Ferredoxin binding site on ferredoxin: NADP+ ... - Wiley Online Library

3 downloads 4198 Views 1MB Size Report
In a first set of experiments, free FNR and the FNR:Fd complex were reacted with the ... binding domain may contain the binding site for Fd. This follows from ...
Eur. J. Biochem. 216, 57-66 (1993) 0 FEBS 1993

Ferredoxin binding site on ferredoxin :NADP' reductase Differential chemical modification of free and ferredoxin-bound enzyme Ilian JELESAROV', Antonio R. De PASCALIS', Willem H. KOPPENOL2. Masakazu HIRASAWA3, David B. KNAFF' and Hans Rudolf BOSSHARD'

'

Biochemisches Institut der Universitat Zurich, Switzerland

* Department of Chemistry, Louisiana State University, Baton Rouge LA, USA

' Department

of Chemistry and Biochemistry, Texas Tech University, Lubbock TX, USA

(Received April 14, 1993) - EJB 93 0538/2

The chloroplast enzyme ferredoxin :NADP' reductase (FNR) catalyzes the reduction of NADP' by ferredoxin (Fd). FNR and Fd form a 1 : 1 complex that is stabilized by electrostatic interactions between acidic residues of Fd and basic residues of FNR. To localize lysine residues at the Fd binding site of FNR, the FNR:Fd complex (both proteins from spinach) was studied by differential chemical modification. In a first set of experiments, free FNR and the FNR:Fd complex were reacted with the N-hydroxysuccinimidyl ester of biotin. Biotinylated peptides and non-biotinylated peptides were separated on monovalent avidin-Sepharose and purified by high-performance liquid chromatography. Two peptides containing Lysl8 and Lys153, respectively, were less biotinylated in complexed FNR than in free FNR. In a second set of experiments, free and complexed FNR were treated with 4-N,N-dimethylaminoazobenzene-4'-isothiocyano-2'-sulfonicacid (S-DABITC) to obtain coloured lysine-modified FNR. Protection of Lysl53 was again found by modification with SDABITC. In addition, Lys33 and Lys35 were less labelled in the S-DABITC-modified, Fd-bound enzyme. FNR modified in the presence, but not in the absence, of Fd was still able to bind Fd, indicating that the Fd-protected residues are involved in the formation of the Fd: FNR complex. The lysine residues disclosed by differential modification surround the positive end of the molecular dipole moment (558 Debye -1.85 X lopz7Cm) and are located in a domain of strong positive potential on the surface of the FNR molecule. This domain we had proposed to belong to the binding site of FNR for Fd [De Pascalis, A. R., Jelesarov, I., Ackermann, F., Koppenol, W. H., Hirasawa, M., Knaff, D. B. & Bosshard, H. R. (1993) Protein Science 2, 1126-11351. The prediction was based on the complementarity of shape between positive and negative potential domains of FNR and Fd, respectively.

Ferredoxin :NADP' reductase (FNR), an FAD-containing enzyme, catalyzes the terminal step of the chloroplast thylakoid electron transport chain, namely the reduction of NADP' to NADPH by two equivalents of reduced ferredoxin (Fd) (review by Zanetti and Aliverti, 1991). FNR and Fd form an electrostatically stabilized 1 : 1 complex, which has been extensively studied in the past (review by Knaff and Hirasawa, 1991). The FNR:Fd complex studied in vitro may be regarded as a model of the transitory electron transfer complex formed during FNR-catalyzed NADP' reduction in situ. The complex is stable at low ionic strength (K,, 0.1 M. Fd contributes acidic residues and FNR several basic residues to the electrostatic stabilization of the complex (Knaff and Hirasawa, 1991; Chan et ___ Correspondence to H. R. Bosshard, Biochemisches Institut der Universitiit, Winterthurerstrasse '1 90, CH-8057 Zurich, Switzerland F a : +41 1 363 7947. Abbreviations. FNR, ferredoxin :NADP ' oxidoreductase ; Fd, ferredoxin; biotin-ONSuc, d-biotin-N-hydroxysuccinimidyl ester; Fmoc, 9-flurenylmethyloxycarbonyl; Nap,, sodium phosphate buffer; Pth, phenylthiohydantoyl; S-DABITC, 4-N,N-dimethylaminoazobenzene-4'-isothiocyano-2'-sulfonicacid. Enzyme. Ferredoxin:NADP+ oxidoreductase (EC 1.18.1.2).

al., 1983). Electrostatic effects are important to the FNR reaction (Walker et al., 1991). In particular, orientation guided by electrostatics may enhance the rate of reaction (Koppenol et al., 1991; De Pascalis et al., 1993). Several FNR sequences have been published (Knaff and Hirasawa, 1991). The crystal structure of the spinach enzyme has been solved (Karplus et al., 1991). This latter enzyme (Mr 35317) is composed of an FAD-binding domain (residues 19-161) and an NADP'-binding domain (residues 162-314). The FADbinding domain may contain the binding site for Fd. This follows from indirect evidence obtained by chemical modification studies (Medina et al., 1992a,b), by chemical crosslinking of Fd to FNR (Zanetti et al., 1988), by limited proteolysis of FNR (Gadda et al., 1990), and by site-directed mutagenesis experiments (Aliverti et al., 1990). In a previous study, we used differential amidation of carboxyl groups of spinach Fd to map the binding site for FNR on the surface of Fd (De Pascalis et a]., 1993). We found FNR to protect six acidic residues of Fd from chemical modification. The protected residues belong to two domains of negative electrostatic surface potential on either side of the iron-sulfur cluster of Fd. Surface potential calculation revealed complementary domains of positive potential on

58 FNR. These results allowed us to construct a binding model for the Fd:FNR complex in which the negative potential domains of Fd juxtapose to the positive domains of FNR (De Pascalis et al., 1993). In the present study, we present experimental evidence for the proposed Fd-binding site of FNR. Lysine residues of FNR were modified by reaction with lysine-specific reagents in free and in Fd-bound FNR. Fd was found to protect four lysines of which three are located in the larger of the two positive potential domains that we had proposed to contribute to the Fd-binding site of FNR. The fourth protected lysine is located closeby in a peptide segment that is invisible in the FNR crystal structure (Karplus et al., 1991).

