Regulation of aminotransferase-glutamate dehydrogenase interactions by carbamyl phosphate synthase-I, Mg2+ plus leucine versus citrate and malate.

Citrate, malate, and high levels of ATP dissociate the mitochondrial aspartate aminotransferase-glutamate dehydrogenase complex and have an inhibitory effect on the latter enzyme. These effects are opposed by Mg2+, leucine, Mg2+ plus ATP, and carbamyl phosphate synthase-I. In addition, Mg2+ directly facilitates formation of a complex between glutamate dehydrogenase and the aminotransferase and displaces the aminotransferase from the inner mitochondrial membrane which could enable it to interact with glutamate dehydrogenase in the matrix. Zn2+ also favors an aminotransferase-glutamate dehydrogenase complex. It, however, is a potent inhibitor of and has a high affinity for glutamate dehydrogenase. Leucine, however, enhances binding of Mg2+ and decreases binding of and the effect of Zn2+ on the enzyme. Thus, since both metal ions enhance enzyme-enzyme interaction and Zn2+ is a more potent inhibitor, the addition of leucine in the presence of both metal ions results in activation of glutamate dehydrogenase without disruption of the enzyme-enzyme complex. Furthermore, the combination of leucine plus Mg2+ produces slightly more activation than leucine alone. These results indicate that leucine, carbamyl phosphate synthase-I, and its substrate and cofactor, ATP and Mg2+, operate synergistically to facilitate glutamate dehydrogenase activity and interaction between this enzyme and the aminotransferase. Alternatively, Krebs cycle intermediates, such as citrate and malate, have opposing effects.

affinity for the inner surface of the inner mitochondrial membrane (3). Furthermore, there is evidence that complexes can be formed between these three enzymes (4-6). Thus, it seems possible that, as proposed, these enzymes react in a sequence as a multi-enzyme cluster (2-6).
Mitochondrial aspartate aminotransferase can also form complexes with glutamate dehydrogenase (L-glutamate: NAD(P)+ oxidoreductase (deaminating), EC 1.4.1.3) and carbamyl phosphate synthase-I (EC 6.3.4.16) (7-13). Furthermore, carbamyl phosphate synthase-I does not displace but can enhance aminotransferase-glutamate dehydrogenase interactions, even in the presence of substrates of these enzymes (12,13). This, plus the fact that, in the liver mitochondrial matrix, glutamate dehydrogenase can be cross-linked to carbamyl phosphate synthase-I (14), suggests that a complex can also be formed among these three enzymes. These interactions are specific and could be physiologically significant because they take place even at concentrations less than those present in uiuo. Indeed, carbamyl phosphate synthase-I is essentially present only in liver mitochondria (15), where it is the most abundant mitochondrial protein, and its level has been estimated to be as high as 0.4-1.0 mM (16-18). Consequently, this enzyme could play a structural role by regulating interaction between glutamate dehydrogenase and the aminotransferase when the latter enzyme is involved in NHZaspartate metabolism. Since generation of carbamyl phosphate is a unique function of liver mitochondria, this regulatory or structural role could be the reason why carbamyl phosphate synthase-I is present in extremely high levels.
While the aminotransferase-glutamate dehydrogenase complex is stable in the presence of substrates (13) of these enzymes, two Krebs cycle intermediates, citrate and malate, as well as high levels of ATP dissociate this complex (12,13). This is of interest in view of the fact that malate and citrate play key roles in determining whether oxalacetate is converted into citrate, malate, or aspartate. These and other results suggest structural organization of the aminotransferase in opposing heteroenzyme clusters and regulation of these complexes by substrates and allosteric modifiers of the constituent enzymes. Indeed, this might be the case in tumor mitochondria where glutamate oxidation leads to the production of a high level of citrate (2, 19).
The effects of metal ions on these interactions have not been studied. This could be of interest because Mg2+ is a cofactor of the carbamyl phosphate synthase-I reaction, Zn2+ is a potent inhibitor of both carbamyl phosphate synthase-I (20) and glutamate dehydrogenase (21,22), and it would be expected that metal ions would modify the effects of citrate on enzyme-enzyme interactions. Also, as mentioned above, 6069 the aminotransferase has a high affinity for the inner mitochondrial membrane (3, 23). This might restrict the amount of this enzyme available to interact with glutamate dehydrogenase and carbamyl phosphate synthase-I in the liver mitochondrial matrix. Therefore, in this paper, we have studied in more detail the effects of citrate, malate, and metal ions on enzyme-enzyme interactions and binding of the aminotransferase to the inner mitochondrial membrane.
