Nucleotide-binding sites on Escherichia coli F1-ATPase. Specificity of noncatalytic sites and inhibition at catalytic sites by MgADP.

Nucleotide-depleted EcF1 binds a maximum of two GTP, ATP, or ADP at noncatalytic sites, whereas all three sites can only be filled by a combination of nucleoside di- and triphosphates. MgPPi prevents binding of GTP and significantly slows ATP binding, suggesting that non-catalytic sites also bind PPi. No binding of GDP at non-catalytic sites could be detected. The slow rate of GTP dissociation from noncatalytic sites (t1/2 = 175 min) is increased 2-8-fold by EDTA, MgPPi, MgADP, or EDTA/ATP, but 23-fold by conditions for ATP hydrolysis. ATP hydrolysis by EcF1, depleted of both its inhibitory epsilon-subunit and endogenous nucleotides, shows a burst of activity. However, it shows a lag if preincubated with MgADP but not when preincubated with Mg2+ alone. For epsilon-depleted EcF1 containing endogenous inhibitory ADP, preincubation with an ATP-regenerating system results in a burst of activity, whereas the control shows a lag. This same enzyme form shows significant inhibition with increasing concentrations of Mg2+ during ATP hydrolysis but lesser levels of inhibition when other NTP substrates are used. With the five-subunit enzyme, increasing amounts of azide cause an increase in the level of inhibition with a corresponding increase in amount of bound nucleotide resistant to rapid chase. Azide-trappable nucleotide is bound at catalytic sites as shown by covalent incorporation of 2-azido-ADP. The results suggest that ligand specificity may not be a reliable means of distinguishing between catalytic and noncatalytic sites and that MgADP inhibition should be taken into account in the kinetic analysis of EcF1 mutants.

F,-ATPase, the catalytic component of proton-translocating ATP-synthase complexes from eubacteria, mitochondria, and chloroplasts, has a subunit structure of cr,p,y& and a total of six nucleotide-binding sites. Three of the sites exchange bound nucleotide rapidly during catalytic turnover and are considered to be catalytic sites. The remaining three sites do not exchange rapidly and are referred to as noncatalytic sites (for reviews see Futai et al., 1989;Senior, 1990;Fillingame, 1990;Penefsky and Cross, 1991;Allison et al., 1992;Issartel et al., 1992;Boyer, 1993).
For mitochondrial and chloroplast F,, it has been shown that MgADP can bind at a catalytic site to form an inhibitory complex. This is evident as a lag in ATP hydrolysis when these enzymes are preincubated with ADP in the presence of M e (Vasilyeva et al., 1982a;Feldman et al., 1985). Preincubation with Mg2+ alone has no effect if endogenous nucleotides are previously removed (Drobinskaya et al., 1985). Azide prevents the subsequent reacti-* This work was supported by Research Grant GM 23152 from the National Institutes of Health, United States Public Health Service. The costs of publication of this article were defrayed in part by the payment tisernent" in accordance with 18 U.S.C. Section 1734 solely to indicate of page charges. This article must therefore be hereby marked "aduerthis fact.
vation of the enzyme, whereas anions such as sulfite activate the enzyme by facilitating release of the inhibitory MgADP (Vasilyeva et al., 1982b). In the dark, 2-azido-ADP gives inhibition with identical properties to that of MgADP. Photolysis of the inhibitory complex results in the covalent labeling of a catalytic-site peptide on CF,' (Zhou et al., 1988) and MF, (Milgrom and Boyer, 1990;Chernyak and Cross, 1992).
The F,-ATPase from Escherichia coli has been shown to bind Mg2' tightly (Senior et al., 1980) and to be inhibited when magnesium is added in excess ofATP (Kanazawa et al., 1980). It has been argued that this inhibition is not due to inhibitory MgADP bound at a catalytic site since, unlike the mitochondrial and chloroplast enzymes, no dependence on ADP could be demonstrated (Senior et al., 1992). Although potential support for the presence of catalytic site inhibitory MgADP comes from the observation that azide inhibits EcF,, this has been attributed to a direct effect of azide on subunit cooperativity (Noumi et al., 1987). Additional suggestive evidence is that sulfite activates the E. coli enzyme (Dunn et al., 1987). Factors that could mask an ADP-dependent lag include the presence of an inhibitory subunit, E, which can dissociate slowly upon dilution of EcF, into an ATP hydrolysis assay (Laget and Smith, 1979) and the presence of endogenous catalytic site ADP on the isolated enzyme.
