The chloroplast H⁺-ATPase: partial reactions of the proton

The proton side of F-ATPases appears to be even more difficult to assess than the nucleotide processing side. The proton-motive force not only drives ATP synthesis but also regulates enzyme activity (Junge et al. 1970; Bakker-Grunwald and van Dam, 1974; Junesch and Gräber, 1985, 1991). Three protons are translocated per molecule of ATP formed, whether concertedly or each one serving a particular function is unknown. The equivalence of the electrical portion of the proton-motive force, Δ Ψ, and the chemical portion, ΔpH, as driving forces for ATP synthesis is plausible for near-equilibrium situations but not at all trivial far from equilibrium as in kinetic experiments, where it is also experimentally established (Junesch and Gräber, 1991). Specific amino acid residues which are essential for proton transfer have been identified by mutational analysis (Fraga and Fillingame, 1991; Miller et al. 1990) but evidence for their direct involvement is lacking.

Among F-ATPases those in photosynthetic membranes offer themselves for detailed kinetic studies on reactions involving the proton. The proton pumps are driven by light flashes, and transients of the transmembrane voltage and of the pH on both sides of the coupling membrane are detectable spectroscopically by intrinsic and extrinsic indicator dyes (Witt, 1979; Junge, 1982). When chloroplasts of green plants are excited by light the lumen of inner vesicles, thylakoids, is positively charged and acidified. The lumen is equivalent to the exoplasmatic side of unicellular organisms. The basic proton conductance of the thylakoid membrane is rather low. Proton conduction is enhanced when CF₁F₀ synthesizes ATP or more dramatically when CF₀ is exposed by removal of the catalytic portion CF₁. Under pulse stimulation by short flashes of light, proton intake from the lumen, charge transfer across the dielectric barrier and proton release into the suspending medium can be monitored, both for proton flow related to the ATP-synthesizing CF₁F₀ (Junge, 1987) and for proton flow through the exposed channel portion CF₀ (Schönknecht et al. 1986; Lili et al. 1987; Althoff et al. 1989). In both cases the intake of protons from the lumen and the transport of their charge equivalent across the membrane appear to be synchronized. A typical spectrophotometric experiment deals with more than 10¹³ molecules of CF₁F₀ or CF₀ which may operate independently and in an uncorrelated manner even when subjected to a sudden step of the electrochemical driving force. Thus, partial reactions of the proton could be blurred over. They may become apparent, however, if some reactions are frozen and completion of the transport cycle is blocked.

This article first reviews physicochemical properties of proton conduction by CF₁F₀ and by CF₀ and then reports new results on proton binding by CF₁F₀ without concomitant proton transfer across the membrane.

The turnover rate of CF₁F₀ for protons has been inferred from measurements of the rate of ATP synthesis assuming a stoichiometry of 3H⁺/ATP. Values of up to 400 mol ATP mol⁻¹ CF₁F_os⁻¹ have been found in CF₁F_o-reconstituted liposomes during the initial phase after a large pH jump (Schmidt and Gräber, 1987). Even higher figures (500 s⁻¹) can be observed with spinach thylakoids under continuous light and with phenazine methosulfate as cofactor for cyclic electron transport. This result implies maximum turnover numbers for protons of up to 1500 s⁻¹. It is noteworthy, that reported rates for F₁F₀ from all other sources are lower.

Fig. 1 documents proton flow through CF₁F₀ under flashing light, where the turnover number is smaller than in the work cited above because of a smaller proton-motive force. The figure is taken from earlier work (Junge, 1987). Fig. 1A shows transients of the transmembrane voltage and Fig. 1B of the lumenal pH. They are elicited by excitation of thylakoids with two groups of three closely timed flashes of light. In the presence of a specific inhibitor of CF₁, tentoxin, the relaxation of both is slow. The transmembrane voltage (Fig. 1A) decays more rapidly than the acidification of the thylakoid lumen (Fig. IB). This result reflects an electrical leak conductance by ions other than the proton. In the absence of tentoxin and with substrates (ADP and inorganic phosphate) present, ATP synthesis causes an accelerated decay of both the transmembrane voltage and the lumenal acidification. The difference between the two traces in Fig. 1A indicates the extra transmembrane charge flow that accompanies ATP synthesis. The difference between the two traces in Fig. 1B indicates extra proton intake from the lumen by CF₀. Both differences have been normalized and, for a short time segment, plotted on top of each other in Fig. 1C. Their coincidence within noise limits identifies the extra charges which pass the dielectric barrier in CF₁F₀ as protons taken up from the lumen (Junge, 1987). Proton intake and charge transfer appear to be synchronized. The accelerated charge flow stops at a certain driving force and this is due to the gating of enzyme activity by the transmembrane voltage (Junge et al. 1970; Junge, 1970) or, more generally, by the transmembrane proton-motive force (Junesch and Gräber, 1985, 1991).