In some experiments, enzyme-enzyme interaction was determined by measuring coprecipitation of the enzymes in polyethylene glycol. These conditions might mimic the mitochondrial matrix with polyethylene glycol substituting for the noninteracting proteins (24). Polyethylene glycol does not primarily promote complex formation but mainly precipitates complexes which are found in its absence (25). Consequently, we have found that results obtained in polyethylene glycol are consistent with those obtained by other techniques including: enzyme immobilized on Sepharose (IO), divalent cross-linkers (9), and gel equilibrium (7). However, because of conceivable limitations of performing experiments in polyethylene glycol, results are also presented which were obtained in its absence.

MATERIALS AND METHODS
Enzymes and Reagents-Bovine and rat liver mitochondrial aspartate aminotransferase and glutamate dehydrogenase were prepared with previously described methods (26)(27)(28)(29). Rat liver carbamyl phosphate synthase-I was prepared with some modifications of a previously described method (30). Pig heart mitochondrial malate dehydrogenase and citrate synthase were obtained from Boehringer Mannheim and Sigma, respectively. The latter was also the source of bovine serum albumin. All of the enzymes were over 99% pure as estimated from electrophoresis on sodium dodecyl sulfate-polyacrylamide gels. Unless indicated, enzymes were dialyzed extensively uersus 0.02 M potassium phosphate, 0.1 mM EDTA, pH 7.0, before use in experibuffer (minus EDTA) which had been chromatographed on Chelex.
ments. In some experiments, enzymes were dialyzed versus the same Polyethylene glycol (6000) was obtained from Fisher and prepared as a 40% (w/v) solution in 0.02 M potassium phosphate, pH 7.0. Other substrates, enzymes, coenzymes, and reagents were obtained from Sigma. Stock solutions of all reagents used in assays were adjusted to the pH of the assays and prepared as sodium salts. Solutions of coenzyme were prepared fresh daily.
Coprecipitatwn of Enzymes in Polyethylene Glycol-This was done essentially as described previously (4, 18). Solutions of enzymes plus ligands were incubated in 14% (w/v) polyethylene glycol at 25 'C in either 1 ml of 14 mM potassium phosphate, 0.1 mM EDTA, pH 7.0, or 1 ml of this buffer (minus EDTA) which had been chromatographed on Chelex. During the incubation, the turbidities of the solutions were read at 510 nm in an Aminco-Bowman spectrofluorometer. After 20-min incubations, the solutions were centrifuged at 25 "C for 10 min at 20,000 X g in a Sorvall RC 2B centrifuge. The supernatant ml of 0.02 M potassium phosphate, 0.1 mM EDTA, pH 7.0. The solutions were then removed, and the pellets were resuspended in 1 amount of protein and enzyme activity in the original samples, the supernatant solutions, and the dissolved pellets was then determined with the standard assays of enzyme activity described below. In these experiments, the sum of total enzyme units in the supernatant plus the total units in the precipitate essentially equalled the total units added, Thus, the amount of enzyme in the precipitate could be calculated by multiplying the fraction of enzyme units in the precipitate by the total amount of enzyme units incubated.
Standard Enzyme Assays-Aspartate aminotransferase activity was measured in a spectrophotometric coupled assay in the presence of a-ketoglutarate (10 mM), aspartate (10 mM), DPNH (100 pM), and an excess of malate dehydrogenase as described previously (26).
Turbidity Experiments-These experiments were performed in an Aminco-Bowman spectrofluorometer equipped with an E polarizer.
The instrument was adjusted prior to each experiment so that a 0.5mg/ml solution of blue dextran would read 195.