Fd. Since Fd also has amino groups reactive with SDABITC, the concentration of S-DABITC was reduced in experiment F in order to keep the reagenuamino group ratio constant in both labelling experiments. In experiment B (binding conditions) 82 nmol FNR was incubated with 98nmol Fd in 440yl 20mM Hepes pH7.5. S-DABITC (2.87 pmol dissolved in 380 pl of the same buffer) was added (final volume 820 pl). In experiment F, 82 nmol FNR was incubated with 2.52 pmol S-DABITC in 20 mM Hepes pH 7.5 (final volume 820 p1). The molar ratio of reagent/&amino groups was 0.9 in both experiments. The reaction was allowed to proceed for 120 min at 25 "C with gentle stirring. Modification was stopped by fast desalting. FNR and Fd from experiment B were separated (FPLC Superdex 75 HiLoad, 0.2 M Nap, pH 7). Proteins were desalted and concentrated into SO mM NH,HCO, pH 7.8. The extent of modificaEXPERIMENTAL PROCEDURES tion was calculated using an absorption coefficient at 465 nm Materials of 28000 M-' cm-' for S-DABITC-modified lysine (Chang, FNR was purified from fresh spinach leaves (Spinacia 1990). olevacea) as described (Shin and Oshino, 1978) and stored at -80°C in 50 mM Tris/HCl pH 8.0. The Fd preparation followed the procedure of Tagawa and Arnon (1968). Tso- Proteolytic digestion form I of Fd was purified (Ohmori et al., 1989) and stored FNR was partially denatured by incubation at 37°C for at -80°C in 0.2 M sodium phosphate buffer (Nap,) pH 7.0. 6 h in 4 M guanidine/HCl, 50 mM Hepes pH 7.5. Denaturant Concentrations of FNR and Fd were determined using ab- and released FAD prosthetic group were removed by gel sorption coefficients of 9680 M-' cm-' at 420 nm for Fd (Tafiltration on a Bio-Gel P6 column (EconoPac lODG, Biogawa and Arnon, 1968) and 10740 M-' cm-' at 460 nm for Rad), equilibrated in SO mM NH,HCO, pH 7.8. After conFNR (Forti et al., 1970). Purified FNR had an absorbance centration in the same buffer (Centricon 10, Amicon), the ratio A,,JA,,,) of 0.12-0.13. The ratio A,,JA,,, for Fd was biotinylated FNR was digested at 37°C for 8 h with Glu0.45-0.46. Intact 35-kDa FNR may be proteolytically de- specific protease from Staphylococcus aureus V8 (ICN). The graded during purification (Shin et al., 1990). The material protease was added in four portions to a final FNWenzyme we used had no detectable truncated form in SDSPAGE. ratio of 20: 1 (by mass). S-DABITC-modified FNR was diN-Hydroxysuccinimidyl esters of biotin (biotin-ONSuc) gested with protease V8 as above, or with trypsin (Miles) for and 2-iminobiotin were from Fluka and Sigma, respectively ; 8 h at 40°C and at a FNR/enzyme ratio of S O : 1 (by mass). d-[8,9-'H]biotin N-hydroxysuccinimidyl ester was from Am- The trypsin preparation had some chymotryptic activity, ersham. 4-N,N-Dimethylaminoazobenzene-4'-isothiocyano- which explained the production of peptide Phe32-Tyr38. 2'-sulfonic acid (S-DABITC) was kindly provided by Dr J.Y. Chang (Ciba-Geigy AG, Basel, Switzerland). Preparation of monovalent avidin-Sepharose Modification of FNR with biotin-ONSuc The swollen CNBr-activated gel (activated Sepharose FNR (77.3 nmol) was incubated at 25°C with 85.7 nmol Fd (isoform I) in 1.8 ml 20 mM Hepes pH 7.5 (experiment B, binding conditions) or in the same buffer plus 0.4 M NaCl (experiment F, non-binding conditions). To both reaction mixtures was added 770 nmol biotin-ONSuc in 15 pl dimethylformamide. In both experiments, the final concentration of reactants was 43 pM FNR, 47 pM Fd and 424 pM biotinONSuc. The molar ratio of biotin-ONSuc to total modifiable amino groups of FNR plus Fd was 0.29. The reaction was allowed to proceed under N, at 25 "C in the dark with stirring and was stopped with a 100-fold molar excess of ethanolamine after 60 min. Proteins were desalted (disposable PD10 column) and separated under non-binding conditions by gel filtration chromatography on a Superdex 75 HiLoad FPLC column (Pharmacia). Peaks containing FNR and Fd were concentrated (Amicon microconcentration unit, model 8010, membrane YM 10 for FNR and YM 5 for Fd).

Modification of FNR with S-DABITC S-DABITC was found to precipitate under the high-ionic-strength reaction conditions necessary to dissociate the Fd:FNR complex in experiment F. For this reason, experiment F had to be conducted at low ionic strength, omitting

4B, Pharmacia, 5 ml packed volume) was washed with 500 ml 1 mM HCl on a sintered glass filter and then suspended in coupling buffer (100 mM NH4HC03,0.5 M NaCl, pH 8.2). Avidin (15 mg, Fluka) in 2 ml coupling buffer was added and the gel transferred to a small glass column. Buffer was added to allow gentle end-over-end rotation of the slurry in the stoppered column. Coupling was allowed to proceed at 4°C overnight. Free avidin was removed by washing with buffer. The gel was rotated for another 2 h at 4°C in the presence of 1 M ethanolamine in 100 mM NH,HCO1 to block unreacted CNBr-activated groups, followed by ample washing with 10 mM Nap, pH 7. Coupling efficiency was >95 %. Non-covalently bound subunits of tetrameric avidin were removed as described (Kohanski and Lane, 1990). Briefly, 3.5 ml avidin-coupled Sepharose packed in a glass column was treated with freshly prepared and filtered solutions of denaturants as follows; 10 ml 6 M guanidine/HCl in 0.2 M KC1, pH 1.5, was pumped through the column at a flow rate of approximately 6 ml/h, followed by 10 mM Nap, pH 7.0 until the effluent was nearly neutral. Second, 6 ml 3 M guanidine thiocyanate in 0.2 M KCl, pH 1.5, was pumped through. Finally, the column was washed with 500 ml 10 mM Nap, pH 7.0 and stored in 0.02% NaN, at 4°C until used.

59

Affinity chromatography on monovalent avidin-Sepharose Monovalent avidin-Sepharose still contains binding sites with very high affinity for biotin and cannot be used directly for affinity chromatography of biotinylated molecules. The ‘unexchangeable’ sites were blocked by passing through the gel 2 mM biotin in 50 mM Hepes pH 7.5. The lower-affinity ‘exchangeable’ binding sites were then regenerated by washing the gel with 10 vol. 0.1 M glycine/HCl, 0.2 M KCl, pH 1.5 (elution buffer) at a flow rate of 1.8 ml/h. The column was equilibrated in washing buffer (50mM Hepes, 0. 2M KCI, pH 7.5) and the sample (digested biotinylated FNR) in 0.5 nil 50 mM NH,HCO, pH 7.8 was loaded. After elution with 10 vol. washing buffer, bound biotinylated peptides were recovered with elution buffer.

