Interrelation between high and low affinity tentoxin binding sites in chloroplast F1-ATPase revealed by synthetic analogues.

Eight synthetic analogues of tentoxin (cyclo-(L-N-MeGlu1-L-Leu2-N-MeDeltaZPhe3-Gly4)) modified in residues 1, 2, and 3 were checked for their ability to inhibit and reactivate the ATPase activity of the activated soluble part of chloroplast ATP synthase. The data were consistent with a model involving two binding sites of different affinities for the toxins. The occupancy of the high affinity site (or tight site) gave rise to an inactive complex, whereas filling both sites (tight + loose) gave rise to a complex of variable activity, dependent on the toxin analogue. Competition experiments between tentoxin and nonreactivating analogues allowed discrimination between the absence of binding and a nonproductive binding to the site of lower affinity (or loose site). The affinity for the loose site was not affected significantly by the modifications of the tentoxin molecule, whereas the affinity for the tight site was found notably changed. Increasing the size of side chain 1 or 2 and introducing a net electrical charge both resulted in a decrease of affinity for the tight site, but the second change dominated the first one. The activity of different ternary complexes enzyme-tentoxin-analogue depended on the nature of the toxin bound on each site and not only on that bound on the loose site. This demonstrates that the reactivation process results from an interaction, direct or not, between these two binding sites. Possible molecular mechanisms are discussed.

F 0 F 1 proton ATPases (or ATP synthases) are bound to energy-transducing membranes and couple the phosphorylation of ADP into ATP to the dissipation of a protonmotive force. They consist of a transmembrane proton channel (F 0 ) and an extrinsic part (F 1 ) bearing six nucleotide binding sites, catalytic and noncatalytic. The F 1 moiety is composed of five different subunits named ␣, ␤, ␥, ␦, and ⑀ (stoichiometry ␣ [3]

␤[3] ␥[1] ␦[1]
⑀ [1]). Subunits ␣ and ␤ bear the nucleotide binding sites and are disposed as a crown, the ␥ subunit being located in the center of this structure (1)(2)(3)(4). The F 0 moiety basically consists of three or four different subunits (Escherichia coli: a [1] b [2] c [9 -12]; chloroplast: a [1] b [1] bЈ [1] c [9 -12]), the mitochondrial enzyme having additional subunits (5,6). It is proposed that the F 0 moiety would work as a rotative proton-driven motor, the rotor consisting of the c subunits (7), presumably arranged in a crown (8). The rotation would be transmitted to the ␥ subunit of the F 1 moiety (9), which should modify sequentially the three catalytic sites located on ␤ subunits (4) to induce ATP synthesis (10). Experimental arguments have been presented against (11,12) and for (9,(13)(14)(15)) the rotation of ␥. An essential feature of this model is that the cooperative functioning among the three catalytic sites is strictly related to the rotation of the ␥ subunit and thus to the proton pumping activity.
Tentoxin (TTX) 1 is a natural cyclic tetrapeptide (cyclo-(L-MeAla 1 -L-Leu 2 -Me⌬ Z Phe 3 -Gly 4 )), produced by several phytopathogenic fungi of the Alternaria genus (16,17). Under special conditions, this toxin induces a chlorosis in some higher plants (18). It specifically inhibits ATP synthesis in isolated chloroplasts (19). In vitro and at low concentrations (10 Ϫ8 -10 Ϫ7 M), TTX inhibits the isolated chloroplast F 1 -ATPase (19 -22), but at higher concentrations (10 Ϫ5 -10 Ϫ4 M), it strongly stimulates ATPase activity (21)(22)(23). At these same concentrations, the effect observed on membrane-bound ATPase (F 0 F 1 complex) is restricted to a partial release of inhibition, but the reactivated F 0 F 1 complex recovers the ability to couple proton transport to ATP synthesis (24). TTX dramatically disturbs the interactions among different nucleotide sites of ATPase, whatever the toxin concentration range (25,26). Simultaneous perturbation of these interactions and preservation of proton coupling in the TTX-reactivated form are intriguing in the context of rotational catalysis. Understanding the inhibitory and reactivating properties of TTX is therefore one of the elements that may contribute to the elucidation of the mechanism of energy coupling.