Preparation of and Experiments with Inuerted Inner Mitochondrial Membranes-Rat liver inverted inner mitochondrial membranes were prepared exactly as described previously (3,38). The amounts of membranes or organelles are given in terms of micrograms or milligrams of proteins. The inner membranes were washed extensively with 2 mM HEPES' buffer, pH 7.0. For the binding studies, enzyme samples and membranes (1.7 mg/ml) were incubated for 15 min at 0 "C in (1 ml) 2 mM HEPES buffer, 0.5 mM dithiothreitol, pH 6.9. The membranes were then sedimented and washed with the same buffer. The final pellet was then resuspended in 1 ml of the above buffer, as described previously (3), and enzyme activities in the different fractions were measured. Over 95% recovery of added enzyme was found. Thus, the amount of bound protein could be obtained from multiplying the fraction of enzyme units in either the supernatant or the precipitate by the total amount of enzyme units incubated.
Treatment of Enzymes with Chlex 100-Carbamyl phosphate synthase-I and glutamate dehydrogenase were dialyzed versus 0.02 M potassium phosphate, 0.1 mM EDTA, pH 7.0, and then uersus this buffer (minus EDTA) which had been chromatographed on Chelex. These enzymes were then chromatographed on a 0.8 X 11-cm column of Chelex which had been equilibrated with the dialysis buffer (minus EDTA). Mitochondrial aspartate aminotransferase and malate dehydrogenase could not be eluted from Chelex. Therefore, these enzymes were dialyzed versus the Chelex-treated buffer and then versus a large excess of Chelex in the Chelex-treated buffer. Coenzymes and substrates were also prepared in Chelex-treated buffer and then chromatographed on a Chelex column.
Assays for Metals and Equilibrium Dialysis-Prior to assays for metals, glutamate dehydrogenase was dialyzed extensively versus 0.02 M potassium phosphate, 0.1 mM EDTA, pH 7.0, and then versus this same buffer (minus EDTA) which had been chromatographed on a Chelex column. The Caz+ measurements were performed with a Perkin-Elmer Model 303 spectrometer with a Model 200 furnace. The Zn2+ analysis was conducted with a flame atomic absorption spectrophotometric method using an Instrumental Laboratories Model 551 instrument. Prior to neutron activation experiments, glutamate dehydrogenase was dialyzed versus these buffers and then versus doubledistilled water.
Equilibrium dialysis was performed at 4 "C in the Chelex-treated phosphate buffer described above. The concentration of enzyme and ZnClz in the dialysis chamber was 20 p~ and, when indicated, 1 mM leucine and 1 mM TPN was added. The enzyme was then dialyzed versus either 20 p M Zn2+ or 20 p M Znz+, 1 mM TPN, and 1 mM leucine.

RESULTS AND DISCUSSION
Effect of Mg2+ on Aminotransferase-Glutamate Dehydrogenase Interaction-As shown in Fig. 1 (curves A and B ) , M P increases the amount of glutamate dehydrogenase associated with the aminotransferase. This is actually an increase in enzyme-enzyme interaction because M e does not markedly enhance precipitation of either enzyme alone (Fig. 1, curues E and 3'). Furthermore, this effect is not due to increasing ionic strength, which in general tends to decrease interaction The abbreviation used is: HEPES, 4-(2-hydroxyethyl)-l-piperazineethanesulfonic acid.

TABLE I Effect of metal ions on turbidity of enzyme-enzyme complexes
In these experiments, the turbidity was measured 5 min after incubating the enzyme in 10 mM potassium phosphate, pH 7.0, at 25 "C as described under "Materials and Methods." Enzymes and buffer were treated with Chelex as described under "Materials and Methods." The abbreviations used are: GDH, glutamate dehydrogenase; Asp-AT, mitochondrial aspartate aminotransferase; MDH, mitochondrial malate dehvdroeenase. and, as shown in Fig. 1 (curves C and D), NaCl has no significant effect.

Additions
This effect of M$+ on enzyme-enzyme interaction can also be seen in the absence of polyethylene glycol. Thus, as shown in Table I (lines 1-3), in the absence of polyethylene glycol, the turbidity of the mixture of the two enzymes is considerably greater than the sum of the turbidity of either enzyme alone, and 1 mM Mg2' increases this difference. That is, even though M$+ slightly increases the turbidity of glutamate dehydrogenase alone (but not the aminotransferase alone), this in-crease is considerably less than that produced by M e in the presence of both enzymes.
