Subunit interactions within the chloroplast ATP synthase (CF0-CF1) as deduced by specific depletion of CF0 polypeptides.

The proton-linked ATP synthase (CF1-CF0) of chloroplasts consists of a catalytic component (CF1) and a membrane-embedded part (CF0) that interacts with CF1 and contains a proton channel. The subunits of CF0 which are involved in binding of CF1 were studied by examining the effect of selective depletion of subunits I, II, and IV of CF0 from the chloroplast ATP synthase on the association of the remaining CF0 subunits with CF1. Dissociated CF0 subunits were identified by sucrose density gradient centrifugation. Removal of subunit IV alone from CF0-CF1 did not cause dissociation of the other CF0 subunits from CF1. Upon removal of both subunits I and IV from CF0-CF1, subunit II also dissociated, but subunit III was still bound to CF1. Thus, at least two subunits of CF0, I and III, directly associate with CF1. Subunit II is unlikely to bind CF1 directly and may associate with subunit I. Although depletion of subunit IV does not cause dissociation of CF0 from CF1, its interaction with CF1 subunits is uncertain.

The interactions between CFo' and CF1 of the chloroplast ATP synthase may involve several subunits both in CF1 and CFo. In CF, at least two subunits, (Y and 6, probably associate with CFo subunits (l-4). Reconstitution of subunit a-deficient CF, with CF1-depleted thylakoid membranes indicates that subunit 6 is not absolutely required for the binding of CF, to , but in the absence of the 6 subunit, CF1 binding is reduced (5) and the CFo proton channel is not blocked (l-4). Subunit (Y of CF1 may be protected from proteolytic cleavage when CFI is bound to the thylakoid membranes (2). Two subunits of CFo, I and II, have large hydrophilic domains that are proposed to protrude into the chloroplast stroma and possibly interact with CF1 (6-8). Evidence from the studies of the biogenesis of chloroplast ATP synthase in Chlumydomenus reinhordtii, a green alga, shows that subunit III may also be involved in the binding of CF1. Chloroplast ATP synthase in C. reinhardtii contains nine subunits, five in CF, and four in CFo, similar to the subunit composition of chloroplast ATP synthase in higher plants (9). Lemaire and Wollman (10)  a significant amount of CFI in the cells was membrane-bound (5-10% of that in the wild type), but in a mutant lacking both subunits III and IV, binding of CF, to the membranes was abolished. Their studies suggest that subunit III may be involved in the binding of CF1 to the thylakoid membranes while subunit IV may not be absolutely required. However, depletion of any CF, subunits in these mutants could prevent the assembly of CFO or alter CFO structure. It is difficult, therefore, to draw unambiguous conclusions about the subunit interactions in CF,-CF, from these experiments. Another approach to studying the subunit interactions in CFO-CF, employed chemical cross-linking. By this method, cross-links of (Y-II, p-1, p-11, r-11, b-1, and t-111 were reported (11). Similar experiments were done by Rott and Nelson (12). However, the yield of cross-linked product was low, and nonspecific cross-links and aggregation cannot be excluded.
In this paper, we report a different approach to studying the subunit interactions between CFo and CF,. By examining the association of CFo subunits with CF1 after selective removal of particular CFo subunits by procedures reported previously (13), we were able to determine more specifically which subunits of CF, directly interact with CF1. were kept on ice for 30 min and centrifuged at 12,000 x g for 10 min. The precipitates were dissolved in 0.1 mM ATP, 0.5 mM EDTA, 0.1% asolectin, 0.2% Triton X-100, and 30 mM Tris succinate (pH 6.5) to a concentration of about 20 mg of protein/ml. The protein solution (0.5 ml) was layered onto a lo-ml 7-30% sucrose gradient in 0.2% Triton X-100 buffer as described by Pick and Racker (14). After centrifugation in a Beckman SW 41 rotor at 4 "C for 18 h, fractions of 0.6-0.8 ml were collected from the bottom of each centrifuge tube. CFo prepared as previously described (5)

RESULTS
We have developed methods to remove selectively subunits I and IV from spinach CFo-CFi (13). By examining the effect of depletion of particular CFo subunits on the association of the remaining CFo subunits with CF1, we can determine which CFO subunits interact with CFI. Since the molecular masses of CFo subunits are much lower than the mass of CFI, CFo subunits dissociated from CF, can be easily identified following sucrose density gradient centrifugation.
