Report

The Tetrameric Structure of a Glutamate Receptor Channel

See allHide authors and affiliations

Science  05 Jun 1998:
Vol. 280, Issue 5369, pp. 1596-1599
DOI: 10.1126/science.280.5369.1596

Abstract

The subunit stoichiometry of several ligand-gated ion channel receptors is still unknown. A counting method was developed to determine the number of subunits in one family of brain glutamate receptors. Successful application of this method in an HEK cell line provides evidence that ionotropic glutamate receptors share a tetrameric structure with the voltage-gated potassium channels. The average conductance of these channels depends on how many subunits are occupied by an agonist.

Voltage-gated potassium channels are tetramers, and nicotinic acetylcholine receptors are pentamers (1). Brain glutamate receptors are often assumed to be pentamers (2) because they, like the acetylcholine receptor, are ligand-gated. We developed a method to count the number of subunits in the brain glutamate receptor. The key observation upon which our counting method is based is that the mean single-channel current depends on how many of a receptor's binding sites have an agonist bound. The number of binding sites can then be counted by observing the distinct electrophysiological states that a receptor passes through as successively more binding sites become occupied. Successful application of this counting method requires solving three problems. First, because our method counts binding sites, the number of binding sites must equal the number of subunits; in addition, the binding sites must be equivalent so that sites are not missed. We therefore used the α-amino-3-hydroxy-5-methyl-4-isoxazol propionate (AMPA)–receptor GluR3flip (and mutant versions) expressed in a mammalian (HEK) cell line (3), because this receptor forms homomultimers.

Second, at the saturating concentrations of agonist that are needed to ensure full binding-site occupancy, the lifetime of each occupancy state is too brief to resolve. To prolong the lifetime of each state, we slowed down the agonist binding rate by interposing a very slow step, the dissociation of a high-affinity competitive antagonist; thus, we started the receptor with all of its binding sites occupied by a competitive antagonist and then made the sites available for agonist binding, one by one, as the bound antagonist molecules slowly dissociated. We used a rapid superfusion system to change an outside-out patch's environment from a saturating concentration of the high-affinity antagonist 6-nitro-7-sulphamoyl-benzo(F)quinoxalinedion (NBQX) to a saturating concentration of agonist (4).

Finally, if the receptor's normal desensitization mechanisms were intact, the receptor would desensitize long before this progression through the various occupancy states is complete (5). We therefore used single channels of a GluR6/GluR3 chimera in which desensitization is completely absent (Fig.1A) (6).

Figure 1

(A) A response from an outside-out patch containing a single AMPA receptor (GluR6/GluR3) during a 1-s concentration step from control solution into quisqualate (1 mM) and back. Holding potential, –160 mV. Filtered at 2 kHz. (B) Four individual responses [different patch from (A)], with step from NBQX (10 to 30 μM) into quisqualate (1 mM). After some delay, two intermediate (approximately 5 and 15 pS) states preceded the same large conductance seen in (A). Filtered at 1 kHz. Holding potential, –160 mV. (C) The same patch is shown as in (B); Note the time scale. (D) The same patch is shown as in (C), but with a low-affinity competitive antagonist MNQX (300 μM). Note the different time scale compared to (C). Scale bar: 1 pA and 10 ms. (E) NBQX/quisqualate switch for a single GluR3flip channel treated with 100 μM cyclothiazide. The scale bar (1 pA, 100 ms) appears in (D). (F) Amplitude histograms for the states C, S, M, and Laveraged from 10 consecutive episodes of antagonist/agonist switches. Same patch as in (B), (C), and (D). (G) Amplitude histograms for a single nondesensitising channel activated with 1, 2, 6, and 300 μM 1-quisqualate.

Our key observation is that the receptor passes through three distinguishable states, each with a different mean conductance, as all binding sites become occupied in turn. The transition into the first of these states occurs with two time constants, whereas each of the two remaining states relax with a single time constant. The most straightforward interpretation of these data is that the receptor is a tetramer.

Our experiment required recording from single channels. Determining if a patch has only a single channel was easy, because we used saturating agonist concentrations (for example, quisqualate, 1 mM) that consistently caused our noninactivating channels to open to an apparent 23 pS state with 88.3 ± 5% (n = 6) probability (Fig. 1A). The increased probability of an open state that results when channel desensitization is removed has been also reported for native AMPA receptors (7).

Figure 1, B and C, exhibits the basic phenomenon. An outside-out patch containing a single GluR6/GluR3 channel was rapidly switched between saturating concentrations of NBQX (10 to 30 μM) and agonist quisqualate (1 mM). The channel starts in its closed state Cand then progresses “staircase” fashion through three distinguishable conducting states that we call S (small, approximately 5 pS mean conductance), M (medium, 15 pS), andL (large, 23 pS).

The channel proceeds through the same three states in theSML order with each antagonist/agonist switch, but the dwell time in each state varies randomly from one switch to the next. When the agonist is removed or replaced by antagonist, the receptor passes in the reverse LMS order to the closed state (Fig. 1B).

To determine whether these three distinct states are an artifact of the GluR6/GluR3 chimeras used, we did the antagonist/agonist switch on GluR3flip homomultimeric receptors treated with 100 μM cyclothiazide to remove inactivation (7, 8). Channels constructed from native subunits reveal the same “staircase” behavior (Fig. 1E; n = 4).

