Cross-linking of engineered subunit delta to (alphabeta)3 in chloroplast F-ATPase.

Ser → Cys mutations were introduced into subunit δ of spinach chloroplast F0F1-ATPase (CF0CF1) by site-directed mutagenesis. The engineered δ subunits were overexpressed in Escherichia coli, purified, and reassembled with spinach chloroplast F1-ATPase (CF1) lacking the δ subunit (CF1(−δ)). By modification with eosin-5-maleimide, it was shown that residues 10, 57, 82, 160, and 166 were solvent-accessible in isolated CF1 and all but residue 166 also in membrane-bound CF0CF1. Modification of the engineered δ subunit with photolabile cross-linkers, binding of δ to CF1(−δ), and photolysis yielded the same SDS gel pattern of cross-link products in the presence or absence of ADP, phosphate, and ATP and both in soluble CF1 and in CF0CF1. By chemical hydrolysis of cross-linked CF1, it was shown that δS10C was cross-linked within the N-terminal 62 residues of subunit β. δS57C, δS82C, and δS166C were cross-linked within the N-terminal 192 residues of subunit α. Cross-linking affected neither ATP hydrolysis by soluble CF1 nor its ability to reassemble with CF0 and to structurally reconstitute ATP synthesis. Functional reconstitution, however, seemed to be impaired.

F-ATPases synthesize ATP at the expense of protonmotive force (1)(2)(3)(4)(5)(6)(7) or sodiummotive force (8,9). F-ATPase is composed of the membrane-embedded proton (sodium) channel (F 0 ) and the extrinsic, water-soluble F 1 . CF 1 1 consists of five different subunits, ␣ (56 kDa), ␤ (54 kDa), ␥ (36 kDa), ␦ (21 kDa), and ⑀ (15 kDa), in stoichiometric proportion of 3:3:1:1:1. According to the crystal structure (10), six nucleotide-binding sites are present between subunits ␣ and ␤, three catalytical mainly on ␤ and three noncatalytical on ␣. The detailed mechanism by which F-ATPases synthesize ATP and couple ATP hydrolysis to proton pumping is still unknown, but it is generally accepted that two to three of the six nucleotide-binding sites participate cooperatively in the reaction and that ion flux through the F 0 portion of the enzyme causes conformational changes that are relayed into F 1 to drive ATP liberation. The binding change model of ATP synthesis (reviewed in Refs. 2 and 6) envisages a functional cycling between these catalytic sites. Its structural correlate may be a rotation of ␥ relative to (␣␤) 3 . This concept is supported by the recently published structure of F 1 from bovine heart mitochondria at 2.8-Å resolution (10), by electron microscopy (11), and by cross-linking data (12). Most recently, intersubunit rotation of subunit ␥ in (␣␤) 3 was time-resolved by polarized spectrophotometry (13). The identity and linkage of the other "stator" or "rotor" elements of F 1 with their counterparts in F 0 , however, remain to be elucidated.
Subunits ␥, ␦, and ⑀ are thought to function at the interface between the membrane-embedded F 0 and the extrinsic F 1 . They are instrumental for the coupling between ion movements through F 0 and ATP release from F 1 (1-5, 7, 14). Purified subunit ␦ enhances the reconstitutional activity of CF 1 lacking subunit ␦ (CF 1 (Ϫ␦)) in partially CF 1 -depleted thylakoids. This enhancement has been attributed to the plugging of open CF 0 channels. Because of the reduced proton leak, the protonmotive force was restored, which activated both the reconstituted and remaining CF 0 CF 1 (15)(16)(17).
Five Ser 3 Cys point mutants of chloroplast ␦ were overexpressed in Escherichia coli. The engineered single Cys residues were modified with a sulfhydryl-specific dye or with heterobifunctional and photoactivable cross-linking reagents. We studied the topological and functional consequences of cross-linking ␦ to other F 1 subunits. All mutant ␦ subunits were cross-linked to either the ␣ or ␤ subunit under all conditions employed. Cross-linking did not impair ATP hydrolysis by soluble CF 1 . ATP synthesis by CF 0 CF 1 seemed to be impaired.