Spectroscopic measurements Difference absorbance spectra (Hewlett-Packard 8451 A diode array spectrophotometer) were measured as follows : 3.4 nmol FNR or biotinylated FNR were incubated with 3.8 nmol Fd in 0.5 ml of either 20 mM Hepes pH 7.5 (binding conditions) or 20 mM Hepes, 0.4 M NaCI, pH 7.5 (nonbinding conditions). The absorbance spectra in the range 300- 650 nm were recorded and the difference spectra ‘bound minus free’ calculated. Fluorescence measurements (Perkin Elmer fluorimeter MPF-2A) were performed by scanning the tryptophan emission spectrum over 320-400 nm with the excitation wavelength set at 295 nm. For fluorescence quenching measurements, FNR or modified FNR (1 -3 nmol protein in 3 ml 10 mM Tris/HCI pH 7.5) were thermostatted at 20°C. Aliquots of Fd from a concentrated stock solution (0.050.1 mM in the same buffer) were added and the fluorescence at 342 nm was measured after 2 min. Combined additions of Fd did not increase the total volume by more than 2%.

Calculation of molecular dipole moment The dipole moment of FNR was calculated as described for spinach Fd (De Pascalis et al., 1993). It was assumed that all lysine and arginine residues carry a positive charge, and that all glutamic and aspartic acid residues are negatively charged. An additional negative charge was placed at the peptide oxygen of C-terminal Tyr314. The contribution of peptide bonds in a-helices (131-137, 175-186, 213-222, 255-262, 277-290, 296-305) was taken into account with the approximation of 0.5 positive charge at the N-terminus and 0.5 negative charge at the C-terminus (Koppenol and Margoliash, 1982). All other bond dipoles were assumed to have random orientations and were ignored. The location of the first 18 residues is unknown (Karplus et al., 1991) and the charges in that section could not be taken into account. FAD has a diphosphate group which carries two negative charges. Together, FNR has 44 positive and 48 negative charges.

Other methods Gel filtration binding experiments were performed on a Superdex 75 column (FPLC equipment from Pharmacia) eluted at a flow rate of 0.5 or 0.75 ml/min with 20 mM Hepes pH7.5, with or without 0.4M NaCl. Amino acid analysis was done on an Applied Biosystem Inc. instrument, model

420 H/A. Amino acid sequences were determined on a pulsed-liquid-phase sequencer (ABI model 470A) equipped with an on-line phenylthiohydantoyl (Pth) amino acid analyzer (ABI model 120A).

RESULTS Modification with biotin-ONSuc, outline of experiment Two experiments, designated B and F, were conducted in parallel. Experiment B was performed at 0.02 M ionic strength (binding conditions) and experiment F at 0.42 M ionic strength (non-binding or ‘free’ conditions), all other conditions being equal. The dissociation constant of the FNR:Fd complex at 0.02 M ionic strength is below 0.5 pM (Foust et al., 1969). Thus, keeping concentrations of FNR and Fd above 40 pM in experiment B ensured that virtually all FNR was bound to Fd. In contrast, the complex was fully dissociated at 0.42 M ionic strength. In the first reaction step, lysine &-aminogroups were reacted with biotin-ONSuc. Biotinylated FNR was separated from Fd and purified. Biotinylated FNR from both experiments was cleaved into peptides by proteolytic digestion, and biotinylated and non-biotinylated peptides were separated on monovalent avidin-Sepharose. Biotinylated peptides were then separated by reversed-phase HPLC. Finally, peptides that appeared in lesser amount in the HPLC profile of experiment B were sequenced.

Time course of biotinylation and effect of ionic strength Although it is desirable that biotinylation per se does not alter the conformational integrity of the protein under study, this may happen if many biotin groups are incorporated. On the other hand, too low a degree of biotinylation will make the analysis difficult. In preliminary experiments, the degree of biotinylation was tested by modification with [3H]biotinONSuc. Incorporation of 3-5 mol biotin/mol FNR was found optimal. Higher ratios yielded biotinylated FNR that had a tendency to precipitate and which also had altered spectral properties. The change of ionic strength between experiment B and F must only change the state of association of FNR with Fd, but not the general reactivity of FNR with biotin-ONSuc. This was tested by reacting FNR with biotin-ONSuc in the absence of Fd at 0.02 M and 0.42 M ionic strength, respectively. The time course and the overall degree of biotinylation did not depend on ionic strength within the error of the analysis, which was approximately 2 0.2 mol biotin/mol FNR (Fig. 1A). Thus, any overall difference in the degree of biotinylation in experiments B and F can be ascribed to the ligand state of FNR, i.e. free versus complexed. In preliminary experiments we found that, depending on the molar ratio of biotin-ONSuc to total modifiable amino groups, the overall degree of biotinylation was lower by about 1-2 mol biotidmol FNR in the Fd:FNR complex. Fig. 1B shows an experiment in which we followed the time course of the biotinylation reaction in experiment F and B at a molar ratio of biotin-ONSuc/total modifiable amino groups of 0.2. For the main experiment, a molar ratio of 0.29 reagent/modifiable amino groups was chosen. At this ratio, several experiments reproducibly yielded 5 -5.5 rnol biotidmol FNR in experiment F and 3.2-3.6 mol biotidmol FNR in experiment B after 1 h incubation time. The degrees of overall biotinylation

60 7 ,

I

I

I

(

0 0.3 '4 0.2 0.2

L*i c i n l4 4

0.1

2

0.0

0

0.3

1

II

0.2

4

0.1

2

' 1

00

20

40

60

6

0

0.0

0

20

40

60

FRACTION NUMBER

REACTION TIME (rnin)

Fig. 1. Time course of modification of FNR with biotin-ONSuc. (A) Biotinylation of FNR in the absence of Fd at 0.02M ionic strength (binding conditions, 0) and at 0.42 M ionic strength (nonbinding conditions, 0 ) follows the same time course and results in the same degree of biotinylation. (B) Biotinylation of FNR in the presence of Fd is slower and the final degree of labelling smaller at binding conditions (0)than at non-binding conditions ( 0 )The . time course shown was obtained at a molar ratio of 0.2 for biotin-ONSuc/ total modifiable amino groups. In the main experiment the molar ratio was 0.29, leading to a somewhat larger difference after the 60-min reaction time (V,V).

in the main experiment are indicated in Fig. 1B by open and filled triangles.

