fluorescence resonance energy transfer mapping of the fourth of

THE JOURNAL OF BIOLOGICAL CHEMISTRY 0 1991 by The American Society for Biochemistry and Molecular Biology, Inc.

Vol. 266, No. 26, Issue of September 15, pp. 17276-17285,1991 Printed in U.S. A.

Fluorescence Resonance Energy Transfer Mapping of the Fourth of Six Nucleotide-binding Sites of Chloroplast Coupling Factor 1”

(Received for publication, April 13, 1991)

Adam B. ShapiroS, Kenneth D. Gibsons, Harold A. ScheragaQ, and Richard E. McCartySTl From the $Section of Biochemistry, Molecular and Cell Biology and the §Baker Laboratory of Chemistry, Cornell Uniuersity, Ithaca, New York 14853

Equilibrium dialysis measurements of adenine nucleotide binding to chloroplast coupling factor l suggest that the enzyme has six binding sites for ADP, adenylyl-B,r-imidodiphosphate (AMP-PNP), and 2’(3’)-0-2,4,6-trinitrophenyl-ATP (TNP-ATP). High affinity binding at all six sites requires the divalent cation, Mg2+.

Three of the nucleotide-binding sites, sites 1,2, and 3, have been mapped by fluorescence resonance energy transfer distance measurements (see McCarty, R. E., and Hammes, G. G. (1987) Trends Biochern. Sci. 12, 234-237). Using the same technique, we mapped the location of nucleotide-binding site 4, a tight, exchangeable site (Shapiro, A. B., Huber, A. H., and McCarty, R. E. (1991) J. Biol. Chern. 266, 4194-4200). Two arrangements of the energy transfer map of coupling factor 1 were found which are compatible with the available data. The two arrangements make different predictions about which sites interact in cooperative catalysis.

The chloroplast ATP synthase (H+-ATP synthase, CF1- Fo)’ is a complex, multisubunit enzyme exhibiting catalytic cooperativity and regulatory and catalytic conformational changes (Refs. 1-4). The enzyme transduces a chemiosmotic proton gradient across the thylakoid membrane, generated in the light by the photosynthetic apparatus, into chemical bond energy in the form of the P-y-phosphoanhydride bond of ATP. In the dark, when the proton gradient dissipates, ATP hydrolysis by the ATP synthase is prevented by several regulatory mechanisms. These include inhibitory binding of ADP

* This work was supported in part by Grant DMB 88-03608 (to R. E. M.) from the National Science Foundation; in part by a predoctoral fellowship (to A. B. S.) from the Cornell Biotechnology Program, which is sponsored by the New York State Science and Technology Foundation, a consortium of industries, the U. S. Army Research Office and the National Science Foundation; and in part by a research grant (to H. A. S.) from the National Institute of General Medical Sciences of the National Institutes of Health, U. S. Public Health Service Grant GM-14312, National Science Foundation Grant DMB 84-01811, and grants from Hoffmann-La Roche, Inc, and from the Cornell Biotechnology Program. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked “aduertisement” in accord- ance with 18 U.S.C. Section 1734 solely to indicate this fact.

T To whom correspondence should be addressed Dept. of Biology, The Johns Hopkins University, Baltimore, MD 21218.

’ The abbreviations used are: CF,, chloroplast coupling factor 1; AMP-PNP, adenylyl-p,y-imidodiphosphate; AMP-PNH,, adenylyl phosphoramidate; TNP-ATP (-ADP), 2’(3’)-0-2,4,6-trintrophenyl adenosine 5”triphosphate (-diphosphate); Bicine, N,N-bis(2-hydrox- yethy1)glycine; HPLC, high pressure liquid chromatography; Lucifer Yellow VS, 4-amino-N-[3-(vinylsulfonyl)phenyl]-naphthalimide-3,6- disulfonate; CPM, coumarin 5-phenylmaleimide.

(5, 6), oxidation of a dithiol to a disulfide (7, 8) , and altered binding of an inhibitory subunit (9). Energization of the membrane reverses these processes (10-12).

Cooperativity among at least two catalytic sites of CF1, the catalytic portion of the ATP synthase, is well established (13). Binding of the second adenine nucleotide substrate molecule increases the catalytic rate by several orders of magnitude (14). The number of catalytic sites is unknown but is often assumed to be three. The total number of nucleotide-binding sites is thought to be six (15), as in the homologous Esche- richia coli and mitochondrial enzymes (16). Some of these sites are noncatalytic (17-19).

The number and properties of all the nucleotide-binding sites is essential information for understanding the catalytic mechanism and regulation of the chloroplast ATP synthase. Three of the sites were characterized by Bruist and Hammes (17). Sites 1 (20) and 3 (17) are believed to be catalytic, whereas site 2 is noncatalytic (17). We characterized two more sites (18). Our data suggested that one of these may be catalytic and the other, non-catalytic. Boyer’s laboratory has also examined nucleotide binding to CF1. Their data indicate one to two catalytic and three noncatalytic sites (19). They have also found evidence for a regulatory role of noncatalytic sites (21, 22).

Information about the positions of the nucleotide-binding sites on CF1 could be useful for understanding their interactions. Fluorescence resonance energy transfer distance measurements (23) have yielded a map for nucleotide-binding sites 1, 2, and 3, as well as additional sites not involved in nucleotide binding. In this paper, we report fluorescence energy transfer distance mapping of nucleotide-binding site 4, which may be catalytic. We also present data showing the existence of a total of six nucleotide-binding sites of CF,.

MATERIALS AND METHODS

CF, was prepared as described previously (24). Contaminating ribulose bisphosphate carboxylase/oxygenase was removed by affinity chromatography. The CF, was stored as an ammonium sulfate precipitate at 4 “C.

CF, from which the d and c subunits had been removed (CFl(-d,c)) was prepared as described in Ref. 25. It was stored as an ammonium sulfate precipitate at 4 “C. The d and subunit-containing fractions were pooled and concentrated by ultrafiltration with an Amicon PM- 10 membrane. The concentrate was stored a t -20 “C.

Equilibrium Dialysis-CF, was desalted by passage through two consecutive Sephadex G-50 (fine) centrifuge columns (28) equilibrated with 50 mM Tris-HC1, pH 8, 50 mM NaC1, and 2 mM EDTA (TNE-8 buffer). The CF, was reductively activated by incubation with 50 mM dithiothreitol overnight at 23 “C. The dithiothreitol was removed by passage of the enzyme solution through two consecutive centrifuge columns equilibrated with 50 mM Tris-HC1, pH 8, and 50 mM NaCl (TN-8 buffer). The reduced disulfide bond was then prevented from reoxidizing by reacting the SH groups with 5 mM N- ethylmaleimide for 10 min at 23 “C, followed by passage through two

17276

Mapping Nucleotide-binding Site 4 of CF, 17277

consecutive centrifuge columns equilibrated with TN-8 buffer to remove excess N-ethylmaleimide.

Activated CF, (10 nmol) in TN-8 buffer was placed in one 100-pl half-cell of a Lucite equilibrium dialyzer (Scienceware). The nucleotide to be tested (100 nmol), also in TN-8 buffer, was placed either in the same half-cell or in the opposite half-cell. MgC1, (50 nmol) or EDTA (200 nmol) was placed in the same half-cell as the nucleotide. The two half-cells were separated by a 50,000 molecular weight cutoff dialysis membrane disc. The half-cell ports were closed with Teflon- coated screws. Each half-cell contained 100 p1 of liquid. The chamber was rotated for 10-14 h at 23 "C. The samples were then withdrawn and assayed for nucleotide and protein concentration. Equilibration was complete since the same result was obtained when the nucleotide was placed on the same side or the opposite side of the membrane as the CF,. The concentrations of AMP-PNP, ADP, and AMP-PNHz were determined from the absorbance at 259 nm of the protein-free half-cell, using an extinction coefficient for adenine of 15,400 M-'. The concentrations of free TNP-ATP and TNP-ADP were determined from the absorbance at 408 nm of the solution in the protein- free half-cell, using an extinction coefficient for trinitrophenyl of 26,400 M" (26). Control cells containing no CF, were used to determine the initial nucleotide concentrations. Protein concentrations were determined by the method of Lowry et al. (35), using 1 : l O dilutions of the solutions in the protein-containing half-cells.

