Glutamate Receptor Subtypes Evidenced by Differences in Desensitization and Dependence on the GLR3.3 and GLR3.4 Genes 1

Ionotropic glutamate receptors in the central nervous system of animals are tetrameric ion channels that conduct cations across neuronal membranes upon binding glutamate or another agonist. Plants possess homologous molecules encoded GLR genes. Previous studies of Arabidopsis thaliana root cells showed that the amino acids alanine (Ala), asparagine (Asn), cysteine (Cys), glutamate (Glu), glycine (Gly), and serine (Ser) trigger transient Ca 2+ influx and membrane depolarization by a mechanism that depends on the GLR3.3 gene. The present study of hypocotyl cells demonstrates that these six effective amino acids are not equivalent agonists. Instead, they grouped into hierarchical classes based on their ability to desensitize the response mechanism. Sequential treatment with two different amino acids separated by a washout phase demonstrated that Glu desensitized the depolarization mechanism to Gly but Gly did not desensitize the mechanism to Glu. All 36 possible pairs of agonists were tested to characterize the desensitization hierarchy. The results could be explained by a model in which one class of channels contained a subunit that was activated and therefore desensitized only by Glu, while a second class could be activated and desensitized by Ala, Cys, Glu, or Gly. A third class could be activated and desensitized by any of the six effective amino acids. Analysis of knockout mutants indicated that GLR3.3 was a required component of all three classes of channels, while the related GLR3.4 molecule specifically affected only two of the classes. The resulting model is an important step toward understanding the biological roles of these enigmatic ion channels. gene-specific glr3.3


INTRODUCTION
One surprise to emerge from the first comprehensive inventory of a plant genome (The Arabidopsis Genome Initiative, 2000) was the identification of a family of 20 genes unequivocally homologous to mammalian ionotropic glutamate receptors (Lam et al., 1998;Lacombe et al., 2001). Such receptors function as ligandgated ion channels in the mechanism that transmits signals between cells in the central nervous system (Madden, 2002;Mayer, 2005). At the junction between two neurons, or synapse, the presynaptic cell releases the neurotransmitter Glu, which binds to ionotropic glutamate receptors located at the plasma membrane of the postsynaptic cell. Ionotropic glutamate receptors are tetrameric ion channels.
They respond to the binding of Glu by adopting a conformation that permits mixed cation flow into the postsynaptic cell, resulting in membrane depolarization (Madden, 2002). The channels quickly shift to a non-conducting (desensitized) conformation, which can be exited after unbinding of the agonist (Sun et al., 2002). Removal of Glu from the extracellular space promotes agonist unbinding, and therefore recovery of sensitivity.
Depending on the types of subunits constituting a particular glutamate receptor channel, Ca 2+ may accompany Na + and K + ions flowing into the postsynaptic cell (Dingledine, 1999). In these cases, a rise in intracellular Ca 2+ accompanies the membrane depolarization. The rise in Ca 2+ plays a role in regulating synaptic transmission efficiency, which underpins learning and other central nervous system processes (Gosh and Greenberg, 1995). Misregulation of synaptic activities mediated by glutamate receptors is associated with neurodegenerative diseases including Alzheimer's, Parkinson's, and Huntington's diseases (Olney et al., 1998;Schiffer, 2002;Barnes and Slevin, 2003).
Because the ionotropic glutamate receptors are studied mostly in the context of central nervous system signaling, it was a surprise to find genes having all the hallmarks of common ancestry, and even common function, in the genomes of plants (Davenport, 2002). Using hindsight, evidence of their existence can be seen in early plant electrophysiological studies. Etherton and colleagues demonstrated that large, transient membrane depolarizations were triggered by some amino acids, and that the response displayed desensitization (Etherton & Rubinstein, 1978;Novacky et al., 1978). Models based on electrogenic amino acid transport were constructed to explain the results (Kindraide & Etherton, 1980;Kindraide & Etherton, 1982), but the possibility that plants possess amino acid-gated ion channels was not yet raised. After the GLR genes were discovered, the phenomenon of amino-acid-triggered ionic events in plants cells was investigated with a different model in mind. A study of Arabidopsis roots showed that membrane depolarization in response to Glu was accompanied by a large spike in intracellular Ca 2+ (Dennison and Spalding, 2000). A subsequent reverse-genetic study linked GLR3.3, one of the twenty Arabidopsis GLR genes, to the membrane depolarization and the Glu-induced Ca 2+ rise (Qi et al., 2006). Mutation of GLR3.3 severely impaired both the membrane depolarization and the Ca 2+ rise triggered by Glu in the micromolar to millimolar concentration range (Qi et al., 2006). These results leave little room for doubt that the plant GLR molecules are components of ligand-gated ion channels that transiently conduct cations, including Ca 2+ , into plant cells.
While some evidence indicates that plant glutamate receptors function much like their animal homologs, other results point to important differences.
Sequence similarity between the Arabidopsis GLRs and animal iGluRs is high in the transmembrane domains (Chiu et al., 1999;2002) but the N-termini of the Arabidopsis GLRs may have come from ancestral amino-acid-binding G-protein coupled receptors (Turano et al., 2001). Structure modeling indicates that the plant molecules may have an additional amino-acid binding site not present in animal iGluRs (Acher and Bertrand, 2005). Perhaps related to these potentially atypical binding sites, Qi et al. (2006) found that six very different amino acids were capable of triggering membrane depolarization and a Ca 2+ rise. The effective amino acids were Ala, Asn, Cys, Glu, Gly, and Ser. Even glutathione (a tripeptide consisting of Glu-Gly-Cys) triggered a full response. Responses to all six effective amino acids and glutathione were equally impaired by glr3.3 mutations (Qi et al., 2006). These results provided genetic evidence that a GLR molecule was mechanistically related to the ionic events triggered by an amino acid. More specifically, these results indicated that that GLR3.3 is a necessary subunit of channels that respond to all six amino acids and the tripeptide. A question raised by the surprisingly broad agonist profile is whether or not all six effective amino acids are equivalent, and whether or not other GLR family members contribute to the breadth of the profile. For example, do heteromeric channels differing in subunit composition have different agonist profiles? The present study takes advantage of the desensitization phenomenon and mutations in the GLR3.3 and GLR3.4 genes to address these questions.