Plasmids, Bacterial Strains, and Molecular Genetics-We have cloned the gene for spinach ␦ into pET-3d (20) and expressed the protein in E. coli strain BL21(DE3) (21). Mutant recombinant ␦ subunits were obtained by synthesizing mutagenesis primers, followed by two consecutive polymerase chain reaction cycles, one to introduce the mutations into the nucleotide sequence and the other one to obtain full-length genes, followed by transformation and expression (22). Recombinant spinach ␦ was purified from the cytoplasmic fraction by anion-exchange * This work was supported by Grant SFB 171-B3 from the Deutsche Forschungsgemeinschaft and by the Fonds der Chemischen Industrie and the Land Niedersachsen. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Two additional mutants (␦ S112C and ␦ T141C ) were prepared along with the described mutants, but were discarded. Cys-112 was not accessible in nondenatured ␦ (it was titratable with Ellman's reagent only under denaturing conditions), and ␦ T141C did not bind to CF 1 (Ϫ␦) after modification of Cys-141 with TFPAM-3, presumably due to steric hindrance.
Chemical Modification-Chemical modifications were carried out after gel filtrating ␦ SXC (19 M) against 50 mM MOPS, pH 7, immediately followed by the addition of the respective modifying reagent (100 mM eosin-5-maleimide, APDP, or TFPAM-3 or 50 M TIDM/3). The reaction was allowed to proceed at room temperature for 1 h in the dark. Excess reagent was removed by a second gel filtration against 25 mM Tris-HCl, pH 7.8. Photoactivation was achieved by 20 min of illumination in a UV transilluminator shielded by an optical filter (Schott KG2, short wavelength cutoff at 340 nm) or by exposure to 20 flashes (300 mJ/cm 2 ) from a frequency-doubled Ruby laser at 347 nm. Laser excitation was superior in avoiding nonspecific protein breakdown as caused by the continuous UV illumination.
Chemical cleavage of X-Cys peptide bonds in cross-linked CF 1 was carried out after cyanylation of cysteine residues with 2-nitro-5-thiocyanatobenzoate in 6 M guanidinium chloride/Tris, pH 8, followed by acidification and then by incubation at 37°C for 24 h in 6 M guanidinium chloride/sodium borate, pH 9 (25). SDS electrophoresis was carried out with the Pharmacia Phast system on 8 -25% gradient gels stained with silver/silicotungstic acid (26).

Accessibility of Engineered Cys Residues in ␦-
The titration of engineered Cys residues in spinach ␦ with Ellman's reagent (35) gave ratios of close to 1 for each of the five single point mutants. Under nondenaturing conditions (Table I), all Cys residues were exposed to this reagent in isolated ␦.
We studied their accessibility after incorporation of ␦ into CF 1 (Ϫ␦). To this end, CF 1 (Ϫ␦) (250 nM) was complemented with ␦ SXC (500 nM); unbound ␦ SXC was removed by anionexchange chromatography; and labeling was carried out with eosin-5-maleimide (100 M). For CF 0 CF 1 samples, NaBr vesicles (300 M chlorophyll) were incubated with CF 1 (Ϫ␦)ϩ␦ SXC (1.5 ϩ 1.5 M, respectively). After 1 h of incubation, excess CF 1 was removed by centrifugation, and the reconstituted vesicles were labeled with eosin-5-maleimide (100 M). After washing to remove excess label, they were again treated with NaBr to remove the labeled CF 1 from the membrane. The isolated CF 1 was run on SDS gels and photographed under UV illumination. The relative intensities of the bands of labeled ␦ and ␥ were measured, and their ratios were calculated; the results are shown in Table II. It is evident that all introduced cysteines were labeled both in CF 1 and in CF 0 CF 1 . In CF 0 CF 1 , however, the labeling yield of Cys-160 and Cys-166 was decreased. This pointed toward an exposed location of Cys-10, Cys-57, and Cys-82 both in CF 1 and in CF 0 CF 1 . Cys-160 and Cys-166 were less accessible than the other engineered Cys residues in CF 0 CF 1 .