How can we know that the rate-limiting step in the “staircase” response is dissociation of antagonist? We used a lower affinity competitive antagonist, 5,7-dinitro-quinoxalinedion (MNQX, IC50 = 2.2 μM) instead of NBQX (IC50 = 150 nM) (9). As expected if the rate-limiting step is antagonist dissociation, the channel progressed through the same states in theSML order when MNQX was substituted for NBQX, but the rate of progression was increased about 30-fold (Fig. 1, C and D;n = 5).

If the states we identified do indeed correspond to different numbers of bound agonist molecules, then the current amplitude histogram should change in an orderly way as the agonist concentration is increased. At the lowest agonist concentrations, the S state should predominate; at very high agonist concentrations the channel should be always in the L state, and the amplitude histogram should exhibit a mixture of states in between. This prediction is confirmed by the amplitude histograms shown in Fig. 1G for agonist concentrations of 1 to 300 μM (compare Fig. 1, F and G).

Correct counting of binding sites requires an analysis of dwell times at each step along the “staircase.” The distribution of dwell times combined from eight patches is shown (Fig.2A). The second (SM) transition is most rapid (mean dwell time = 224 ± 9.1 ms), the last (ML) is slowest (461 ± 20.3 ms), and the first (CS) transition is intermediate (258 ± 9.9 ms). Note also that the waiting times for the SM and ML transitions are exponential, but that the CS transition exhibits two clear components.

Figure 2

Waiting times reveal four subunits. (A) Histogram of waiting times for CS,SM and MLtransitions from 417 episodes (eight patches). Four-subunit theory superimposed (11) with τ = 462 ms. (B) Cumulative probability versus waiting times with predictions from the four-subunit model (11). Same observations and theory as in (A). Inset: expanded data and the same theory. (C) Cumulative probability for the CS waiting time for different models as indicated by numbers associated with expanded graph in inset (11). (D) Four-subunit model and alternatives. We calculated the sum of the squared deviations between predicted and observed mean waiting times such that the sum of waiting times adds up to the observed 943 ms (ordinate); the numbers on the abscissa indicate the various models (11).

A first guess might be, “Three states, three subunits: it's a trimer.” This initial notion is inconsistent with our data, however, because it can explain neither the two-component waiting time distribution for CS nor the fact that theSM transition is fastest.

The simplest theory consistent with a two-component waiting time distribution for the CS transition and single-component waiting time distributions for the subsequent transitions would have the first (CS) transition involve the dissociation of two antagonists (and the binding of two agonists) and to have each of the other transitions require only a single antagonist dissociation. This model implies four subunits (10). According to this view, occupancy of two binding sites is necessary for channel opening, but each additional occupancy increases mean single-channel current. If binding sites are identical and independent, then this theory makes specific predictions (10) about all of the waiting time distributions (illustrated in Fig. 2, A through C) with only a single free parameter τ, the average time an antagonist remains on its binding site. As can be seen from the smooth curves in Fig. 2, A through C, the simple theory provides a satisfactory fit to the data (Kolmogorov-Smirnov test, P > 0.2). We examined similar alternative schemes, like three subunits or five subunits with two or three occupancies required for a channel opening, and find that they do not fit the waiting time data satisfactorily (Fig. 2, C and D).

Our model accurately predicts the relative mean dwell times in the various states with no parameters estimated from the data (10). The predicted ratio of the waiting times (CS)/(SM) is (1/4 + 1/3)/(1/2) = 7/6 = 1.17, and the observed ratio for eight patches is 1.15 ± 0.07. The predicted ratio for the (ML)/(SM) waiting times is 1/(1/2) = 2, and the observed ratio (same patches) is 2.06 ± 0.13.

We conclude that the glutamate receptor we studied is most likely a tetramer. This conclusion is not, perhaps, completely unexpected for several reasons, despite the common assumption that the glutamate receptors are pentamers (2). First, biophysical and biochemical studies on the N-methyl-D-aspartate (NMDA)–type glutamate receptor suggest four binding sites, although a pentameric structure has recently been proposed by Premkumar and Auerbach (11). Second, using a biochemical approach, Mano and Teichberg (12) report a tetrameric structure for glutamate receptors. Finally, the glutamate receptor pore structure may be like that of the potassium channel, which is known to be tetrameric (1). Our data support a tetramer, but we cannot, of course, exclude more elaborate schemes with more than four subunits that interact in whatever complicated way necessary to look like four independent subunits.

Perhaps the most interesting observation we made is that the average conductance of this channel, like that of the cyclic nucleotide–gated channel (13), depends on the number of binding sites occupied by agonist molecules. Because agonist binding seems to be required for the channel-opening conformational change, our observation leads to a model in which a single subunit can open the receptor's pore a certain amount and conformational changes in multiple subunits can open it more. Chapman et al. proposed a similar picture for delayed rectifier channels drk1 (14). In addition to the differences in single-channel conductance based on the subunit composition (15), the notion that conductance levels represent the conformational state of GluRs helps to explain why so many conductance levels are seen in natural single-channel currents at synapses (16).

  • * To whom correspondence should be addressed. E-mail: cfs{at}salk.edu

REFERENCES AND NOTES

View Abstract

Navigate This Article