M e increases the turbidity of glutamate dehydrogenase alone (Table I) and inhibits this enzyme (not shown) over the same concentration range required for enhancement of enzyme-enzyme interaction. Mg2+ has no effect on the turbidity (Table I) or the activity of the aminotransferase (not shown). Thus, M e is bound to glutamate dehydrogenase, and this apparently increases enzyme-enzyme interaction.
Interactions with Malate Dehydrogenase-In both the presence (not shown) and absence of polyethylene glycol (Table  I), M$+ had considerably less effect on the complex between malate dehydrogenase and glutamate dehydrogenase. Therefore, M$+ is another factor (5, 18) which favors interaction between glutamate dehydrogenase and the aminotransferase over interactions between glutamate dehydrogenase and malate dehydrogenase.
Specificity of Metal Zons-The effects of M e on aminotransferase-glutamate dehydrogenase interactions and glutamate dehydrogenase activity are not specific for this divalent cation. Mn2+ (not shown) and Ca2+ have effects comparable to M$+. Zn2+ is effective at quite low levels where neither Mg2+ nor Caz+ have an effect (Tables I and I1 and Fig. 2), and it increases the amount of both enzymes in the complex, while M e and Caz+ mainly enhance the amount of glutamate dehydrogenase associated with aminotransferase ( Table 11). Zn", as the other metal ions, did not precipitate either enzyme alone (Table 11) and did not alter aminotransferase activity (not shown).
The effects of metal ions on enzyme-enzyme interaction and glutamate dehydrogenase activity (Tables I and I1 and Fig. 2) were observed in the absence of EDTA, in experiments performed with enzymes, and other constituents of the assays which were treated with Chelex. Thus, the effect of metal ions does not result from displacing EDTA from the enzyme. In addition, treating both enzymes with Chelex (not shown) and adding EDTA to Chelex-treated enzymes do not markedly alter enzyme-enzyme interaction (Table 11) or glutamate dehydrogenase activity (not shown). The slight observed effect of EDTA could result from binding of this chelator to glutamate dehydrogenase (21,22).
The glutamate dehydrogenase used in these experiments was crystallized four times in 0.1 mM EDTA, dialyzed versus

Modification of enzyme-enzyme interaction by metal ions
In these experiments, 1.8 nmol of glutamate dehydrogenase (GDH) and 1.1 nmol of mitochondrial aspartate aminotransferase (Asp-AT) were incubated either alone or together with the indicated concentrations of metal ions or EDTA in 1.0 ml of polyethylene glycol (14%, w/v), 14 mM potassium phosphate, pH 7.0, at 25 "C. The solutions were then centrifuged and assayed as described under "Materials and Methods" and in the legend to Fig. 1. Enzymes and buffer were treated with Chelex as described under "Materials and Methods." When glutamate dehydrogenase, prepared by this method, was assayed with neutron activation (both short-and longterm irradiation schedules) and atomic absorption spectroscopy, only 0.3 eq of Ca2+/enzyme polypeptide chain was found. The former method screens for 50-60 metals including Zn". Thus, it can be concluded that only a small amount of Ca2+ was associated with the glutamate dehydrogenase used in these experiments.
Effect of Leucine-While metal ions enhance enzyme-enzyme interaction, they also reversibly inhibit glutamate dehydrogenase (Fig. 2). However, leucine can reverse this inhibition (Fig. 3, curves A and D). Also, Mg2+ slightly enhances activation of this enzyme by leucine in either the presence or absence of Zn2+ (Fig. 3, curves A and B versus C ) . That is, the rate in the presence of Mg2+ and leucine is actually higher than it is in the presence of leucine alone.
While leucine can reverse inhibition by and even activate glutamate dehydrogenase in the presence of metal ions, it does not decrease the ability of metal ions to enhance enzymeenzyme interaction (Table 111). Consequently, both glutamate dehydrogenase activity and enzyme-enzyme interaction are optimal in the presence of leucine and metal ions. This would not be the case if leucine were functioning as a chelator which removed the metal ion from glutamate dehydrogenase or if leucine competitively displaced the metal ion from this enzyme. Furthermore, ornithine and alanine, which have a low affinity for glutamate dehydrogenase, do not decrease inhibition by Zn2+ (Fig. 4, curves C and D ) , even though the latter is even a better chelator of ZnZ+ than leucine (39). Alternatively, valine, which is a considerably poorer chelator of Zn2+  than alanine (39) but is bound to the same site on bovine glutamate dehydrogenase as leucine (40), decreases inhibition but has less of an effect than leucine (Fig. 4, curue B ) .
