Mechanisms of Substrate Binding with Glutamine Synthetase EQUILIBRIUM ISOTOPE EXCHANGES

SUMMARY Equilibrium isotope exchange kinetics were used to in-vestigate the sequences of substrate binding with ovine brain, pea seed, and Escherichia coli glutamine synthetases. Without exception, the relative rates of exchange are (gluta-mate % glutamine) > (NH3 % glutamine) > (Pi % ATP) N (ADP % ATP). This suggests that the rate of net turn-over at saturating substrate levels is limited more strongly by the rate of nucleotide release than by the rate of covalent interconversion. With the ovine brain enzyme, the kinetic patterns of equilibrium exchange are consistent with a partially ordered sequence of substrate binding, where ATP binds before NH,, glutamine leaves before ADP, but glutamate and phosphate are bound and released randomly on their respective sides of the reaction. This particular binding order does not permit observation of those partial exchange reactions expected if an enzyme-bound y-glutamyl phosphate intermediate were formed. Upon variation of the concentration of all substrates in constant ratio at equilibrium, the pea seed, the Mn2+-activated adenylylated and the Mg2+- activated unadenylylated E. coli enzymes exhibit exchange kinetic patterns


Mechanisms
This suggests that the rate of net turnover at saturating substrate levels is limited more strongly by the rate of nucleotide release than by the rate of covalent interconversion.
With the ovine brain enzyme, the kinetic patterns of equilibrium exchange are consistent with a partially ordered sequence of substrate binding, where ATP binds before NH,, glutamine leaves before ADP, but glutamate and phosphate are bound and released randomly on their respective sides of the reaction.
This particular binding order does not permit observation of those partial exchange reactions expected if an enzyme-bound y-glutamyl phosphate intermediate were formed. Upon variation of the concentration of all substrates in constant ratio at equilibrium, the pea seed, the Mn2+-activated adenylylated and the Mg2+-activated unadenylylated E. coli enzymes exhibit exchange kinetic patterns consistent with random substrate addition and release.
The kinetic data also suggest synergism of substrate binding between glutamate and ATP. A brief section on theory and interpretation of kinetic patterns of, equilibrium exchanges as related to substrate-binding order is presented.
Recent studies on glutamine synthetase in Escherichia coli (14) have led to an understanding of the central role this enzyme plays in regulation of nitrogen metabolism.
The research described here is part * This work was supported in part by Grants GB 34751 and GU 3182 of the National Science Foundation, and by Grant 2402-Gl from the Petroleum Research Fund of the American Chemical Society.
of an investigation of differences in mechanism of action among the enzymes from bacterial, plant, and mammalian sources. Specifically, this study investigates whether the mechanisms of substrate binding differ markedly, as suggested by earlier data. With the ovine brain enzyme, the observed enhancement of glutamate binding by added ATP has led Meister and co-workers (10, 111 to propose that reactant' binding is ordered: first ATP binds, then glutamate, then ammonia. The tight (E ATP .Glu) complex in ammonia-free solution appears to exist as (E.ADP. Glu-P) . ADP is not released, however, and hence an independent ADP + ATP exchange is not observed.
This suggests that product release also may be ordered.
With the pea seed enzyme, the equilibrium exchange data of Graves and Boyer (la), observed upon variation of substrate pairs, indicated a partially compulsory order of binding: preferential binding of NH, after L-glutamate and compulsory release of ADP before glutamine and possibly before Pi. The kinetic patterns that led to these conclusions might be attributable in some cases to competitive substrate interactions, however, as found in similar experiments with the E. coli enzyme (13).
Previous kinetic data (13) with adenylylated E. coli glutamine synthetase (Eg) were consistent with a random order 'substratebinding scheme, but were obtained with a mixture of Mg* and MnZ+. Random substrate binding is verified in the present study for both the adenylylated and unadenylylated forms, activated purely by Mn* and Mg"+, respectively.
The technique employed in thii study is isotopic exchange kinetics at chemical equilibrium, a powerful and incisive approach to enzyme mechanisms (see under "Theory").