For a while, published figures of the proton conductance of the channel portion, F_o, have fallen short of the above turnover numbers. Values ranging from 10 to 100 s⁻¹ have been obtained mainly with reconstituted archaebacterial and eubacterial channels in liposomes and by pH electrode (Negrin et al. 1980; Sone et al. 1981; Schneider and Altendorf, 1987). These low values are incompatible with the proposed role of F₀ as a low-impedance access to the coupling site in the enzyme (Mitchell, 1966). One reason for these shortcomings might be an insufficient time resolution of the pH electrode (G. Althoff and G. Dekkers-Hebestreit, personal communication), another reason might be the survival of only a few active channels during reconstitution.

We investigated the protonic conduction of CF₀ by flash spectrophotometry. The catalytic portion, CF₁, was removed from thylakoid membranes by treatment with EDTA. The number of CF₁ molecules removed from the membrane was determined by electroimmunodiffusion. Thylakoids were excited by short light flashes and the relaxation of the transmembrane voltage and of the pH transients in the lumen and in the suspending medium was monitored as shown in Fig. 1 (Althoff et al. 1989; Lill et al. 1987; Schönknecht et al. 1986). To label charge flow mediated by CF₀, signals were recorded pairwise, with and without blocking agent (DCCD, venturicidin or organotins; Linnett and Beechey, 1979). The area-specific electrical capacitance of thylakoid membranes and the specific number of CF₁ molecules removed were taken into account (Althoff et al. 1989). Assuming that any CF₁ which was removed leaves a conducting CF₀ behind, the average protonic conductance per exposed CF₀ was about 10 fS. The average was taken over two domains, the ensemble of those CF₀ that had lost their CF₁ counterpart and the time (counting open and closed intervals of channels). At moderate driving force, 100 mV, this conductance is equivalent to a turnover number for protons of about 6000 s⁻¹. Under 200 mV driving force the turnover number is 12000 s⁻¹, ohmic behaviour assumed. This number is fully compatible with the proposed role of CF₀ as the proton channel of the enzyme. As a time-averaged figure, it represents only a lower limit for the turnover number of active channels.

We attempted to determine the conductance of the subset of conducting CF₀ channels. The rationale was to use vesicles so small as to contain 0, 1, 2 or very few active channels only and to determine the time-averaged conductance of single channels by a statistical analysis of electrochromic absorption transients. Thylakoids were fragmented into very small vesicles and the analysis technique was calibrated with gramicidin as a channel with known properties. Its conductance in thylakoids (Lili et al. 1987; Althoff et al. 1991 ; Schönknecht et al. 1992) was in the range of values published for its conductance in artificial bimolecular lipid membranes. The same type of analysis applied to CF₀ (without gramicidin present) led to the conclusion that only one out of 100 exposed CF₀ molecules per vesicle was conducting. Accordingly, the proton conductance of the few functional channels was much greater than the average conductance, namely 1 pS instead of 10 fS (Schônknecht et al. 1986). This result was in line with another report of a very large proton conductance of CF₀, namely of 3.5 pS, which was obtained for CF₁F₀ by the (dip stick) patch-clamp technique (Wagner et al. 1989). We did not find any clear-cut mistake in our spectrophotometric analysis, particularly as it worked well with gramicidin. Still, the result is not compatible with theoretical and model studies. A proton conductance of 1 pS, particularly around neutral pH, exceeds by orders of magnitude (a) the calculated convergence conductance to a pore mouth of reasonable diameter (say 1 nm) (Peskoff and Bers, 1988), (b) calculated rates of net proton transfer through short hydrogen-bonded chains (Brünger et al. 1983), and (c) measured rates of proton transfer through other proton-specific channels (Veatch et al. 1975; Lear et al. 1988). Charge flow enhancing factors, a short selectivity filter and large channel mouth, as discussed in the context of K⁺ maxi channels (Hille, 1984), the drag force in the coulomb cage (Peskoff and Bers, 1988) and mobile buffers (Decker and Levitt, 1988; Nunogaki and Kasai, 1988) do not really account for this discrepancy.