Affinity purification of biotinylated peptides Because of the extremely high affinity of avidin for the biotin group, recovery of biotinylated peptides from an avidin-affinity column was very low. We therefore tried modification with 2-iminobiotin, which is known to dissociate from avidin more easily at pH 4 (Orr, 1981). Unfortunately, modification with the N-hydroxysuccinimdyl ester of 2-iminobiotin yielded insoluble material when the degree of modification was more than 5-6 mol iminobiotdmol FNR. Moreover, in our hands dissociation of iminobiotinylated peptides from the avidin-Sepharose column was incomplete at pH 4. Dissociation at pH 2 was effective, but partial dissociation of the non-covalently bound subunits of tetrameric avidin from avidin-Sepharose took place at pH 2. Satisfactory separation of biotinylated from non-biotinylated peptides was finally achieved on monovalent (monomeric) avidin-Sepharose to which binding of the biotin group seems to be somewhat less tight (Kohanski and Lane, 1990). Of the peptides obtained by digestion of FNR with Gluspecific protease from Staph. aweus V8 10-15% were retained by the affinity column, indicating that this percentage of the digest contained one or more biotinyl group/peptide molecule. More than 85% of the total radioactivity was bound (Fig. 2). The small amount of radioactive material that did not bind (first peak in Fig. 2) also did not bind upon rechromatography, indicating that avidin binding sites were not limiting and that the unbound radioactive material was probably a radioactive impurity. Over 95% of the bound radioactivity was re-eluted. The small amount of tightly bound radioactivity may have been composed of undigested

Fig. 2. Separation of biotinylated and non-biotinylated peptides on monovalent avidin-Sepharose. Peptides obtained by digestion with Glu-specific protease from S. aureus V8 of 15 nmol biotinylated FNR from experiment B (A) and experiment F (B) were separated. Loading buffer: 50 mM Hepes, 0.2 M KCl, pH 7.5. Elution buffer (arrow): 0.1 M glycine/HCI, 0.2 M KCl, pH 1.5. Absorption at 220nm ( 0 )and radioactivity in 10-1.11 aliqot of each 0.75-ml fraction (open bars).

biotinylated FNR and very tightly bound large peptides with several biotinyl groups.

Analysis of biotinylated peptides Two major and several minor differences were observed in the HPLC pattern of the biotinylated peptides from experiment B and F (Fig. 3). In experiment B, peptide 1 was missing and peptide 2 was present in much lower amount. Amino acid composition (Table 1) and sequence analysis of peptide 1 from experiment F corresponded to the sequence Xaa-HisSer-Lys-Lys-Met-Glu-Glu. Peptide 2 had the sequence ValLys-Leu-Thr-Gly-Pro-Val-Gly-Xaa-Glu. The two peptides correspond to Lysl8-Glu25 and Va1145-Glul54 of FNR, respectively. To prove that Xaa is biotinylated lysine, the octapeptide Tyr-Gly-Gly-Phe-Leu-Arg-Lys-Tyr (a-neoendorphin) was biotinylated and cleaved with a-chymotrypsin to yield a fragment after HPLC purification that had the sequence Arg-Xaa-Tyr and the amino acid composition Arg+Lys+Tyr after acid hydrolysis. On the Pth analyzer, the Pth-Xaa of this peptide co-eluted with the Pth-Xaa of peptides 1 and 2. Hence, Lysl8 and Lys153 were protected in the FNR :Fd complex.

Modification with S-DABITC Reaction of FNR with S-DABITC yielded a strongly coloured derivative with a broad absorption band in the visible region. The reaction with S-DABITC was quite slow. Only two or three groups were modified after 2 h, using a molar ratio of S-DABITC/amino groups of one. Moreover, SDABITC was found to precipitate at the high-salt concentration necessary to dissociate the FNR: Fd complex (non-binding conditions in experiment F). For this reason, experiment F had to be conducted at low ionic strength, omitting Fd. The concentration of S-DABITC was decreased in experiment F to keep the reagenthimino group ratio constant in both labelling experiments. FNR modified in the presence of Fd con-

61 0.06

0.04

0.05

0.03

0.04

Experiment B

E

5

0

In b.

0 N

0.04

0.02

In 0.03 +

0.01

0.02

0

w

4

:0.00

0.01

N fi 4

2 W

z

2 CY

Q W

W 0

0

5m

0.02

0

m

I Y

0

0

tn m

5

4

8 4

0.00

0.06

0

20

40

I

I

I

Experiment F

E

0.00

60

I

0

20

40

0

20

40

ELUTION TIME (MIN)

Fig. 4. Rechromatography of two peptides from the protease V8 digest of S-DABITC-modifiedFNR. (A) Peptide from experiment B. (B) Peptide from experiment F. Upper trace, 220 nm, lower trace, 546 nm.

tides could not be physically separated. Monitoring the HPLC eluate at 220nm and 546nm indicated several SDABITC-modified peptides despite the fact that the overall degree of modification was small. As a consequence, the RETENTION TIME (min) amount of an S-DABITC-modified peptide was usually small Fig. 3. Separation by HPLC of avidin-purified biotinylated pep- compared to the amount of the corresponding non-modified tides from the protease V8 digest of FNR. Peptide 1, miscing in peptide. In addition, retention times of corresponding modithe pattern of experiment B, and peptide 2, present in reduced fied and non-modified peptides differed in many cases. amount in experiment B, are indicated. Amino acid analyses of these Therefore, separation and identification of S-DABITC-modipeptides are shown in Table 1 . Separation was achieved on a C, reverse-phase column with a linear gradient (0-60%, 70 min) of fied peptides was difficult and a comprehensive analysis of the differences between experiments B and F could not be acetonitrile in aqueous trifluoroacetic acid. accomplished. One modified peptide from the protease V8 digest was Table 1. Amino acid composition of peptides 1 and 2 of the recovered in an amount sufficient for rechromatography and HPLC chromatogram shown in Fig. 3. Values were not corrected sequence analysis. The 546-nm absorbance of this peptide for oxidative loss during hydrolysis, Met was not determined. A'was high in experiment F and low in experiment B (Fig. 4). Biotinyllysine was hydrolyzed to free lysine during acid hydrolysis. The peptide had the sequence Val-Lys-Leu-Thr-Gly-Pro-ValExpected values in parentheses are for FNR peptides Lys'*-His-Ses- Gly-Xaa-Glu, identical to peptide 2 of the biotinylation exLys-Lys-Met-Glu-Gl~~~ (peptide 1) and Val'"-Lys-Leu-Thr-G1y-Pro- periment, i.e. Va1145-Glu154 of FNR (Fig. 3). The Pth-Xaa Val-Gly-Lys-Glu"4 (peptide 2). was identified as S-DABITC-modified Lys by comparison with an S-DABITC-modified control peptide: FmocAmino acid Peptide 1 Peptide 2 Trp-His-Trp-Leu-Gln-Leu-Lys-Pro-Gly-Gln-Pro-Met-Tyr, Se__________~ (expt F) quencing of this peptide after Fmoc-deprotection produced a (expt B) (expt F) Pth-Xaa in cycle 7 that had the same retention time as the Glu 2.4 (2) 1.25 (1) 1.4 (1) Pth-Xaa obtained from peptide 2. 2.5 (2) GlY - (0) 2.07 (2) HPLC separation of the tryptic peptides of S-DABITCHis 0.75 (1) - (0) - (0) modified FNR is shown in Fig. 5. S-DABITC modification Leu - (0) 1.0 (1) 1.0 (1) of lysine residues alters the pattern of tryptic peptides. The LYS 3.0 (3) 1.93 (2) 1.94 (2) 220-nm profiles of experiments B and F were more similar Pro - (0) 0.94 (1) 0.82 (1) (Fig. 5 A, C) than the 546-nm profiles (Fig. 5 B, D). One Ser 0.83 (1) - (0) - (0) peak, which was significantly seduced in experiment B, is Thr - (0) 1.0 (1) 0.85 (1) marked by an arrow in each panel of Fig. 5. The peptide Val - (0) 1.79 (2) 1.55 (2) isolated from this peak had the sequence Phe-Xaa-Pro-XaaThr-Pro-Tyr where Xaa is S-DABITC-modified Lys. The peptide corresponds to FNR sequence Phe32-Tyr38. Iontained 1.3 mol modified amino groups/mol, whereas 2.5 mol/ spray mass spectrometry of the peptide gave a strong signal mol were modified in the FNR derivative from experiment for the (M,+H)'+ ion at 1243.2. This mass peak corresponds to the mass of the peptide Phe32-Tyr38 with a single SF. DABITC group. From this it follows that the peak marked by an arrow in Fig. 5 contained a mixture of both mono-SAnalysis of S-DABITC-modified peptides DABITC derivatives. Hence we conclude that both Lys33 One batch of S-DABITC-modified FNR was digested and Lys35 were less reactive in the FNR:Fd complex. with protease V8, another batch with trypsin. Unlike biotinAnalysis of peptides in areas marked by an asterisk in ylated peptides, S-DABITC-modified and non-modified pep- Fig. 5B and D was also attempted, as these areas exhibited