While the combination of leucine plus metal ions results in optimal enzyme-enzyme interaction, there is no precipitation of either of these enzymes alone under these conditions (Table  111). In these experiments (Table 111 and Figs. 3 and 4), Ca2+ and Mg2C had similar effects.
According to the above results, leucine is not competitive with metal ions for glutamate dehydrogenase. Consequently, M e facilitates activation of glutamate dehydrogenase by leucine. This synergy between leucine and Mg2+ is also shown in Fig. 5. It can be seen that both leucine and M$+ aione These experiments were performed in the presence of 10 mM potassium phosphate, 10 p~ EDTA, pH 7.0, at 25 "C as described under "Materials and Methods." favor association of glutamate dehydrogenase, but leucine has a considerably greater effect in the presence than in the absence of M 2 + (Fig. 5, left, curue A versus C). Furthermore, leucine is not competitive with Zn2+ because it decreases dissociation by saturating levels of Zn2+ (Fig. 5, right, curue  B ) , and Zn2+ decreases association by saturating levels of leucine (Fig. 5, left, curue C uersus D). However, M$+ is probably competitive with Zn2+ because it, unlike leucine, does not prevent dissociation of glutamate dehydrogenase by saturating concentrations of Znz+ (Fig. 5, right, curve C uersus Dl. The above suggests that leucine is not competitive with either metal ion but enhances binding of Mg2+ and decreases binding and the effect of Zn2+. Consequently, when leucine is added in the presence of both metal ions, the effect of Zn2+ is decreased, and glutamate dehydrogenase activity and polymerization are restored to the level observed in the absence of Zn2+ (Figs. 3 and 5, curue A uersus B). Similarly, when Zn2+ is added in the presence of leucine plus Me+, it produces little dissociation of the enzyme (Fig. 5, right, curue A). In these experiments (Fig. 5), the effects of Ca2+ and Mg2+ were again quite similar.
The concept that leucine decreases binding of Zn2+ is consistent with equil~brium dialysis experiments. The addition of 1.0 mM leucine to these experiments (see "Materials and Methods") decreased the amount of Zn2+ bound to glutamate dehydrogenase from 0.9 to 0.6 eq/enzyme polypeptide chain.
Effect of ATP, Malate, and Citrate-ATP and other purine nucleotides have no significant effect on aminotransferase activity (not shown). However, GTP and ADP are bound to nonidentical but overlapping sites on glutamate dehydrogenase in the vicinity of the leucine site. At these sites, ADP activates and GTP inhibits enzyme activity (40-42). GTP and higher levels of ADP are also bound to additional sites on this enzyme (41,43). The effects of ATP, which is a substrate of carbamyl phosphate synthase-I, have not been studied in much detail. As shown in Fig. 6 ( It has been proposed that the second ADP site on glutamate dehydrogenase is the pyridine nucleotide active site because ADP was competitive with DPN (43). However, DPN is bound to both the active and ADP allosteric sites. This accounts for the nonlinearity of double-reciprocal plots of velocity uersus DPN concentration (42) and, possibly, the competition with ADP. TPN, which is not bound to the allosteric site (42), gives linear double-reciprocal plots, and we have found that at pH 7.0, TPN and high levels of ATP were not competitive (not shown). These results indicate that the second ATP site is not the active site. Furthermore, even a high (1 mM) level of DPNH (which is bound to both an active and a substrate inhibition site (41, 42)) does not itself markedly alter either enzyme-enzyme interaction or the effect of ATP on these interactions (Fig. 7 , curues D and E ) .
The predominant form of ATP in uiuo and the actual substrate of carbamyl phosphate synthase-I is Mg-ATP (44). As shown in Fig. 6 (curue B ) and Fig. 7 (curues A and B ) , Mg2+ decreases the ability of ATP to dissociate the enzymeenzyme complex without preventing ATP from activating glutamate dehydrogenase. This is apparently because a high level of free ATP is required for dissociation of the complex, while only a low level is required for enzyme activation. Furthermore, free Mg2+ and a high level of free ATP have opposing effects on the stability of the enzyme-enzyme complex, while neither (in the presence of 100 p~ EDTA) has a marked effect on enzyme activity (Fig. 6). Mg-ATP is apparently not bound to glutamate dehydrogenase (45).