Other properties of this channel as determined by flash spectrophotometry (Althoff et al. 1989) are independent of the absolute value of the conductance. CF₀ is proton-specific even at pH 8 against a background of 300 mmol l⁻¹ NaCl or KC1, or 30 mmol l⁻¹ MgCh in the medium. The selectivity filter is therefore a property of CF₀ and not of the ratelimiting (coupling) step in CF₁F₀. The proton specificity of over more than seven orders of magnitude is unparalleled by other known proton channels (e.g. gramicidin, selectivity for protons over other cations less than 100; Decker and Levitt, 1988). It is the more astounding, as a homologous enzyme to CF₁F₀ has been described for Propionigenium modestum that can act both as a sodium-and as a proton-translocating ATPase (Dimroth, 1991). The conductance of CF₀ is pH independent in the range between 5.6 and 8. In the same pH range there is a constant hydrogen/deuterium isotope effect of 1.7. Addition of glycerol decreases the conductance and abolishes the isotope effect. This finding suggests that the isotope effect may be caused by events in the channel and that the channel operates close to limitation by events in the water phase. The Arrhenius activation energy of proton conduction by CF₀ is 42 kJ mol⁻¹, intermediate between the respective figures of a channel-type (30 kJ mol⁻¹, gramicidin) and a carrier-type antibiotic (65 kJ mol⁻¹, valinomycin) (see Althoff et al. 1989, for details).

In conclusion, CF₀ is a very effective and selective proton channel. Its turnover number greatly exceeds that of the coupled enzyme. It can act as a low-impedance access channel to the coupling site. CF₀ is endowed with an extremely specific selectivity filter for protons, unparalleled by other types of proton channels.

Three subunits of F₁, namely, γ, δ and ϵ, of the thermophylic bacterium PS3 were reported to block proton flow through the respective F_o channel (Kagawa, 1978). We asked whether single subunits of CFi can interact with exposed CF_O in such a way as to block proton flow. Triggered by the observation that δ may remain on CF₀ after removal of CF₁, keeping CF₀ non-conducting (Junge et al. 1984), we found that isolated δ, when added back to CF₁-depleted thylakoids, can throttle proton flow through exposed CF₀ (Lili et al. 1988). Thereby, it restores photophosphorylation by those CF₁F₀ that have remained intact on the membrane (Engelbrecht and Junge, 1988). Any ‘stopcock’ action of δ has to be overcome in the operating enzyme in favour of a controlled operation as valve, admitting protons from the channel further onto the coupling site, or as part of the conformational transducer between protons and ATP (Engelbrecht and Junge, 1990).

Following the same line of research we also found subunit β to be effective in throttling proton conduction by CF₀ (monitored by electrochromic absorption transients), and restoring photophosphorylation in CF₁-depleted thylakoids (S. Engelbrecht and W. Junge, in preparation). In this series of experiments, subunit a could not be purified in sufficient amounts. The particular samples of γ, δ and ϵ were only marginally effective by themselves, but the combination of γ, δ and ϵ was, as if distorted δ was reshaped by its association with γ and ϵ (see also Engelbrecht and Junge, 1990). Interestingly, the combination of γ, δ and ϵ was equally effective on spinach CF₀, when subunit γ was taken either from spinach or from the thermophilic bacterium PS3. The finding that the small subunits of CF₁ act in concerted manner and that subunit β by itself specifically interacted with the exposed proton channel, CF₀, qualifies the concept of δ just acting as a plug.

Instead, it seems as if there are several essential contacts between CF₀ and CFi. Following the hypothesis of a rotational mechanism of catalysis (Boyer, 1989) one may speculate that both portions of F₁, bearings (β?) and rotating shaft elements (at least γ and δ?), may interact with CF₀ so as to block and, in CF₁F₀, so as to control, proton conduction.