The amount of bound nucleotide was calculated as follows. The concentration of nucleotide in the half-cell without CF, was multi- plied by the volume of liquid in the whole cell (200 pl), giving the amount of unbound nucleotide. This amount was subtracted from the initial amount of nucleotide added to the cell, as determined from the nucleotide concentration in protein-free control cells, giving the amount of bound nucleotide. This amount was divided by the amount of CF, in the cell, giving the mole of nucleotide/mole of CF,.

An additional 1.5 mol of nucleotide/mol of CF, was added to this figure when the test nucleotide was ADP, AMP-PNP, or AMP-PNHz for the following reason. The CF, contained an endogenous 1.5 mol of ADP/mol of CF, at the start of the dialysis, which was indistin- guishable from the test nucleotides at 259 nm. Some of this ADP may not have exchanged with the test nucleotide. If it did not exchange with the added nucleotide, it must be included among bound nucleotides and added to the mole of nucleotide/mole of CF,. If the endogenous ADP exchanged with the added nucleotide, then the added nucleotide became bound in its place, but no reduction in the free nucleotide was observed. In both cases, the binding sites originally occupied by the endogenous ADP are occupied at the end of the experiment. In neither case, however, is there a change in the free nucleotide concentration measured at 259 nm due to the occupancy of the 1.5 sites originally containing the endogenous ADP.

When the test nucleotide was TNP-ATP or TNP-ADP, the amount of bound TNP-nucleotide/mol of CF, was determined by the absorbance at 408 nm, at which ADP has no absorbance. It was possible, therefore, to calculate the amount of test nucleotide bound. The amount of unexchanged, endogenous ADP was not determined, but separate experiments (see "Results") showed that about 0.5 mol of ADP/mol of CF, remained bound after incubation with TNP-nucleotides.

Labeling the Lucifer Yellow Site (27)"CF1(-6,e) was desalted by passing it through two consecutive centrifuge columns equilibrated with 50 mM Bicine-NaOH, pH 9, and 3 mM MgCl,. Lucifer Yellow vinyl sulfone (Aldrich) dissolved in water was added to 50 p~ and the mixture incubated for 20 min at 23 "C. The excess label was removed by passing the reaction mixture through two consecutive centrifuge columns equilibrated with TNE-8 buffer. The disulfide bond of the y subunit was reduced by incubating the CF, with 50 mM dithiothreitol for 1 h at 23 "C. The dithiothreitol was removed by passage through two consecutive centrifuge columns equilibrated with TNE-8 buffer. Reoxidation of the disulfide bond was prevented by reaction with 5 mM N-ethylmaleimide for 5 min at 23 "C. Excess N- ethylmaleimide was removed by passage through two consecutive centrifuge columns equilibrated with TNE-8 buffer.

Labeling the y Subunit Disulfide Bond Sulfhydryls with CPM (31)- CFI(-6,t) was desalted by passage through two consecutive centrifuge columns equilibrated with TN-8 buffer and was incubated with 50- 150 p M CuC12 for 1 h at 23 "C to assure that the subunit dithiol/ disulfide was fully oxidized (29). Excess CuClz was removed by passage through two consecutive centrifuge columns equilibrated with 50 mM Tris-HC1, pH 7, 50 mM NaCI, and 2 mM EDTA (TNE-7 buffer). Exposed sulfhydryls were blocked by reacting with 10 mM N- ethylmaleimide for 10 min at 23 "C, followed by passage through two

consecutive centrifuge columns equilibrated with TNE-8 buffer. The disulfide bond was reduced by incubating with 5 mM dithiothreitol for 1 h at 23 "C, followed by passage through two consecutive centrifuge columns equilibrated with TNE-7 buffer. The two sulfhydryls of the disulfide bond (Cys-199 and Cys-205 (32,33)) were reacted with 50 p~ CPM dissolved in N,N-dimethylformamide (less than 1% final volume) for 5 min at 23 "C. N-Acetyl-L-cysteine (2 mM) was then added to eliminate unreacted CPM. After 5 min, the reaction mixture was passed through two consecutive centrifuge columns equilibrated with TNE-8 buffer.

Labeling the y Subunit Cys-322 with CPM (30)-cF1(-6,t) was desalted and oxidized with CuC12 as described above. The "dark" site sulfhydryl (Cys-322 (32, 33)) is the only accessible sulfhydryl residue in oxidized CFl(-6,c). It was labeled with CPM, followed by N-acetyl- L-cysteine addition as described above. The enzyme was then incubated with 50 mM dithiothreitol for 1 h and 5 mM N-ethylmaleimide for 5 min. The centrifuge columns used after each step were the same as above except that the reactions with dithiothreitol and N-ethylmaleimide were followed by passage through centrifuge columns equilibrated with TNE-8 buffer instead of TNE-7 buffer.

Labeling the c Subunit Sulfhydryl (Cys-6 (34)) with CPM and Reconstitution with CFl(-6,e) (9)-The mixture of 6 and c subunits in elution buffer (25) was reacted with 50 p~ CPM dissolved in N,N- dimethylformamide (less then 2% final volume) for 10-12 min at 4 "C. Sodium dodecyl sulfate-polyacrylamide gel electrophoresis showed that the t subunit was labeled and the 6 subunit was unlabeled. The mixture of 6 and t subunits was reconstituted with excess CF1(- 6,c) in TNE-8 buffer. The CF1(-6,t) had previously been reacted with 50 mM dithiothreitol for 1 h at 23 "C followed by 5 mM N-ethylmaleimide for 5 min at 23 "C. Centrifuge columns equilibrated with TNE- 8 buffer were used for each desalting step. The volumes were adjusted so that the final ethanol concentration was 5% (v/v). After 1 h, the reconstitution mixture was passed through two consecutive centrifuge columns equilibrated with TNE-8 buffer.

Labeling with Both Lucifer Yellow and Eosin Maleimide-CF1(-6,c) was labeled with Lucifer Yellow as described above, passed through two consecutive centrifuge columns equilibrated with TNE-8 buffer, reacted with 5 mM N-ethylmaleimide for 5 min at 23 "C, passed through two more centrifuge columns equilibrated with TNE-8 buffer, reacted with 50 mM dithiothreitol for 1 h at 23 "C, and passed through two consecutive centrifuge columns equilibrated with TNE-7 buffer. The protein was then reacted with various concentrations of eosin maleimide (Molecular Probes) dissolved in N,N-dimethylformamide for 5 min at 23 "C and passed through two consecutive centrifuge columns equilibrated with TNE-8 buffer. Control samples were also prepared that were not labeled with either Lucifer Yellow or eosin maleimide.

Incorporation of TNP-ATP into Nucleotide-binding Sites 1 and 4- The various CF, preparations were incubated with TNP-ATP, up to 5 mM, for 2 h at 23 "C in TNE-8 buffer containing an additional 10 mM EDTA. Excess and loosely bound TNP-ATP was removed by passage through two consecutive centrifuge columns equilibrated with TNE-8 buffer.

Labeling Specificity-Specificity of labeling by covalent fluorescent probes (CPM, Lucifer Yellow, and eosin maleimide) was assessed by reversed phase high pressure liquid chromatography of trypsin digested protein. CaCl, (5 mM) and tosylphenylalanyl chloromethyl- ketone-treated trypsin (Sigma) dissolved in 1 mM H2S0, (20 pg/ml) were added to the labeled CFI sample (1 mg/ml). The sample was incubated at 23 "C at least 2 h, heated in boiling water for 30 s, and cooled. A second addition of trypsin was made, and the sample was incubated overnight at 23 "C. After centrifuging at 100,000 X g for 5 min, 100 p1 was injected into an Ultrasphere ODS 4.6 X 250-mm reversed phase column attached to a Beckman 342 gradient chromat- ograph equipped with a model 160 absorbance detector set at 214 nm, and a fluorescence detector attached to a Spectraphysics SP4270 integrator. The flow rate was 2 ml/min. For Lucifer Yellow-labeled CF1(-6,c) the gradient was from 2 to 45% (v/v) acetonitrile in water in 60 min. For CPM-labeled CF1(-6,c) the gradient was from 20-45% acetonitrile in 65 min. For Lucifer Yellow- and eosin maleimide- labeled CFl(-6,t) the 2-45% gradient was used. Trifluoroacetic acid was present at 0.1% (v/v) in both water and acetonitrile.