Cross-linking of ␦ with CF 1 and with CF 0 CF 1 -␦ SXC was modified with one of three heterobifunctional cross-linkers, APDP, TFPAM-3, or TIDM/3. APDP is a cleavable, rather long (1.9 nm) cross-linker specifically reacting with sulfhydryl groups by a disulfide exchange reaction; the photoactivable group is an azide. TFPAM-3 was introduced by Capaldi and co-workers (18) in cross-linking studies of E. coli F-ATPase. It reacts by its maleimide function with sulfhydryl groups; the spacer is 0.9 nm long, and the photolabile group is a perfluorophenylazide. TIDM/3 is also 0.9 nm long, with a maleimide as the sulfhydryl-reactive group and with a diazirine as the photolabile group.
Figs. 1 and 2 document the cross-link products of engineered ␦ after its modification with the first function of the respective cross-linker, reincorporation into CF 1 (Ϫ␦), and photolysis to activate the second function of the cross-linker. Figs. 1 and 2 show the results for the cross-linker TIDM/3. Fig. 1 shows a silver-stained SDS gel, and Fig. 2 the respective Western blots with monospecific rabbit antisera directed against spinach CF 1 -␣, CF 1 -␤, and CF 1 -␦. ␦ S10C was cross-linked to subunit ␤; the other four mutant ␦ subunits were cross-linked to subunit ␣. Additional bands in Fig. 1 (protein stain) represent degradation products caused by UV illumination. These were always observed at slightly different positions depending on the crosslinking reagent running between subunits ␥ and ␦. These breakdown products probably originated from subunit ␤ since bands at corresponding spots were visible only in the Western  blot with anti-␤ antibodies. Blots with primary antibodies directed against subunits ␥ and ⑀ revealed only bands at the expected positions of monomeric ␥ and ⑀ (data not shown). It was noteworthy that identical cross-link products were observed under rather different conditions, namely in the absence and presence of ADP, phosphate, and ATP. TFPAM and TIDM/3 gave identical results upon cross-linking, whereas APDP cross-linked ␦ to both large subunits, ␣ and ␤ (data not shown). Cross-linking by photolysis of chemically derivatized proteins relies both on the specificity of the cross-linker and on the specific rebinding of the modified subunit to its "host." The specificity of the derivatization reaction was ensured by the chosen pH (7), which favors the attack of maleimides at sulfhydryl groups over the reaction with primary amines by a factor of 1000 (36). Since cysteine-free wild-type ␦ did not yield cross-link products after UV illumination, nonspecific side reactions could be ruled out.
Specific rebinding of the derivatized ␦ subunits to CF 1 (Ϫ␦) was checked by titrating CF 1 (Ϫ␦) with chemically modified, engineered ␦. We observed the same cross-link products (␦ 7 ␣ or ␦ 7 ␤) irrespective of the molar ratio of ␦ SXC over CF 1 (Ϫ␦) (2:1). Moreover, the binding of wild-type ␦ followed by the addition of derivatized ␦ SXC completely prevented the formation of cross-links (Fig. 3). This implied that the engineered subunits bound to the same domain on ␣ and ␤ as their wildtype counterpart.

FIG. 2. Western blot of the samples shown in
Activity of Cross-linked CF 1 -The activity of CF 1 (Ϫ␦) reconstituted with engineered or wild-type ␦ and before and after chemical modification and cross-linking was studied 1) by the ability of soluble CF 1 (Ϫ␦)ϩ␦ SXC to hydrolyze ATP and 2) by the degree of reconstitution of ATP synthesis in CF 1 -depleted thylakoid membranes recombined with CF 1 (Ϫ␦)ϩ␦ SXC . Table III shows that the Ca 2ϩ -and Mg 2ϩ -ATPase activities of soluble CF 1 were largely unaffected by the formation of cross-links.