Over the past 15 years, many studies of the nucleotide-binding sites on F,-ATPases have relied on ligand specificity to distinguish between catalytic and noncatalytic sites (Harris, 1978;Harris et al., 1978;Perlin et al., 1984). Catalytic sites are known to be nonspecific toward purine nucleotide substrates. They bind and hydrolyze GTP and ITP nearly as well as ATP. In contrast, noncatalytic sites have been considered to be highly specific for adenine nucleotides. However, recent results have shown that CF, (Guerrero et al., 1990;Milgrom et al., 1991) and MF, (Milgrom and Cross, 1993) can bind GTP at noncatalytic sites.
In this study we have reexamined the specificity of nucleotide sites and the question of inhibitory catalytic site-bound MgADP. The results indicate that both GTP and PP, can bind to noncatalytic sites on EcF, and that the enzyme is inhibited by binding MgADP at a catalytic site. [P,y3'P12-Azido-ATP was synthesized as described by Melese and Boyer (1985). Enzyme for Noncatalytic Site Specificity Studies-EcF, was isolated from E. coli strainAN1460 (Downie et al., 1980) by the method of Senior et al. (1979aSenior et al. ( , 1979b as modified by Wise et al. (1981). Purified EcF, was stored at -70 "C at a concentration of about 20 mg/ml in buffer containing 50 mM Tris-HC1, pH 7.4, 2 mM EDTA, 1 mMATP, 1 mM dithiothreitol, 40 mM E-aminocaproic acid, and 10% glycerol. Just before use, EcF, was diluted to a concentration of about 2 mg/ml with buffer containing 20 mM MOPS/Tris, pH 8.0, 150 mM sucrose, and 0.2 mM EDTA and passed through two sequential 1-ml Sephadex centrifuge columns (Penefsky, 1977) equilibrated with the same buffer. This form of the enzyme (referred to as native or EcF,[2,11) contained 2.8 mol of endogenous ANP/ mol, of which 2.0 mol/mol was not chaseable by GTP in agreement with Beharry and Bragg (1992). The nonchaseable nucleotides consisted of 0.9 mol of ADP and 1.1 mol of ATP. Nucleotide-depleted enzyme (EcF,[O,OI) was prepared by published procedures (Garrett and Penefsky, 1975;Senior et al., 1992). Two ammonium-sulfate precipitation steps preceded gel filtration. The enzyme was eluted from a 1.0 x 100-cm column of Sephadex G-50 (medium) at a flow rate of 0.6-1.2 mVh with 100 mM Tris-H,SO,, pH 8.0,4 mM EDTA, 100 mM Na,SO,, and 50% glycerol. Selected fractions containing EcF, were combined and concentrated using a Centricon-10 microconcentrator (Amicon). The enzyme was diluted 20-fold and reconcentrated twice in order to transfer it to storage buffer containing 100 mM Tris-H,SO,, pH 8.0, 4 mM EDTA, and 50% glycerol. EcF,[O,Ol was stored a t a concentration of about 20 mg/ml at -20 "C. The nucleotide content of EcF,[O,Ol was less than 0.2 mol/mol of enzyme, as determined by subjecting a neutralized perchloric acid extract to ion-exchange HPLC (Milgrom and Cross, 1993).

EXPERIMENTAL PROCEDURES
Enzyme for MgADP Inhibition Studies-EcF, was isolated from strain JP17 containing the wild-type plasmid pJWl and prepared to the solubilization stage as described by Senior et al. (1979a, 197913). Chromatographic purification employed a Productiv DE cartridge, and the enzyme was judged to be greater than 95% pure by SDS-polyacrylamide gel electrophoresis analysis. Enzyme was stored at 20-30 mg/ml at -70 "C. Epsilon-depleted enzyme, ~d-EcF,[2,11, was prepared as described (Dunn, 1986). Unlike the 5-subunit enzyme, we found that ed-EcF, could be depleted of endogenous nucleotides simply by ammonium sulfate precipitation. This involved adding 2 volumes of saturated ammonium sulfate to Ed-EcF, in TE buffer (50 mM Tris-C1, pH 8.0, 0.5 mM EDTA), incubating at 0 "C for 2.5 h, pelleting 10 min in a Beckman microfuge, and redissolving in TE buffer followed by passage through a 1-ml centrifuge column equilibrated in the same buffer. A luciferase assay of neutralized, perchloric acid extracts showed ~d-EcF,[2,11 to contain 2.5 mol ANP/mol of enzyme, whereas ~d-EcF,[O,01 contained less than 0.1 mol ANP/mol.