Subunit specificity of labeling was also checked by examining the fluorescent protein bands on sodium dodecyl sulfate-polyacrylamide gel electrophoresis. Fluorescent bands were visualized by placing the unstained gel on an ultraviolet transilluminator. When CPM was used to label the y subunit, a small amount of labeling occurred on the a and (3 subunits. CPM labeling of the e subunit was completely

17278 Mapping Nucleotide-binding Site 4 of CF1 specific. Lucifer Yellow labeled the /3 and y subunits slightly. Eosin maleimide was completely specific for the y subunit.

ured with a Shimadzu RF-5000U spectrofluorometer. For fluorescence Fluorescence Measurements-Steady-state fluorescence was meas-

resonance energy transfer measurements, the energy transfer efficiency was taken as the quotient of the fluorescence of a sample containing donor and acceptor probes divided by the fluorescence of a sample containing only the donor probe. Samples were diluted sufficiently to make the inner filter effect negligible, usually about 0.5 FM protein. Correction was made for differences in protein concentration between samples. Protein concentration was measured by the method of Lowry et al. (35). The molecular weights of CF,(-b,t) and CF, are 364,000 and 400,000 (36), respectively.

The corrected fluorescence emission spectrum of Lucifer Yellow was measured and integrated with an SLM Aminco SPF-5OOC spectrofluorometer. The absorbance spectrum of eosin maleimide was taken with a Shimadzu UV16OU spectrophotometer. The overlap integral for the Lucifer Yellow-eosin maleimide pair (Fig. 1) was calculated according to Ref. 37. The quantum yield of Lucifer Yellow attached to CF, was taken as 0.36 (27). The TNE-8 buffer had a refractive index of 1.335. The Ro for the Lucifer Yellow eosin maleimide pair was calculated according to (37) to be 52.8 A, assuming a value of 2/3 for K'.

Values of RO for the Lucifer Yellow-TNP-ATP and CPM-TNP- ATP pairs for each donor site were taken from the literature. For Lucifer Yellow and TNP-ATP in site 1, the Ro is 34.8 A (27). For CPM and TNP-ATP in site 1, the R, for the y subunit disulfide bond sulfhydryls is 45.2 ii (31), for Cys-322 of the y subunit is 45.6 A (30), and for the t sulfhydryl is 44.8 A (9). TNP-ATP in site 4 was assumed to give the same Ro values as TNP-ATP in site 1.

Distances between donor and acceptor were calculated from the R, and energy transfer efficiency according to Ref. 37.

Acceptor stoichiometries (TNP-ATP and eosin maleimide) were calculated by subtracting the absorbance of the protein bearing the donor only from the absorbance of the protein bearing both the donor and acceptor at the absorption maximum of the acceptor and cor- recting for differences in protein concentration. The corrected acceptor absorbance was divided by the extinction coefficient for the bound acceptor (til" = 25, 100 M" .cm" for TNP-ATP (38) and ts30 = 96,000 M". cm" (Molecular Probes catalog) for eosin maleimide) to get the acceptor concentration, which was then divided by the protein concentration to get the acceptor/CF, stoichiometry.

Donor/CF1 stoichiometries were not calculated, but conditions were chosen to yield stoichiometries less then 0.4 mol of donor/mol of CF, in order to maintain high labeling specificity.

RESULTS

Equilibrium Dialysis-There are believed to be six nucleotide-binding sites on CFI, as in mitochondrial and bacterial

2 100000

450 500 550

wavelength (nm)

FIG. 1. Spectral overlap of Lucifer Yellow fluorescence emission and eosin maleimide absorbance. -, corrected emission spectrum of Lucifer Yellow VS covalently bound to CFI(-G,c) at Lys-378 of a single a subunit. Excitation was a t 440 nm. Fluorescence is in arbitrary units. - - -, absorbance spectrum of eosin maleimide bound to CF,(-b,c) at the y subunit disulfide bond.

Fls (16). Non-equilibrium binding measurements have re- vealed the presence of four sites which bind nucleotides tightly under appropriate conditions (18, 19). These sites are labeled 1, 2, 4, and 5 (17, 18). Photoaffinity labeling studies showed that each of the three p subunits of CF1 can bind two nucleotides (19). Using the Hummel and Dreyer gel filtration method, Girault et al. (15) found that approximately six nucleotides bound to CF1 under equilibrium conditions. Those experiments were restricted to nucleotide concentrations less than 150 p~ so that the ultraviolet absorbance of the nucleotides did not saturate the detector, and so that a difference due to nucleotide binding to CF1 (typically 0.6 nmol) could be observed. Here we show a similar result from equilibrium dialysis experiments. We used high concentrations of nucleotides and enzyme in order to saturate all available nucleotide- binding sites with dissociation constants less than a few hundred micromolar. We also showed that complete saturation requires a divalent cation. The results are shown in Table 1.

Equilibrium dialysis with ATP was not performed because the ATP would have been rapidly hydrolyzed to ADP. AMP- P N P is a non-hydrolyzable analog of ATP which binds in the same fashion as ATP (18). AMP-PNP (ICN Biochemicals) + endogenous ADP filled nearly six sites when 100 nmol of AMP-PNP was incubated with 10 nmol of CF1 in the presence of 50 nmol of MgC1,. If the MgC12 was replaced with 200 nmol of EDTA, binding was much reduced.

Some spontaneous degradation of AMP-PNP to AMP- PNH, occurs during prolonged incubation at room tempera- ture. This may reduce the amount of bound nucleotide, since AMP-PNH, does not bind well to CF1 (18). We found previously that CF1 does not have any sites which bind tightly to M$+-AMP-PNH,, whereas it has four tight binding sites for M$+-AMP-PNP and two for Mg2"ADP. (A tight binding site is defined here as one which retains hound nucleotide when the CFI is passed through Sephadex G-50 centrifuge columns.) As expected, M$+-AMP-PNH2 fills only 2.1-3.6 sites during equilibrium dialysis (Table 1). We showed previously that Mg2+-AMP-PNH2 does not displace any of the tightly bound, endogenous ADP (18), so only 2.1 sites/CFl were occupied by Mg"-AMP-PNH, under equilibrium dialysis conditions.

CF, has only two tight binding sites for ADP or Mg"-ADP (18). Under equilibrium dialysis conditions, however, CF1 binds 6 mol of M$+-ADP/mol of CF,. In the absence of Mg2+, only 4 mol of ADP bind/mol of CF,. Interestingly, Mg2+ does not influence the number of tight ADP-binding sites on CF1, whereas M$+ increases the number of tight AMP-PNP- and

TABLE 1 Equilibrium dialysis experiments

The calculations are described under "Materials and Methods." Total mole of nucleotide bound/mole of CF, includes an additional 1.5 mole of endogenous ADP/mole of CF, for the experiments with AMP-PNP, ADP, and AMP-PNH, and an additional 0.5 mole of endogenous ADP/mole of CF, for the experiments with TNP-ADP and TNP-ATP added to the mean mole of nucleotide bound/mole of CF,. Values are means f S.D.; n = 4.

Nucleotide mole nucleotide Total mole nucleotide bound/mole CFI bound/mole CF,

Mg'+-AMP-PNP 3.9 ? 0.3 5.4 AMP-PNP 1.0 k 0.3 2.5

M e - A D P 4.9 f 0.4 6.4 ADP 2.5 r 0.2 4.0

M$'-AMP-PNH, 2.1 f 0.5 3.6 M$+-TNP-ATP 7.0 f 0.9 7.5 M e - T N P - A D P 6.3 k 0.1 6.8


ATP-binding sites from two to four (including unexchanged endogenous ADP) (18).