It has been demonstrated (27) that CF 1 binds two molecules of TTX on two sites of different affinities, which could be related to the inhibitory and reactivating effects of this molecule. These binding sites have not yet been identified, and the reasons for the specificity of TTX for the CF 1 -ATP synthase of some higher plants remain obscure (23,28). TTX stabilizes and enhances the ATPase activity of an ␣ 3 ␤ 3 complex from spinach CF 1 (29), which proves that the ␥, ␦, ⑀ subunits are not required for the stimulation effect of TTX but suggests that they could be necessary for the inhibition.
We have shown recently (30) that a very limited change in the molecule of TTX (replacement of L-MeAla 1 by L-MeSer 1 ) resulted in a dramatic loss of the reactivating effect at high concentrations, although the inhibitory effect at lower concentrations was unaffected. This led to the idea that it was possible to discriminate inhibitory and activating effects by an appropriate set of molecules derived from TTX. Because high concentrations of MeSer 1 -TTX were able to prevent CF 1 -ATPase reactivation by high concentrations of TTX, we proposed that MeSer 1 -TTX could bind the reactivating site competitively with TTX, giving rise to a poorly active form of the enzyme. However, it cannot be excluded that MeSer 1 -TTX prevents reactivation simply by chasing TTX from the high affinity site. The question was whether the stimulation by TTX only involves the low affinity binding site or the two binding sites of CF 1 . To get information about possible cooperation among TTX binding sites, we have used a kinetic approach consisting of studying the catalytic properties of ternary complexes formed by CF 1 -⑀ and different TTX analogues. This approach involves various combinations of analogues of different affinities for the inhibitory site and able or not to reactivate the enzyme at high concentrations. The results suggested that both binding sites participate in the formation of the reactivated state. At the same time, we were able to characterize the binding and effector properties of the set of TTX analogues modified in various positions. This allowed us to make hypotheses about the domains of the TTX molecule which are important for binding, inhibition, and stimulation.

MATERIALS AND METHODS
Preparation and Assay of Solubilized, Activated CF 1 -⑀-The soluble chloroplast ATPase (CF 1 ) was extracted and purified from spinach (Spinacia oleracea L.) leaves in the active form, devoid of the inhibitory ⑀ (31). Storage conditions and determination of the concentration of CF 1 -⑀ were described previously (31). The assays were modified slightly with respect to previous conditions (30). The enzyme (80 g ml Ϫ1 ) was activated by incubation for 3 h, at room temperature, in a medium containing 20 mM Tricine and 3 mM dithiothreitol, pH 8. Its activity, measured as described below, was constant for the entire experiment (5-8 h), ranging between 4 and 6 mol of hydrolyzed ATP/mg of protein/ min. For assays of ATP hydrolysis, the activated enzyme was diluted 40-fold in the reaction medium containing 50 mM Tris-SO 4 , 0.18 mM MgSO 4 , 40 mM KHCO 3 , pH 8.0. This medium was supplemented with toxins at the indicated concentrations. After 5 min of incubation at 37°C, the reaction was triggered by adding 1 mM ATP (final concentration). Aliquots were taken up at different time intervals and analyzed for nucleotide contents by high performance liquid chromatography, as described (30). The ADP concentration increased linearly with time, which allowed measurement of the rate of ATP hydrolysis. All of the rates displayed in the figures were normalized to that of the control (not toxin-treated). A rates versus concentration plot was fitted by a nonlinear iterative algorithm using Microcal Origin 3.54 (Microcal Software). With respect to our previous experimental conditions (30), the main change consisted in replacing the 1-h preincubation stage with toxin at 80 g ml Ϫ1 , at room temperature, by a direct 5-min incubation in the reaction medium, at 37°C. This time was sufficient to get the maximal effect of toxins. Suppression of the preincubation stage with TTX or derivatives had the advantage of controlling the toxin/enzyme concentration ratio strictly.
Synthesis of TTX and Derivatives-All toxins, including TTX, were synthesized by Drs. Florine Cavelier and Jean Verducci, Laboratoire des Amino acides, Peptides et Protéines, Université Montpellier II, Montpellier, France. The synthesis and structural properties of TTX and MeSer 1 -TTX have been reported already (30,32). Synthesis and structural features of the other derivatives will be published elsewhere.