Methods
Protein concentration was determined by a modified Lowry method according to Peterson (1977). The concentrations of free and complexed metals and nucleotides were calculated using the program Bound and Determined (Brooks and Storey, 1992).
Noncatalytic Site Specificity St~dies-[~H]Nucleotide binding was measured by a centrifuge column assay. Columns containing 1 or 2 ml of swollen Sephadex were equilibrated with MTSEM buffer (20 mM MOPS/Tris, pH 8.0, 150 mM sucrose, 0.2 mM EDTA, and 2 mM Mg(CH,COO),). When a n NTP-regenerating system was present during incubation with L3H]GTP or [3H]ATP, it consisted of 50 mM KC1, 10 mM phosphoenolpyruvate, and 1 mg/ml pyruvate kinase. When the EcF, concentration was less than 3 PM, 1.1 mg/ml bovine serum albumin was included in all buffers to enhance recovery of the enzyme through centrifuge columns (Cross and Nalin, 1982). Under these conditions, the recovery from 100-pl samples was 90% when applied to 1-ml columns and 85% when applied to 2-ml columns. When column effluents were to be counted, they were collected directly in liquid scintillation vials. Data were corrected for radioactivity eluting in the absence of EcF, (usually less than 5%). The standard deviation for replicate measurements is *lo% using these procedures.
The composition of [3H]nucleotides bound to EcF, was determined by anion-exchange HPLC using a Partisil PX3 10/25 SAX column (4.6 x 250 mm). Bound nucleotides were extracted by perchloric acid precipitation of protein in the presence of carrier ADP, ATP, GDP, and GTP. Extracts were neutralized with KHCO,, centrifuged, and diluted prior to loading on the HPLC column. Nucleotides were eluted using a phosphate gradient as described (Milgrom and Cross, 1993). Aliquots of the fractions were counted by liquid scintillation using 5-10 volumes of Bio-Safe I1 (Research Products International).
MgADP Inhibition Studies-Enzyme stocks were diluted to 1-2 mg/ml with either TE buffer or TM buffer (50 mM Tris-C1, pH 8.0,0.5 mM MgCl,) and passed through two sequential 1-ml centrifuge columns equilibrated in the same buffer. When the enzyme concentration was less than 0.5 mg/ml, buffers were supplemented with 1 mg/ml bovine serum albumin (TEB and TMB buffer (TE or TM buffers containing 1.0 mg/ml bovine serum albumin)).
Except where noted, hydrolysis assays were performed by dilution of the preincubated enzyme into 50 mM Tris-C1, pH 8.0, 50 mM KCl, 1 mg/ml bovine serum albumin, 0.5 mg/ml pyruvate kinase, 2 mM phosphoenolpyruvate (reaction buffer). plus MgC1, and [y3'P1ATP to give the indicated concentrations of MgATP and free Mg2'. The amount of inorganic 32P released was determined after removal of residual [Y-~~PIATP by adsorption to charcoal. Aliquots were subjected to Cerenkov counting with correction for 32P1 present in controls lacking EcF,.
Binding of Ipy-32P12-azido-ATP was initiated by addition of 100 pg of EcF, in TM buffer to reaction buffer containing 0.05 mg/ml pyruvate kinase and 5 mM phosphoenolpyruvate plus MgC1, and the nucleotide affinity probe to give final concentrations of 260 nM EcF,, 460 nM Mg(2azido)ATP, and 1 mM free Mg". Binding of the probe was terminated after 20 min by exposure to a 30-s cold chase containing 50 p MgATP, 1 mM free Mg2+, and either 1 mM azide or 7.5 mM sulfite followed by passage of 0.5-ml aliquots through 5-ml centrifuge columns equilibrated in TM buffer plus 0.5 mM azide. Column efiluents were irradiated for 10 min with a Mineralight lamp set at short wavelength. Noncovalently bound probe was removed by acid precipitation in the presence of ATP, followed by two brief washes with 80% acetone (-20 "C). The pellets were dissolved in 75 p1 of 8 M urea made fresh in 100 mM EPPS, pH 8.0, diluted 8-fold, and digested by 4 pg of trypsin for 4 h followed by a second equal addition of trypsin overnight. Samples were diluted 2-fold with 0.2% trifluoroacetic acid, filtered, and applied to a Vydac C, reversed-phase column. Peptides were eluted with the indicated gradient of acetonitrile in 0.1% trifluoroacetic acid at 1 mumin with 2-ml fractions collected and subjected to Cerenkov counting.

