Differential Expression of Glutamate Receptor Subunits in the Nervous System of Caenorhabditis elegans and Their Regulation by the Homeodomain Protein UNC-42

Both NMDA and non-NMDA classes of ionotropic glutamate receptors are expressed in C. elegans

We have identified by sequence similarity searches (BLAST) 10 genes encoding putative glutamate receptor subunits in C. elegans (see Materials and Methods). glr-1 has been described and encodes a subunit most similar to the non-NMDA receptor subtypes (Hart et al., 1995; Maricq et al., 1995). We have cloned partial cDNAs for the remaining nine receptor subunits. These genes were located on linkage groups I, II, III, V, and X (see Table 2). Although some genes were closely spaced (glr-1 andglr-2 were separated by ∼0.1 map units), the receptors did not reside in clusters. We designed oligonucleotide primers to amplify cDNAs predicted to encode a highly conserved region of ∼400 amino acids for each of the nine new receptor subunits. This region includes the predicted transmembrane segments, pore region, and ligand-binding regions (Stern-Bach et al., 1994). Sequence alignments of these predicted polypeptides revealed many of the conserved regions and specific amino acid residues that are found in ionotropic glutamate receptors (Fig. 1). Overall, each of the predicted C. elegans subunits has the highest amino acid identity with members of a particular class of vertebrate glutamate receptor subunits (Table 1; indicated in bold). For example, GLR-2 has the highest identity to AMPA subtypes, GLR-3 to kainate subtypes, and NMR-1 to NMDA subtypes.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1.

Partial sequence alignments of predicted C. elegans ionotropic glutamate receptors. Sequence alignment of a highly conserved region of ∼400 amino acids of GLR-1 and the additional nine putative C. elegans glutamate receptor subunits is shown. This region includes the four hydrophobic domains (TMI–TMIV, solid underline) and two ligand-binding domains (S1, S2, dashed underline witharrows) that are similar in sequence to bacterial amino acid-binding proteins (Nakanishi et al., 1990; Stern-Bach et al., 1994). The polypeptide sequences were derived from partial cDNAs amplified from mixed-stage first-strand cDNA and are numbered on theleft beginning with the first amino acid included in the alignment. Identical or similar residues are shaded in blackor gray, respectively. Ce, C. elegans.

The topological arrangement of each protein is predicted to be similar to that of other NMDA and non-NMDA receptors. We identified four hydrophobic regions, three of which are predicted to be transmembrane domains (Fig. 1, TMI, TMIII, and TMIV). TMII, the pore-forming region, is a reentrant loop that does not traverse the membrane (Hollmann et al., 1994) and contains amino acids that have been shown to modulate channel permeability in mammalian ionotropic glutamate receptors (Hollmann, 1999). In mammals, a conserved glutamine in TMII is a key determinant of rectification and calcium permeability in non-NMDA receptors (Fig. 2) (Hume et al., 1991;Verdoorn et al., 1991). In a subset of receptor subunits, the conserved glutamine can be converted to an arginine by a process of RNA editing (Sommer et al., 1991). Receptors that contain a subunit with an arginine at this position have a significantly reduced calcium permeability (Burnashev et al., 1992). In C. elegans, most of the non-NMDA-type subunits contain the conserved glutamine in TMII. Because both the genomic and cDNA sequences encode for the same amino acids in TMII, RNA editing of the conserved glutamine is unlikely or rare. However, within the genomic sequence two of the receptors have a basic amino acid substituted for the conserved glutamine in TMII; GLR-5 has a lysine and GLR-6 has an arginine, suggesting that in these cases the alternate edited form is genomically encoded. In contrast, GLR-7 has a proline substituted for glutamine. The effect of these changes is difficult to predict but is likely to affect the conductance of the channel. Like other glutamate receptors, two separated regions in the protein (S1 and S2) have a strong identity to bacterial amino acid-binding proteins and are believed to be required for ligand binding (Nakanishi et al., 1990; Kuryatov et al., 1994; Stern-Bach et al., 1994; Armstrong et al., 1998) (Fig. 1). Other regions show conserved features of glutamate receptors, including a nine amino acid sequence in TMIII (SYTANLAAF) that is found in essentially all ionotropic glutamate receptors (Fig. 2). The third alanine in this sequence is conserved among all known receptor subunits. Mutating this alanine to a threonine in selected subunits results in a constitutively conducting ionotropic receptor (Zuo et al., 1997; Zheng et al., 1999).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