Inhibitory and Reactivating Effects of TTX and L-MeSer 1 -TTX-We have shown previously that MeSer 1 -TTX inhibits CF 1 -ATPase with the same efficiency as TTX at low concentrations, but it reactivates CF 1 -ATPase poorly at high concentrations. This analogue also prevents the reactivation of ATPase by TTX (30). We have reinvestigated the effects of these two toxins to estimate quantitatively the binding and catalytic parameters. Fig. 2 shows the effect of TTX and MeSer 1 -TTX on the ATPase activity of CF 1 -⑀. Although much less pronounced than with TTX, the reactivation by MeSer 1 -TTX at high concentrations was effective. It was possible to fit the data with a simple model involving two independent binding sites for the toxin, the high affinity site being responsible for the inhibitory effect and the low affinity site being responsible for the reactivation (see Equation 12 under "Appendix"). TTX and MeSer 1 -TTX were found to have exactly the same affinity for the first site (K d1 ϭ 0.038 M) and also for the second site (K d2 ϭ 39 M for TTX, K d2 ϭ 41 M for MeSer 1 -TTX). The only difference between the two toxins was the ATPase activity of the complex having its two sites occupied: 220% of the control in the case of TTX but only 27% in the case of MeSer 1 -TTX.
Effect of L-MeSer 1 -TTX in the Presence of TTX at Reactivating Concentrations-To know whether the activity of these ternary complexes is governed by the nature of the toxin bound on the low affinity site, on the high affinity site, or on both sites, we have carried out the following experiment. MeSer 1 -TTX at various concentrations was first mixed with TTX at a constant concentration (30 M) in the reaction medium, then CF 1 -⑀ was added. After incubation, MgATP was added and the ATPase activity measured. Fig. 3 (closed squares) shows the continuous decrease of the rate of ATP hydrolysis caused by the addition of increasing concentrations of MeSer 1 -TTX. The reaction rate actually depends on the proportions and on the catalytic activities of the following four ternary complexes (see "Appendix"): that bearing two molecules of TTX (ET 1 T 2 ), that bearing two molecules of MeSer 1 -TTX (EX 1 X 2 ), that bearing TTX on the tight site and MeSer 1 -TTX on the loose site (ET 1 X 2 ), and that bearing MeSer 1 -TTX on the tight site and TTX on the loose site (EX 1 T 2 ). The activities of ET 1 T 2 and EX 1 X 2 (respectively, 220 and 27% of the control) were known from the data of Fig. 2 fitted with Equation 12 under "Appendix" as well as the dissociation constants K d1 (0.038 M) and K d2 (40 M), which are identical for the two toxins. This allowed us to fit the activity versus concentration data points of Fig. 3 with the model developed in Equation 27 under "Appendix." Since the concentrations of ET 1 X 2 and EX 1 T 2 are always identical because of the identical affinities of TTX and MeSer 1 -TTX for the enzyme, only the average of the activities of these two ternary complexes can be determined. It was experimentally found to be 85% of the control. If the catalytic activity was governed only by the nature of toxin bound on the loose site, ET 1 X 2 should have the same activity as EX 1 X 2 , and EX 1 T 2 should have the same activity as ET 1 T 2 . Thus, their average activity would equal 125% of the control rather than 85% (the same result should be found if the activity of the ternary complex depended only on the nature of the toxin bound on the first site, i.e. if EX 1 X 2 and ET 1 T 2 had, respectively, the same activity as EX 1 T 2 and ET 1 X 2 ). This is depicted by the dashed curve in Fig. 3, which represents the theoretical relationship between the rate of ATP hydrolysis and the MeSer 1 -TTX concentration with the average activity of ET 1 X 2 and EX 1 T 2 set at 125%: it is clearly not superimposable onto the experimental data.