The ATPase activity of the CFI samples in the equilibrium dialysis experiments was not measured. Therefore, the effect of the nucleotide and Mg2+ incubations on the ATPase activity was not determined.

Equilibrium binding of the 2'-(3')-0-2,4,6-trinitrophenyl analogs of ADP and ATP was tested. TNP-ATP is a useful analog for fluorescence resonance energy transfer distance measurements in which it is used as an acceptor. Because its concentration can be measured by its absorbance a t 408 nm, the amount bound to CF, can be determined without inter- ference from endogenous ADP. In the presence of Mg", 6-7 mol of TNP-ATP or TNP-ADP bound/mol of CFI. In addition, up to about 0.5 mol of endogenous ADP/mol of CF1 may remain bound. After 2 h at 23 "C with 3 mM TNP-ATP and 1 mM MgCl', reductively activated CFl retained 0.5 mol of endogenous ADP/mol of CF1 (not shown).

Fluorescence Resonance Energy Transfer Distance Measure- ments-The relative distances between the nucleotide-binding sites could be useful information for understanding their interactions. In particular, nearby sites or sites located on the same 01/p pair may be more likely to interact than distant sites or those located on different CY/@ pairs. Furthermore, the locations of the nucleotide-binding sites with respect to regulatory regions of CF1, such as the y and t subunits, could suggest regulatory mechanisms. Therefore, we undertook fluorescence resonance energy transfer distance mapping of one of the newly characterized tight nucleotide-binding sites, site 4.

Site 4, like site 1, binds ADP and ATP tightly, does not require Mg2+ for tight binding, and exchanges with medium nucleotides. Because of the similarity between the properties of site 4 with those of site 1, and because of our data showing ATP hydrolysis in site 4 in the absence of medium nucleotides (18), we suggested that site 4 may be catalytic. If so, it probably interacts cooperatively with one or more other nucleotide-binding sites.

Incubation of CF, with low concentrations of TNP-ATP, followed by removal of excess and loosely bound nucleotide by gel filtration, results in incorporation of TNP-ATP into site 1. Site 2 will also be filled by TNP-ATP if Mg2+ is included. At higher TNP-ATP concentrations, in the absence of M$+, the TNP-ATP/CF1 stoichiometry approaches 2.0, suggesting that both sites 1 and 4 are filling. Saturation requires about 3 mM TNP-ATP. (This is the nominal concentration. Samples of TNP-ATP usually consist of a mixture of TNP-ATP and TNP-ADP. Both bind to sites 1 and 4.) The concentration of TNP-ATP required for saturation of sites 1 and 4 is about the same as the concentration of ADP, ATP, or AMP-PNP required to saturate tight binding sites (18). Some endogenous ADP does not exchange with TNP-ATP, so complete saturation of sites 1 and 4 with TNP-ATP is not possible. This was also observed with Mg2+-AMP-PNP (18).

TNP-ATP is a useful acceptor probe for fluorescence energy transfer. Although site 4 alone can not be filled with TNP- ATP, distances from various donor sites to site 4 can still be derived from experiments in which TNP-ATP inhabits sites 1 and 4. First, the distance between the donor site and site 1 must be known. Second, site 1 must be saturated with TNP- ATP and the stoichiometry of TNP-ATP filling of site 4 must be known. The first condition is easily met because abundant distance information is available for site 1 (9, 27, 30, 31, 38). T o meet the second condition we incorporated TNP-ATP to several stoichiometries and approached saturation as closely as possible, then extrapolated the energy transfer efficiency

l o j n

FIG. 2. Representative HPLC fluorescence elution profiles of trypsin-digested CFl(-6,e) or CF, labeled at each donor site. A , CFl(-6,c) labeled with CPM at the y subunit disulfide bond sulfhydryls (Cys-199 and Cys-205). B, CFI(-6,t) labeled with CPM a t Cys- 322 of the y subunit. C, CF1(-6,c) labeled with Lucifer Yellow VS a t Lys-378 of a single a subunit. D, CF1(-6,t) reconstituted with 6 subunit and t subunit labeled with CPM at Cys-6. E, CFl(-6,t) labeled with Lucifer Yellow VS at Lys-378 of a single a subunit and eosin maleimide at the y subunit disulfide bond sulfhydryls. Fluorescence is in arbitrary units.

versus TNP-ATP/CF, curve to a stoichiometry of 2.0. I t was not feasible to use other nucleotide-binding sites as

donor sites for energy transfer to site 4 because it was not possible to assure site specificity. The donor sites used were (i) the y subunit disulfide bond sulfhydryls (Cys-199 and Cys- 205), (ii) the y subunit Cys-322 sulfhydryl, (iii) the t subunit sulfhydryl (Cys-6), and (iv) the Lucifer Yellow vinyl sulfone labeling site (Lys-378)' present on only one of the three 01

subunits. For the sulfhydryl sites, CPM was the donor probe. For each site, the Ro, the distance between donor and acceptor at which the energy transfer efficiency is 0.5, has already been reported (9, 27, 30, 31, 38).

For each donor site labeling reaction, specificity of the covalent reaction was tested by high pressure liquid chromatography of complete trypsin digests of the labeled protein. Representative fluorescence elution profiles are shown in Fig. 2. For the y disulfide-labeled CF1, the two peaks represent labeling of the 2 cysteine residues. For the t sulfhydryl-labeled CF1, the double peak probably represents oxidation of the NH2-terminal methionine residue, since the relative sizes of the two peaks changed with the age of the labeled protein (not shown). The CPM probe appears to contain three isomers with slightly different mobilities on HPLC, as judged by the elution profile of CPM-labeled N-acetyl-L-cysteine (not shown). This produces the frequently observed separated or merged triplet peaks.

The energy transfer efficiency versus TNP-ATP/CFI stoichiometry graphs are shown in Fig. 3. It is clear from Fig. 3A that site 1, which is very far from the disulfide bond (31), begins filling with TNP-ATP first, as indicated by the small slope of the curve at low TNP-ATP/CF1 stoichiometry. Site 4 begins filling after site 1 has filled to a stoichiometry of about 0.5 TNP-ATP/CFI. Site 1 is completely filled with TNP-ATP by a stoichiometry of about 1.5 TNP-ATP/CF,, as indicated by the end of the curved portion of the graph and the start of the steep linear portion. Linear extrapolation of the data for TNP-ATP/CFI stoichiometries higher than 1.5 to a stoichiometry of 2.0 yields the energy transfer efficiency when both sites 1 and 4 are saturated with TNP-ATP.

' A. B. Shapiro and R. E. McCarty, unpublished result.

17280 Mapping Nucleotide-binding Site 4 of CFI 55

B 50 A . 45.

. 0- . 40 ~ .

e. 0 . .

L 35

P > '

- I- 25 z

30- . VI . ' 2 - L

+ a . . * C ..

I-

>. Ln . .* - c 20 C

1 5 - . 0 . . 8-

z .* . 10 - . 5 - . .

FIG. 3. Energy transfer effi- '. ..** ciency as a function of TNP-ATP acceptor/CF1(-6,c) stoichiometry. A ,

. unit disulfide bond sulfhydryls (Cys-199 TNP-ATPICF-1

CFI(-6,e) labeled with CPM at the y sub-

and Cys-205). B, CFI(-6,e) labeled with CPM at Cys-322 of the y subunit. C,

cifer Yellow VS a t Lys-378 of a single cy

CFI(-6,e) (0) and CF, labeled with Lu-

0 . 8 0 CPM at Cys-6. with 6 subunit and e subunit labeled with

. subunit (0). D, CF1(-6,c) reconstituted

-5 0 0.5 1 .o 1.5 2.0 0 0.5 1 .O 1.5 2.0

TNP-ATPICF-1

C 0 .

14 - 'D

25 - . . ; 1 0 -

0 0. O bo 20 -

+ VI

m C a

P 6 - + >.

C 0

. - . VI 0'

z 1 5 - : '. 0 > . . .

C : w .I .

8- c 1 0 -

s ..