62 0.2

I

A

F

0.1

I

I

I A,. I

I

I

0.12

B E

0.06 (0

0 N N

t

0

6 + W

0

z m

b4

W

0.0

0.00

g Q

Q

m

w

[L

0

m

0

m m 6

m 4

0.1

0.06

0. 0 .

0.00

ELUTION TIME

(rnin)

Fig. 5. Reverse-phase HPLC purification of tryptic peptides of FNR that had been reacted with S-DABITC in the presence of Fd (A, B ; experiment B) and in the absence of Fd (C, D; experiment F). Peptides in peaks marked by arrows were sequenced and assigned to the amino acid sequence of FNR. Other areas of difference between the chromatograms of experiment B and F are marked by asterisks. Modified peptides from these areas could not be unequivocally assigned to the sequence of FNR. Separation was achieved as described in Fig. 3.

_ _

In contrast, biotinylated FNR from experiment F did not exhibit the characteristic difference absorbance spectrum of the FNR :Fd complex. This loss of binding capacity of FNR from experiment F codld be due to a higher overall number of W 0 z biotinylated lysine residues, 5 -5.5 mol biotin/mol FNR in W DI experiment F versus 3.2-3.6 mol biotidmol FNR in experiw LL ment B. To rule out this possibility, free FNR was biotin'L n ylated to 3.7 mol biotin/mol FNR, which was the same overw 0.01 0 all degree of biotinylation as in experiment B. This derivative z 6 m bound significantly less well than the FNR sample from exDI 0 periment B. It follows that different residues were biotinm 4 ylated in free and in Fd-bound FNR and that the residues protected in the FNR:Fd complex seem to be involved in binding of Fd. This conclusion was further supported by fluorescence 0.00 300 400 500 600 quenching experiments. Binding of Fd to FNR quenches the WAVELENGTH (nrn) intrinsic tryptophan fluorescenc of FNR (Davis, 1990). ConFig. 6. Binding of Fd to FNR detected by difference absorbance stant amounts of native FNR and of biotinylated FNR from spectroscopy. Difference absorbance spectra for binding of Fd to experiments B and F were titrated with Fd and the decrease of the tryptophan fluorescence was followed (Fig. 7). Titranative FNR (-), to FNR biotinylated in the FNR:Fd complex (-----), to FNR biotinylated at non-binding conditions (.....), and to tion of native FNR quenches fluorescence by about 40%. FNR that was biotinylated to the same overall extent as in experiSignificant quenching was observed also when biotinylated ment B but in the absence of Fd (-.-.-). Concentrations were 6.8 pM FNR from experiment B was titrated. Both titration curves FNR and 1.6 pM Fd. saturate at a molar ratio of Fd/FNR of about 1 : 1. In contrast, the fluorescence of biotinylated FNR of experiment F was very significant differences in the 546-nm profiles. Although not quenched by Fd, indicating that this FNR derivative was some peptides were found to match the FNR sequence, we no longer able to bind Fd (open triangles). As a control, FNR biotinylated in the absence of Fd to a degree of 3.7 rnol biocould not unequivocally assign modified lysine residues. tidmol FNR, i.e. as in experiment B, was quenched by only a very small extent on titration with Fd (filled triangles). Binding properties of modified FNR Finally, the FNR :Fd complex was separated from free To support our presumption that the lysine residues sin- FNR and Fd by gel filtration (Table 2). The native FNR:Fd gled out by biotinylation and S-DABITC modification are complex eluted at 7.13 ml, well separated from free FNR located at the Fd binding site, modified FNR from experi- (8.01 ml) and free Fd (7.93 ml). The unusual elution position ments B and F was analyzed for binding to Fd. Binding of of 7.93 ml for native Fd was found only under low-ionicFd to FNR changes the absorbance spectrum of FAD (Foust strength binding conditions. At high-ionic-strength non-bindet al., 1969). The binding difference absorbance spectra of ing conditions, Fd eluted well after FNR. Mixtures of biotinylated FNR from experiment B and native FNR and of FNR that had been modified in the FNR:Fd complex (experiment B) were very similar (Fig. 6). either native Fd or biotinylated Fd from experiment B 0.02

0)

63 I

I

I

w

K

0 J 3

LL

60 -

\

-

A 0.25

0.00

The strong absorption of the diazobenzene group precluded spectroscopic binding studies with S-DABITC-modified FNR. However, in gel filtration S-DABITC-modified FNR from experiment B co-eluted with Fd as a single peak and at the same position as the native FNR: Fd complex, well separated from free FNR and Fd. In contrast, no FNR:Fd complex was observed with S-DABITC-modified FNR from experiment F (not shown).

0.50

Fd CONCENTRATION (pM)

Molecular dipole moment of FNR The distribution of charged residues of FNR is asymmetric, which leads to a large dipole moment of 558 Debye (1.85 X lo-’’ Cm). The dipole moment was calculated based on the published X-ray structure of FNR (Karplus et al., 1991). The dipole vector crosses the surface of FNR near Val151 (positive end) and near Phe231 (negative end) (Fig. 8).