Cooperative transient trapping of protons was first observed in thylakoids treated with very low concentrations of EDTA (Junge et al. 1984). This effect is documented in Fig. 2. Thylakoids isolated from spinach chloroplasts according to standard procedures, but without added magnesium in the medium, were incubated for 2 min in distilled water with very low concentrations of EDTA added (typically 10 μmol l⁻¹ at 10 μmol l⁻¹ chlorophyll, but see below). The exposure to low ionic strength was stopped by addition of NaCl to 10 mmol l⁻¹.

Fig. 2 shows voltage transients (left) and pH transients (right) in the lumen for thylakoids that had undergone mild EDTA treatment. Although the extent and the decay of the voltage were unaffected by addition of venturicidin (blocker of F_o channels binding to the proteolipid, Galanis et al. 1989), the acidification of the lumen was seemingly increased. We attributed this increase to the abolition by venturicidin of a transient trapping of protons at the lumenal side of CF₁F₀ or of CF₀. The extent of proton intake is shown in the lower part of Fig. 2 (right). It is simply the difference between the two traces at the top. On an enlarged time scale (Fig. 3) proton intake appeared biphasic, rapid trapping preceding a slow phase attributable to proton leakage through a few exposed CF₀ channels. The pH dependence of proton trapping was remarkably steep (see Fig. 3). Whereas rapid trapping was absent at pH 7.2, it was fully expressed at pH 7.7. In contrast, the comparatively slow proton leakage was practically constant over this narrow pH range. The steep pH-dependence of proton trapping conformed with a hexacooperative binding isotherm (Hill coefficient around 6) (Griwatz and Junge, 1992; Junge et al. 1984). It is noteworthy that the proton-trapping capacity was saturated by the first flash of light; further flashes did not produce further trapping. It was inhibited by various inhibitors of proton conduction through CF₀, namely, venturicidin, DCCD and organotins, and thus was clearly attributable to the ATPase. Trapping was also eliminated by adding very small concentrations of divalent cations (typically 10 μmol l⁻¹), concentrations too small to produce rebinding of possibly solubilized CF|.

The pH difference which is generated by a single turnover of both photosystems is very small, about 0.05 units (Junge et al. 1979). What was the driving force for proton trapping? We found that rapid trapping could be eliminated if the flash-induced transmembrane voltage was even more rapidly shunted by an ionophore (e.g. valinomycin) (Griwatz and Junge, 1992). This finding led us to interpret these effects in the framework of Mitchell’s concept of a proton well (Mitchell, 1977). Consider a protonspecific domain that reaches from the lumenal surface of the membrane into CF₁F₀. The lumenal surface still represents an electrical equipotential surface, but the electrochemical equipotential surface stretches out into this proton-conducting domain. Potentially buffering groups at its top sense a pH that is more acid than the one at the lumen surface. The pH decrease is equivalent to the voltage drop between the lumen and the location of these groups in the membrane. This concept for the transformation of electrical into chemical (entropie) force could account for the observed kinetic equivalence of Δ Ψ and ΔpH for ATP synthesis (Gräber et al. 1984; Junesch and Gräber, 1987). Without a proton well the kinetic equivalence is not at all trivial.

Proton trapping was highly cooperative. In spinach thylakoids the pK of the trapping groups was 7.3 and the Hill coefficient around 6. This behaviour suggests the presence of at least six identical and strongly interacting groups. The candidates are two essential residues, Glu61 and Arg41 (see Fraga and Fillingame, 1991; Miller et al. 1990, for E. coli) on subunit III. Only this subunit of CF₀ is present in sufficient numbers (9–12 in E. coli). Derivatization of a single copy by DCCD is sufficient to block ATPase activity in E. coli (Hermolin and Fillingame, 1989).

Transient trapping of protons has only been observed with distorted CF|F_O. Is the trapping state an intermediate of the operational cycle of this enzyme? In the same type of flash spectrophotometric experiments described above trapping has not been detected during ATP synthesis by CFiF_o(Junge, 1987) (and see Fig. 1) or proton leakage through exposed CF₀ (Althoff et al. 1989). In both cases the respective transporter has carried out multiple turnovers. Each experiment was an average over very many enzyme cycles. Time resolution of partial reaction steps was not expected.