2 - 0

0 0 0

0 0

5 -

-2 0 0.5 1 0 1.5 2 0 0 5 1 0 1.5 2.0

TNP-ATPICF-1 TNP-ATPICF-1

24 7

:L 0 0.2 0 4 0.6 0.8 1.0

EM/CF-1

FIG. 4. Energy transfer efficiency between Lucifer Yellow VS at Lys-378 of a single a subunit and eosin maleimide at the y subunit disulfide bond sulfhydryls (Cys-199 and Cys- 205) attached to CFI(-G,c) as a function of eosin maleimide acceptor/CF,(-b,e) stoichiometry.

Two slopes are also evident in each of the other curves in Fig. 3. In Fig. 3B, site 1 is closer to the donor probe at Cys- 322 of the y subunit than site 4. The slope at low TNP-ATP/ CF1 stoichiometry is high, whereas the slope at high TNP- ATP/CF, stoichiometry is low. In Fig. 30, the slopes in the two regions are similar, but the slope representing the distance from Cys-6 of the 6 subunit to site 4 is slightly larger than the

slope representing the distance to site 1. Thus site 4 is closer to Cys-6 than is site 1. The only exception to this pattern is the result for the Lucifer Yellow site (Fig. 3C), for which it appears that the majority of site 4 filling occurred between TNP-ATP stoichiometries of 0.5 and 1.0, beyond which additional filling of distant site 1 occurred. This result may be a statistical artifact produced by the difficulty of measuring very low energy transfer efficiencies. Fortunately, the distance measurement is not very sensitive to small errors in the extrapolated energy transfer efficiency.

Fig. 3C shows that there is no difference, within experimen- tal error, in the Lucifer Yellow to TNP-ATP energy transfer efficiencies between CF1 and CF1(-6,c), indicating that removal of the 6 and 6 subunits does not greatly alter the shape of the remainder of the enzyme. Mitra and Hammes (29) showed that removal of the 6 and 6 subunits has no measurable affect on the distances between Cys-322 on the y subunit and disulfide bond and between Cys-322 and nucleotide-binding sites 1, 2, and 3.

The extrapolated energy transfer efficiencies a t a TNP- ATP/CFI stoichiometry of 2.0 from Fig. 3 are 0.54 (Fig. 3A), 0.46 (Fig. 3B) , 0.13 (Fig. 3C), and 0.26 (Fig. 30).

To derive the distances between site 4 and the donor sites when both sites 1 and 4 are filled with acceptor TNP-ATP, the distances between site 1 and the donor sites must be known. The measurements were made previously (9, 27, 30, 31, 38), but in some cases the distances were so great that only a lower limit could be determined. Modeling was done to determine the distances from the available measurements

Mapping Nucleotide-binding Site 4 of CFI 17281

TABLE 2 Intersite distances in CF, measured by fluorescence resonance energy transfer

Measured distances/best-fit distances (A) generated from measured distances by computer algorithm. The five measurements indicated by (*) are described in this paper. The other measurements are taken from (9, 27, 30, 31, 38). N1, N2, N3, and N4 are nucleotide binding sites 1, 2, 3, and 4, respectively. D, L, and SS are the Cys-322 (dark site), Cys-89 (light site), and disulfide bond sulfhydryls of the y subunit, respectively. LY is t.he Lucifer Yellow VS-labeling site of one 01 subunit. t is the sulfhydryl of the I subunit.

N1 N2 N3 D L ss I,Y 6

N2 44/43.3 N3 48/48.4 D

36/36.7 48/49.5

L 43/44.5 47144.7

52/51.6 51151.4 51151.2 SS

t2717.9

LY >74/82.0 >74/84.8 62162.6 41-46/48.1 42-47146.3 >57/63.7 36/35.0 37137.6 30/30.2 38137.6 66*/61.2

I 62/60.9 66164.7 49150.0 26/25.1 -/22.8 23/24.3 N4

42145.9 -/100.6 -/81.7 -161.5 61*/61.9 -/65.4 44.5*/46.9 49*/49.9 58*/54.4

(23). A model for CF1 was generated from all the distance measurements made prior to this paper for the sites within the CF, molecule, using algorithm for least-squares optimi- zation described in the Miniprint. The computations yielded two equally likely solutions, which were nearly identical except for the position of the y subunit disulfide bond. This occurred because that site had an insufficient number of well- defined distances measured between it and other sites.

This problem was approached by making an additional measurement involving the disulfide site. A previous attempt to measure the disulfide bond-Lucifer Yellow site distance had yielded only a lower limit (27) because no donor-acceptor pair was available with a sufficiently large Ro. We solved this problem by using eosin maleimide as an acceptor for Lucifer Yellow fluoTescence. The R,) for this pair was an exceptionally large 52.8 A, thanks to an excellent spectral overlap, a high quantum yield for Lucifer Yellow fluorescence and an ex- tremely high extinction coefficient for eosin maleimide.

The energy transfer efficiency a t an eosin maleimide stoichiometry of 1.0 was calculated to be 0.21 from a linear regression fit to the energy transfer versus eosin maleimide/ CF, stoichiometry plot (Fig. 4). This efficiency corresponds to a distance of 66 A.

After adding the Lucifer Yellow-disulfide bond distance to the other distance measurements, the least-squares optimi- zation yielded a single model. The best-fit distances between the four donor sites and site 1 0(73.0 A for the y subunit disulfide b p d sulfhydryls, 48.6 A for Cys-322 of the y subunit, 63.8 A for the Lucifer Yellow site, and 62.7 A for the t subunit sulfhydryl) were used to calculate the distances between the donor sites and site 4 (Table 2). All of the distance measurements including the new ones involving site 4 were used to generate new least-squares best-fit models. A single model resulted. All measured distances, and the best-fit distances computed from the model, are shown in Table 2. Despite significant uncertainties in the energy transfer measurements, in particular the true value of K’ and the often large scatter in the data, no distance in the final model deviated by more than 7.5% from the measured distance, only six distances deviated by more than 4%, and only one distance bound was violated.

A set of Cartesian coordinates showing the relative positions of the nine energy transfer sites in the final model is presented in Table 3.

DISCUSSION:^ The equilibrium dialysis experiments confirm the results of

Girault et al. (15) that CF, contains six nucleotide-binding

Portions of this paper (including part of “Discussion” and Table

sites. In addition, the experiment shows that the divalent cation MgL’ substantially increases the binding affinities of some of the these sites. Four of the six sites have been characterized already as tight binding sites (18, 19) and a fifth as a loose binding site (17). The sixth site must therefore be another loose binding site which we shall call site 6. The equilibrium dialysis experiments are relevant to the energy transfer mapping because they prove the existence of a sixth nucleotide-binding site having loose nucleotide binding properties like site 3.

Only one concentration of CF, and one of nucleotide was tested in the equilibrium dialysis experiment, so no dissociation constants can be calculated.

Although our data indicate as many as 7.5 binding sites for TNP-nucleotides, we do not think that there are really more than six nucleotide-binding sites on CF,. There may be some nonspecific T N P nucleotide binding to the protein or apparatus due to the hydrophobic nature of the trinitrophenyl moiety.

The three-dimensional coordinates generated by the modeling program (Table 3) were used to make plots of the energy transfer model. The view in Fig. 5A is in the same orientation as the energy transfer model of CFI presented previously (23). Nucleotide binding sites 1, 2, and 3 lie in a plane parallel to the thylakoid membrane surface. Site 4 is much closer to the membrane. The dark (Cys-322), “light” (Cys-89), and disulfide (Cys-199 and Cys-205) site sulfhydryls of the y subunit are 48, 45, and 31 A from the membrane, respectively (31, 39). These data are from energy transfer measurements of the distance of closest approach between these sites and lipid-soluble probes.

