High Affinity Interactions of GTP (cid:103) S with the Heterotrimeric G Protein, Transducin EVIDENCE AT HIGH AND LOW PROTEIN CONCENTRATIONS*

A well known difference in nucleotide binding char- acteristics between heterotrimeric G proteins and small GTP binding proteins of the Ras superfamily is that the former bind GTP or guanosine 5 (cid:42) - O -(3-thiotriphosphate) (GTP (cid:103) S) with a much lower affinity ( K d (cid:59) 10 8 –10 7 M ) than the latter ( K d (cid:59) 10 (cid:50) 11 –10 (cid:50) 10 M ). We report here that the (cid:97) subunit of the heterotrimeric G protein transducin (G t ) binds GTP (cid:103) S with an affinity comparable to that of Ras. High affinity binding was suggested by GTP (cid:103) S titrations of rod outer segment samples with G t concen- trations in the range of 7 n M to 300 n M ; the results were more consistent with a dissociation constant for GTP (cid:103) S in the subnanomolar range, than with one in the 10 (cid:50) 8 – 10 (cid:50) 7 M range typically reported for heterotrimeric G proteins. Equilibrium binding experiments with G pro- tein concentrations in the subnanomolar to nanomolar range confirmed this conclusion and revealed a dissoci- ation constant of 50 p M . Thus, transducin’s affinity for GTP (cid:103) S, and by inference, for GTP, appears to be approximately three orders of magnitude higher than previ- ously reported. These results raise the possibility that some results obtained with high concentrations of nu- cleotide analogues may be due Equilibrium Binding Assays— Equilibrium binding experiments carried out using of M M , m M 0.1 m M ovalbumin (for or Buffer B m M MOPS, 7.4, m M KCl, m M 2 m M MgCl 2 , 1 m M DTT, 0.1 mg/ml ovalbumin (for comparison to previous studies done at similar pH values). No significant differences were detected in binding results obtained with these two buffers. Samples of freshly bleached rod outer segments (3–150 n M R*) were mixed with GTP (cid:103) 35 S ( (cid:59) 100 Ci/mmol) to various final concentrations and incubated on a rotating mixer at 23 °C (1 or 2 ml, final volume) for periods of time ranging from 3 to 17 h. Then they were filtered through 0.22- (cid:109) m pore-size nitrocellulose filters, in a vacuum filter manifold, and washed with 1 ml of ice-cold buffer. Filters were removed with the vacuum on to remove solution from the edges. The solution passing through the filters was collected using funnels and scintillation vials positioned below the filters. Radioactivity bound to the filters and in the solu- tions was measured by scintillation counting. binding was determined by 1 (cid:109) M nonradioactive GTP (cid:103) S to the 30 to addition of GTP (cid:103) 35 S, and the radioactivity bound to filters in these samples (typically 0.95% (cid:54) 0.05% of the total radioac- tivity added) was subtracted from the total radioactivity bound to the other filters to calculate specific binding. The sum

ments (ROS) and the relative ease of its purification, G t has served as a paradigm for many studies of G protein function. Obviously, one of the most important properties of G proteins is their ability to bind GTP with sufficiently favorable energetics to allow rapid and dramatic changes in their functional properties, so they can switch from inactive to active states when stimulated to release GDP by activated receptors. In the case of G t , the differences between the GDP and GTP forms and the speed with which the switch occurs are remarkable. The GTP form shows a dramatic decrease in affinity for R* and G t␤␥ as compared to the GDP form (3,4). Activation by GTP can occur on a millisecond time scale at physiological levels of GTP (5)(6)(7). Most importantly, perhaps, while the GDP form of G t␣ appears to be capable of activating the effector PDE only at concentrations on the order of 50 M (8), the GTP or GTP␥S form interacts with it much more efficiently, with an apparent affinity for membrane-bound holo-PDE of Ͻ0.2 nM determined by activation (9), or of 2.5 nM determined by light scattering (7). The K d of G t␣ -GDP for the inhibitory PDE ␥ subunit of PDE has been reported to be 3 nM, while the GTP␥S form binds with a K d of Յ0.1 nM (10), corresponding to a Ն30-fold higher affinity.