Thus, it is conceivable that mild EDTA treatment stabilizes and thereby exposes to observation one particular intermediate of the normal operating cycle of the ATP synthase. The respective residues are placed on the lumenal side, beyond the selectivity filter, but before the main dielectric barrier in CFiF₀. It is tempting to associate this barrier with the main coupling step in CF₁F₀.

Another situation where proton binding precedes conduction is related to proton slip through CF₁F₀. There are at least four mechanisms of proton conduction by thylakoid membranes: (1) coupled proton translocation by CF₁F₀ during ATP synthesis; (2) proton channeling through exposed CF₀; (3) proton slip through CF₁F₀; and (4) any other proton leakage. While the first two are self-explanatory, proton slip differs from proton leakage (through the bilayer or through proteins other than CF|F₀) in its sensitivity to nucleotides and/or specific blockers of CF₁. In contrast to the proton leak in the cristae membrane of mitochondria, which is insensitive to blockers of F₁F₀ (Brown and Brand, 1991), several authors have reported proton slip in thylakoid membranes (McCarty et al. 1971; Underwood and Gould, 1980; Gräber et al. 1981; Strotmann et al. 1986; Evron and Avron, 1990). It is particularly evident at very low nucleotide concentrations. Fig. 4 demonstrates this effect. Thylakoids were incubated in nucleotide-free medium and illuminated by a constant background intensity of green light. They were excited with a group of three short light flashes creating a voltage transient (Fig. 4A) and a pH transient (Fig. 4B) in the lumen on top of the steady acidification. The electrical leakage was greater in the absence of ADP than with ADP added (see Fig. 4A). The attribution of this slip to CF₁F₀ was corroborated by its inhibition by venturicidin. Fig. 4B demonstrates that the slip was due to proton conduction. There was apparently less proton release into the lumen in the absence of venturicidin. As in the context of Fig. 1, this diminution was attributed to increased proton intake. Interestingly, the addition of ADP eliminated charge slip (Fig. 4A) but not proton intake (Fig. 4B). Still venturicidin eliminated both. Proton trapping was more rapid than proton conduction. This discrepancy is evident from a comparison of the respective differences between the traces shown in Fig. 4, which are plotted in Fig. 5. It has still to be determined how this example of proton intake by CF₁F₀ without concomitant proton conduction relates to the result after EDTA treatment.

Other properties of the proton slip, as will be detailed elsewhere (G. Groth and W. Junge, in preparation), are as follows. The inhibition of the electrical slip by ADP occurred with an apparent K₁(ADP) of 200 nmol l⁻¹. It required inorganic phosphate; K_i(P_i) of 45 μmol l⁻¹. GDP was also effective. There was an energetic threshold for the slip. More importantly, when a pH difference was created by continuous light, such that the electrical potential difference induced by a single additional flash of light was still below threshhold, there was no slip. The slip became evident, however, when the electrical potential difference was increased (by firing three flashes instead on only one, see Fig. 4). This showed that the slip threshold was defined in terms of proton-motive force. The electrical component, ΔΨ, was additive to the chemical one, ΔpH.

A simplistic structural model based on the above kinetic studies on proteolytic reactions is given in Fig. 6. To avoid the impression that we can pin down any of the functional elements to certain protein domains, the enzyme is sketched as a ball and rod. For protons, the CF₁F₀ entity represents a structure with three elements, an entry channel, a major dielectric barrier and an exit channel. Protons enter CF₁F₀ through an extremely proton-selective filter, which is part of CF₀. On the cis side of the dielectric barrier, but after or within the filter, there are highly cooperative proton binding groups. They are freely accessible only in structurally distorted CF₁F₀ (induced by EDTA treatment or, perhaps also, at low nucleotide concentration). Then a small portion of the total transmembrane electrical potential difference is converted into a pH difference at these groups, as postulated in Mitchell’s (1977) concept of a proton well. It is tempting to identify the major dielectric barrier with the coupling step. Whether the proton trapping groups are involved in the reaction cycle of the undistorted ATP synthase remains to be seen. It is hoped that the above speculations are of heuristic value, at least until the atomic structure of CF|F₀ is available as a time-resolved movie.

This work was supported financially by the Deutsche Forschungsgemeinschaft (SFB171 –TPB2) and by Fonds der Chemischen Industrie.