The effects of ATP on enzyme-enzyme interaction can also be seen in the absence of polyethylene glycol. ATP alone decreases the ratio of the turbidity of the combination of the two enzymes to the sum of the turbidity of each enzyme alone (Fig. 8, curue E ) . This is because ATP produces a greater decrease in the turbidity of aminotransferase-glutamate dehydrogenase (Fig. 8, curue C) than it does in the turbidity of glutamate dehydrogenase alone (Fig. 8, curve F). In the presence of M8+, ATP increases this ratio (Fig. 8, curue B ) because it has considerably less of an effect on the turbidity of the mixture of the two enzymes (Fig. 8, curue A ) than it does on the turbidity of Mg2+ plus glutamate dehydrogenase alone (Fig. 8, curve D). This is again consistent with ATP dissociating and Mg2+ facilitating association of the enzymeenzyme complex. Neither ATP nor M e alters the turbidity of the aminotransferase alone (Fig. 8, curue G).
Malate is bound to the active site of glutamate dehydrogenase and decreases interaction between this enzyme and the aminotransferase (46). Even though leucine or low levels of ATP or ADP are bound to allosteric sites, they antagonize the effects of malate on this enzyme and enzyme-enzyme interaction (46). Consequently, while low levels of ATP have no effects on enzyme-enzyme interaction in the absence of malate, in the presence of DPNH or TPNH plus malate, they enhance these interactions apparently by displacing malate from glutamate dehydrogenase (Table IV,

IV
Effect of ligands on aminotransferase-glutamate dehydrogenase interactions In these experiments, 2.2 nmol of mitochondrial aspartate aminotransferase (Asp-AT) was incubated with 1.8 nmol of glutamate dehydrogenase (GDH), with the indicated additions in 1.0 ml of polyethylene glycol (14%, w/v), 14 mM potassium phosphate, 0.1 mM EDTA, pH 7.0, at 25 "C. After 20 min, the solutions were centrifuged for 10 min at 25 "C, and the precipitates and supernatant solutions were assayed. Details of the assay and incubation conditions are described under "Materials and Methods." Less than 10% of the aminotransferase or glutamate dehydrogenase precipitated when incubated alone.

GDH
AsP-AT these conditions, there is less evidence of interaction between it and the ATP allosteric site. Consequently, low levels of ATP increase activity but do not completely abolish inhibition by malate (Fig. 9, curve C) and do not prevent malate from dissociating the enzyme-enzyme complex (Fig. 9, curues E and F ) . In the presence of TPN (unlike TPNH or DPNH), both Mg2+ and low levels of ATP are required for these effects (Fig. 9, curues A, B, and D). M$+ alone, like ATP alone, has little effect on dissociation of the enzyme-enzyme complex by malate, and M P alone slightly enhances inhibition of glutamate dehydrogenase by malate (Fig. 10, top, and Fig. 11, right, curues E and F ) . Thus, when TPN rather than TPNH is the coenzyme, the combination of ATP plus Mg2+ and not ATP alone is required to decrease the effects of malate.
The level of citrate in liver mitochondria can be as high as 8 mM (47). As shown in Fig. 10 (bottom), levels of citrate considerably lower than this produce a marked decrease in enzyme-enzyme interaction. Furthermore, citrate has a considerably greater effect on enzyme-enzyme interaction than malate (Fig. lo), even though 4 mM levels of citrate have little effect on either aminotransferase or glutamate dehydrogenase activity (not shown). Citrate, unlike malate, is probably not bound to the active site of glutamate dehydrogenase because the small amount of inhibition observed is noncompetitive with glutamate (not shown). Furthermore, the effects of citrate, unlike malate (46) on aminotransferase-glutamate dehydrogenase interactions in the presence of DPNH or TPNH, are not decreased by a-ketoglutarate (Table IV, line 5 versus  9). In addition, citrate, unlike malate, has essentially the same effect in the absence of coenzymes as in the presence of either oxidized or reduced pyridine nucleotide (Table IV) however, bound to glutamate dehydrogenase because it markedly decreases the activity of this enzyme in the presence of ATP or leucine (Fig. 11, left, curve A uersus E, and right, curue B versus D). This apparently does not result from binding of citrate to the overlapping ATP-leucine site, because citrate does not completely eliminate the effect of these activators (Fig. 11) and, in the presence of DPNH, low levels of ATP do not decrease dissociation of the aminotransferaseglutamate dehydrogenase complex by citrate (Table IV, line 5 uersus 6). Therefore, citrate is apparently bound to a site on glutamate dehydrogenase which is distal from both the active and ATP allosteric sites. Binding of citrate to this site has little effect on enzyme activity but decreases activation by ATP or leucine and decreases enzyme-enzyme interaction.