The observation of K d values for G t␣ -GTP␥S binding PDE or PDE ␥ in the low nanomolar or subnanomolar range seems hard to reconcile with earlier reports of K d values for formation of the G t␣ -GTP␥S complex itself (i.e. G t␣ -GTP␥S) on the order of 50 -100 nM (4,11). This apparent paradox raised the question of whether the G t␣ complex with GTP␥S could be capable of activating its effector at a concentration so dilute that the activator itself should fall apart. Furthermore, titrations measuring PDE activity in rod outer segments as a function of added GTP␥S yielded curves whose shapes were more consistent with stoichiometric binding of a very high affinity ligand than with reversible association of a ligand with a K d value similar to those reported for G t and other G proteins (9).
A noteworthy feature of the literature on affinities of GTP␥S and other GTP analogues for G proteins (including several reports that are conveniently tabulated in Ref. 4) is that most of the experiments reported have been conducted at concentrations of total G protein on the same order as the apparent K d values that were determined. Lower K d values have tended to be reported from experiments at lower protein concentrations, while higher K d values have been reported at higher protein concentrations. For example G s at concentrations of 12 nM was reported to bind GTP␥S with a K d of 5-10 nM (12), while for the same protein assayed at 400 nM (13) a K d of 700 nM was reported. In addition, in most cases where plots of bound versus added nucleotide have been presented, no attempt has been made to distinguish between total and free (i.e. not bound to protein) ligand. It has been well established in the case of the small GTP-binding proteins of the p21 ras superfamily that K d values in the picomolar range were mistaken for values in the high nanomolar range when total protein concentrations in the nanomolar range were assayed (reviewed in ref. 14). From a comparison of the three-dimensional structures of p21 ras (15) and G t␣ (16) with nonhydrolyzable GTP analogues bound, it is far from obvious why the affinity for GTP␥S of the former should be 1000 times stronger than that of the latter.
High affinity interactions frequently go unrecognized because the low concentrations of binding sites (near K d ) required to measure them accurately often present experimental difficulties. However, even at higher than ideal concentrations of binding sites, these interactions can be recognized by careful analysis of the shapes of binding curves, and least squares analysis of fits of the curves to either simplified or more complete binding equations (17).
We present here a reexamination of the affinity of GTP␥S for G t␣ using experimental conditions that allow determination of low (picomolar) values of K d . These consist principally of the use of low G t␣ concentrations and direct measurement, rather than calculation or assumption, of the values of free [GTP␥S]. In addition we present a comparison of the 2 surfaces obtained from least-squares fitting of data obtained at high and low protein concentrations. The dependence of this measure of goodness of fit, on the parameters used to calculate theoretical binding curves, reveals that important information about high affinity interactions can be obtained even at higher-than-ideal protein concentration. This analysis also reveals the limits of such information.

EXPERIMENTAL PROCEDURES
Materials-GTP␥S was obtained from Boehringer Mannheim (purity of 85% estimated by supplier). GTP␥ 35 S was obtained from DuPont NEN, and its radiochemical purity by TLC (18) was verified to be Ͼ95% for experiments reported here. Stocks were routinely checked for degradation, which was often found to be Ͼ90% in stored samples. Rod outer segments (19) and purified transducin ␤␥ (G t␤␥ ) subunits (9,20) were prepared as described; G t␤␥ was used to test its effects on GTP␥S binding.
Equilibrium Binding Assays-Equilibrium binding experiments were carried out using Buffer A consisting of 20 mM MOPS, pH 8.2, 150 mM KCl, 2 mM MgCl 2 , 1 mM dithiothreitol, 0.1 mM EDTA, 0.1 mg/ml ovalbumin (for comparison to conditions used in pH-based PDE assays), or Buffer B consisting of 10 mM MOPS, pH 7.4, 60 mM KCl, 30 mM NaCl, 2 mM MgCl 2 , 1 mM DTT, 0.1 mg/ml ovalbumin (for comparison to previous studies done at similar pH values). No significant differences were detected in binding results obtained with these two buffers. Samples of freshly bleached rod outer segments (3-150 nM R*) were mixed with GTP␥ 35 S (ϳ100 Ci/mmol) to various final concentrations and incubated on a rotating mixer at 23°C (1 or 2 ml, final volume) for periods of time ranging from 3 to 17 h. Then they were filtered through 0.22-m pore-size nitrocellulose filters, in a vacuum filter manifold, and washed with 1 ml of ice-cold buffer. Filters were removed with the vacuum on to remove solution from the edges. The solution passing through the filters was collected using funnels and scintillation vials positioned below the filters. Radioactivity bound to the filters and in the flow-through solutions was measured by scintillation counting. Nonspecific binding was determined by adding 1 M nonradioactive GTP␥S to the ROS 30 min prior to addition of GTP␥ 35 S, and the average radioactivity bound to filters in these samples (typically 0.95% Ϯ 0.05% of the total radioactivity added) was subtracted from the total radioactivity bound to the other filters to calculate specific binding. The sum of the radioactivity in the filter and flow-through samples was found to correspond to essentially all the radioactivity added.