The view in Fig. 5B is of the same orientation but seen from above the membrane. This “top” view, scaled appropri- ately, can be used for comparison with the map of CFI obtained by electron microscope single particle image aver- aging (40). Several criteria may be used for attempting to align the energy transfer map with the electron microscope image. (i) The nucleotide-binding sites should be on p subunits or at NIP interfaces (1, 3, 36). (ii) The t sulfhydryl site should be in or near the region of the electron microscope image assigned to the t subunit (40). (iii) The Lucifer Yellow site should be on an N subunit (27). (iv) The sulfhydryls of the y subunit should be in the V-shaped region of the electron microscope image associated with the y subunit (40). (v) All the sites should fall within the boundaries of the electron

3 ) are presented in miniprint at the end of this paper. Miniprint is easily read with the aid of a standard magnifying glass. Full size photocopies are included in the microfilm edition of the Journal that is available from Waverly Press.

17282 Mapping Nucleotide-binding Site 4 of CFl

A ON1 ON2 ON3

membrane I

10 A

FIG. 5 . Two views of the fluorescence energy transfer map of CFI arranged so that nucleotide binding sites 1, 2, and 3 lie in a plane parallel to thylakoid membrane surface. A , “side” view with the membrane surface perpendicular to the plane of the page. B, top view with the membrane surface in the plane of the page superimposed upon the CF, electron microscope average image (40). The positions of the energy transfer sites are marked with closed circles. N l , N2, N3, and N4 are nucleotide-binding sites 1-4, respectively. D, L, and SS are the Cys-322 (dark site), Cys-89 (light site), and disulfide bond sulfhydryls of the y subunit, respectively. LY is the Lucifer Yellow VS-labeling site of one a subunit. c is the sulfiydryl of the e subunit. The open circles represent the putative positions of nucleotide-binding sites 5 and 6.

microscope image. The energy transfer model may be some- what distorted as the result of the accumulated errors of many measurements. Similarly, the electron microscope image may be distorted as the result of negative staining. Furthermore, the view of the CFI molecule seen in the electron microscope image may not be a true top view if the molecule tends to rest on the electron microscope grid in a tilted fashion. Therefore, overlay of the two maps may not be perfect. A maximum diameter of th! CFI molecule in the electron microscope studies of 115 A (40) was used to scale the energy transfer map.

A search was made for additional orientations of the energy transfer map to fit the electron microscope image. Two criteria in addition to the above were applied. First, the y subunit sulfhydryls must be at similar distances from the thylakoid membrane as in the orientation in Fig. 5A to agree with the measured distances of closest approach (31, 39). Note that some flexibility is possible since these distances need not be perpendicular to the plane of the membrane; the protein chains of the ATP synthase may exclude membrane lipid from directly beneath the y sulfhydryls.

Second, the four mapped nucleotide-binding sites must lie in at most two planes parallel to, or nearly parallel to, the plane of the membrane. We assume that the pseudo %fold symmetry axis of CF1 is normally oriented perpendicularly to the thylakoid membrane surface. Then the distance above the membrane of a nucleotide binding site on one @ subunit or at one cy/p interface is the same no matter which b subunit or

interface it is on when the sites are formed from the same amino acid residues. Since each cy/@ pair contains two nucleotide-binding sites (19) consisting of different amino acid residues, the six sites must occupy at most two planes. Assuming no major distortion of the CF1 molecule due to the presence of the single copy subunits, both planes should be parallel to the thylakoid membrane surface.

The orientation of the energy transfer model shown in Fig. 6A fits the above criteria. Sites 2 and 3 lie in one plane parallel to the membrane surface and sites 1 and 4 lie in another. Sites 1 and 4 are closer to the membrane than sites 2 and 3. The y subunit sulfhydryls are at appropriate distances from the membrane. The view from above the membrane is shown in Fig. 6B. This orientation may be superimposed upon the electron microscope image as well as the orientation shown in Fig. 5B, using the several criteria stated above, except that the t sulfhydryl falls in the area identified as containing the b subunit rather than the nearby area containing the c subunit (40). This is not a serious flaw, however, because it is possible that part of the t subunit is in the b subunit area.

For both orientations, when the Cys-322 and Cys-89 of the y subunit are placed on or near the central y subunit density, all four nucleotide-binding sites may be placed on @ subunits or at cy/@ interfaces. The subunit sulfhydryl and y subunit disulfide bond may also be placed in appropriate positions. The Lucifer Yellow site does not fall on the main density of an cy subunit, however. Instead, it is on the edge of an cy

subunit, near the central y subunit density. If the Lys-378 residue that labels with Lucifer Yellow is on a part of the a subunit polypeptide near the central part of the y subunit, it would explain why only 1 Lys-378 of 3 labels with Lucifer Yellow. The nearby y subunit provides the unique environ- ment that lowers the pK, of the t-amino group of this partic-

A @N2,N3

0 N1 0 LY

0 NL

OD

L*E.SS

I

108 membrane

FIG. 6. Two views of the fluorescence energy transfer map of CF, arranged so that nucleotide-binding sites 2 and 3 lie in one plane parallel to the thylakoid membrane surface and sites 1 and 4 lie in another plane parallel to the thylakoid membrane surface. A , side view with the membrane surface perpendicular to the plane of the page. B, top view with the membrane surface in the plane of the page superimposed upon the CF, electron microscope average image (40). Symbols are as in Fig. 5 .

Mapping Nucleotide-binding Site 4 of CF1 17283

ular lysyl residue. Indeed, it would be more difficult to under- stand how single-site labeling by Lucifer Yellow could occur if the site were on the main body of the a subunit, where it would be less likely to have a substantially different environ- ment from one a subunit to the next.

Nucleotide-binding sites lying in the same plane should occupy similar positions on their respective p subunits or a t their respective a l p interfaces because the sites should consist of the same amino acid residues. Presumably, the three p subunits or a l p pairs have similar if not identical tertiary structures. Three nucleotide-binding sites sharing a plane parallel to the membrane plane should also be symmetrically arranged around the center of the CFI molecule at the vertices of an equilateral triangle, although the asymmetry of the CF1 molecule may cause the triangle to be distorted. In Fig. 5B, sites 1, 2, and 3 are arranged in a small triangle near the middle of the CF, molecule. Sites 4, 5, and 6 must then form a large triangle with vertices near the periphery of the molecule. In Fig. 6 B , sites 2 and 3 form the small, central triangle with site 5 or 6, whereas sites 1 and 4 form the larger triangle with site 6 or 5 . Otherwise the two orientations place the nine energy transfer sites similarly within the electron microscope image. Viewed parallel to the plane of the membrane (Figs. 5 A and 6 A ) , however, there are substantial differences. Energy transfer distance measurements between the membrane surface and the nucleotide-binding sites may be able to distin- guish between the two models.

Boekema4 has recently obtained data to suggest that the subunit at the twelve o’clock position in the electron micro- scopic single particle average image is an a subunit. If that is so, then the orientation of the energy transfer map shown in Fig. 6 B is more likely than the one shown in Fig. 5B.

The two orientations of the energy transfer map suggest different interactions between pairs of nucleotide-binding sites. According to the alternating site binding change hy- pothesis of cooperative catalysis by the ATP synthase (41), two or three sites interact cooperatively, alternating nucleotide binding properties sequentially during each catalytic cycle. At any given moment during catalysis, one nucleotide- binding site in a cooperative unit will have tight nucleotide binding properties and another site will have loose binding properties. It has frequently been assumed, because of the 3- fold symmetry of the CF1 molecule and the good fit of three- site models to kinetic data (13, 41), that three sites are involved in a single cooperative unit, with all three sites engaged in catalysis. The pairwise similarity of properties of the six nucleotide-binding sites (18) suggests, however, that two cooperative units may exist, each consisting of a tight binding, exchangeable site (sites 1 and 4) and a loose binding site (sites 3 and 6) . Each unit may also be regulated by a tight binding, nonexchangeable, noncatalytic site (sites 2 and 5) . Noncatalytic sites have recently been shown to regulate the rate of catalysis under certain conditions (21,22). The energy transfer map and electron microscope image may suggest which pairs of sites are most likely to interact. Those which are located close together in space or on the same a l p pair may be more likely to interact cooperatively than those at a large distance.