Fig.7. FNR tryptophan fluorescence quenching by Fd. FNR (0.33 pM) was titrated with Fd and the change of fluorescence followed at 342 nm. The following FNR derivatives were titrated: native FNR (0),hiotinylated FNR from experiment B (O), biotinylated FNR from experiment F (O), FNR biotinylated to the same overall degree as in experiment B hut in the absence of Fd (V), control titration with buffer (0). Titration with Fd of native FNR and FNR from experiment B saturates at a ratio of 1 : 1 (crossed DISCUSSION lines). Experimental approach Table 2. Gel chromatography of FNR and Fd and their biotinylated derivatives at low ionic strength (binding conditions). Molar ratios of Fd/FNR are indicated in parentheses. See Experimental Procedures for further details. Sample composition

Elution volume ml

Native FNR Native Fd Native FNR :Fd complex

8.01 7.93 7.13

FNR (expt B) + native Fd (1 : 1) FNR + Fd (both from expt B) (1 : 1) Native FNR + Fd (expt B) (1: 1) Native FNR + Fd (expt F) (1 : 1)

7.08 7.13 7.13 7.09

FNR (expt F) + native Fd (1 : 1) FNR + Fd (both from expt F) (1 : 1)

7.44“ 7.41‘

Mean of duplicate experiments.

yielded single peaks eluting at the same position as the native FNR: Fd complex. In contrast, mixtures of biotinylated FNR from experiment F and either native Fd or Fd isolated from experiment F yielded single peaks at about 7.4 ml. This elution volume is difficult to explain. Clearly, it is larger (smaller molecular size) than that of the native FNR:Fd complex (7.13 ml), but smaller than the elution volumes of 8.01 ml and 7.93 ml for native FNR and Fd, respectively. One interpretation could be that biotinylated FNR was still able to bind Fd weakly and that, during gel filtration, an equilibrium between free and complexed FNR and Fd was established resulting in a broad peak halfway between the peaks of native FNR:Fd complex and individual proteins. We think this is unlikely as the peak at 7.4 ml was as narrow as that of the native FNR:Fd complex at 7.13 ml. It should be remembered that gel filtration was performed under lowionic-strength binding conditions and that low ionic strength may change the elution behaviour of proteins in gel pernieation chromatography.

Differential chemical modification has been employed by us (Rieder and Bosshard, 1980; Bechtold and Bosshard, 1985; Burnens et al., 1987; Bosshard et al., 1987) and by others (Pettigrew, 1978 ; Hitchcock-DeGregori et al., 1985 ; Omen and Kaplan, 1987; Wei et al., 1988; Yadav and Brew, 1990) to map intermolecular contact sites in protein-protein complexes. The method makes a direct comparison of the properties of a complex and its free components. It is possible to obtain information about several residues by a single experiment. One difficulty with this method is that chemical modification per se of one functional group may influence the reactivity of a neighbouring group (Bosshard, 1979). Hence, observed changes in the degree of labelling of a particular side chain may, in principle, reflect events other than binding of a ligand, e.g. a conformational change induced by chemical modification. To overcome this problem, the overall degree of labelling is best kept so low that the number of multiply modified protein molecules is negligible (Kaplan et al., 1971 ; Bosshard, 1979). In the case of the biotin-avidin system, probably for the first time used here to map a binding site by the differential protection technique, the degree of biotinylation had to be greater than one molecule biotidprotein molecule because the amount of FNR available in the experiment was limited. We therefore have taken great care to examine the properties of FNR that was biotinylated in the FNR:Fd complex (experiment B). By three independent methods (difference absorbance spectroscopy, fluorescence quenching, and binding behaviour in gel filtration) it was confirmed that biotinylated FNR from experiment B retained the ability to bind Fd. The observation that the extent of fluorescence quenching was lower with FNR from experiment B than with native FNR is not inconsistent with this conclusion. For both native FNR and FNR from experiment B, the quench curve saturated at one molecule Fd/FNR molecule. Subtle changes in the microenvironment of tryptophan residues may have altered the overall degree of fluorescence quenching in biotinylated FNR. Discrepancies between absorbance and fluorescence data in the FNR:Fd complex have been observed by others (Davis, 1990). A second set of control experiments was conducted with FNR that had been biotinylated to the same degree as FNR from experiment B, but in the absence of Fd. Binding of Fd

64 to this derivative was very poor. Finally, FNR biotinylated to a degree of 5.5 mol biotin/mol from experiment F had completely lost the ability to bind Fd. We may hence conclude that a particular set of lysine residues is important to the interaction with Fd and that these residues were protected in the FNR:Fd complex. The second modification method was introduced by Chang (Chang, 1988, 1990). S-DABITC is specific for aand e-amino groups of proteins. Modification by S-DABITC yielded a derivative of FNR the properties of which could not be as carefully analyzed as those of biotinylated FNR because of the strong ultraviolet and visible absorbance of the diazobenzene group. Still, native and S-DABITC-modified FNR from experiment B, but not from experiment F, exhibited identical binding properties in the gel filtration experiment, strongly indicating that again lysine residues at the intermolecular interface of the FNR:Fd complex were protected from reaction with S-DABITC.

Why did the two reagents not reveal protection of the same residues? Ideally, biotinylation and S-DABITC modification should have revealed the same protected residues. We can see two reasons why, with the exception of Lysl53, different residues were found with the two reagents. First, it was not possible to analyze all the differentially modified peptides of experiment B and F. Close inspection of Fig. 3 reveals additional peaks, apart from peaks 1 and 2, that differed in the two experiments. We could not assign these peaks to the sequence of FNR. Similarly, our analysis failed for several S-DABITC-modified peptides. Second, the differential reactivity of an individual lysine residue with the two reagents need not be the same because the two reagents have quite different structures. In SDABITC, the reactive isothiocyanate group is attached to a large, rigid and flat structure, while the reactive ester of biotin-ONSuc is on a flexible four-carbon chain. Depending on the microenvironment of a lysine residue in the FNR:Fd complex, access of one reagent may have been more sterically hindered than of the other.

Location of protected residues in the crystal structure of FNR Fig. 8 shows the crystal structure of FNR (Karplus et al., 1991). The FAD-binding domain is on the right, the NADP’ binding domain on the left. The redox-active isoalloxazine ring of the FAD prosthetic group is in the cleft between the two domains (arrow). Protected residues Lys33, Lys35, and Lys153 are indicated. Protected Lysl8 belongs to the flexible N-terminal peptide 1-18 which was not visible in the crystal structure since this starts only at residue His19 (Karplus et al., 1991). Judged from the position of Hisl9, Lysl8 may be near to Lys153. The protected lysines are located below the exposed edge of the isoalloxazine ring. Lys33 and Lys3S are to the right, LyslS3 and possibly also Lysl8 to the left of the positive pole of the molecular dipole moment (asterisk). The large dipole moment of 558 Debye will provide electrostatic guidance and facilitate orientation of Fd. In the case of cytochrome c it had been shown that the rate of the electron transfer reaction of singly lysine-substituted derivatives with cytochrome c oxidase and ubiquinol :cytochrome c oxidoreductase closely followed the strength and direction of the molecular dipole moment. This observation was taken to