Analysis of Covalent Incorporation of 35 S-Three different approaches were taken to test whether thiophosphorylation of G t␤ (21,22) might contribute to the filter-bound radioactivity. 1) In equilibrium binding experiments such as those described above, control samples were treated with 8% (final, w/v) trichloroacetic acid at 0°C just prior to filtration to denature the protein and release noncovalently bound radioactivity. 2) Identical samples to those used in binding assays (not treated with acid) were washed from nitrocellulose filters with SDS sample buffer, applied to a gel for SDS-PAGE (23), and subjected to fluorography, along with standards of known amounts of GTP␥ 35 S to estimate detection limits. 3) In an attempt to force the thiophosphorylation reaction, ROS (5, 15, or 29 M R) were treated with GTP␥ 35 S (1000 Ci/mmol) at a concentration of 100 nM for 40 min at 23°C. The membranes were separated from soluble proteins by centrifugation, and both supernatant and pellet fractions were subjected to SDS-gel electrophoresis and fluorography to locate radiolabeled bands. Radioactivity incorporated into protein was quantitated on dried gels using a Beta Scope 603 Blot Analyzer (␤etagen).
Other Assays-PDE assays were carried out as described previously (9), using suspensions of freshly bleached rod outer segments and Buffer A with ovalbumin and DTT omitted. Activity was measured before and after the addition of GTP␥S to the indicated concentrations, and GTP␥S-induced activity is expressed as a percentage of the maximal activity observed at the highest [GTP␥S]. To determine the G t content of ROS, a titration with GTP␥S was performed, using PDE activity to monitor formation of the G t␣ -GTP␥S complex at an R concentration of at least 5 M to ensure stoichiometric complex formation. The concentration of added GTP␥S giving maximal PDE activity was taken as the concentration of total G t␣ competent to bind GTP␥S. GTP␥S concentrations were determined spectrophotometrically using ⑀ 254 nm ϭ 13,700 M Ϫ1 cm Ϫ1 . Previous studies (9,20) have shown that spontaneous exchange of GTP␥S for GDP on transducin, and R*-catalyzed exchange give rise to the same product at equilibrium, and that G t␣ -GTP␥S is capable of fully activating PDE without the presence of R*. In these studies and many unpublished experiments in this laboratory, we have consistently found agreement between G t concentrations determined by nucleotide titrations and those determined by dye binding assays to be typically within Ϯ20%, and in the very worst cases to differ by a factor of two.
Data Analysis-The activation and direct binding titrations (Figs. 1-3) were analyzed using two equations for equilibrium binding. In the full equation (Equation 1), the amount of bound radioactivity or PDE activity (Y) for any values of K d and total added ligand is calculated using the expressions for equilibrium binding and conservation of mass. Only total ligand concentration (C L ), and S t , the total concentration of ligand binding sites (i.e. total G t␣ in this case), need to be known independently.
, Y max is activity or binding at saturating ligand, S t is as described above and C L is the total concentration of ligand. Equation 2, the standard equilibrium binding equation, can be used instead of Equation 1 when the free ligand concentration, L, is known explicitly, or can be used as an approximation to Equation 1 for the special case when free ligand is approximated by total ligand i.
Here L is the concentration of free ligand and Y max is as described above. Note that calculated K d values can be in error by orders of magnitude if total ligand is substituted for free ligand, L. These equations provide three alternative ways to determine a K d value from any data set: using total ligand concentration and Equation 1, using measured free ligand concentration and Equation 2, or using total ligand concentration and Equation 2 with the assumption that total ligand and free ligand are equivalent. This last procedure, although not valid in many cases, is the method that has been most commonly used for measuring GTP␥S binding to G proteins.