Specificity of Noncatalytic Nucleotide-binding
Sites-Incubation of nucleotide-depleted enzyme (EcF,[O,OI) with 2 mM MgL3H1GTP for 2 min results in the binding of about 3 mol of [3HIGNP/mol of enzyme (Table I, Experiment 1). If the enzyme is then incubated for 1 min without further addition, all of the [3H]GNP remains bound during passage through a second centrifuge column (Experiment 2). However, if millimolar concentrations of unlabeled GTP (Experiment 3 ) or ATP (Experiment 4 ) are present during the second incubation, 1 mol of GNP/mol of enzyme is released. L3H]GNP displaced from catalytic sites during this cold-chase step consists of nearly equal amounts of GDP and GTP. In contrast, the 2 moVmol remaining bound at noncatalytic sites is mostly GTP (-90%).
When the experiments described above were repeated using native enzyme (EcF,[2,11), about 2 mol of L3H1GNP boundmol of enzyme during incubation with 2 mM [3HlGTP (Table I, Experiments 5 and 6 ) . Again, close to 1 mol of labeled nucleotide is displaced from catalytic sites during subsequent incubation with unlabeled GTP (Experiment 7 ) or ATP (Experiment 8). However, native enzyme retained only 1 mol of labeled nucleotide at noncatalytic sites. Again this was mostly GTP. In summary, the results presented in Table I show that nucleotidedepleted enzyme can bind 2 GTP at noncatalytic sites, whereas the native enzyme binds 1 GTP at a noncatalytic site.
A time course for binding GNP to EcF,[O,Ol during incubation with 100 VM L3H]GTP in the presence of a GTP-regenerating system is presented in Fig. 1. A total of 2.7 mol of L3H1GNP boundmol of enzyme (circles), with 1.7 moVmol bound at noncatalytic sites (triangles), and 1.0 mol/mol bound a t catalytic sites (squares). Binding reached maximal levels within 3-4 min. Bound nucleotides were extracted from aliquots removed at 10 min, and their composition analyzed by anion-exchange  1.1 mg/ml bovine serum albumin without (circles) or following (tri-(squares) was calculated as the difference between total and nonchaseable (noncatalytic) binding. An additional incubation (inverted triangles) contained enzyme that was preincubated for 5 min with 100 p~ unlabeled ATP in the presence of an ATP-regenerating system and passed through a centrifuge column prior to incubation with r3H1GTP.
HPLC. Again, nearly all of the r3HIGNP bound a t noncatalytic sites was GTP (95%), whereas catalytic sites contained a mixture of GDP and GTP.
In contrast to the results obtained with GTP, we were unable to detect significant GDP binding at noncatalytic sites. Incubation of EcF, [O,O] with 100 p [3HIGDP for up to 90 min gave a total of 1.6 moVmol bound with only 0.2 moVmol bound at nonchaseable sites. This experiment demonstrates the efficacy o f the conditions used to chase catalytic site-bound nucleotide.
In order to compare these results to the binding of adenine nucleotides, EcF, [O,O] was incubated with 100 VM F3H1ATP in the presence of an ATP-regenerating system (Fig. 2) or with 100 p ['HIADP (Fig. 3). Binding stoichiometries were very similar in all cases, although the rate of binding adenine nucleotides to noncatalytic sites (Figs. 2 and 3, triangles) appears to be faster than the rate of binding GTP (Fig. 1, triangles). Bound nucleotides were extracted from aliquots, taken after 3 min of incubation of EcF,[O,OI with L3HIATP and a regenerating system (Fig. 21, and their composition was analyzed by HPLC. ATP accounted for 96% of the nucleotide bound at noncatalytic sites. To determine whether prior binding of ATP at noncatalytic sites on EcF, [O,O] blocks subsequent binding of GTP, we preincubated the enzyme with 100 p~ ATP and an ATP-regenerating Nucleotide binding was measured as described for Fig. 1, except that 3.7 m~ MgATP was used in the cold chase step. An additional incubation (inverted triangles) contained enzyme that was preincubated for 5 min with 100 PM ATP in the presence of an ATP-regenincubation with l3H1ADP.