Inhibitory and Reactivating Effects of Different Synthetic Analogues of TTX-To discriminate better the functional role of the low and high affinity TTX binding sites, we have used a another set of molecules differing from TTX by a single residue. The effect of these analogues on ATPase activity is shown in Fig. 4. All of these compounds inhibited the ATPase but only at higher concentrations compared with TTX. The reactivation was not observed in the case of toxins modified in position 1 (Fig. 4a). Its level was not significant in the case of Lys 2 -TTX and was moderate in the case of Lys(Z) 2 -TTX (40% of the control, Fig. 4b). Reactivation was marked in the case of Tyr 3 -TTX (about 75%) and maximal in the case of Tyr(Me) 3 -TTX (about 290%, Fig. 4c).
For reactivating toxins (TTX, MeSer 1 -TTX, Lys(Z) 2 -TTX, Tyr 3 -TTX, and Tyr(Me) 3 -TTX), the data were fitted satisfactorily using the same two-sites model, with the assumption that the complex bearing a single molecule of toxin was fully inactive (V 1 set to zero, see Equation 12 under "Appendix"). When no reactivation occurred (MeSer(Bn) 1 -TTX, MeGlu 1 -TTX, and (MeGlu(tBu) 1 -TTX), a simpler model, involving only one binding site (Equation 5 under "Appendix"), could fit easily the data. In this latter model, to account for possible incomplete inhibition, the enzyme-toxin complex was allowed to have an activity different from zero, which was derived from the fit. In the case of Lys 2 -TTX, the two models (two sites and single site) have been used to fit the data because this toxin did not reactivate the ATPase significantly at high concentration but was nevertheless shown to bind to the reactivating site (see below). In all cases, the dissociation constant for the inhibitory site (K d1 ) could be determined. The values are summarized in Table I, first column.
Effect of L-Lys 2 -TTX, a Nonreactivating Compound, in the Presence of TTX at Reactivating Concentrations-The question is now to know whether the absence of reactivation by MeSer(Bn) 1 -TTX, MeGlu 1 -TTX, MeGlu(tBu) 1 -TTX, and Lys 2 -TTX results from a default of binding or a nonproductive binding on the low affinity site. The experiments carried out to address this question were based on the same principle as that used in Fig. 3 for the poorly reactivating analogue MeSer 1 -TTX. The effect of high concentrations of nonreactivating analogues on the ATPase activity was checked in the presence of reactivating concentrations of TTX. Fig. 5 shows the effect of Lys 2 -TTX on the ATPase activity in the presence of three different concentrations of TTX in the reactivating range. The data show that the addition of Lys 2 -TTX decreases the activity, by limiting the reactivation by TTX, and that the concentration of Lys 2 -TTX needed to prevent the enzyme reactivation increases with the concentration of TTX (compare the three curves of inhibition). This effect is consistent with a competition on the loose site. To determine the binding parameters of Lys 2 -TTX and the activity of the ternary complexes, we have fitted the data of Fig. 5 with the two-sites model described above (see also Equation 26 under "Appendix"). In this model, the enzyme can exist under the following states: E (without toxin), ET 1 (with TTX bound at the high affinity site), ET 1 T 2 (with TTX bound at both sites), ET 1 X 2 (with TTX bound at the high affinity site and Lys 2 -TTX bound at the low affinity site), and EX 1 X 2 (with Lys 2 -TTX bound at both sites). These states are the only ones present at significant levels, when one considers the values of K d1 for TTX and for Lys 2 -TTX. Fitting of the competition data allowed a refined of the "Appendix" (competition between two toxins of identical K d1 and K d2 ). Dashed curve, theoretical values of the rate V calculated from the same equation but assuming that V TX ϩ V XT ϭ 2.6. For details, see "Results." determination of the binding constant of Lys 2 -TTX for the high affinity site and the estimation of its binding constant for the low affinity site. The corresponding values of K d1 and K d2 have been found close to 2 M and 55 M, respectively. We also determined the ATPase activities of the ternary complexes EX 1 X 2 and ET 1 X 2 (called, respectively, V XX and V TX ). There were found almost negligible (about 5% of the control).