To determine values of Y max , S t and K d that best fit the data for each method, nonlinear least squares fitting was used with all parameters varied to minimize the sums of the squared deviations of the calculated from the measured values, or equivalently, to minimize 2 . Parameter space was searched using a manual or automated grid search, or using the method of Marquardt (24,25). Plots of 2 versus parameter values were examined either as three-dimensional representations, or as twodimensional cross-sectional plots, to determine the steepness of the surfaces near minima, and to verify that the fitting routines were not trapped in local minima. Values reported for the fit parameters were those giving the minimum 2 values for each equation used. Fig. 1 shows the characteristic response of cGMP phosphodiesterase activity in bleached rod outer segments to the addition of GTP␥S: a linear increase to a plateau, with a good correspondence between the total GTP␥S concentration required to reach the plateau, and the estimated total transducin. Because only a fraction (Ͻ50%) of PDE is activated at saturating [GTP␥S] in ROS suspensions, and PDE can be further activated by addition of exogenous activated G t␣ , it is reasonable to assume that PDE activity in this curve reports in a linear way on the fraction of G t activated by GTP␥S.

Activation by GTP␥S at High [G t ]-
There are two important features of this curve. One is that the concentration of GTP␥S at which a half-maximal response is observed is 3.5 nM, a value more than an order of magnitude lower than the previously reported K d for binding of GTP␥S to G t . The other is that a good fit to the data is obtained with Equation 1 by assuming that K d is much lower than the total concentration of binding sites (i.e. total G t Ͼ Ͼ K d , Fig. 1A), while assuming that free GTP␥S is equal to the GTP␥S added and using Equation 2 gives a much poorer fit (Fig. 1B). Fig. 1, C-F, shows the results of nonlinear least squares analysis using these two approaches; the best 2 value for the "low affinity model" (Equation 2) is an order of magnitude higher than that for the "high affinity model" (Equation 1). However, Fig. 1C shows that a reliable estimate of K d cannot be obtained from a fit to Equation 1, exactly because G t Ͼ Ͼ K d . While the best fit value is 33 pM, any value below ϳ150 pM fits about as well. In fact, simply drawing two straight lines through the data points in Fig. 1A, which is equivalent to assuming an infinitely tight binding (K d ϭ 0), gives a curve that fits the data quite well. The fit does make it possible to establish an upper limit for a K d value consistent with these data; 200 pM would be a conservative estimate.

GTP␥S Binding at Low [G t ]-
The obvious solution to the problem of determining K d when its value is apparently much lower than the concentration of G t commonly assayed is to carry out assays at much lower concentration. While PDE assays are not as convenient at low concentrations of ROS, direct binding assays using radiolabeled GTP␥S can be used instead. The kinetics in this case might be much slower (although not necessarily so for membrane-confined reactants) so we carried out long incubations to ensure that equilibrium was reached. Fig. 2 shows the results of a titration carried out at 30 nM R. It shows unequivocally that K d is on the order of 50 pM. A technical difficulty with this type of experiment is that prolonged incubations under dilute conditions tend to lead to some loss of total binding activity (20); in these earlier studies it was confirmed, however, that G t␣ -GTP␥S formed by spontaneous nucleotide exchange or by R*-catalyzed exchange are functionally indistinguishable. In this particular experiment, about half (51%) of the GTP␥S binding activity observed at higher ROS concentrations using PDE assays like the one described above, was retained after a 3 h incubation at 23°C and 30 nM R. Fig. 2B shows that at appropriate protein concentrations, the method of analyzing the data is not nearly as critical as at protein concentrations well above K d . Roughly equivalent estimates of K d with similar 2 values are obtained using either Equation 1 (short dashes in Fig. 2A, inset, and in Fig. 2B), which implicitly calculates the value of free [GTP␥S] from the total, or using Equation 2 (solid lines in Fig. 2, A and B) with the measured free [GTP␥S] as the independent variable. Simply making the invalid but common assumption that free and total ligand are the same (long dashes in Fig. 2A, inset, and in Fig. 2C), again gives a much poorer fit, but in this case, the value of K d obtained, 280 pM, is somewhat more realistic than FIG. 1. GTP␥S activation of PDE in rod outer segment membranes, 500 nM R. GTP␥S was added at the indicated concentrations to a suspension of bleached ROS (0.5 M R) and PDE activity was monitored by pH recording. PDE activity is plotted as a percent of the maximal activity, determined by averaging the observed hydrolytic rates at the 10 highest GTP␥S concentrations. The theoretical curves represent the best fit curves derived A, using Equation 1 (K d ϭ 33 pM, S t ϭ 6.95 nM) or B using Equation 2 assuming free [GTP␥S] ϭ total GTP␥S, with maximal activity in the fits either fixed at the observed value (V max ϭ 100%, K d ϭ 3.9 nM, short dashes) or allowed to vary to obtain the best fit (V max ϭ 122%, K d ϭ 1.9 nM, long dashes). C-F. 2  the one estimated using this technique at higher [ROS]. An experiment at even lower [ROS], 3 nM R, is shown in Fig. 3. A very similar value of K d , 30 pM is obtained, regardless of the method of analysis. In this case, even using Equation 2 and the assumption that free and total ligand are equal (long dashes, Fig. 3) gives the correct result.