Reconstitution of ATP synthesis by the addition of CF 1 to EDTA-treated and thereby partially CF 1 -depleted thylakoids can be simply due to the rebinding of CF 1 and plugging of the proton channel CF 0 (without formation of fully functional CF 0 CF 1 pairs). This "structural reconstitution" relies on the enzymatic activity of the residual CF 0 CF 1 that remained on the membrane upon EDTA treatment. "Functional reconstitution," on the other hand, implies that the reassembled CF 0 CF 1 pairs are active. A differentiation between both types of reconstitution is possible after irreversible blocking of residual CF 1 molecules by tentoxin, followed by the addition of untreated CF 1 (31).
Data from such reconstitution experiments revealed the following. The reconstitutional activity (ATP synthesis) of engineered ␦ was in the range of 94 -100% in comparison with wild-type ␦ both before and after modification with the crosslinking reagent. This holds for ␦ S10C , ␦ S57C , and ␦ S82C . ␦ S160C and ␦ S166C were not quite as efficient as the other ␦ SXC mutants, especially after modification with the cross-linking reagent (59 -79%). This agreed with the observation that residues ␦ 160 and ␦ 166 were partially shielded in CF 0 CF 1 , but not in CF 1 , and might indicate steric hindrance caused by the cross-linker. Photolysis-induced cross-linking did not affect structural reconstitution in EDTA vesicles in comparison with the effectiveness of the modified, but not yet photolyzed samples. The average loss of functional reconstitution of tentoxin vesicles for all five ␦ SXC mutants ranged between 14 and 28% and thus grossly matched the estimated yield of cross-linking in the range of 10 -20%.
Titration of engineered ␦ with Ellman's reagent before and after the reaction with maleimides yielded a ratio of Ͼ10:1, thus indicating a nearly quantitative derivatization of Cys  ␣ (right panel) of spinach chloroplast CF 1 . wt*, wild type (non-illuminated); wt, wild type (illuminated); 10, 57, 82, 160, and 166, the respective ␦ SXC mutants. 100 g of CF 1 (Ϫ␦) ϩ 5 g of ␦ SXC were incubated for 20 min at room temperature and then cross-linked (␦ SXC was previously labeled with TFPAM-3) by a 10-min UV illumination. Samples were gel-filtrated through BioSpin 30 columns (Bio-Rad) in order to remove unbound ␦, reduced with 0.5 mM dithiothreitol at 37°C for 30 min, and then diluted with 8 M guanidinium Cl and 0.1 M Tris acetate, pH 8, followed by a 30-min incubation at 37°C in the presence of 1 mM 2-nitro-5-thiocyanatobenzoate. Samples were acidified by the addition of acetic acid and then gel-filtrated (Pharmacia NAP-5) against 6 M guanidinium Cl and 0.1 M sodium borate, pH 9.0, followed by a 24-h incubation at 37°C. Thereafter, they were gel-filtrated against 10% (v/v) formic acid and precipitated by trichloroacetic acid. SDS electrophoresis was carried out on an 8 -25% Pharmacia Phast gel with 1.5 g of protein/lane. Samples were blotted semi-dry in the Pharmacia Phast system and reacted with anti-␦ antibodies (1:4000) or with anti-␣ antibodies (1:2000). Visualization of IgG was by peroxidase-conjugated secondary antibodies and chemiluminescence (Boehringer Mannheim); exposure was for 5 min (left panel) and 1 min (right panel). (data not shown). The efficiency of photo-cross-linking, on the other hand, was low. It was conceivable that a subset of noncross-linked CF 1 bound preferentially to CF 0 . We checked whether cross-linked CF 1 per se rebound to CF 0 as follows. NaBr vesicles were incubated with CF 1 (Ϫ␦)ϩ␦ SXC ; the supernatant containing unbound CF 1 was removed; and the reconstituted vesicles were re-extracted with NaBr. This "second" NaBr extract contained those CF 1 molecules that had rebound to CF 0 . Fig. 6 shows the results of such an experiment with CF 1 (Ϫ␦)ϩ␦ 160 . Lanes 1 show SDS electrophoretic separations of the unbound fractions; lanes 2 show the re-extracted fractions. The left panel (CF 1 ) shows controls, in which NaBr vesicles (150 g of chlorophyll) were incubated with 300 g of CF 1 ; the right panel shows the same experiment with cross-linked CF 1 (Ϫ␦)ϩ␦ 160 . It was evident that even the ␦ 7 ␣ cross-linked CF 1 was bound to CF 0 . Its characteristic band appeared in both the unbound and the re-extracted fractions. We chased bound cross-linked CF 1 (Ϫ␦)ϩ␦ 160 with CF 1 (300 g). It did not change the result (cf. lanes ϩCF 1 ). This implied that both cross-linked CF 1 and native CF 1 bound to CF 0 with similar affinity. DISCUSSION We engineered cysteines into subunit ␦ of spinach chloroplast CF 0 CF 1 at sequence positions (Ser) 10, 57, 82, 160, and 166. The cysteines served as anchors for the maleimide function of photoactivable cross-linking reagents. The aim was 2-fold: 1) to gain information on the location of subunit ␦ in the complex and 2) to reveal the functional consequences of crosslinking on ATP hydrolysis with CF 1 and ATP synthesis with CF 0 CF 1 .