Effect of L-Lys(Z) 2 -TTX in the Presence of TTX at Inhibitory Concentrations-Contrary to the enzyme bearing two mole-cules of Lys 2 -TTX, the enzyme bearing two molecules of Lys(Z) 2 -TTX (EX 1 X 2 complex) exhibited a significant ATPase activity, about 30% (Table I, last column). The affinity of Lys(Z) 2 -TTX for the high affinity site is low compared with TTX (K d1 ϭ 1 M instead of 0.04 M), but its affinity for the low affinity site is somewhat better than that of TTX (K d2 ϭ 11 M versus 39 M). This means that, starting from the complex where the high affinity site is occupied by TTX, one can fill specifically the low affinity site with Lys(Z) 2 -TTX, forming the ET 1 X 2 complex. This was achieved by adding various concentrations of Lys(Z) 2 -TTX to an enzyme sample already containing TTX at a micromolar concentration. Fig. 6 shows the ATPase activity measured in such conditions. Two different concentrations of TTX were used. The addition of Lys(Z) 2 -TTX to CF 1 -⑀ inhibited by TTX did not restore any significant activity. The theoretical curve displayed on Fig. 6, drawn using the same activity for ET 1 X 2 and EX 1 X 2 , i.e. 30% of the control, does not fit the experimental data. The activity of the ET 1 X 2 complex can indeed be estimated to be less than 10% of the control,  Table I. The fitted curve obtained with TTX in Fig. 2 was redrawn (dots) in panels a, b, and c for comparison.  Figs. 2 and 4 for all toxins and also from the data of Fig. 5 for Lys 2 -TTX, second line. K d1 ; dissociation constant from the high affinity site; K d2 , dissociation constant from the low affinity site; V I , relative activity of the complex bearing a single molecule of toxin (assumed to be zero in the case of a two-sites model); V A , relative activity of the complex bearing two molecules of toxin. The equations used were: Ϫ1 for the two-sites model, where V is the activity normalized to that of the nontreated enzyme. See "Appendix" for details.  Effect of Different Nonreactivating Toxins in the Presence of TTX at Reactivating Concentrations-At high concentrations, MeSer(Bn) 1 -TTX, MeGlu 1 -TTX, and MeGlu(tBu) 1 -TTX did not reactivate the ATPase activity of CF 1 -⑀. By competition experiments similar to those carried out with Lys 2 -TTX and depicted above, we have obtained the following results (data not shown): (i) MeGlu 1 -TTX and MeGlu(tBu) 1 -TTX did not compete at all with TTX for binding at the low affinity site; (ii) MeSer(Bn) 1 -TTX competed efficiently with TTX, with a K d2 comparable to that of the natural toxin, and its binding to the low affinity site gave rise to a fully inactive complex.

DISCUSSION
The Two TTX Binding Sites Are Related to Inhibition and Reactivation of CF 1 -ATPase Activity-In a preceding study (27), we detected the presence of two TTX binding sites on isolated CF 1 and CF 1 -⑀, using synthetic 14 C-labeled TTX, and we related these two sites to the TTX properties of inhibition of the CF 1 -ATPase activity below 0.1 M and reactivation above 5-10 M. The present study, using TTX synthetic analogues, fully confirms the existence of these two independent TTX binding sites and their direct connection to the inhibition and stimulation processes. More, the binding parameters have been determined carefully using the two-sites formalism. For TTX, the value of K d2 derived from kinetic experiments (40 M) is consistent with that derived from equilibrium dialysis experiments (70 M). The K d1 values (0.04 M from kinetic experiments and 1 M from 14 C binding studies) are different, but this can be explained by the low temperature (4°C) at which the dialysis experiments were carried out. It has to be noticed that, although the two TTX properties can be discriminated with some analogues, the reactivation process was never observed solely, i.e. without a prior inhibition resulting from the active binding to the first site.