The sorts of kinetic patterns that can result from synergistic substrate binding, as opposed to those characteristic of ordered substrate binding, are defined by computer-derived model studies. Both enzymes showed single bands upon electrophoresis on polyacrylamide gels at pH 8.3 and also in procedures with enzvme in 0.1% sodium dodecyl sulfate (16). The freezing or lyophjlization steps were replaced by storage at 4" in 50yo glycerol-buffer (pH 7, 20 mre Tris, 2 mM EDTA, and 2 rnM dithiothreitol).
One unit of activity was defined as in the preparative procedures (14,15). E. coli glutamine synthetase, adenylylated and unadenylylated, was purified to homogeneity by the procedures of Shapiro and Stadtman (2).
The reaction components L-glutamate, L-glutamine, ADP, and ATP were of highest quality obtainable, supplied by Sigma Chemical Co. Radioactive substrates were obtained from New England Nuclear.
['6N]Ammonia (99%) was from Bio-Rad Laboratories. DEAE-cellulose (formate) used for separation or reaction components was prepared from Whatman DE52 by batchwise washing in 0.5 M formic acid. then washing to neutrality with double distilled water.
' Methods-The procedures for preparing a series of reaction solutions at chemical equilibrium. maintaining ionic strength and pH constant, have been described elsewhere (13). The series of solutions, usually 1.0 ml each, was then thermostated, a constant amount of enzyme added to each, and allowed to equilibrate for 10 min to establish chemical equilibrium exactly.
The isotopic exchange reaction was then initiated upon addition of a low level (<l% of the total pool size) of high specific activity substrate, e.g. L-I"C]glutamate.
The reactions were then incubated for a _ .
specific time period, usually 30 min or less, then the reactions were stormed bv addition of EDTA (20'% excess over metal ion present), dilu;ion to 3.0 ml, then chilling'to.0" or freezing prior to separation of the reactants undergoing isotopic exchange.
Reactions involving ovine brain and pea enzymes were carried out at 37" and 30", respectively, all at pH 6.5 with 50 mM MgC12, since the results of Monder (17) and Elliott (15) indicate that each of these systems exhibit optimal activities under these conditions. The equilibrium constant for the reaction at 30" and 37" was verified by procedures described elsewhere (13) to be 460 f 30. Kinetics with the unadenylylated E. coli (I%) enzyme employed reaction conditions satisfying a K,, = 1209 (18) at pH 7.0, 37", with Mge+ present.
The E= enzyme exhibits -5Oo/, activity under initial velocity conditions at pH 7.0, compared to that at pH 7.5, the pH optimum (l-4).
The gradient elution procedure described by Wedler and Boyer (13) was modified to three separate stepwise elution procedures.
Each stopped and diluted reaction mixture was annlied to a column (0.7 X 13 cm) of DEAE-cellulose.
Elution w<th first 20 ml of water, then 20 ml of 0.5 N HCl, separated alutamine and NH, from glutamate.
Phosphate and ATP were separated by washes with SO ml of 0.05 N HCl, then 20 ml of 0.5 N HCl, respectively.
ADP and ATP were separated by washes with 25 ml of 0.05 N HCl and 20 ml of 0.5 N HCl, respectively. Samples of the pools thus separated were counted by liquid scintillation techniques.
From calculated disintegrations per min values, the number of micromoles exchanged between pools per unit time was calculated by equations presented previously (13).
Exchange reactions involving 15N-labeled ammonia were carried out with L-['*C]glutamate also present for comparison of exchange rates. At chemical equilibrium, one can observe flux of isotopic label from one pool to another with no net change in pool size. Equations for the rates of exchange between A and P (A ti P, called R) and B and Q (B % Q, called R') have been derived (21,22).
One may use equilibrium exchange rates to distinguish between random and particular compulsory sequences of substrate binding by observing the responses of R and R' to variations in levels of substrates above and below K, values. One may vary either substrate pairs (one reactant, one product) or all substrates in constant ratio at equilibrium.
As an example, consider the exchange pattern of Scheme 1.
A,P const. R (BsQ) RATE CB* Ql &2HEME 1 This pattern is one expected for the binding mechanism of .Equation 3. Note that R' rises smoothly to a maximum, but R peaks then falls toward zero. The basis for this behavior is that in a random order system A and P can escape from the complexes EA, EP, EAB, and EPQ, but, in the ordered system of Equation This lack of inhibition can be used to exclude the binding of A after B or the release of Q before P.