Tests for Covalent Radiolabeling-Because the binding we observe is so much stronger than has been reported previously for any heterotrimeric G protein, we were concerned that covalent attachment of 35 S might be mistaken for high affinity noncovalent binding. It has been reported (21,22) that G t␤ is covalently labeled by GTP␥S in bleached ROS. Brief treatment (Ͻ1 min) with trichloroacetic acid at 0°C resulted in nearly complete release of the tightly bound GTP␥ 35 S (data not shown). When autoradiography was performed on gels from SDS-PAGE of samples of bleached ROS incubated with GTP␥ 35 S under the conditions used for measuring binding, no labeling of transducin subunits was observed; from estimates of the detection limits of this procedure, less than 10% of the radioactivity should have been readily detected if covalently incorporated into proteins. Attempts to force thiophosphorylation by using high [ROS] and high specific activity GTP␥ 35 S, resulted in no detectable thiophosphorylation of transducin subunits. However, in the membrane pellet fraction, thiophosphorylation of R, probably catalyzed by rhodopsin kinase, could be detected at trace levels. DISCUSSION It is important to ask how reliable is the 50 pM value of K d estimated for the reaction: G t␣ ϩ GTP␥S º G t␣ -GTP␥S. There are two kinds of artifacts or errors that must be considered: those that cause an underestimation of K d and those that cause an overestimation.
Those that could cause overestimation of K d include inhibition of GTP␥S binding by GDP, R*, G t␤␥ , and membranes, all of which favor either the GDP or nucleotide-free states of G t␣ . However, corrections for these competing ligands are probably not very significant. As we report in a companion study (26), GDP competes extremely poorly with GTP␥S for G t␣ binding; it must be present in at least 200-fold excess for significant inhibition of GTP␥S binding to be observed. High affinity binding to R* requires association of G t␣ and G t␤␥ , which have been estimated to bind one another with a K d 150 -300 nM as detected by fluorescence intensity changes or resonance energy transfer (27). This estimate is consistent with the dependence on G t␤␥ concentration of pertussis toxin catalyzed ADP-ribosylation of G t␣ (28) and the dependence on G t␣ concentration of 125 I-G t␤␥ binding to illuminated ROS membranes (29). Thus, under the conditions of Figs. 2 and 3 essentially no G t␣␤␥ (with GDP or GTP␥S bound) or R*-G t␣␤␥ should be present, although the latter presumably exists in small amounts as a transient intermediate in nucleotide exchange. If there were significant amounts of these complexes, then the measured K d would be higher than the actual value, as these complexes favor the GDP form or the nucleotide-free form, respectively, of G t␣ . The effect of G t␤␥ on the equilibrium was tested directly under the conditions of Fig. 2A, by varying the added G t␤␥ from 0 to 238 nM at 85 pM free GTP␥S. The average value of bound 35 S at 238 nM added G t␤␥ was within 1% of the value at 0 added G t␤␥ and the maximum range of variation over the whole range of added G t␤␥ (total of 10 measurements) was Ϯ5% (data not shown).