TTX Does Not Aggregate at High Concentrations-The twosites formalism used here is relevant only if the formation of some toxin aggregates does not compete with the binding of toxin to the enzyme. Earlier NMR experiments suggested a possible self-association of TTX in aqueous solution, at 3.5 mM, at Ϫ5°C, and in the presence of salts (32). However, only a part of the data was consistent with this view: the modification of the chemical shift of the leucyl proton and the decrease of the transverse relaxation time T 2 of given backbone protons. The first change could as well be ascribed to an intramolecular Leu-⌬Phe stacking, especially since we failed to observe other effects expected from intermolecular dipole-dipole interactions (nuclear Overhauser effects). The T 2 decrease can hardly be attributed univocally to self-association of TTX molecules. It could indeed be biased by chemical exchange (maybe not fully abolished in the NMR experimental conditions), or it may reflect more complex hydrodynamic properties than those generally assumed for small molecules. Time-resolved fluorescence data were also supposed to reveal the formation of tentoxin aggregates, in the 10 M range (32), on the basis of a heterogeneous decay of fluorescence, but the two lifetimes thus determined (1.8 and 0.7 ns) were actually too close and the method not sensitive enough to distinguish effectively between a heterogeneous and a homogeneous decay. To overcome the ambiguities of past studies, we carried out further investigations using different spectroscopic techniques expected to detect stable aggregates via modifications of the conformation of peptide bonds (CD) or of the stretching vibrational modes of carbonyl groups (FTIR). No change in CD spectra of TTX in aqueous solution was observed in the 5-1,000 M range at temperatures between 5 and 37°C (not shown). FTIR difference spectra of TTX were recorded at 12°C in deuterated water solutions. Likewise, there were no variations observed in the absorption domain of the stretching vibrational modes of carbonyl groups (1,656 and 1,636 cm Ϫ1 ) in the 100 -4,000 M TTX concentration range (not shown). The stretching mode of CϭO groups is indeed sensitive to hydrogen bonding, and if intermolecular associations took place or were stabilized via the formation of hydrogen bonds, they would produce a detectable shift of the wave number. These data do not support the existence of a micellar equilibrium involving TTX molecules in the concentration range where we observed the reactivation (10 -1,000 M). Finally, definitive evidence for the absence of aggregates at high concentrations of TTX in our experimental conditions came from simple dialysis experiments using 14 Clabeled TTX. These experiments were carried out in the same medium as that used for ATP hydrolysis activity measurements at 35°C. The use of 14 C-labeled TTX allowed the detection of very low concentrations of the toxin in the dialysis buffer. The initial diffusion rate of 14 C-labeled TTX through the dialysis tube, chosen with a cut-off at 1 kDa (experimentally checked with molecules of different sizes), was strictly proportional to its internal concentration ranging from 2 to 500 M. This demonstrates that the molecular mass of the diffusing species was constant and always lower than 1 kDa. The molecular mass of TTX being 0.414 kDa rules out the possibility of concentration-dependent aggregate formation. The concentration of available monomers of TTX is identical to the total concentration, which validates the enzymologic approach used in this work.
Effects of Modifications of the TTX Molecule on Its Inhibitory Properties-All the analogues of TTX studied in the present work have retained their inhibitory properties. In all cases, the ATPase activity of CF 1 -⑀ was completely lost when a single molecule of toxin was bound. The only effect of the substitutions was to decrease to various extents the affinity of the molecule for the tight site, moderately for certain modifications (MeAla 1 3 MeSer 1 , Phe 3 3 Tyr 3 , Phe 3 3 Tyr(Me) 3 ) and more drastically (up to 200 times) for others (in the increasing K d1 order: MeAla 1 3 MeSer(Bn) 1 , Leu 2 3 Lys(Z) 2 , MeAla 1 3 Me-Glu(tBu) 1 , Leu 2 3 Lys 2 , MeAla 1 3 MeGlu 1 ). Different molecular factors can account for this K d1 increase, such as the introduction of longer chains, giving rise to additional steric hindrance (MeAla 1 3 MeSer(Bn) 1 , MeAla 1 3 MeGlu(tBu) 1 , Leu 2 3 Lys(Z) 2 ), or the introduction of a net electrical charge on the molecule (MeAla 1 3 MeGlu 1 , Leu 2 3 Lys 2 ). The introduction of charges seems to be more determining, since the replacement of a charged residue by a neutral residue, even larger (MeGlu 1 3 MeGlu(tBu) 1 , Lys 2 3 Lys(Z) 2 ), led to some recovery of the affinity.