Structural Considerations-The engineered cysteine residues reacted with the fluorescent reagent eosin-5-maleimide when ␦ was solubilized, but also when ␦ was incorporated into CF 1 . All but Cys-160 and Cys-166 were just as reactive when ␦ was incorporated into CF 0 CF 1 . This points to a rather peripheral location of ␦, with most of the engineered Cys residues remaining solvent-exposed even after binding of ␦ to (CF 0 )CF 1 . The largely ␣-helical structure of subunit ␦ (84%) (37) with many of the side chains pointing outwards and the expected exposed location of the polar hydroxyl group of Ser residues are compatible with this view. Our interpretation relied on two assumptions. 1) Eosin-5-maleimide does not penetrate into the protein interior; and 2) the reagent does not induce the disso-ciation of ␦ from the remainder of the complex. The hindered access of eosin-5-maleimide to Cys-166 in CF 0 CF 1 makes the first assumption plausible, and the fact that ␦ was isolated together with CF 1 from reconstituted washed NaBr vesicles supports the second.
␦ was cross-linked exclusively to ␣ and ␤, and this allowed us to narrow down its position in CF 0 CF 1 . ␦ participates in the coupling of proton translocation through F 0 and substrate conversion in F 1 . It functionally interacts both with F 0 and with F 1 (reviewed in Ref. 14). Was ␦ cross-linked to one particular ␣/␤ pair, or was it "bridging" two or even all three ␣ subunits, depending on the position of the engineered Cys residue? In view of the recently published structure of MF 1 (10), the length of the cross-linker and the distribution of the cross-links (one between ␦ and ␤ and four between ␦ and ␣) would seem to exclude the latter possibility. It is unlikely that ␦ traverses one complete ␤ subunit, a distance of at least 50 Å (cf. Fig. 3 in Ref. 10). Instead, ␦ probably interacts with one single ␣/␤ couple.
According to the atomic structure of F 1 (10), most of these portions of ␣ and ␤ are located at the periphery and at the top third of F 1 (cf. Fig. 7). Taking the ␦ 7 CF 0 -I cross-link (38) into account, this would imply the following. Subunit I (␣ in E. coli) protrudes from the membrane around the outside of F 1 up to the point where it contacts ␦. In view of the ␦ S10C 7 ␤ crosslink, Cys-10 is expected rather at the top of the F 1 molecule. The mass of subunit ␦ (21 kDa) and its elongated shape (14) in comparison with known proteins of similar size may allow for a length of ϳ45 Å. ␦ then could reach down to just below the height of the nucleotide-binding sites. Beckers et al. (38) concluded that ␦ is cross-linked to the C-terminal end of subunit I of CF 0 , which is built from one transmembrane stretch plus a hydrophilic headpiece. If its hydrophilic head is stretched out, FIG. 6. SDS electrophoresis of unbound and bound re-extracted fractions from reconstituted vesicles. NaBr vesicles (150 g of chlorophyll) were incubated with CF 1 (300 g; left panel) or cross-linked CF 1 (Ϫ␦)ϩ␦ 160 (300 g;, right panel). The supernatant containing unbound CF 1 was removed (lanes 1), and the reconstituted vesicles were re-extracted with NaBr (lanes 2). Lanes ϩCF 1 indicate an additional incubation of NaBr vesicles with 300 g of CF 1 after prior incubation with cross-linked CF 1 (Ϫ␦)ϩ␦ 160  subunit I would be long enough to contact ␦. In our view, ␦ by itself is not part of the stalk linking F 0 and F 1 .