The presence of phospholipids may also affect the results, as there are differences in membrane binding of G t␣ -GDP and G t␣ -GTP␥S even to dark-adapted ROS membranes. 2 Phospholipids present in the assays also probably do not influence GTP␥S binding strongly, as binding of G t␣␤␥ -GDP to phospholipid vesicles is half-maximal at approximately 60 M accessible phospholipid (determined at 0.75 M G t , 3 a value more than 50-fold higher than the accessible phospholipid in the binding experiments described here.
Degradation of GTP␥ 35 S can also lead to overestimation of the value of K d . While this seems a likely source for at least some of the reported findings of K d values Ͼ 10 nM for G proteins binding GTP␥S, our practice of routinely analyzing the integrity of GTP␥ 35 S stocks by thin layer chromatography and discarding those with low radiochemical purity makes this an unlikely source of error in our experiments.
Possible artifacts and errors leading to underestimation of K d include possible binding to PDE, PDE␥, or the GTPaseactivating protein, GAP (18,30), the only reactions known to favor the GTP␥S or GTP form of G t␣ . In addition, underestimation of K d could result from contamination with a high affinity GTP binding protein other than G t␣ or from covalent radiolabeling of G t␤␥ (21,22). The low total membrane concentration means that G t␣ binding to PDE is unlikely to shift the equilibrium significantly because accessible ROS phospholipids are well below the concentration of 10 -13 M required for half-maximal interactions of PDE with G t␣ -GTP␥S; determined at 10 nM PDE and either 0.3 M or 2 M G t␣ -GTP␥S (9). The K d values for PDE␥ binding to PDE catalytic subunits; ϳ10 pM (31) and to G t␣ -GTP␥S; ϳ0.1 nM (10) imply that G t␣ -GTP␥S binding to PDE␥ is also unlikely to affect the equilibrium substantially. GAP interactions with G t␣ are also sufficiently weak as to be negligible at such low membrane concentrations, based on the failure of endogenous GAP in ROS to elicit noticeable GTPase acceleration at R concentrations of 10 M or lower (18,30).
Thiophosphorylation of G t␤ , a reaction that has been re-2 Z. Yang and T. Wensel, unpublished results. 3 T. Melia and T. Wensel, unpublished results. ported to occur in the presence of R* (21,22), is rendered unlikely by our control experiments. The ready loss of the apparent binding following denaturation by low pH or SDS-PAGE argues strongly against a significant component of the bound radioactivity being due to covalent modification. It might also be suggested that the binding is not due to G t at all, but rather to some previously unidentified high affinity GTP binding protein in ROS. This is very unlikely the case, because the stoichiometry of binding is too high, e.g. 51% of the total activable G t␣ measured at high [ROS] bound GTP␥S in the experiment depicted in Fig. 2, with the discrepancy between total GTP␥S binding observed at low and high [ROS] likely due to surface-induced and other losses of active protein (20). The amount of bound GTP␥S is too high to be accounted for by minor unknown proteins. Only G t , R*, and arrestin are abundant enough to account for such a high stoichiometry (3,32), and of these only G t␣ binds GTP␥S.
An additional concern might be that the long incubation times required to ensure equilibrium at low ROS concentrations could conceivably lead to overestimation of K d through thermal denaturation and loss of native binding activity. Alternatively, prolonged incubations might lead to underestimation of K d through some slow binding reaction distinct from the G t␣ activation by GTP that occurs rapidly under physiological conditions. It is reassuring in this regard that: 1) G t␣ loaded with GTP␥S by incubation for many hours in the absence of R* is as functional in PDE activation as G t␣ activated with R* (20) and 2) the essentially instantaneous binding reactions observed in the experiments of Fig. 1 give nearly identical K d values to those resulting from long incubations (Figs. 2 and 3).
Thus, while appropriate caution must be exercised because of reactions with other proteins, experimental variation, and losses of activity during prolonged incubations, it is probably safe to conclude that the true K d for G t␣ binding to GTP␥S is greater than 10 pM and less than 100 pM. This conclusion means that transducin binds GTP␥S with an affinity 2-3 orders of magnitude stronger than has been reported previously for heterotrimeric G proteins.