RESULTS
The first report of plant GLR genes presented pharmacological evidence that GLR channels may play a role in light signal transduction during de-etiolation (Lam et al., 1998). A second study provided more evidence of a connection between GLR function and light-controlled hypocotyl growth (Brenner et al., 2000). It had been previously established that light-induced membrane depolarizations in hypocotyls (Spalding and Cosgrove, 1989;Spalding and Cosgrove, 1992;Cho and Spalding, 1996;Parks et al., 1998)  both organs were studied with similar techniques. Figure 1A shows that application of Gly to light-grown hypocotyls caused a large, transient membrane depolarization, similar to previously reported measurements in roots (Qi et al., 2006) and reminiscent of recordings made in mesophyll cells (Meyerhoff et al., 2005). Figure 1B summarizes 88 similar measurements of Gly-triggered transient depolarizations. The average initial, resting potential from was -141 mV and the response to Gly peaked at a value of -49 mV. The difference of 92 mV would be slightly greater if the eight individuals that did not respond beyond -100 mV for unknown reasons were disregarded. Rather than report the difference between initial and peak potential as a measure of the response, the less variable peak potential was used in subsequent quantifications of response magnitude. The Ca 2+ rise triggered by Gly in isolated hypocotyl segments ( Fig. 1C) was also similar to the Ca 2+ response measured in whole seedlings (Dennison and Spalding, 2000;Dubos et al., 2003) and roots (Qi et al., 2006). Hypocotyl cells depolarized similarly to the same six amino acids as roots -namely alanine, asparagine, cysteine, glutamate, glycine, and serine -and, as in roots, the response depended strongly on the GLR3.3 glutamate receptor (Fig. 2).
Therefore, the basic ionic response to amino acids in the light-grown hypocotyl is similar to that in the root.