erating system and passed through a centrifuge column prior to system. After unbound nucleotides were removed, the enzyme was incubated with 100 p L3H1GTP in the presence of a GTPregenerating system. The results show that prior exposure to ATP prevents ['HIGTP binding at noncatalytic sites (Fig. 1, inverted triangles versus triangles), while having no effect on EcF,[O,Ol was incubated at 60 n M for 10 min in MTSEM buffer containing 100 VM C3H]GTP and a GTP-regenerating system. T3H]GNP was displaced from catalytic sites by incubating with 3.3 m M MgATP for 1 min and unbound nucleotides were removed on a centrifuge column as described ("Experimental Procedures"). The effluent was incubated without further addition (Experiment 1) or with the additions shown. Half-times were calculated from the measured rate constants. binding a t catalytic sites (data not shown). This suggests that GTP binds at the same two noncatalytic sites that bind ATP or that cooperative interactions prevent the binding of more than two nucleoside triphosphates at the three noncatalytic sites.

Experiment
Preincubation of EcF, [O,O] with unlabeled ATP was also very effective in slowing L3H1ADP binding at noncatalytic sites (Fig.  3, inverted triangles versus triangles), again having no effect on catalytic site binding (data not shown). It was of interest, however, that in contrast to L3H1GTP (Fig. l, inverted triangles), [3H]ADP appears either to slowly exchange for noncatalytic site-bound ATP or to bind at the third noncatalytic site (Fig. 3,  inverted triangles).
In order to distinguish between these possibilities, EcF,LO,Ol was incubated briefly with 100 p~ ATP and an ATP-regenerating system to bind 1.4 mol ATP at noncatalytic sites. Unbound ligand was removed, and the enzyme was further incubated with 100 1.1~ ADP for 1 h. This resulted in the binding of 1.5 mol of ADP at nonexchangeable sites, with loss of only 0.1 mol of the ATP to give a total of 2.8 mol ofANP bound at noncatalytic sitedmol EcF,. It was also shown in Table I (Experiments 7 and 8) that native enzyme, which has one noncatalytic site filled with ADP and one with ATP ("Experimental Procedures"), can still bind one GTP at the third noncatalytic site. Hence, the results show that although only 2 mol/mol of GTP (Fig. l), ATP (Fig. 2), or ADP (Fig. 3) can bind at noncatalytic sites during incubation with 100 p~ nucleotide for up to 1 h, all three sites can be filled by a combination of nucleoside di-and triphosphates.
GTP bound at noncatalytic sites on EcF, is quite stable. In MTSEM buffer, the half-time for dissociation is 175 min ( Table  11). The addition of EDTA in excess of Mg2' causes a 2-fold increase in the rate of release. A 5-8-fold increase is obtained by addition of MgPP,, MgADP, or EDTNATP. The greatest increase (23-fold), however, is obtained with the addition of 100 p~ ATP and an ATP-regenerating system. This decreases the half-time for dissociation of L3H]GTP to 8 min (Table 11). This is very close to the value for L3H1GTP dissociation from noncatalytic sites on the mitochondrial enzyme (tu2 = 10 min) obtained under similar conditions (Milgrom and Cross, 1993).
We have recently found that MgPP, effectively competes with nucleotides for binding at noncatalytic sites on MF, (Milgrom and Cross, 1993), and this also appears to be the case for the E.
coli enzyme. At a concentration of 1 mM, MgPP, nearly completely prevents GTP binding at noncatalytic sites on EcF,[O,Ol (Fig. 4, triangles versus Fig. 1, triangles) and native enzyme (data not shown), and significantly slows ATP binding at noncatalytic sites on EcF,[O,Ol (Fig. 2, inverted triangles  during hydrolysis of 100 p~ L3H]GTP by more than 2-fold (data not shown). As with MF, (Milgrom and Cross, 19931, MgPPi has little effect on the binding of either GNP (Fig. 4, squares) or ANP (data not shown) a t catalytic sites.

Inhibition by MgADP Bound at Catalytic
Sites-Studies of MgADP inhibition of MF, and CF, have shown that a steadystate level of the inhibitory complex exists during ATP hydrolysis. Increased free Mg2' shifts the distribution of enzyme forms toward the inhibited state. A similar effect of Mg2' is shown for EcF, in Fig. 5 (circles). The fact that considerably less inhibition is observed over the same concentration range of Mg2' during GTP and ITP hydrolysis (Fig. 5 , squares and triangles) indicates that MgNDP rather than Mg2' alone is responsible for the inhibition.