A first conclusion of our work is that the nature of residues 1 and 2 (N-MeAla and Leu in the natural molecule) is not so critical for the inhibitory power of the molecule once it is bound to its site. This can be related to previous structural results obtained by NMR (30,32). Indeed, TTX and MeSer 1 -TTX exhibited the same conformation of the cyclic backbone (cis-transcis-trans configuration of the amide bond sequence) and the same interconversion among four conformers. These structural properties have also been observed for all of the analogues of the residue 1. 2 The conservation of the conformational features of the molecules can account for the ability of the molecule to inhibit the enzyme (V I ϭ 0) totally, and the variations of side chain can account for the changes in the affinity for the tight site. However, changes in residue 3 did not result in an increase in K d1 . In fact, these modifications were confined to a substitution of the para proton of the benzyl group, without any consequence on the rigid configuration of the double bond of the ␣,␤-dehydroamino acid. Despite the introduction of polar groups on this side chain, there was no repercussion on the affinity of the molecule for the tight site. This result is not so unexpected if the binding of the molecule inside the hydrophobic site involves a stabilization by an aromatic ring stacking.
Effects of Modifications of the TTX Molecule on Its Activating Properties-Whereas the only change in the inhibitory properties of toxin derivatives is a variation of their affinity for the tight site, the situation is quite different with regard to reactivating properties. Noteworthy is that with the exception of MeGlu 1 -TTX and MeGlu(tBu) 1 -TTX, all the analogues were found to bind to the low affinity site with a K d2 comparable to that of natural TTX, whether they were reactivating or not. The differences lie in the effect of the molecule once bound to this site. It is possible that the loose site has a more open configuration than the tight one, which makes it less sensitive to steric hindrance variations and also to electrical charges (that can be shielded by water molecules). The exceptions of MeGlu 1 -TTX and MeGlu(tBu) 1 -TTX, however, remain to be explained.
In the absence of a structural model of the chloroplast ATPase species, which differs from the mitochondrial species in its sensitivity to TTX and in various structural features (notably regarding the ␥ subunit), it is still difficult to explain the various levels of reactivation obtained with the different analogues bound on the two sites. For the complexes bearing two molecules of the same toxin (homogeneous complexes), a change of hydrophobicity of the molecule (30) can result in a significant change in the reactivation level. Thus, the replacement of a proton by a hydroxyl group (Ala 3 Ser; Phe 3 Tyr) dramatically decreases the activity of the ternary complex when made in position 1 and significantly when made in position 3. In the latter case, this activity is restored when the labile proton is replaced again by a more hydrophobic group (Tyr 3 Tyr(Me)). The comparison of the reactivating properties of MeSer 1 -TTX and Tyr 3 -TTX, two molecules slightly modified with unchanged affinities for both sites, suggests that the nature of residue 1 is more important than that of residue 3 in conferring to TTX its reactivating properties. But also, as discussed below, the level of reactivation was found to be dependent on the combination of the two toxins bound (hybrid ternary complexes).
Importance of the Two TTX Binding Sites in the Reactivation-It is tempting to speculate whether the two TTX binding sites are homologous domains of two different ␣␤ pairs, put into different states by the asymmetry of the ATPase complex. Such a situation has already been stated in the case of the binding of two molecules of aurovertin to the bovine heart MF 1 complex (33), with the noteworthy difference that aurovertin is never reactivating. If the two binding sites of TTX were located on different ␣␤ pairs, the TTX molecule bound to the first ␣␤ pair could block the enzyme, and the TTX molecule bound to the second pair could unlock it.
An original result of this work deals with hybrid ternary complexes, bearing natural TTX on the tight site and a synthetic derivative on the loose site. By applying a simple model, with two independent binding sites, we could estimate the activities of some of these hybrid complexes and compare them with the activities of ternary complexes bearing the same molecule on both sites (Table II). It is not possible, at the present time, to understand the molecular rules that would determine the activity of all of these ternary complexes, but there are two main points. First, the activity depends on the toxin bound on the loose site; compare, for example, the CF 1 ⅐TTX⅐TTX complex (220% activity) with the CF 1 ⅐TTX⅐Lys(Z) 2 -TTX and CF 1 ⅐TTX⅐Lys 2 -TTX complexes (negligible activities). Second, the activity also seems to depend on the toxin bound on the tight site; compare the CF 1 ⅐TTX⅐Lys(Z) 2 -TTX complex (negligible activity) with the CF 1 ⅐Lys(Z) 2 -TTX⅐Lys(Z) 2 -TTX complex (30% activity), and see also the competition between TTX and MeSer 1 -TTX. So, although the CF 1 -ATPase enzymes complexed with a single molecule of TTX or with a single molecule of an analogue cannot be discriminated on the basis of their activity, they are potentially different. This difference is revealed only when a second molecule of the analogue is bound on the loose site. All of our data are consistent with an interaction between the two TTX binding sites. 2 F. André, unpublished results.