The data presented here are compatible with and expand results from other laboratories. Studies with E. coli F 1 revealed an ␣ 7 ␦ disulfide cross-link (39,40) involving ␦ Cys-140 (equivalent to Cys-141 in spinach ␦) (41). E. coli ␣ contains Cys residues at positions 47, 90, 193, and 243. These residues correspond to positions 48,91,194, and 244 in spinach chloroplast ␣ and to positions 47, 90, 201, and 251 in mature bovine heart mitochondrial ␣, respectively. We homology-built the E. coli sequence of subunit ␣ into the Leslie-Walker structure (10) 2 ; modeling was performed with WHATIF 4.99 (42). The modeled positions of the four Cys residues in E. coli ␣ are as follows. Residues 47 and 90 are close together and rather exposed at the ␤-sheet on top of ␣, with Cys-90 sticking out just a little bit farther than Cys-47; Cys-193 and Cys-243 are both buried in the central domain not far from the nucleotide-binding region. The disulfide cross-link between EF 1 -␦ Cys-140 and EF 1 -␣ thus pins the reactive sites on ␣ down to residue 47 or 90. In a different approach, it has been shown that proteolytic digestion of the amino-terminal portion of both EF 1 -␣ (43) and MF 1 -␣ (44) results in loss of the capability to bind EF 1 -␦ or oligomycin sensitivity-conferring protein. Taken together, these results identify the amino-terminal third of subunit ␣ as the prime candidate for the binding of ␦ and oligomycin sensitivity-conferring protein.
The rather exposed position of ␦ on the outside of the upper half of the (␣␤) 3 assembly contrasts with conclusions inferred from immunological and proteolysis studies. A rather hidden location of ␦ within intact CF 0 CF 1 has been postulated (27,45). It is conceivable, though, that antibodies are sterically or electrostatically excluded from the thylakoid membrane surface. An exposed location of ␦ within CF 0 CF 1 is compatible both with the recently published 2.8-Å structure of mitochondrial F 1 (10), which leaves little if any space inside the (␣␤) 3 ␥ core, and with previous findings that oligomycin sensitivity-conferring protein is cross-linked to subunit ␤, but not to the other subunits that presumably form the stalk in MF 0 MF 1 (46).
Outlook-The yield of cross-linking between ␦ and ␣ or ␤ was dependent to a small degree on the state of CF 1 , solubilized or bound to CF 0 , with or without added nucleotides. When engineered and chemically modified ␦ was added to CF 1 (Ϫ␦), the activities of isolated CF 1 were unaffected. The hydrolysis activity by solubilized and chemically modified CF 1 was not impaired even after photolysis of the second function of the crosslinker. Cross-linked CF 1 rebound to CF 0 with about the same affinity as native CF 1 , and it restored photophosphorylation in partially CF 1 -depleted thylakoids (EDTA vesicles) to the same extent as non-cross-linked CF 1 . Preliminary data suggested that the ability of modified ␦ to functionally reconstitute ATP synthesis in NaBr vesicles might be impaired by photo-cross-linking.
Recently, a large-scale (Ͼ200°) rotational motion in a few 10 ms of subunit ␥ relative to immobilized (␣␤) 3 of spinach CF 1 was demonstrated (13). In terms of a rotatory mechanism of catalysis (2, 6) with ␥ acting as a rotor relative to (␣␤) 3 , we visualize ␦ together with subunits I and II rather as elements of the stator or counterbearing.