As shown in Fig. 10 (bottom), citrate decreases enzymeenzyme interaction only when its concentration exceeds that of Mg2+. Furthermore, Mg2+ enhances glutamate dehydrogenase activity in the presence of citrate plus either ATP or leucine (Fig. 11, left,  It is unlikely that the effects of citrate, malate, or high levels of ATP in polyethylene glycol result from these ligands increasing the solubility of either enzyme alone rather than dissociating the enzyme-enzyme complex. Under these experimental conditions and even when the level of enzyme is 5fold higher, both glutamate dehydrogenase alone and the aminotransferase alone are essentially completely soluble in polyethylene glycol (less than 5% precipitated; Refs. 5, 13, and 18), and these ligands did not increase the solubility of either enzyme alone (not shown). In addition, it has been shown that, in the absence of polyethylene glycol, malate decreases cross-linking between these two enzymes (46), and ATP decreases the turbidity of the enzyme-enzyme complex (Fig. 8).
Citrate and high levels of ATP are similar in that both decrease enzyme-enzyme interaction, and this is reversed by MgZ+. Also, both decrease activation of glutamate dehydrogenase by low levels of ATP, both are chelators, and both markedly increase ionic strength. However, it is unlikely that chelation plays a major role in the effect of citrate and ATP on either glutamate dehydrogenase activity or enzyme-enzyme interaction in the absence of added metal ions. This is because, as mentioned above, only a small amount of Ca2+ is associated with glutamate dehydrogenase. Furthermore, both citrate and high levels of ATP slightly decrease glutamate dehydrogenase activity, while metal ions inhibit this enzyme. Thus, if citrate and high levels of ATP functioned as chelators, they would be expected to be activators. In addition, treating both enzymes and buffer with Chelex did not markedly decrease enzyme-enzyme interaction (not shown), and added EDTA had Iittle effect on Chelex-treated enzymes (Table 11). Furthermore, isocitrate, which is a 10-fold poorer chelator of Mg2+ than citrate (48), has an effect almost equal to that of citrate (Table IV). Finally, dialyzing both enzymes uersus 4 mM citrate had less of an effect on enzyme-enzyme interaction than adding citrate (Table V).