To obtain subunit IV-deficient CFo-CF1, CFo-CF1 was bound to a DEAE-Trisacryl column. Subunit IV was dissociated from the complex and eluted from the column by 0.08% Zwittergent 3-12 in the presence of 50 mM sodium phosphate buffer (13). Subunit IV-deficient CFo-CF, recovered from the column was applied to a sucrose density gradient. Triton X-100 (0.2%) was included in the sucrose gradient buffer solution to prevent aggregation of polypeptides. Triton X-100 was used to keep CFo-CFi soluble during purification by the sucrose gradient centrifugation procedure developed by Pick and Racker (14). This detergent does not cause dissociation of CFO subunits from CFI. Fig. 1 shows the SDS-polyacrylamide electrophoresis pattern of the sucrose gradient fractions containing either subunit IV-deficient CF,-CF, or CF,. Because of its lower molecular mass, CF, sediments slower than CF, when subjected to radial acceleration in an ultracentrifuge and moved to a sucrose gradient zone containing 8-11% sucrose (Fig. lB), whereas subunit IV-deficient CFo-CF1 sediments at a zone containing 20-22% sucrose (Fig. L4). In the absence of subunit IV, the other CFo subunits remained associated with CF,. The sucrose density gradient peak fraction of subunit IV-deficient CFo-CFI was electrophoresed on a SDS-polyacrylamide gel (Fig. 2). The gel was lightly stained with silver and scanned with a densitometer to quantitate the residual subunit IV in subunit IV-deficient CFo-CF1. In comparison with control CFO-CF1, there was only 18% residual subunit IV in subunit IV-deficient CFo-CF,. These experiments clearly demonstrate that depletion of subunit IV does not cause dissociation of the other CFo subunits from CF1.
Subunits I and IV can be depleted together from CFo-CFi on a DEAE-Trisacryl column by elution of the column with 20 mM Zwittergent 3-12 at a low concentration of sodium phosphate (5 mM the column, subunit I-and IV-deficient CFo-CF, was eluted with a high salt buffer. The subunit-deficient CFo-CF1 thus obtained was partially purified. In addition to the subunitdeficient CFo-CF1, ribulose bisphosphate carboxylase/oxygenase and some other minor contaminants were present (Fig.  3). This preparation was applied onto a sucrose density gra- The subunit I-and IV-deficient CFo-CF, preparation was obtained by the chromatographic method described under "Materials and Methods" and fractionated on a sucrose density gradient. After centrifugation, 18 fractions were collected from the sucrose gradient, and samples were electrophoresed on a 15% acrylamide gel. Lane I, CFo-CF, (30 rg); lane II, subunit I-and IV-deficient CFo-CF, (30 rg) obtained by DEAE-Trisacryl chromatography; lanes l-21, sucrose gradient fractions (fractions l-11) of subunit I-and IVdeficient CFo-CF1. The gel was stained by Coomassie Blue. SW, small subunit of ribulose bisphosphate carboxylase/oxygenase. dient and subjected to centrifugation. Fig. 3 shows that subunit II (shown in Fig. 3, fractions 2 and 3) was separated from CF, (shown in Fig. 3, fractions 10 and ll), sedimenting in the 8-10% sucrose region. Fractions 2 and 3 of the sucrose gradient also contained some subunits /3 and y and a trace amount of subunit 6 of CF1 besides subunit II. It is unlikely that subunit II is associated with these CF, subunits because the relative amount of subunit II is much higher. Thus it appears that in the absence of subunits I and IV, subunit II does not bind to CFi.