Fig. 8. Stereo model of C,-trace of FNR. The modul is taken from Karplus et al. (1991) and the Brookhaven Protein Data Base (entry 2FNR). The FAD-binding domain (right) and the NADP ‘-binding domain (left) are separated by a cleft into which the isoalloxazine ring of the FAD prosthetic group protrudes (arrow). Protected residues Lys33, Lys35 and Lys1.53 are indicated. Because protected Lysl8 is in a mobile segment invisible in the crystal structure, His19 is indicated to mark the approximate area where Lysl8 may be located. Also marked are Lys85 and Lys88 found to be cross-linked in a covalent FNR:Fd complex (Zanetti et al., 1988). The calculated molecular dipole moment is shown as a straight line with the positive pole marked by an asterisk. This view of FNR is similar to that shown in Fig. 5B of De Pascalis et al. (1993).

support the conjecture that the dipole moment facilitates and accelerates optimal orientation of cytochrome c’s reaction partners (Koppenol and Margoliash, 1982; Koppenol et a]., 1982, 1991). The complementary electrostatic surface potentials of FNR and Fd further help the molecules to orient their redox prosthetic groups optimally, which may approach to within a few hundred picometers between the dimethylbenzene ring of the isoalloxazine group and the exposed edge of the ironsulfur cluster (De Pascalis et al., 1993). There are three domains of positive surface potential on FNR (see Fig. 5 of De Pascalis et al., 1993, for surface potential distribution; Fig. 8 of this paper and Fig. SB of De Pascalis et al. are similarly oriented). One small domain of positive potential is found around the NADP+-binding site. This domain is removed from the Fd-binding site (Karplus et al., 1991 ; De Pascalis et al., 1993). Another small positive domain is in the NADP’binding half of the molecule, to the left of and at about the same level as Lys153. The third and largest domain of positive potential is in the FAD-binding half below the isoalloxazine ring. This domain contains protected Lys33, Lys35 and LyslS3, in full agreement with our previous proposal that the large positive potential domain contributes to the Fd-binding site. A role for Lys33 and Lys3S in Fd binding also agrees with experiments performed with FNR derivatives shortened at the N-terminus and implicating the segment Lys22 - Phe32 in Fd binding (Gadda et al., 1990; Aliverti et al., 1990). Other basic residues in the large positive potential domain are Lys91 and Arg93. Arg77 of FNR from Anabaena sp. PCC 7119 contributes to Fd binding by this enzyme (Medina et al., 1992a). Arg77 is equivalent to Arg93 in the large positive domain of spinach FNR. Two residues in the small positive domain to the left of the isoalloxazine ring were also proposed possibly to contribute to the Fd-binding site: Lys304 and Lys30.5 (De Pascalis et al., 1993). We did not see protection of either of these residues. In the enzyme from Anabaena, Lys294, equivalent to Lys305 of spinach FNR, seems to be involved in Fd binding (Medina et al., 1992b). Evidence for a role of Lys8S and Lys88 in Fd binding was obtained by carbodiimide-mediated cross-linking of Fd

65 and FNR (Zanetti et al., 1988). Although residues 85 and 88 are not immediately adjacent to residues 33, 35 and 153 (Fig. 8), the cross-linking data do not contradict the present results. To cross-link FNR with Fd, the carbodiimide reagent must have access to carboxyl groups of Fd. Access is hindered if the groups are buried at the intermolecular interface of the FNR:Fd complex. Therefore, the cross-linking data complement the present results. Very recently, the crystal structure of phthalate dioxygenase reductase from Pseudomunus cepacia was solved (Corell et al., 1992). An FMN and a [2Fe-2S] cluster are brought together by the folding of the single polypeptide chain of this enzyme. The overall shape of the FMN domain of phthalate dioxygenase reductase is similar to the structure of spinach FNR; the shape of the enzyme's iron-sulfur cluster domain is similar to cyanobacterial Fd (sequence similarities are less than 20%). The proposed Fd-binding site of spinach FNR (De Pascalis et al., 1993; and this paper) and the area around the FMN prosthetic group to which the iron-sulfur cluster abuts in phthalate dioxygenase reductase are similar, except that no acidic and basic residues seem to contribute to the superposition of the redox prosthetic groups in phthalate dioxygenase reductase. In conclusion, the present results give experimental support to the location of the Fd-binding site on FNR and provide a cogent guideline for future modeling of the complex as well as for site-specific mutations of FNR. We thank Dr J.-Y. Chang (Ciba-Geigy, Basel) for a gift of SDABITS, Wladislaw lvancew for performing the mass-spectrometric analysis of peptides, R. Zinn for help with dipole moment calculations, and P. Andrew Karplus for supplying the coordinates of FNR. This work was supported by the Swiss National Science Foundation (grant 31-30815.91 to H. R. B.), by a grant from the Council for Tobacco Research, USA (to W. H. K.) and by the US National Science Foundation and the US Department of Energy (grants INT8822574 and DE-FG05-90ER20017 to D. B. K.)

REFERENCES Aliverti, A,, Jansen, T., Zanetti, G., Ronchi, S., Herrmann, R. G. & Curti, B. (1990) Expression in Escherichia coli of ferredoxin :NADP * reductase from spinach: bacterial synthesis of the holoflavoprotein and of an active enzyme form lacking the first 28 amino acid residues of the sequence, Eur: J. Biochem. 191, 551-555. Batie, C. J. & Kamin, H. (1984) Ferredoxin:NADP+ oxidoreductase. Equilibria in binary and ternary complexes with NADP and ferredoxin, J. Biol. Chem. 259, 8832-8839. Bechtold, R. & Bosshard, H. R. (1985) Structure of an electron transfer complex. I1: Chemical modification of carboxyl groups of cytochrome c peroxidase in presence and absence of cytochrome c, J. Biol. Chenz. 260, 5191 -5200. Bosshard, H. R. (1979) Mapping contact areas in protein-nucleic acid and protein-protein complexes by differential chemical modification, Methods Biochem. Anal. 25, 273 - 301. Bosshard, H. R., Wynn, R. M. & Knaff, D. B. (1987) Binding site on Rhodospirillum rubium cytochrome c2 for the Rhodospirillum rubrum cytochrome bc, complex, Biochemistr?, 26, 7688-7693. Bumens, A., Demotz, S., Corradin, G., Binz, H. & Bosshard, H. R. (1987) Epitope mapping by differential chemical modification of free and antibody-bound antigen, Science 235, 780-783. Chan, T. M., Ulrich, E. L. & Markley, J. L. (1 983) Nuclear magnetic resonance studies of two-iron-two-sulfur ferredoxins. 4. Interactions with redox partners, Biochemistry 22, 6002-6007. Chang, J.-Y. (1988) A complete quantitative N-terminal analysis method, Anal. Biochem. 170, 542-556. Chang, J.-Y. (1990) The hirudin-binding site of human alpha-thrombin. Identification of lysyl residues which participate in the com-