In addition, a comparison of relative equilibrium exchange rates at saturating substrate levels also can be used to exclude certain orders of substrate binding. In an ordered binding situation, the exchange rate between the first substrates to dissociate from the central complexes (B G Q in Equation 3) must be as fast as or faster than all other exchanges.
If all exchange rates are equal, then covalent interconversion is likely to be dehnitively rate-limiting.
If the exchange rates are unequal, as is e This nomenclature differs from that of Cleland (20) in that Q is released before P, but is used elsewhere (25) deliberately in equilibrium exchange derivations to indicate that P is the first product to bind in the reverse reaction or at equilibrium. most often observed with actual enzyme-catalyzed exchanges (13, 23), covalent interconversion cannot be the slowest step, and individual substrate dissociation rates determine exchange rates. Finally, it has been shown (24) that feedback modifiers may produce distinctive changes in R and R' with random versus compulsory order-binding systems. In some cases, therefore, modifier-induced inhibition patterns can be used as evidence to support or exclude ordered binding (25).

RESULTS
Ovine Brain Enzyme-Variation of all substrates in constant ratio at equilibrium produces the kinetic patterns seen in Fig. 1.
Substrates were raised to values above their published K, values (26). The rate of glutamate % glutamine exchange is faster than the rates of Pi % ATP or ADP + ATP at all levels of substrates, and the rate of glutamate ti glutamine exchange rises smoothly toward a maximum.
This argues that none of the other four substrates bind after glutamate or glutamine in a compulsory manner.
However, the Pi % ATP and ADP % ATP exchanges first rise to a maximum then fall toward zero as substrate levels increase.
These patterns may result from ordered substrate binding, or alternatively from substrates acting as negative modifiers at noncatalytic sites. Variation of all substrates in constant ,ratio blocks out competitive effects at the catalytic site. If ordered binding is the principal effect, then either NH8 or n-glutamate bind after ATP, or glutamine is released before ADP or Pi, or both may be true. Alternative explanations for such peaking and inhibition patterns, other than ordered substrate binding, are presented under "Discussion." The saturation curves in Fig. 1 show sigmoidal character at low substrate, especially in the glutamate ti glutamine exchange.
It was found that by repeating the experiment with different absolute levels and ratios of glutamate, glutamine, and Pi it was possible to enhance or abolish the sigmoidicity in glutamate it glutamine and increase or decrease the maximal rate as well. These phenomena were not investigated further, although a likely explanation seems to be competitive interactions between glutamate and glutamine, leading to dead-end complexes such as (E.Glu.ADP. Pi) or (E.NH,.Gln.ATP). Other explanations for the sigmoidal effects are considered below. The absence of inhibition patterns in any exchange rate curve even at high substrate implies that the sequences of binding of glutamate relative to ATP and of glutamine relative to Pi, are completely random.
The data in Fig. 2 also indicate that there may be some mutual synergism of binding between amino acid and nucleotide substrates under dynamic conditions. Initial velocity kinetic data indicate a K,,, value for ATP of 1.5 mu at pH 6.5, 37", with 50 mM MgCl, present.
In the experiment of Fig. 2A, ATP was used at fixed levels~ above and below this value. This altered the half-saturation level of glutamate and glutamine in the glutamate ti glutamine exchange.
In this same figure, Pi * ATP shows a half-saturation value below that for glutamate + glutamine.
In Fig. 2B, upon varying Pi and ATP with fixed glutamate and glutamine levels, the glutamate % glutamine exchange half-saturates at a lower substrate level than does Pi + ATP. In Fig. 3A, Pi * ATP rises smoothly to a maximum, indicating that ADP is not released before Pi. The glutamate + glutamine exchange, however, rises to a peak, then is partially inhibited.