When ed-EcF1[2,1] is assayed for ATP hydrolysis activity, a lag in product release is seen (Fig. 6, closed squares). In contrast, enzyme preincubated with pyruvate kinase and phosphoenolpyruvate to remove endogenous exchangeable ADP exhibits a burst in product release (open squares) at the same Mg2' concentration.
To further confirm the ability of MgADP t o inhibit EcF,, the epsilon-depleted enzyme was stripped of endogenous nucleotides by ammonium sulfate precipitation ("Experimental Procedures"). A burst in activity of ~d-EcF,[O,01 was seen after preincubation with either EDTA (Fig. 7, open

squares) or Mg2' (open circles), whereas MgADP resulted in a lag (closed circles).
Preincubation with MgATP under conditions that allowed complete conversion to ADP also produced a lag (data not shown).
Azide has been shown to trap the MgADP complex formed at catalytic sites on MF, and CF, (Vasilyeva et al., 1982b;Murataliev et al., 1991). Preliminary experiments with native EcF, indicated that retention of labeled ANP during a cold chase could be enhanced by azide. With increase in azide concentration during cold chase, an increase in azide-trappable ANP showed a linear correlation with the loss of hydrolytic activity (Fig. 8). Extrapolation to complete inactivation corresponded to 0.45 mol of azide-trappable ANP/mol of enzyme.
In order to identify the nucleotide site occupied by the azidetrappable ADP, 2-azido-adenine nucleotide was used. In the dark, this photoafflnity analog is trapped by azide during a cold chase in a similar manner to ADP. The trapped azido-adenine nucleotide was covalently incorporated into the protein by photolysis, and reversed-phase HPLC elution profiles of digested samples were determined (Fig. 9). Photoincorporation following a cold chase in the presence of azide resulted in significant labeling of a catalytic site peptide (Cat) eluting at 23% acetonitrile (top panel). In contrast, a cold chase in the presence of sulfite resulted in near complete disappearance of the catalytic site peptide (bottom panel). In both cases, the amount of labeled noncatalytic peptide ( N C ) eluting at 18% was small and nearly constant. The positions of catalytic and noncatalytic site peptides in the elution profile were assigned based on controls where EcF, was selectively labeled at either catalytic (pY331) or noncatalytic (pY354) sites as described by Wise et al. (1987). Also, using anion-exchange HPLC which separates peptides labeled by the di-and triphosphate forms of the probe, it was found that the azide-trappable catalytic site peptide is labeled exclusively by 2-azido-ADP (data not shown), DISCUSSION Past studies of the nucleotide-binding sites of F,-ATPases have relied on three different criteria for distinguishing between catalytic and noncatalytic sites. The first was based on affinity. Noncatalytic sites were thought to be very tight and catalytic sites to be loose (Harris, 1978). The three tightly bound endogenous nucleotides retained during purification of MF, were assumed to be bound at the three noncatalytic sites. Aliquots of 100 pl were applied to 2-ml centrifuge columns equilibrated with TMB buffer containing 0.5 mM azide. Binding was determined by Cerenkov counting of the column effluents. Azide-trappable ADP was calculated by subtracting the amount bound in the absence of azide (noncatalytic site binding) from the amount bound in the presence of azide. Activity assays were performed with the same reaction buffer using unlabeled ATP and supplemented with 0.25 mg/ml lactate dehydrogenase and 300 p~ NADH. The activity of the 50 p~ MgATP chase reaction was followed spectrophotometrically at 340 nm and was linear for 5 min.
However, this was later shown not to be the case. One of the endogenous nucleotides exchanges rapidly during catalytic turnover (Kironde and Cross, 19861, and the first catalytic site to bind ADP (K,(ADP) = 1 nM, Cunningham and Cross, 1988) is tighter than the loosest noncatalytic site (K,(ADP) = 50 nM, Kironde and Cross, 1987). Another reason for not considering catalytic sites to be loose comes from the finding that the first catalytic site on MF, to bind ATP does so with a remarkably high affinity (K,(ATP) = 1 0 " ' M, Grubmeyer et al., 1982). Hence, in recent years few laboratories have relied on affinity as a means for distinguishing between catalytic and noncatalytic nucleotide sites. A second widely used criterion is based on the nucleotide specificities of the sites. Catalytic sites clearly tolerate variations in the structure of the base moiety as GTP and ITP are hydrolyzed at rapid rates. In contrast, noncatalytic sites have been thought to be highly specific for adenine nucleotide (Harris, 1978;Harris et al., 1978;Perlin et al., 1984). However, it was recently shown that CF, binds GTP a t noncatalytic sites (Guerrero et al., 1990;Milgrom et al., 19911, and we find the same is true for MF, (Milgrom and Cross, 1993) and EcF, (Table I and Fig. 1; see also Weber et al., 1994). In addition, it appears that pyrophosphate can bind at noncatalytic sites on EcF, (Figs. 2 and 4) and MF, (Milgrom and Cross, 1993).