One difficulty with any orientation of the energy transfer map with respect to the electron microscope image is that, for at least one of the three pairs of similar nucleotide-binding sites, one of the two sites will be in the inner triangle of sites and one will be in the outer triangle. It is reasonable to expect that the nucleotide-binding sites in the two environments consist of different amino acid residues. This difference would

E. J. Boekema, personal communication.

seem to preclude the two sites having identical properties. Note, however, that although the two sites in each pair have similar properties, they are not identical. For example, site 4 requires higher TNP-ATP concentrations than site 1 in order to bind TNP-ATP tightly. There is no a priori reason why two nucleotide-binding sites cannot have similar properties despite being comprised of different portions of the tertiary structure of the protein.

Based on the proximity criterion, the orientation of the energy transfer model shown in Fig. 5 B predicts that sites 3 and 4 form one catalytic pair and sites 1 and 6 form another. The orientation shown in Fig. 6 B predicts that sites 1 and 3 form one catalytic pair and sites 4 and 6 form the other.

If there really are two cooperative catalytic units, each consisting of two alternating catalytic sites and one regulatory noncatalytic site, the question arises as to whether they interact or are independent. Given the highly cooperative nature of interactions within the CF, molecule, interacting catalytic units seem likely. If so, the rate of a communicating conformational change between two cooperative catalytic units, for example, must be included in kinetic analyses of cooperative catalysis by CF1. It would be of interest to see whether the available data (13, 41) would fit such a model as well as a three-site-binding change mechanism.

Using fluorescence resonance energy transfer between Lu- cifer Yellow and TNP-ATP in isolated CF1, we showed previously (24, 42) that site 3 switched its properties from loose binding to tight binding upon binding Mg2’-ATP or Mg’+- AMP-PNP. Simultaneously, another site, distant from the Lucifer Yellow site switched from tight binding to loose binding. We concluded that site 3 was engaged in alternating site catalysis with site 1, as the properties of sites 4, 5, and 6 had not yet been characterized. The energy transfer distance map will make it possible to use a similar technique to test whether site 4 is also involved in alternating site catalysis and determine which pairs of sites interact.

Fluorescence energy transfer distance mapping, along with characterization of the properties of all six nucleotide-binding sites of CF1, continue to reveal important structural and functional features of the chloroplast ATP synthase. This information will eventually contribute to a detailed understanding of the catalytic and regulatory mechanisms of this complex enzyme.

Acknowledgments-We thank Dr. P. A. Karplus for help with exploring orientations of the energy transfer model; Dr. Jane Gibson for advice on equilibrium dialysis; Dr. Eghert Boekema for the use of the electron microscope average image of CF,; Marylee Noden for preparing anti-ribulose hisphosphate carhoxylase/oxygenase antihod- ies; Patricia Soteropoulos for preparing the immunoaffinity column; Dr. Karl-Heinz Suss and Dr. Mark Richter for helpful discussions and advice; Dr. Michael Dann, Patricia Soteropoulos, Carolyn Wetzel, and Xenia Young for additional assistance; and Lysa Nemec for preparing CF, and CFI(-6,c). (The computations were performed on the Cornel1 National Supercomputer Facility, a resource of the Cor- ne11 Theory Center and the IBM Corporation.)

REFERENCES

1. Strotmann, H., and Bickel-Sandkotter, S. (1954) Annu. Reu. Plant Physiol. 35, 97-120

2. Jagendorf, A. T. (1981) in Photosynthesis II: Electron Transport and Photophosphorylation (Akoyonoglou, G., ed) pp. 719-730, Balahan International Science Services, Philadelphia

3. McCarty, R. E., Shapiro, A. B., and Feng, Y. (1988) in Light- Energy Transduction in Photosynthesis: Higher Plant and Bac- terial Models (Stevens, S. E., Jr., and Bryant, D. A., eds) pp. 290-304, American Society of Plant Physiologists, Rockville, MD

4. Gogol, E. P., Johnston, E., Aggeler, R., and Capaldi, R. E. (1990) Proc. Natl. Acad. Sei. U. S. A . 87, 9585-9589

17284 Mapping Nucleotide-binding Site 4 of CFl 5. Bruist, M. F., and Hammes, G. G. (1982) Biochemistry 21,3370-

6. Du, Z., and Boyer, P. D. (1989) Biochemistry 28,873-879 7. Nalin, C. M., and McCarty, R. E. (1984) J. Biol. Chem. 259,

8. Ketcham, S. R., Davenport, J. W., Warncke, K., and McCarty,

9. Richter, M. L., Snyder, B., McCarty, R. E., and Hammes, G. G.

10. Schumann, J., Richter, M. L., and McCarty, R. E. (1985) J. Biol.

11. Strotmann, H., and Bickel-SandkAtter, S. (1977) Biochim. Bio-

12. Richter, M. L., and McCarty, R. E. (1987) J. Biol. Chem. 262,

13. Cross, R. L. (1988) J. Bioenerg. Biomembr. 20, 395-405 14. Andralojc, P. A., and Harris, D. A. (1990) Biochim. Biophys. Acta

15. Girault, G., Berger, G., Galmiche, J.-M., and Andre, F. (1988) J.

16. Senior, A. E. (1988) Physiol. Rev. 68, 177-231 17. Bmist, M. F., and Hammes, G. G. (1981) Biochemistry 20,6298-

18. Shapiro, A. B., Huber, A. H., and McCarty, R. E. (1991) J. Biol.

19. Xue, Z., Zhou, J.-M., Melese, T., Cross, R. L., and Boyer, P. D.

20. Leckband, D., and Hammes, G. G. (1987) Biochemistry 26,2306-

21. Xue, Z., and Boyer, P. D. (1989) Eur. J. Biochem. 179,677-681 22. Guerrero, K. J., Ehler, L. L., and Boyer, P. D. (1990) FEBS Lett.

23. McCarty, R. E., and Hammes, G. G. (1987) Trends Biochem. Sci.

24. Shapiro, A. B., and McCarty, R. E. (1990) J. Biol. Chem. 266,

25. Xiao, J., and McCarty, R. E. (1989) Biochim. Biophys. Acta 976,

26. Hiratsuka, T., and Uchida, K. (1973) Biochim. Biophys. Acta

27. Nalin, C. M., Snyder, B., and McCarty, R. E. (1985) Biochemistry

28. Penefsky, H. S. (1977) J. Biol. Chem. 252, 2891-2899 29. Mitra, B., and Hammes, G. G. (1988) Biochemistry 27, 245-250

3377

7275-7280

R. E. (1984) J. Biol. Chem. 269,7286-7293

(1985) Biochemistry 24, 5755-5763

Chem. 260,11817-11823

p h y ~ . Acta 460, 126-135

15037-15040

1016,55-62

Biol. Chem. 263,14690-14695

6305

Chem. 266,4194-4200

(1987) Biochemistry 26, 3749-3753

2312

270, 187-190

12,234-237

4340-4347

203-209

320,635-647

24,2318-2324

30. Snyder, B., and Hammes, G. G. (1985) Biochemistry 24, 2324-

31. Snyder, B., and Hammes, G. G. (1984) Biochemistry 23, 5787-

32. Miki, J., Maeda, M., Mukohata, Y., and Futai, M. (1988) FEBS

33. Moroney, J. V., Fullmer, C. S., and McCarty, R. E. (1984) J. Biol.

34. Zurawski, G., Bottomley, W., and Whitfield, P. R. (1982) Proc.

35. Lowry, 0. H., Rosebrough, N. J., Farr, A. L., and Randall, R. J.

36. McCarty, R. E., and Moroney, J. V. (1985) in The Enzymes of Biological Membranes (Martonosi, A., ed) 2nd ed., pp. 383-413, Plenum Publishing Co., New York

37. Hammes, G. G. (1981) in Protein-Protein Interactions (Frieden, C., and Nichol, K. W., eds) pp. 257-287, John Wiley & Sons, New York

38. Cerione, R. A., and Hammes, G. G. (1982) Biochemistry 21,745- 752

39. Cerione, R. A., McCarty, R. E., and Hammes, G. G. (1983) Biochemistry 22,769-776

40. Boekema, E. J., Xiao, J., and McCarty, R. E. (1990) Biochim.

41. Boyer, P. D. (1989) FASEB J. 3 , 2164-2178 42. Shapiro, A. B., and McCarty, R. E. (1988) J. Biol. Chem. 263,

43. Braun, W. (1987) Q. Reu. Biophys. 19, 115-157 44. Crippen, G. M., and Havel, T. F. (1988) Distance Geometry and

Molecular Conformation, Research Studies Press, Taunton, United Kingdom

2331

5795

Lett. 232,221-226

Chem. 259, 7281-7285

Natl. Acad. Sci U. S. A. 79, 6260-6264

(1951) J. Biol. Chem. 193, 265-275

Biophys. Acta 1020,49-56

14160-14165

45. Wattrich, K. (1989) Acc. Chem. Res. 22, 36-44 46. Hendrickson, B. A. (1990) Ph. D. thesis, Cornel1 University 47. Hadwiger, M. A,, and Fox, G. E. (1989) J. Biomol. Struct. Dymm.