Such partial inhibition is usually explained in terms of a random sequence of binding (see Equation  2) with one branch kinetically preferred. However, since (glutamate + glutamine) > (Pi % ATP) under all conditions, it is not logically consistent that glutamate binds preferentially before ATP, or that glutamine is released after ADP (see under "Theory"). Tight complexation of amino acid and nucleotide substrates under saturating conditions, due to the mutual synergism of binding noted above, might be another possible explanation for this partial inhibitory effect. Kinetic models designed to test this possibility are derived and discussed below. Yet another explanation is that nucleotide binds to a noncatalytic, separate modifier site. This is supported by observations of Wellner and Meister (27) that the stoichiometry of ADP binding to ovine brain enzyme subunits is at least 2: 1. The mode of modifier action likely to produce the pattern in Fig. 3A is a differential, partial inhibition of glutamate or glutamine dissociation.
Present data do not support or disprove this idea.
In Fig. 3B, upon raising the levels of NH3 and glutamine, the tion.
This could, a priori, result from NHz binding after ATP, or glutamine being released before Pi, or both. The data of Fig. 2, however, showed that Pi and glutamine bind randomly relative to each other.
Thus, the inhibition of Pi + ATP in Next, experiments were conducted to determine whether either glutamine or Pi is released in an ordered manner before ADP on the product side of the reaction.
Thus, the ADP * ATP exchange kinetics were observed upon varying the levels of the substrate pairs (A) ATP and Pi, (B) n-glutamate and L-glutamine, and (C) NH, and n-glutamine. Fig. 4 shows the results of these experiments.
In Fig. 4A, varying the levels of Pi and ATP simply allows the ADP -ATP exchange to rise smoothly to a maximum. This argues that Pi is not released,before ADP in a compulsory manner. Also, the data in Fig. 3A showed that ADP is not released before Pi. Thus, Pi and ADP are released randomly relative to each other.
As noted above, Pi is also released randomly relative to glutamine. glutamate + glutamine exchange rises smoothly to a maximum, Interestingly, Fig. 4B shows that glutamate and glutamine which indicates that NH3 does not bind after glutamate. The produce a definitive inhibition pattern in the ADP % ATP Pi s ATP exchange, however, shows a rise, then strong inhibi-exchange. This could result from glutamate binding after ATP or from glutamine being released before ADP.
The first of these two possibilities is excluded, since it was shown in Fig. 2A that glutamate does not bind after ATP in a compulsory manner. This strongly suggests compulsory release of glutamine before ADP.
The reciprocal plot for fractional inhibition (l/i), versus (Glu)-1 indicates complete inhibition of nucleotide release at infinite glutamate and glutamine levels. In Fig. 4C, variation of NH3 and glutamine also produces strong, complete inhibition of ADP 6 ATP at saturation, as would be expected for NH, binding after ATP and glutamine release before ADP. This partially ordered binding is the one most consistent with all present data.
A logical consequence of this particular binding pattern is that the independent exchanges indicative of y-glutamine-P will not be observable.
The observation of a nucleotide-independent, Pi-dependent NH, % glutamine exchange is not possible since ATP must bind before NH3 in the forward reaction, or ADP must bind before glutamine in the reverse reaction.
The observation of an NH,-and glutamine-independent, glutamate-dependent ADP % ATP exchange is not possible, since glutamine must be released before ADP, and glutamine is not formed in the NH8free partial system. These predictions were tested by our attempting to observe these exchanges with the complete system and a variety of partial reaction systems. No exchange reactions were catalyzed by the ovine brain enzyme at appreciable rates unless all substrates were present.
These results agree with an earlier report (10) of failure to observe independent ADP + ATP or Pi + ATP exchanges under somewhat different conditions. Pea Seed Enzyme-Equilibrium exchange data reported earlier by Graves and Boyer (12) were interpreted as consistent with a partially ordered substrate binding mechanism for this enzyme (see introductory remarks). Competitive effects among substrates might account for some of these inhibition patterns, as shown with the E. coli enzyme (13). Therefore, to block out such effects at the active site, the concentration of all substrates was varied in constant ratio at equilibrium.
The results of this experiment on the kinetics of glutamate ti glutamine, Pi +ri ATP, and ADP * ATP are shown in Fig. 5A. Substrate levels were at or above those used previously (12), and above those reported as K, values (15,28).
All exchange rates rise smoothly to a maximum without inhibition.