Citrate and high levels of ATP also do not decrease enzymeenzyme interaction and inhibit glutamate dehydrogenase as a result of increasing ionic strength. NaCl, in comparable ionic strengths, has considerably less of an effect than citrate on either glutamate dehydrogenase activity in the presence of ATP or leucine or enzyme-enzyme interaction (Figs. 11 and 12). F u~h e~o r e , in the absence of coenzyme, 1 mM citrate (ionic strength, 5 X produces a greater decrease in enzyme-enzyme interaction than 4 mM malate (ionic strength, 12 X (Table IV). Also, citrate itself does not markedly inhibit glutamate dehydrogenase in the absence of leucine or ATP (Figs. 11 and 12), and 1 mM levels of DPNH, TPNH, TPN, and DPN have considerably less effect than ATP on enzyme-enzyme interaction (Table IV)  In these experiments, 2.2 nmol of mitochondrial aspartate aminotransferase (Asp-AT) and 1.8 nmol of glutamate dehydrogenase (GDH) were incubated either together or with either 2 nmol of carbamyl phosphate synthase (CPS) or 8 nmol of citrate synthase (CS), plus the indicated ligands in 1.0 ml of polyethylene glycol (14%, w/v), 14 mM potassium phosphate, 0.1 mM EDTA, pH 7.0, at 25 "C. After 20 min, the solutions were centrifuged and assayed as described under "Materials and Methods." Chelex-treated enzymes refers to enzymes which had been either chromatographed on a column of Chelex or dialyzed versus Chelex as described under "Materials and Methods." The buffer was chromatographed on Chelex, and all ligands were dissolved in the Chelex-treated buffer. Where indicated, enzyme was dialyzed twice uersm 500 mi of 4 mM citrate in the above DhosDhate buffer.  that this effect of carbamyl phosphate synthase-I results from release of a metal ion associated with this enzyme. However, 2 p M of this enzyme enhance aminotransferase-glu~mate dehydrogenase interaction even in the presence of chelators such as 4 mM citrate and 0.1 mM EDTA (Table V, line 4  versus 5). Under these conditions, even 0.5 mM Mg+ has no effect on aminotransferase-glutamate dehydrogenase interaction (Table V, line 4 versus 8). Furthermore, treating carbamyl phosphate synthase-I with Chelex does not decrease but slightly enhances its effect (Table V, line 5 versus 6). Therefore, a metal ion associated with carbamyl phosphate synthase-I is unlikely to be responsible for its ability to enhance aminotransferase-glutamate dehydrogenase interactions. The aminotransferase can also form a complex with carbamyl phosphate synthase-I (13). However, enhancement of aminotransferase-glu~mate dehydrogenase interaction by carbamyl phosphate synthase-I is apparently not the result of nonspecific absorption of glutamate dehydrogenase to the aminotransferase-carbamyl phosphate synthase-I complex (13). If this were the case, then it would be more likely to take place in the presence of malate than citrate, because only the latter markedly decreases interaction between carbamyl phosphate synthase-I and the aminotransferase (12, 13). Another reason for believing that there is interaction among all three enzymes is that considerably lesser amounts of anionic proteins, other than giu~mate dehy~ogenase, associate with these two enzymes (131, and high levels of citrate synthase have considerably less of an effect than carbamyl phosphate synthase-I in enhancing aminotransferase-glutamate dehydrogenase interactions (Table V, line 5 versus 7). Thus, these plus results, which demonstrate cross-linking between glutamate dehydrogenase and carbamyl phosphate synthase-I in the mitochondrial matrix (14), suggest that a complex is formed between all three enzymes.

Binding to Inner ~~t o c~n d r~l
Membrane-It has previously been shown that mitochondrial malate dehy~ogenase and aspartate aminotransferase have a high affinity for the inner surface of the inner mitochondrial membrane (3, 23). The binding process is saturable and specific in that a series of nonmatrix proteins, including cytosolic malate dehydrogenase and aspartate aminotransferase, have a low affinity for these vesicles (3). Citrate synthase was bound more poorly than the other two mitochondrial enzymes, and binding of this enzyme was decreased by malate dehydrogenase (3). In addition, as shown in Table VI (lines Id), the aminotransferase can displace malate dehydrogenase. These and previous results (3) indicate that there is a specific site on the membrane for these enzymes and the order of specificity is: aminotransferase, malate dehydrogenase, and citrate synthase. Glutamate dehydrogenase has a low affinity for the membrane, and its binding is also slightly decreased further by the aminotransferase (Table VI, lines 5-6). Carbamyl phosphate synthase-I is essentially not bound (not shown). Only a trace (less than 1%) amount of these enzymes was associated with the membrane in the absence of added enzymes.
As shown in Table VI (lines 2 and 7), 2 mM NaCl has little effect on binding of the aminotransferase to the membrane. This is similar to previous results (3) which indicated that 5-10 mM KC1 or NaCl was required to produce a 2-fold decrease in binding of malate dehydrogenase. However, 1 mM MgClz produces a marked decrease in binding of the aminotransferase (Table VI,

Binding of mitochondrial enzymes to inner mitochondrial membrane
In these experiments, enzymes and inverted inner mitochondrial membranes were incubated for 15 min at 0 "C, sedimented, and washed, and the pellets and supernatant fractions were assayed for enzyme activity as described under "Materials and Methods." The abbreviations used are the same as given in the legend to Table I.

Enzyme(s) added Additions
Enzyme bound