Comparisons with Previous Results-As discussed in the Introduction, most previous measurements of GTP␥S binding to G proteins have been carried out with total G protein concentrations considerably higher than the K d values observed here, and free [GTP␥S] was not determined directly, so they are difficult to compare to the present results. However, in experiments with G o and G i (33,34) it was reported that the concentration of free GTP␥S was known and was at least nine times the concentration of G protein-bound nucleotide; G ␣ concentrations ranged from 5-330 nM. In those studies, the reported K d for binding of GTP␥S to both G ␣ s was ϳ30 nM, a value almost three orders of magnitude higher than the one we observe for G t␣ , and one which would lead to binding of at least 15% of the GTP␥S even at the lowest protein concentration of 5 nM. This value is also intermediate between the low and high extremes of protein concentrations used. It would be interesting to determine whether there is any dependence of the apparent K d on the concentration of G ␣ used, or, alternatively, if there is actually a large difference in GTP␥S affinity for G o and G i as compared to G t . In the case of G t␣ , previous reports indicating apparent lower affinity binding of GTP␥S or GTP (nucleotide concentration for half-maximal binding or K d Ն 50 nM, e.g. Refs. 4,11,35,36) used high concentrations of G t␣ , did not directly monitor the concentration of free GTP␥S or GTP, and are thus consistent with our results. Fig. 4 shows clearly that if our results are analyzed without taking into account the difference between total nucleotide and free nucleotide (open symbols) the apparent K d values calcu-lated increase in an approximately linear way with protein concentration. Taking account of this difference explicitly (Fig.  5, closed symbols) gives nearly identical K d values over protein concentrations spanning three orders of magnitude.
Structural Correlations-A high resolution structure of G t␣ complexed to GTP␥S determined by x-ray crystallography (16) reveals clearly the many contacts between the protein and GTP␥S. Although there are some differences in detail even within the conserved GTPase domain, an overall comparison between the nucleotide binding sites of G t␣ and p21 ras (15) indicates that all the p21 ras binding interactions are either found in G t␣ or replaced by similar interactions, and that G t␣ makes additional contacts from its insert domains unique to the G ␣ family. With all these interactions, the surprise is that we measure a GTP␥S K d for G t␣ that is only comparable to, and not much stronger than, binding of GTP␥S to p21 ras .
Implications for G t␣ Function-Given the millimolar concentration of GTP in rod cells (37), it is reasonable to ask what functional consequences may have favored the evolution of such a high affinity binding site. Certainly the probability of occupancy would be identical even if the K d value were micromolar. Two possibilities seem worth considering. The first is that the free energy of binding serves to drive rapidly and efficiently the conformational changes in G t␣ -GTP␥S and its subsequent reactions: dissociation from R*, dissociation from G t␣␤␥ , and binding to PDE. Another possibility is that viselike positioning of GTP in the binding site of G t␣ facilitates precise control in the timing of GTP hydrolysis, making it an event that occurs only once in 20 s when G t␣ is not in contact with its GAP and PDE␥ (18,30,38,39) but that happens rapidly once these interactions occur.
Experimental Implications-While the consequences of high affinity binding for G t␣ function in vision require further study, some experimental ramifications are immediately apparent. One is that in kinetic experiments conducted at low concentrations of [GTP] or [GTP␥S], G t␣ can be expected to bind virtually all of the nucleotide present, as long as it is present at concentrations above that of the nucleotide and above ϳ1 nM, and dissociation of [GTP␥S] for practical purposes will not occur (an observation already made in several laboratories). Another is that studies with analogues of GTP can be greatly confused by the presence of very small amounts of contaminating GTP, GTP␥S, or GTP␣S (26).
Recognizing High Affinity Interactions-Another important conclusion concerning experiments involving high affinity binding sites in general is that their high affinity nature can be detected even at concentrations far from those that allow them to be measured accurately (Fig. 1). Recognizing them is facilitated by 1) critically comparing the shapes of binding curves to low affinity and high affinity models, 2) experimentally testing for dependence on protein concentration of apparent half-maximal binding concentrations, and 3) independently measuring the free and bound ligand concentrations. It will be interesting to determine what will be revealed by application of these approaches to other heterotrimeric G proteins, and to discover whether transducin's high GTP affinity is unique, or a common feature of this protein family.