Another possibility is that the dissociation of subunit II from CF, is caused by treatment of CFo-CF1 with 20 mM Zwittergent 3-12, rather than depletion of subunits I and IV. We reported previously that four-subunit CFo and subunit 6 were dissociated from CFo-CFI by elution of the DEAE-Trisacryl column with 20 mM Zwittergent 3-12 in the presence of 30 mM sodium phosphate (5). Zwittergent 3-12 treatment in the presence of 5 mM sodium phosphate might cause not only the dissociation of subunits I and IV but also subunits II, III, and 6. Because of the low concentration (5 mM) of sodium phosphate, however, subunits II, III, and 6 might fail to elute until the column is washed with a buffer of higher ionic strength. Richter et al. (19) reported that both the c and 6 subunits may be dissociated from CF1 by 20% ethanol. When CF1 was bound to a DEAE-cellulose column, however, elution of the column with 20% ethanol in the presence of a low concentration of salt only removed the c subunit from the column. Upon removal of ethanol, the 6 subunit reassociated with CF, (19). Similarly, after removal of Zwittergent from the DEAE-Trisacryl column, the majority of the 6 subunit recovered from the column appears to be associated with the other CF, subunits since it sediments with CFI to a 20% sucrose region after density gradient centrifugation (Fig. 3). In contrast, free 6 subunit sediments to a 10% sucrose region under these conditions.* Similar to the 6 subunit, the majority of subunit III is also found to be associated with CFI after removal of Zwittergent 3-12 (Table I)  Samples obtained by the DEAE-Trisacryl chromatography followed by sucrose density gradient centrifugation were analyzed on 15% acrylamide gels as described under "Materials and Methods." The amounts of subunits I and II were determined by scanning the Coomassie Blue-stained gels with a densitometer. Subunit IV was quantitated on silver-stained gels by densitometric scanning. The amount of subunit III was determined by ["C]DCCD labeling. Amounts of CFo subunits in control CF,-CF, were assumed to be 100%.

CFo subunits depleted
Subunits that remain associated with CF, Zwittergent, however (Fig. 3), so it is unlikely that subunit II binds to CFI by itself. The sucrose density gradient fractions of the subunit I-and IV-deficient CFo-CFI preparation were treated with ['"Cl DCCD to determine whether subunit III in this preparation was associated with CFI. The labeled fractions were then run on a SDS-polyacrylamide gel. Fig. 4 shows the ['YJ]DCCD labeling patterns of the sucrose density gradient fractions containing subunit II and CF,. Both fractions contained a ["'CIDCCD-labeled polypeptide with a molecular mass of 8 kDa. The distribution of the 8-kDa DCCD-binding polypeptide in the sucrose density gradient is shown in Fig. 5. The majority of radioactivity was present in the fractions contain- ing CFi, indicating that the DCCD-binding protein (subunit III) was associated with CF1. A small portion of radioactivity was in the fractions containing 8-10% sucrose and may represent free subunit III or subunit III associated with other CFo polypeptides.
The co-sedimentation of subunits II and III does not necessarily mean that subunits II and III are associated together, since several subunits of CFo-CF1, including subunits I, II, IV, p, y (Fig. lB), 6, and e2 sediment at a similar rate under these conditions. The ['4C]DCCD-labeled 8-kDa polypeptide which sedimented in the 8-10% sucrose region (Figs. 4 and 5) could be some minor contaminant rather than the true subunit III of CFo. We tried to label the CFo fractions obtained by sucrose density gradient centrifugation (shown in Fig. 1B) to localize CF1-dissociated subunit III in the sucrose gradients. This attempt was not successful because CFo in the sucrose density gradient fractions was very dilute, and the protein was lost during frequent washing after the ['4C]DCCD treatment. Table I shows a quantitative measurement of CF, subunits which remained associated with CF1 when particular CFo subunits were depleted. The amounts of subunits I, II, and III did not change significantly upon removal of subunit IV from CFo-CF1. Subunit III was quantitated by [14C]DCCD labeling. Since the recovery of ['*C]DCCD-labeled subunit III was somewhat variable after trichloroacetic acid precipitation and organic solvent washing for removing unreacted [14C]DCCD, the assay is only semiquantitative.
Nevertheless, it is clear that subunit III appears to interact with CF, directly. Even after all other CF, subunits were depleted from CFO-CF1, greater than 60% of subunit III still remained bound to CF1.