Depolarization-independent desensitization
Hypocotyl cells also underwent desensitization, a phenomenon much like in root cells. Figure 3A demonstrates that exposure to Glu desensitized a hypocotyl cell to a second application, even though the ligand had been removed by a 2-min washout period and the membrane had fully repolarized. Gly also desensitized the cell to a subsequent treatment of Gly (Fig. 3B). Unlike in root cells, an effective amino acid did not depolarize the cell much when the experiment was performed at pH 7.7 instead of pH 5.7 (Fig. 3C) desensitized the cell to a second application of Gly presented at the permissive pH 5.7 (Fig. 3C). The pH change by itself did not cause desensitization (data not shown). Thus, desensitization occurred even in the absence of a large membrane depolarization.

Asymmetric desensitization
Experiments in which two different amino-acid treatments were separated by a 2min washout period produced evidence that not all effective amino acids are equivalent with respect to desensitization. Figure 4 shows an example of double agonist treatments (x→y). When Ala preceded Glu (Ala→Glu), two full depolarizations were observed. But when Glu preceded Ala (Glu→Ala), Ala was ineffective; the cell displayed a desensitized state. Thus, Glu desensitized the mechanism to Ala but Ala did not desensitize the mechanism to Glu. Each of the 36 possible permutations of amino acid pairs was tested in at least 3 independent trials. Table 1 shows the average peak potential for each treatment. Values in the vicinity of -40 mV (a high peak potential) indicate that the second treatment produced a large response despite the first treatment. Values in the vicinity of -100 mV (a low peak potential) indicate that the first treatment desensitized the mechanism to the second treatment. Recordings typical of each unique treatment pair are shown in Supplemental Figure S1. The results were essentially binaryinspection of the shape of the response time course and the average peak potential showed that the mechanism was unambiguously either responsive or desensitized following the first treatment. Treating the data qualitatively permitted assignment of each amino acid to a rank in a hierarchical scheme. Glu desensitized the mechanism to each of the five other amino acids. Ala, Cys, and Gly formed the next level of desensitizers. Ser and Asn formed the lowest tier in the desensitization hierarchy.
One possible explanation for the hierarchical behavior of the six different effective amino acids is that the top tier ligand, Glu, was more potent than the lower tier ligands. A dose-response analysis was performed to address the question of relative potency among the effective amino acids. The aequorinreported Ca 2+ response was chosen for this purpose because of its sensitivity, because the integral of the Ca 2+ rise increased sigmoidally over three orders of magnitude of agonist concentration as expected for a simple ligand/receptormediated process, and because in every respect it was shown to match depolarization measurements made with intracellular microelectrodes (Qi et al., 2006). The results, expressed in two different forms in Figure 5, showed that all six agonists were variously effective over the 10-1000 µM range. The most potent of the six agonists was Cys and the least potent was either Ser or Asn.
These data indicate that the desensitization hierarchy is probably not due to differences in agonist potency because Glu occupies the highest position in the desensitization hierarchy but is not significantly more potent than the others.