bining site of hirudin-thrombin complex, J . Biol. Chem. 264, 7141 -7146. Corell, C. C., Batie, C. J., Ballou, D. P. & Ludwig, M. L. (1992) Phthalate dioxygenase reductase: a modular structure for electron transfer from pyridine nucleotides to [2Fe-2S], Science 258, 1604-1610. Davis, D. J. (1990) Tryptophan fluorescence studies of ferredoxin: NADP' reductase indicate the presence of tryptophan at or near the ferredoxin binding site, Arch. Biochem. Biophys. 276, 1-5. De Pascalis, A. R., Jelesarov, I., Ackermann, F., Koppenol, W. H., Hirasawa, M., Knaff, D. B. & Bosshard, H. R. (1993) Binding of ferredoxin to ferredoxin: NADP' oxidoreductase : the role of carboxyl groups, electrostatic surface potential and molecular dipole moment, Protein Sci. 2, 1126-1135. Forti, C., Melandri, B. A,, San Pietro, A. & Ke, B. (1970) Studies on the photoreduction of ferredoxin and the ferredoxin: NADPH reductase flavoprotein by chloroplast fragments : effect of pyrophosphate, Arch. Biochem. Biophys. 140, 107- 112. Foust, F. G., Mayhew, S. G. & Massey, V. (1969) Complex forniation between ferredoxin triphosphopyridine nucleotide reductase and electron transfer proteins, J. Bid. Chem. 244, 964-970. Gadda, G., Aliverti, A,, Ronchi, S. & Zanetti, G. (1990) Structurefunction relationship in spinach ferredoxin :NADP' reductase as studied by limited proteolysis, J. Biol. Chem. 265, 1195511 959. Hitchcock-DeGregori, S. E., Lewis, S. F. & Chou, T. M. (1985) Tropomyosin lysine reactivities and relationship to coiled-coil structure, Biochemistry 24, 3305-3314. Kaplan, H., Stevenson, K. J. & Hartley, B. S. (1971) Competitive labelling, a method for determining the reactivity of individual groups in proteins, Biochem. J. 124, 289-299. Karplus, P. A., Daniels, M. J. & Herriott, J. R. (1991) Atomic structure of ferredoxin:NADP' reductase: prototype for a structurally novel flavoenzyme family, Science 251, 60-66. Knaff, D. B. & Hirasawa, M. (1991) Fen-edoxin-dependent chloroplast enzymes, Biochim. Biophys. Acta 1056, 93 - 125. Kohanski, R. A. & Lane, M. D. (1990) Monovalent avidin affinity columns, Methods Enzymol. 184, 194-200. Koppenol, W. H., Ferguson-Miller, S., Osheroff, N., Speck, S. H. & Margoliash, E. (1982) The relation between the dipole moment of cytochrome c and the activity with cytochrome c reductase and cytochrome c oxidase, in Oxidases and reluted redox systems, Proceedings of the third internationul symposium (King, T. E., Mason, H. S. & Morrison, M., eds) pp. 1037-1053, Pergamon Press, Oxford. Koppenol, W. H., Rush, J. D., Mills, J. D. & Margoliash, E. (1991) The dipole moment of cytochrome c, Mol. Biol. Evol. 8, 545558. Koppenol, W. H. & Margoliash, E. (1982) The asymmetric distribution of charges on the surface of horse cytochrome c. Functional implications, J. Biol. Chem. 257, 4426-4437. Medina, M., Mendez, E. & Gomez-Moreno, C. (1992a) Identification of arginyl residues involved in the binding of ferredoxin: NADP ' reductase from Anabaena sp. PCC 71 19 to its substrates, Arch. Biochem. Biophys. 299, 281 -286. Medina, M., Mendez, E. & G6mez-Moren0, C. (1992b) Lysine residues on ferredoxin:NADP' reductase from Anabaena sp. PCC 7119 involved in substrate binding, FEBS Lett. 298, 25-28. Ohmori, D., Hasumi, H., Yamakura, F., Murakami, M., Fujisawa, K., Taneoka, Y. & Yamamura, T. (1989) Studies on the molecular structure of spinach ferredoxin. I. Comparison of two molecular species of ferredoxin, Biochim. Biophys. Acta 996, 166- 172. Omen, R. P. & Kaplan, H. (1987) Competitive labeling as an approach to defining the binding surface of proteins: binding of monomeric insulin to lipid bilayers, Biochemistry 26, 303 -308. On, G. (1981) The use of the 2-iniinobiotin-avidin interaction for the selective retrieval of labeled plasma membrane components, J. Bid. Chem. 256, 761 -766. Pettigrew, G. W. (1978) Mapping an electron transfer site on cytochrome c, FEBS Lett. 86, 14-16.

Rieder, R. & Bosshard, H. R. (1980) Comparison of the binding sites on cytochrome c for cytochrome c oxidase, cytochrome c reductase, and cytochrome c,, J. Bid. Chem. 255, 4732-4739. Shin, M., Tsujita, M., Tomizawa, H., Sakihama, N., Kamei, K. & Oshino, R. (1990) Proteolytic degradation of ferredoxin: NADP reductase during purification from spinach, Arch. Biochem. Biophys. 279, 97-103. Shin, M. & Oshino, R. (1978) Ferredoxin-Sepharose 4B as a tool for the purification of ferredoxin :NADP ' reductase, J. Biochem. (Tokyo) 83, 457-361. Tagawa, K. & Amon, D. I. (1968) Oxidation-reduction potentials and stoichiometry of electron transfer in ferredoxins, Biochim. Biophys. Acta 153, 602-613. Walker, M. C., Pueyo, J. J., Navarro, J. A,, G6mez-Moreno, C. & Tollin, G. (1991) Laser flash photolysis studies of the kinetics of reduction of ferredoxins and ferredoxin:NADP ' reductases from Anabaena PCC 7119 and spinach: electrostatic effects on intra-

complex electron transfer, Arch. Biochem. Biophys. 287, 351 358. Wei, Q., Jackson, A. E., Pervaiz, S., Caraway, K. L., Lee, E. Y. C., Puett, D. & Brew, K. (1988) Effects of interaction with calcineurin on the reactivities of calmodulin lysines, J. Biol. Chem. 263, 19541 -19544. Yadav, S. & Brew, K. (1990) Identification of a region of UDPga1actose:N-acetylglucosamine P4-galactosyltransferase involved in UDP-galactose binding by differential labeling, J. Biol. Chem. 265, 14163-14169. Zanetti, G., Morelli, D., Ronchi, S., Negri, A,, Aliverti, A. & Curti, B. (1988) Structural studies on the interaction between ferredoxin and ferredoxin :NADP' reductase, Biochemistry 27, 3753-3759. Zanetti, G. & Aliverti, A. (1991) Ferredoxin:NADP' oxidoreductase, in Chemistry and hiochemistly of flavoenzymes (Muller, F. P., ed.) pp. 305-315, CRC Press, Boca Raton FL.