TABLE II
ATPase activity of ternary complexes CF 1Ϫ⑀⅐ toxin toxin with respect to the activity of nontreated CF 1Ϫ⑀ Conditions are as described under "Materials and Methods." The first two columns indicate which toxins are bound to the high and low affinity sites. Activities of homologous complexes, bearing two molecules of the same toxin, were calculated from the magnitude of the maximal reactivation (V A in Table I, see also Figs. 2 and 4). Activities of heterologous complexes, containing one molecule of TTX and one molecule of an analogue, were estimated from competition experiments (Figs. 3, 5, and 6). In the case of MeSer 1 -TTX, the two different heterologous complexes cannot be discriminated, and only the average of their activities is given. For details, see "Results." Cooperative Binding, an Alternative Hypothesis-To fit our data, we have considered only a simple model with two preexisting and absolute K d values, which assumes that binding of a first toxin molecule to the complex has no influence on the affinity of the complex for a second molecule. The activities of ternary complexes are then the only way to detect interactions between the two sites. However, it would be also possible to fit them with a model of cooperative binding, and in this case the K d2 of a given toxin would depend on the toxin bound on the tight site. For example, data of Fig. 6 could be fitted satisfactorily with a model assuming that all of the ternary complexes have a negligible activity, provided Lys(Z) 2 -TTX binds to the loose site with a K d2 equal to 100 M instead of 11 M, the value drawn from Fig. 4. This means that replacing Lys(Z) 2 -TTX by TTX at the tight site would dramatically decrease the affinity of Lys(Z) 2 -TTX for the loose site. Even though the mechanism differs from that assumed in our first model, this would demonstrate even more directly the interaction between the two binding sites of the toxin. Understanding this interaction should be an important element in the knowledge of the catalytic mechanism of CF 0 CF 1 -ATP synthase. If v 0 and v I are the rates of the reaction catalyzed by the forms E and ET, respectively, the total reaction rate is The rate V normalized to the control reaction rate v 0 is (Eq. 5) with V I ϭ v I /v 0 . In the experimental plots, the free concentration of toxin, [T], will be identified to the total concentration of toxin [T]ϩ[ET], since the toxin is always in large excess under our conditions. This approximation also applies to the two-sites model.
The Two-sites Model-Let us consider the binding equilibria where two molecules of toxin are successively bound to sites called 1 and 2 (Scheme 1, where ET 1 refers to the complex with one molecule of toxin bound to the high affinity site (K d1 ), ET 2 to the complex with one molecule of toxin bound to the low affinity site (K d2 ), and ET 1 T 2 to the complex with the toxin bound to the two sites. One has and ͓E t ͔ ϭ ͓E͔ ϩ ͓ET 1 ͔ ϩ ͓ET 2 ͔ ϩ ͓ET 1 T 2 ͔. (Eq. 8) Because experimentally K d1 Ͻ Ͻ K d2 , one can consider that [ET 2 ] is negligible. The relative concentrations of the different states are therefore and (Eq. 11) To simplify, one assumes that binding of a single toxin molecule fully inhibits the enzyme, i.e. the activity of the state ET 1 is 0. The reaction rate V (with toxin) normalized to that of the control (without toxin) becomes where V A is the ratio between the catalytic activities of the states ET 1 T 2 and E. Competition between Two Different Toxins at the Two Binding Sites-Let us consider the same equilibria where two different toxins (T and X) are bound. One has By convention, T refers to TTX and X to an analogue. Indexes 1 and 2 still refer to the high and low affinity sites, respectively. KЈ d1 and KЈ d2 are the dissociation constants corresponding to toxin X. The concentrations of all of the possible complexes are given by Tentoxin Binding Sites in Chloroplast F 1 -ATPase