This excludes ordered binding among the components involved in these exchanges: glutamate, ATP, ADP, glutamine, and Pi. The NH3 % glutamine exchange was not observed in this experiment, so it is possible that ATP or glutamate bind after NH,.
However, since NH, 3 glutamine is faster than ADP * ATP (see Table I), ATP cannot bind after NHa. Again, at subsaturating substrate levels, some sigmoidal character is apparent in the exchange rate curves of Fig. 5A. This was not observed upon variation of substrate pairs (12). E. coli Enzyme-The order of substrate binding to unadenylylated (Er;7) glutamine synthetase from %. coli was probed by variation of all substrates at equilibrium, 37", pH 7.0, with [Mgz+] = [ATP] + [ADP] + 10 mM. The results are shown in Fig. 5B. The saturation curves show no inhibitory patterns indicative of ordered binding among these substrates.  b Fig. 4 and Tables I and II of Ref. 13.
c Taken at f = 0.5 in Fig. 4.
Glutamate % glutamine is faster than Pi + ATP, and some sigmoidicity in Pi % ATP rate is seen.
In an earlier study with adenylylated enzyme (&) the substrate binding order was determined to be random at pH 6.50 and 37" (13). A mixture of metal ions was used in these experiments, so as to produce mainly MgATP and Mn-enzyme.
The use of mixed metal ions, however, might introduce some doubt regarding detailed kinetic behavior of this complex system, since, e.g. Mg* can noncompetitively block adenylylated enzyme activity.
Thus, the experiment involving variation of all substrates at equilibrium was repeated using Mn* only, with [MnH] = [ADP] + [ATP] + 1 mM and Ec! enzyme.
The kinetic pattern obtained was essentially the same as that reported earlier (13). Neither exchange is inhibited, and some sigmoidicity is seen in Pi * ATP at f 2 0.2. The relative maximal values of (glutamate + glutamine)/(Pi + ATP) rates are unchanged from the previous experiment. Rate-limiting Steps-A comparison of the relative rates of equilibrium exchange between substrate pools catalyzed by various glutamine synthetase enzymes (Table I) reveals several effects. Without exception, glutamate * glutamine and NH8 % glutamine are faster than Pi + ATP or ADP + ATP. This suggests that the rate of nucleotide release contributes more significant.ly to the rate of net runover than do the covalent interconversion steps under saturating substrate conditions. Second, the ratio of rates (glutamate % glutamine)/(Pi * ATP), differs among these enzymes.
This may suggest subtle but important differences in active site steric adaptation to bound substrate, resulting in different rates of substrate release.

DISCUSSION
Ovine Brain Enzyme-The observed equilibrium exchange kinetic patterns can be interpreted to exclude unambiguously certain substrate binding mechanisms.
Glutamate and phosphate do not bind in an ordered manner relative to other substrates, and ADP is not released before glutamine, as argued under "Results," due to lack of inhibition patterns diagnostic of such ordering.
The inhibitions observed in Figs. 1 to 4, however, strongly suggest but do not absolutely prove that NH2 binds after ATP, and glutamine must be released before ADP. This partially ordered binding scheme may be represented by  Fig. 2B) may, however, provide one example of this effect.
(b) Substrates bound to noncatalytic modifier sites could potentially block either catalytic steps or the association-dissociation steps for certain substrates.
As already shown in detail by model kinetics (24), inhibition of catalysis effectively removes active enzyme and thus inhibits all exchanges in constant ratio. This is not observed in Figs. 1 to 4. However, if a modifier differentially blocks dissociation or both association and dissociation for one substrate but not others, in a random binding system, one exchange would be suppressed preferentially. Present equilibrium exchange kinetic techniques are not capable of distinguishing this effect from ordered binding effects. Under initial velocity conditions the pea seed enzyme exhibited substrate inhibition by ammonia that might be interpreted in this manner (28). Thus, for the ovine brain system, Equation 4 represents the substrate binding mechanism most consistent with all present data. The above discussion is not intended to indicate proof of a single substrate binding order for the ovine brain enzyme.
As with analogous systems, e.g. succinyl-CoA synthetase (29), a complex subarray of minor pathways is almost certain to exist.