DISCUSSION
To investigate which subunits in CFo are involved in the binding of CF1, we adopted the straightforward approach of selectively removing particular CFo subunits from CFo-CF1 and then examining the association of the remaining CFo subunits with CF1 by sucrose density gradient ultracentrifugation. Density gradient centrifugation has been widely used to separate macromolecules based on their molecular masses (20). CFo has a total molecular mass of about 170 kDa (21) which is lower than that of CF1 (400 kDa) (22). Free CFo subunits sediment in the sucrose gradient slower than CF1, whereas CF1-associated CFo subunits move with CF1. This provides a convenient method to study the subunit interactions between CFo and CF1. In comparison with Lemaire and Wollman's experiments (10) in which C. reinhardtii mutants lacking CFo subunits were employed to study the subunit interactions between CFo and CFI, the advantage of our approach is that we study the subunit interactions in assembled CFo-CF1 and thus avoid uncertainties resulting from the possible effects of subunit depletion on assembly of the complex.
Our results show that removal of subunit IV by treatment of CF,-CF, with 0.08% Zwittergent does not cause dissociation of other CFo subunits from CFI. All remaining CFo subunits sediment with CF,. Although we do not know at this stage whether subunit IV directly interacts with CF,, it is clear that association of the other CFo subunits with CF, is not dependent on the binding of subunit IV. Subunit III appears to interact directly with CFI. It associates with CF, even in the absence of other CFo subunits.
Our results regarding the requirement for subunits III and IV in the binding of CF, are in agreement with the conclusions made by Lemaire and Wollman (lo), who studied the assembly of CF,-CF1 in C. reinhardtii mutants in which subunit IV or both subunits III and IV were depleted. An opposite observation in regard to the interaction of subunit III and CF, came from the study of binding of CF, to butanol-extracted chloroplast subunit III (23). Subunit III thus obtained did not bind CR when incorporated into liposomes. This could be due to improper orientation or organization of subunit III in the liposomes or possible denaturation of subunit III upon butanol extraction. The subunit interactions involved in the binding of subunits I and II to CF1 seem more complicated than that of subunit III. These two subunits form a stable association with CFi in the absence of subunit IV, but in the absence of both subunits I and IV, subunit II by itself can be dissociated from CF1. This indicates that there is likely to be a direct interaction between subunits I and II. Association of subunit II with CF1 could be mainly through subunit I which is proposed to have a CF,-binding site at its large hydrophilic domain and to have a hydrophobic domain inserted into the membrane (6, 7, 24). Although we have shown that binding of subunits I and II to CFi does not require subunit IV, possible interactions between subunit IV and subunits I and II cannot be excluded, and such interactions may stabilize the whole complex.  Cozens and Walker (29) suggested that the cyanobacterial ATP synthase could contain single copies of subunits I and II instead of two identical b subunits as in the E. coli ATP synthase. The 32 NHz-terminal residues of CFo subunit II of spinach show 41% homology to that of the cyanobacteria (30). Immunochemical evidence also indicates that the topography of subunits I and II is similar and that both have a large hydrophilic carboxylterminal portion protruding into the stroma (7). It is possible that subunits I and II of CFo perform the function that the b dimer of E. coli F0 does.
An interaction between subunit III and CF1 was demon-strated in our experiments. This interaction could be a common feature in F,-F, type ATP synthases. In E. coli. subunit c is postulated to fold in the membrane like a hairpin with two nonpolar a-helices separated by a polar loop region which is accessible to antibodies (8, 31). The Gln4' residue in the putative polar loop region of subunit c when substituted with certain amino acids results in altered F1 binding and uncoupling (32, 33). In the accompanying paper (13) we reported that DCCDsensitive ATPase activity of CFo-CFI after incorporation into asolectin liposomes was relatively insensitive to removal of subunit IV. Subunit IV may thus not directly participate in proton translocation. It will be interesting to test whether CFI associated only with subunit III of CFo expresses DCCDsensitive ATPase activity and other proton translocationrelated activities.
The results that we described here show that at least two subunits of CFo, I and III, are involved in CFi binding, but the CF1 subunits with which they interact are unclear. The (Y and 6 subunits of CF, may be involved in binding to CFO (1,2,4,34). We have depleted the 6 subunit from subunit IIIassociated CF1.3 It will be of interest to examine the association of subunit III with this CF1 (-6). Resolution of Fo into isolated component subunits followed by reconstitution of proton translocation activity by recombination of isolated subunits has so far been most successful with the E. coli Fo (35). We have purified subunit IV of CFo (13), and other CF, subunits may be resolved by Zwittergent 3-12 on DEAE-Trisacryl columns by the procedures reported here and in the accompanying paper (13).