48. Connolly, M. L. (1983) J. Appl. Cryst. 16,548-558 49. Richmond, T. J. (1984) J. Mol. Biol. 178,63-89 50. Perrot, G., Cheng, B., Gibson, K. D., Vila, J., Palmer, K. A.,

Nayeem, A., Maigret, B., and Scheraga, H. A. (1991) J. Comp. Chem., in press

51. Wako, H., and Scheraga, H. A. (1981) Macromolecules 14, 961- 969

52. Dennis, J. E., Jr., Gay, D. M., and Welsch, R. E. (1981) ACM Trans. Math. Software 7, 348-368

7,749-771

~ l u o r r s c ~ ~ c c R E Y ~ ~ ~ E C Energy Transfer Mapping of the Fourth of Six Nuclrolidc Binding Supplementary Material fox

Sites of Chlomplart Coupling Factor 1

Adam B Shapiro. Kenneth D. Gibson, Harold A. Scheraga, and Richard E. MrCsrty

by

distance measurements approxlmateiy, and then refining this structure using no"-linear The algorithm proceeds ~n two steps. by first buildmg up a structure salirfying the

least squares. The first step. finding an approximate ~tructure. uses a subset Of the ret of measured distances to define a rigid structure (47). A set of four rites, for whlch six

distance measurements are available, is used to set up a base tetrahedron. Right-handed Cartesian coordinates are chosen IO that the first poinl is at the origin, the x-axis pamtr towards the second point, and the third point lies in the x.y-piane IO that ~ t s y-oordinale is

positwe. The fourth point can occupy two positions; we arbitrarlly choose the posltion I*

which it has a parihve r-coardmate. The position of each subsequent pomt that IS added is defined by its distances from three or four of the points whose positions have already been defined. Three distances are charen. as Outlined below, to define two porrible portlions for

between these pontions; alternately. I f upper and lower bounds are known for one or more the new point. I f L fourth distance involving the new point is known, it II used to chwse

dstancer between the new point and the points already placed, they can somet>meS be used to make the choice. When there are only three known distances between the new potnt and those placed already, no choice can be made between the Iwo possible porltions. and the pomt i s raid to be "ambiguous." ti there are N points and n of there ran be classfled as

ambiguous, 2" approximate starting s t r u ~ t u r e ~ can be built in thm manner: however. rnmimization of the least rqvmrr functlon prerenled below, from each of these startlng structwes, need not lead to 2" l ind s t r ~ ~ t ~ r e s .


of the rather high degree of accuracy of distances measured by fluorescence energy hanrfer. Thls build-up procedure for obtaining a starting smcNre is effective mainly because

Let A. 8. and C be three points whose positions have bean settied, and let a, b, and c be the distances of a new point P from A. 8. and C, respectively Then P is one of the two pornts of intersection 01 three spheres centered at A, 8, and C. wtth radii a, b, and c Method) for cornputrig the porilmnr of there points. knowing the positions of the centers of the spheres and their radii, have been presented by several authors (48-50): we have chosen to "re the method in Ref. 8. The distances of the two

possible porihonr for P from a fourth point D whore position has been set are computed and mmpared w t h the measured distance from P to D. If ail the distances between the poinu A. 8, C, D and P were exact. one of these computed distances would be exactly equal to the measured distance; however, because of expenmental error this is highly improbable. In the present study, we have found the following criteria for rhoosng between the two porihonr to be satisfactory: (1) if the two computed dstancer are almost equal (wxthm 0.1 A), P is placed hallway between the two powble positions; (ii) if one

distance IS less than the other and is difference from the measured dirtanre does not exceed 10 0 A. it is "red to define the position of P. If no point D can be found lor which either (i) or (11) is satisfied. the two p i n o n s for P are examined to determine whether one of them satisfies all measured upper and lower boundr for dxstancer that have been ret previously, while the other door not. I1 this test also fails, the point is considered to be ambiguous.

ieart squares function: *cture relinement The approxtmate ~tru~ture is refined by minimization of the

F = T ( l ~ , - ~ i d ~ , ) ~ + T [ ( I ~ - ~ l - u ~ ) 1 + 1 2 + ~ ~ l , - I ~ - ~ I ) I + 1 2 iq k 4 m<n

Here 5 is the mrd ina te vector 01 the it" site, di, is the measured distance between the ith and jth lites, ULI is an upper b u n d for the distance between the kth and Ith sites, and 1," i s a lower bound lor the distance htween the mth and nth sms. The l int term is summed over all p a m of points for which the distances have been measured, the second term IS summed over all pairs of points for which an upper bound to the distance is known, and the third term is summed over all paxrs of points for which a lower bound to the distance is known. The symbol I + indicates that the term in parentheses takes only positive or ~ e r o vaiyes. Far N points, at least IN-10 dsetances must be known if the ItTUctUre is to be determinate (47, 51). Since there are only 3N-6 independent vanabies, the number of residuals exceeds the number of variables and the problem is well-pored. The independent varmbles arc allotted as follows: the first point IS taken to be fired, the second point has a variable x-mrdinate but ftxed y- and r-cwrdinator, and the third paint has

Cartesian coordinates. "ariable x- and y- cwrdinater but a fired r-cwrdinate. AI1 other points have three variable

routine NL2SOL (52). In every ease, a solution was obtained after at most 30 lunclion 1; All least squares mmlmizationr were perlomed using the

evaluations and 24 evaluatioru of the Jacobian.

loiiowrng order: rite NI, site N2. site N3, site e, site D, site LY, ste L, site N4, site 55. There Starting ~ t r u c ~ r e e for CFI were built by determining positions for the rites in the

are enough distance mearurernents betweon the first seven sites to define a vnique approximate starting smcture for these sites. Only three distance meawrements connect site N4 to the first seven rites; hence, the starting position for N4 wa.9 ambiguous. T h e SS site was initially connected by distance me&surements to sites N3, e, and N4 only, and was therefore ambiguour (the uppr and lower bounds far the distances of site SS from other sites laled to define it3 positmn) Thus there were four porrible starting structures. east-squares refmement of these starting strn~hlres ied to two distinn modeis, which had neariy the same residual rum of squares. After the addition of the drrtsnce measurement between we, LY and 5s. as d e w r h d in the text, only two distinct starting slmctures could be built by the method desnibrd above. Least-squarer refinement of either starting structure led to the model presented in the text. A set of mrdinater ir given in Table 3'. It

should be noted that the handedness of the model cannot be determmed from the distances alone; a model ~n which the last SIX sites are reflected in the piane of the NI, N2. and N3 Wes is equally hkely.

The program d e s c i h d here IS avrdable upon request from the second author.

Table 3. Cartesian cwrdlnates for the nine fluorescence energy transfer sites, derived from the distance measurements by least-squarer aptnmltation

%E

NI

N2

N3

D

L

55

LY

E

N4

& x . & 0 0 0.0

29.7 31 5

a . 4 O D

28.9 -10 0

24.7 6.9

53.3 -27.9

51.5 26.1

36.7 -12.5

89.7 6.05

0 0

00

0.0

3.9

448

55.8

26.9

47.0

456

fluorescence resonance energy transfer mapping of the fourth of

Documents