There are several reasons why this lack of specificity was not previously detected. The first is that GTP binding is not as tight as ATP binding. Under conditions of catalytic turnover, GTP bound at noncatalytic sites on EcF, dissociates with a half-time of 8 min (Table 11), whereas little dissociation of ATP could be detected (Pagan and Senior, 1990). In testing for GNP binding at noncatalytic sites, Perlin et al. (1984) used a 30-min incubation with unlabeled ATP to chase [3HlGNP from catalytic sites. In light of the data presented in Table 11, it can be predicted that GTP bound at noncatalytic sites would also have dissociated under these conditions. A lower affinity for GTP than for ATP may also explain why binding to the isolated a-subunit, which appears to retain part of the noncatalytic binding domain (Dunn and Futai, 19801, was not detected (Senda et al., 1983;Perlin et al., 1984).
A second reason is that the affinity of noncatalytic sites for GDP is much lower than for GTP, just as ADP binds less tightly than ATP. This may preclude detection of GDP binding by a centrifuge column assay ("Results," and Kironde and Cross, 1986) or when the enzyme is incubated with GTP over a period of hours under conditions where the GTP is rapidly converted to GDP (Perlin et al., 1984).
A final explanation for the failure t o readily detect guanine nucleotide binding at noncatalytic sites relates to the very slow rate of dissociation of adenine nucleotides from these sites. With EcF, having ANP prebound at noncatalytic sites, Senior and co-workers (Perlin et al., 1984;Senior et al., 1992) observed no exchange for medium GTP. Even when GTP is maintained by a regenerating system, no exchange is seen over a 20-min period if enzyme is pretreated with ATP ( Fig. 1, inverted triangles).
A third criterion for distinguishing between catalytic and noncatalytic sites is based on the fact that catalytic sites on fully active MF, exchange bound ligand rapidly during turnover, whereas noncatalytic sites do not (Cross and Nalin, 1982). The validity of this criterion is supported by the finding that 2-azido-ATP labels a single, specific p-tyrosyl residue when loaded at exchangeable sites and a single, but different, p-tyrosyl residue when loaded at nonexchangeable sites on MF, (Cross et al., 1987), CF, (Xue et al., 19871, and EcF, (Wise, et al., 1987).
However, this criterion also has limitations. The first is due to the fact that noncatalytic sites are not nonexchangeable. The noncatalytic site on MF, having the lowest affinity for ADP releases bound ligand with a half-time of 3 min during catalytic turnover (Kironde and Cross, 1987). However, as originally defined (Cross and Nalin, 19821, exchangeable (catalytic) sites exchange with medium ATP very rapidly during a brief episode of catalytic turnover. Typically ATP is added in a 1OOO:l molar ratio to F,, and cleavage is complete within a few seconds. Under these conditions, noncatalytic sites do not exchange. However, it should be noted that pH values below 7 (Harris et al., 1978) and nonaqueous solvents such as glycerol (Garrett and Penefsky, 1975) can accelerate release of nucleotide from noncatalytic sites.
A second limitation results from the fact that inhibitory MgADP bound at a single catalytic site on CF, (Feldman and Zhou et al., 1988) or MF, (Drobinskaya et aE., 1985;Milgrom and Boyer, 1990;Chernyak and Cross, 1992) can dissociate slowly, under certain conditions, upon addition of MgATP. Hence, nucleotide bound at catalytic sites does not always exchange rapidly. However, with MF, this can be avoided simply by use of buffers containing 1 mM Pi (Grubmeyer et al., 1982).
In contrast, the formation of an inhibitory catalytic site MgADP complex has not been thought to play a significant role in the kinetics of EcF,, making this enzyme unique. Instead it has been argued that Mg2' inhibits directly and that azide inhibition of the enzyme is related to changes in subunit cooperativity.