Different contributions of GLR3.3 and GLR3.4
A systematic electrophysiological screen of many glr T-DNA insertion mutants identified GLR3.4 as a component of the amino-acid response mechanism in hypocotyl cells. Two independent glr3.4 alleles were subjected to the same agonist profile screen that demonstrated a key role for GLR3.3 in the response to all six effective amino acids ( Figure 2). Typical traces of membrane potential changes recorded from wild-type, glr3.4-1, and glr3.4-2 are shown in Figure 6A. Figure 6B shows the average peak potentials reached in each treatment for the two glr3.4 alleles, the wild type, and the two alleles of glr3.  and in Table 1 provide some critical information germane to the question. If each effective amino acid activated a specific channel type, then one ligand should not affect the sensitivity of the membrane to another. A Gly→Glu treatment should produce the same result as a Glu→Gly treatment. However, a Gly→Glu treatment elicited two full depolarizations while a Glu→Gly treatment produced one full depolarization and a second response suppressed by desensitization.
The first example is consistent with desensitization being due to a specific agonist-channel pair undergoing activity-based, or homologous desensitization (Gainetdinov et al., 2004), but the second example is inconsistent with each agonist independently activating and desensitizing a specific channel.
The other scenario, in which all six effective amino acids are equivalent agonists of a channel, predicts that the first treatment should always desensitize the mechanism to the second treatment regardless of which amino acid was delivered first. The observed asymmetry in the cross-desensitization relationships appears to rule out this scenario. A caveat requiring investigation was the possibility that multiple agonists of a common channel differed by degree of potency in a manner that created a desensitization hierarchy. The doseresponse curves in Figure  It is possible to accommodate almost all of the data obtained in this study with a model based on three types of qualitatively different, heteromeric GLR channel subtypes that undergo homologous desensitization. As shown in Figure   7, type A channels are activated and desensitized only by Glu. Type A channels contain at least one GLR3.3 subunit, without which the channel cannot function.
This is based on the severe impairment of all responses by glr3.3 mutations. observation that also appears to support the uniqueness of Asn and Ser action is that depolarizations induced by these two were narrower in shape, i.e. shorter duration, on average, than responses to the other amino acids even though the peak potentials attained were similar. This difference was quantified by recording the membrane potential 60 s after application of the agonist. As shown in Supplemental Figure 3, the more negative values for Asn and Ser indicated that the membrane had repolarized back toward its initial condition more so than for the other four amino acids, which on average showed a more prolonged depolarization.
A derivative of this model, equally consistent with the data, has each GLR subunit possessing two binding sites -an LAOBP domain (traditional glutamate binding site) and an N-terminal LIVBP binding domain for other amino acids (Felder et al., 1999;Bouché et al., 2003;Acher and Bertrand, 2005). In this case, all subunits would be able to bind Glu as well as another agonist, and type A channels would be an unnecessary feature of the model. objective in the overall effort to learn the biological function of this intriguing family of receptor channels.

Plant Growth
Arabidopsis thaliana seeds (Columbia ecotype) were surface sterilized with 75% ethanol and sown on 0.6% agar Petri plates (3.5-cm diameter) containing 1 mM KCl, 1 mM CaCl 2 , 5 mM MES, pH 5.7 adjusted with Bis-Tris Propane. The plates were maintained at 4°C in darkness for 48 to 72 h before transfer to a growth chamber with a 16/8 h light/dark photoperiod and grown vertically for 4 days.

Membrane Potential Measurement
Measurements of membrane potential were made as previously described (Dennison and Spalding, 2000). Data acquisition was performed with a PCI-MIO-16XE-10 analog-to-digital converter (National Instruments, Austin, TX) controlled by custom software written in the Labview computer language (National Instruments). The sample rate was 20 Hz. Perfusion of the recording chamber (7 mL min -1 ) was driven by a peristaltic pump (Dynamax RP-1, Rainin Co.). The initial (control) perfusion solution was the same as the growth medium (minus the agar) and could be switched to one supplemented with an amino acid by electronic valves via the data acquisition software. The inflow tube was positioned as close as possible to the site of impalement to produce an abrupt (within approximately 2 s) exchange of solutions, which aided reproducibility of the resulting ionic responses. Experiments proceeded only if a stable membrane potential more negative than -120 mV was obtained in the control condition.

Measurement of intracellular Ca 2+ with Aequorin
Seeds of Arabidopsis plants expressing aequorin, described previously (Lewis et al., 1997), were sown and grown on agar as described above for 4 d before Integrating this large, luminescence transient produced a value that was used to normalize the experimental response, taking into account any variability in tissue content and coelenterazine uptake among trials. The first 25 s of each normalized recording, which included the entire response to agonist, were integrated and the integrals for each trial were averaged ( Figure 5A). To express these same data relative to the maximum possible response for each agonist ( Figure 5B), each result was divided by the mean response to a saturating (10 mM) agonist treatment and then multiplied by 100. The small Ca 2+ response to buffer-only treatment was also subtracted from each trial value.
Data shown in Figure 1C was obtained in a slightly different manner.