Pea Seed and E. coli Enzymes-Upon comparison with model kinet.ic patterns (cf. "Theory"), the data of Fig. 5 clearly show that these enzymes exhibit random binding of glutamate, glutamine, ATP, Pi, and probably ADP.
These results argue against a previous interpretation for the pea seed enzyme, for which a partially ordered system had been postulated (12). With the E. coli enzymes, it is now clear that the random nature of substrate binding is not a function of pH, metal ion, or state of adenylylation.
Explanations for the sigmoidal responses of exchange rates observed in Figs. 1 and 5, A and B, should be considered.
The apparent cooperativity of binding of substrates is not attributable to either homotropic (30) or heterotropic interactions, nor to modifiers acting as positive modifiers at noncatalytic sites, since such effects do not occur in experiments in which pairs of substrates were varied (as in Figs. 2 to 4, and the figures in Refs. 12 and 13). Also, initial velocity kinetics carried out in this laboratory failed to detect any homotropic cooperativity in the binding of either L-glutamate or ATP with any of the four enzymes of this study.
Several lines of evidence are consistent with special active site conformations induced by bound substrates with several of the glutamine synthetase enzymes studied here. The binding of glutamate to the ovine brain enzyme was previously shown to be great,ly improved in the presence of ATP (lo), although (E.Glu. ATP) yields the nondissociating complex (E'Glu-P.ADP) in ammonia-free solution.
Under dynamic conditions the variation in half-saturation levels for exchanges involving glutamate, glutamine, ADP, and ATP seen in Figs. 1 to 3 are interpretable as mutual synergism of binding: glutamate with ATP, and glutamine with ADP.
Such synergistic binding may also exist in the bacterial and plant enzymes, based on similar comparisons of kinetic binding curves (12,13). With the mechanistically analogous succinyl-CoA synthetase, cofactor nucleotide has been observed to induce an enzyme conformation that enhances catalysis rather than substrate binding processes (29,31,32).
The probable existence of synergistic substrate binding under dynamic conditions raises several basic questions.
(a) If the synergism occurs so as to strongly decrease substrate dissociation rates, could it produce inhibitions of exchange rates that would resemble the peaking and inhibition patterns diagnostic of ordered binding?
(b) Are the sigmoidicities observed in the binding curves of Figs. 4 and 5 mainly indicative of multiorder kinetics at subsaturating levels of substrates or also of intrasite synergism of substrate binding?
In an attempt to answer the above questions, the following model for synergism of substrate binding was assumed.
The random order, two-reactant, two-product system of Equation 3 can also be written as shown in Equation 5.

(5)
Equations for the exchange rates R(A fir P) and R' (B % Q) are derived elsewhere (22,23). Consider that bound B and Q make the binding of A and P tighter by slowing the dissociation rates from the central complexes EAB and EPQ at the k-4 and k-7 steps. Fig. 6 shows computer calculations for the A * P exchange rate as a function of [ At the point where k--,l = k-7 = 0, the compulsory ordered system of Equation 4 occurs.
First, with a st,rongly preferred order of substrate release (k--l and k-r reduced lo-to loo-fold), levels of B and Q must be about 5 times their K, values to produce any inhibition in R, the A * P exchange rate. Thus, preferential orders of binding would be difficult to detect.a Further, substrate synergism in complexation, unless extremely tight, cannot produce equilibrium kinetic patterns diagnostic of ordered binding.
To produce the definitive peaking and inhibition patterns seen in Figs. 1 to 4, the model shows that k-d and k-7 must be essentially zero relative to k-1 and k-s so that an essentially compulsory order binding mechanism is operative.
Second, upon variation of all substrates in the model (Fig. 6A) In conclusion, the present data indicate that important differences exist in the mechanisms of substrate binding and interaction in the active sites of glutamine synthetases from mammalian, plant, and bacterial sources. Other differences in mechanism are the subject of future investigations.

AcknozoZe~mentsThanks
are due to Mr. Steven Kowalczykowski, Miss Terry Lerner, Mr. Tim Flavin, and Mrs. JoAnn Cencula for their valuable technical assistance during various phases of this work.
Dr. Robert Reeves of this department is thanked for providing access to and help with the mass spectrometer.
I also thank Dr. P. D. Boyer for a critical reading of the manuscript.