The ability to demonstrate MgADP inhibition of EcF, has been complicated by the presence of the inhibitory €-subunit. When EcF, is diluted into an ATP hydrolysis assay to a concentration below the Kd (2-10 nM) for the €-subunit (Sternweis and Smith, 1980;Wood et al., 1987), its slow dissociation results in a steady increase in activity (Laget and Smith, 1979). This lag would mask a delay caused by the slow release of inhibitory MgADP such as that observed with MF, (Minkov et al., 1979;Drobinskaya et al., 1985) and CF, (Feldman and Boyer, 1985).
The successful demonstration of MgADP inhibition of EcF, obtained in the present study was aided by the use of €-depleted enzyme, With four-subunit EcF,, elimination of endogenous ADP by preincubation with an ATP regenerating system (Fig.  6) or by removal of all bound nucleotides (Fig. 7) resulted in a burst of ATP hydrolysis activity which could be prevented by MgADP but not by Mg2' alone. In addition, only ~d-EcF,[O,01 gave equal rates for hydrolysis and the loading of a catalytic site with labeled nucleotide (results not shown). All other enzyme forms loaded one catalytic site at a rate too slow t o be consistent with a population of fully active enzyme. This reinforces the proposal that native EcF,[2,11 is a heterogeneous mixture of active and inactive enzyme as suggested by kinetic studies from Allison's laboratory (Muneyuki et al., 19911, as shown by the lag in Fig. 6 (closed squares) and as quantitated by the extrapolation to 0.5 mol of azide-trappable ANP/mol of enzyme at zero activity (Fig. 8). Presumably the other half of the enzyme is already inhibited by unlabeled endogenous ADP.
A second criterion for demonstrating MgADP inhibition of MF, and CF, has been its nucleotide specificity. The ability of different nucleotides to differentially inhibit Ed-EcF, with in- creasing free M e (Fig. 5) is a strong indication that M e is not solely responsible for the observed inhibition. In a similar manner, submitochondrial particles are inhibited by preincubation with ADP but not by GDP or IDP (Vasilyeva et al., 1980). A third criterion to show the presence of inhibitory MgADP at catalytic sites has been the photolabeling of a catalytic-site peptide by Mg[a-azido]ADP (Zhou et al., 1988;Milgrom and Boyer, 1990;Chernyak and Cross, 1992). Using azide to trap 2-azido-ADP on EcF,, photolysis of the inhibitory complex resulted in the modification of a catalytic site peptide (Fig. 9).
These observations can be explained by the scheme presented in Fig. 10. Under normal conditions, a heterogeneous mix of enzyme forms includes both €-inhibited (E.*F,) and MgADP-inhibited enzyme (*F,.ADP). Removal of E (step 5) and inhibitory MgADP (step 7) will generate a population of fully active enzyme (steps 14). Sulfite (Vasilyeva et al., 198213) or the filling of the noncatalytic sites with ATP (Jault and Allison, 1993;Milgrom and Cross, 1993) facilitates removal of catalyticsite bound MgADP (step 7). Azide has the opposite effect by stabilizing the inhibitory complex (step 8). We exploited this property t o trap inhibitory MgADP (Fig. 8) and Mg(2azido)ADP ( Fig. 9) under cold chase conditions. Structural and mechanistic features of EcF, have been extensively studied by site-directed mutagenesis. Knowledge of the fact that EcF, is inhibited by catalytic site-bound MgADP is thus of major importance in addition to its sharing this property with MF, and CF,. The relative susceptibility of each mutant enzyme to inhibition by MgADP should be taken into account before binding or kinetic analyses are interpreted. One mutant EcF, analyzed in our laboratory, PESllD, shows significantly lower activity than wild-type enzyme under normal assay conditions, but this has been shown to be due to a heightened sensitivity to MgADP inhibition rather than to a catalytic defect.2 When assaying ATP hydrolysis, we currently minimize the fraction of enzyme in the e*F, form by diluting below the Kd for E and by adding lauryldimethylamine oxide which helps reverse E inhibition (Lotscher et al., 1984). To minimize the level of *F,.ADP, we prefill noncatalytic sites with ATP, add sulfite or selenite to the assay mixture, and maintain low concentrations of free M e .
In summary, with the proper precautions, measuring differences in the exchangeability of bound ligands can provide a reliable means of distinguishing catalytic from noncatalytic sites on F,-ATPases, whereas differences in substrate specificities or affinities may not. Also, in analyzing EcF, mutants, it is important to distinguish between alterations in sensitivity to inhibition by MgADP and changes in catalytic efficiency.
T. M. Duncan and R. L. Cross, unpublished data.