross-linking approach used here offers the advantage of a stable derivative submissive to protein-chemical analysis. However, it is most important to demonstrate that the covalent complex is of a structure similar to that of the native noncovalent complex. Ideally, cross-' At least one additional cross-linked tryptic fragment is necessary to account for the proposed structure of CNBr fragment C2 which, based on size and amino acid composition, contains two peptides of cytochrome c cross-linked to cytochrome c peroxidase 1-119. linking should "freeze-in" the transitory electron transfer complex.
There are several observations to indicate that cross-linking by EDC is a specific process. First, the dissociation constant of the noncovalent complex is around 1-5 ~L M under the conditions of cross-linking (34,42). Since equimolar mixtures of the two proteins at a concentration of 126 pM each were present, at most 5% of cytochrome c and cytochrome c peroxidase was in the free form during reaction with EDC. No covalent complex was formed at high ionic strength where the noncovalent complex dissociates and almost none when cytochrome c peroxidase was preactivated with EDC before addition of cytochrome c. Both observations suggest that only the preformed noncovalent complex is being cross-linked by EDC.
Second, the purified covalent complex had about 5% residual peroxidase activity toward exogeneous ferrocytochrome c. This means that free ferrocytochrome c was mostly prevented from entering the electron acceptor site on cytochrome c peroxidase. Two possible explanations can be offered for the 5% residual activity. Either we are dealing with a mixed population of complexes, in a small percentage of which cytochrome c is cross-linked outside of the electron transfer site, or else, cytochrome c is properly positioned at its binding site on cytochrome c peroxidase, yet the molecule is tethered to cytochrome c peroxidase at a site peripheral to that domain which is crucial for electron transfer. In the latter case, exogeneous ferrocytochrome c should be able to donate an electron to cytochrome c peroxidase, albeit at a slower rate, since cytochrome c is expected to be somewhat mobile when cross-linked at a single site. The second explanation of the 5% residual activity could be tested. As the noncovalent interaction between the two proteins is governed largely by ionic bonds, increasing the ionic strength should render the cross-linked cytochrome c molecule more mobile and, hence, increase residual activity toward exogeneous cytochrome c. This has not been observed (Table I). Incidentally, crosslinking by dithiobis(succinimidylpropionate), a 11-A linker, gave a complex with 10% residual activity at low and 40% at high ionic strength (16, 23). Also, the complex with the 11-A linker catalyzed the oxidation of ferrocytochrome c1 by H202, suggesting that the covalently linked cytochrome c was mobile enough to mediate between cytochrome cl, itself no substrate of the peroxidase, and cytochrome c peroxidase (16). The present EDC cross-linked complex did not accept electrons from ferrocytochrome c~.~ Third, the rate of reduction of heme c by ascorbate is much slower in the covalent complex as compared to free cytochrome c, in agreement with a slower rate of reduction of heme c in the noncovalent complex (51). This suggests again that the ascorbate site on cytochrome c is quite inaccessible in the complex. Like its free counterpart, covalently linked cytochrome c is not autooxidizable. Finally, the reaction with ascorbate shows a simple first-order progress curve, suggesting that we are dealing mainly with a single population of complexes with but one orientation of cytochrome c on cytochrome c peroxidase present.
Location of Cross-links and the Intermolecular Interface-A single derivative of defined structure would have been necessary for the unequivocal identification of every cross-link. However, the electrophoretic analysis under nondenaturing conditions indicated that the purified 1:l complex may have contained some molecular species differing in the site and number of cross-links (Fig. 3). Nevertheless, we could expect H. R. Bosshard, unpublished experiment.

Cytochrome c
Peroxidase-Cytochrome c Complex to locate those cross-links which occur most frequently. Our previous and tentative analysis of the covalent complex indicated that the cross-links are located in the regions 1-119 and 172-229 of cytochrome c peroxidase (21). This expectation was based only on the apparent molecular weights, as judged by SDS-polyacrylamide gel electrophoresis, of three new CNBr fragments obtained from the covalent complex. Cross-linking to the sequence 172-229 could not be confirmed by the present more rigorous analysis.
To locate the sites of cross-linking, the three CNBr fragments were digested with trypsin, and tryptic peptides were separated. The aim was to account for all of the sequence of cytochrome c and for positions 1-163 of cytochrome c peroxidase by either cross-linked or noncross-linked peptides. Two cross-linked peptides were found. Their likely structure is shown in Fig. 11. The exact site of the cross-link could not be located, but in cytochrome c the bonds originate from Lys 13 and Lys 86, respectively. For one, EDC forms amide bonds between side chain amino groups of lysines and carboxyl groups of aspartic and glutamic acid (39-41). Second, from earlier studies it is known that Lys 13, Lys 86, and Lys 87 of cytochrome c are most crucial for complex formation with cytochrome c peroxidase (43-45). Third and most important, the bonds at Lys 22 and Lys 87 of cytochrome c were cleaved by trypsin, and, therefore, these two residues cannot have been cross-linked.
The site of cross-linking in cytochrome c peroxidase might be at any one of 5 different residues, as indicated in Fig. 11. However, these are peculiar residues as they are clustered together. Indeed, Yonetani and co-workers, who determined the amino acid sequence of the protein, already predicted that these residues are a possible "site for the interaction with cytochrome c, the protein substrate, which is strongly basic" (32). Inspection of the three-dimensional structure of cytochrome c peroxidase shows that the 5 clustered carboxyl groups are in an exposed location near the area where the heme propionates come close to the molecular surface (2, 31, 46). Ig their hypothetical model, Poulos and Kraut predicted a 2.8-A ionic bond between Asp 37 of cytochrome c peroxidase and Lys 13 of cytochrome c (1, 2). Also, a Lys 13 arylazido derivative of cytochrome c was found to link to some position in the sequence 1-51 (18). Finally, in the noncovalent complex, residues 33, 34, 35, and 37 are being shielded by cytochrome c, as detected by differential chemical modification of carboxyl groups (38). Therefore, the acidic region 32-37 must be of paramount importance in the electron transfer complex.
We were not able to detect additional cross-linked peptides besides X1 and X2. As discussed above, the existence of such peptides is to be expected because the complex may have been heterogenous and because not the entire sequence of cytochrome c and cytochrome c peroxidase 1-163 was accounted for by the tryptic peptides identified.4 The present information obtained from the cross-linking studies is complemented by our alternative approach of differential chemical modification of carboxyl groups of cytochrome c peroxidase. The results are reported in the following paper (38). . Simondsen, R. P., Weber, P. C.. Salemme, F. R., and Tollin, G. (1982) . Poulos

O A P G P T Y T O A B K B K G I T W K E E T L M E Y L E N~K K Y I P G T K U~
...