Comparison of C. elegans,Drosophila, and rat ionotropic glutamate receptor subunits. Sequence alignment of C. elegans,Drosophila, and rat ionotropic glutamate receptor subunits beginning immediately at TMI and terminating at TMIII is shown. C. elegans subunits have signature features found in both vertebrate and invertebrate subunits. These include predicted hydrophobic domains (solid underline), a conserved glutamine residue in TMII (filled circle; a target of RNA editing in a subset of vertebrate non-NMDA subunits), and a sequence of nine amino acids in TMIII (dashed line) found in almost all known ionotropic glutamate receptor subunits. This region also contains a conserved alanine residue (filled triangle) that when mutated to threonine causes the ion channel to be constitutively open (Zuo et al., 1997; Zheng et al., 1999). Amino acids are numbered on the left beginning with the first residue in TMI. Identical or similar residues are shaded inblack or gray, respectively. Dro,Drosophila.

The relationship between the putative C. elegans and known vertebrate glutamate receptor subunits was analyzed by generating a neighbor-joining tree of receptor subunits (Fig.3). Clear groupings of the subunits can be distinguished. Thus, GLR-1 and GLR-2 group together and are most similar to rat AMPA receptors. GLR-3 and GLR-4, and more weakly GLR-5 and GLR-6, also group together. GLR-3–GLR-7 are clearly non-NMDA receptor subunits but cannot be obviously classified into pharmacological subtypes. GLR-8 is the farthest outlier and is even more divergent than glutamate receptors identified inArabidopsis. NMR-1 groups with rat NR1 andDrosophila NR1, and NMR-2 groups with rat NR2 receptors.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3.

Phylogenetic tree of the amino acid sequences for glutamate receptor subunits. Unrooted phylogenetic tree of the amino acid sequences for the 10 identified C. elegans putative glutamate receptor subunits and selected rat,Drosophila, and Arabidopsis subunits (see Materials and Methods) is shown. The predicted C. elegans polypeptides are based on the cloned partial cDNAs (Fig. 1). The schematic shows clusters of relationships between the amino acid alignments of the glutamate receptor subtypes. The total length of the horizontal lines between different receptors is proportional to the difference between sequences; distance along the vertical axis has no significance. Five hundred bootstrap replicates were performed, and the bootstrap values are indicated by dots on the supporting branches.No dot indicates a value <50%. At,A. thaliana.

Many neurons in C. elegans express multiple glutamate receptor subunits

To determine which cells in C. elegans expressed a particular glutamate receptor subunit, we examined transgenic worms that expressed chimeric fusion proteins to the reporter molecule GFP (Chalfie et al., 1994). Typically, GFP was fused in-frame to the glutamate receptor subunit following the predicted second transmembrane domain (Fig. 4; see Materials and Methods). The expression of each chimeric protein was controlled by ∼5 kb of upstream sequence. In previous experiments we found that the expression pattern of GFP controlled by 5 kb of glr-1upstream sequence did not differ from the pattern observed when GFP was fused to a partial or full-length GLR-1 protein (Maricq et al., 1995;Zheng et al., 1999). However, to minimize the risk of disrupting possible intragenic control elements, we included most of the predicted introns in each construct (Fig. 4). For nmr-1, we show the expression pattern of a partial fusion to NMR-1. The neurons identified by this construct were identical to those that expressed a full-length fusion protein that showed considerably weaker GFP expression (data not shown). Using these constructs, we were able to observe reproducible GFP expression in transgenic strains.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 4.

Glutamate receptor–GFP fusion proteins. Schematic representation of the genomic fusions to GFP for each of the glutamate receptor subunits. Boxes represent exons, andhorizontal lines between boxes represent introns. Black and gray boxes represent genomic sequence that was either included or omitted from the GFP fusion construct, respectively. The underlined regionsrepresent intron and exon boundaries that have been confirmed by cDNA sequence that was used to predict the amino acid sequence shown in Figures 1 and 2. The amount of upstream regulatory sequence included in each fusion is indicated as is the site of GFP fusion (inverted triangle).

Ionotropic glutamate receptor subunits were expressed in a locomotory control circuit and in a thermosensory circuit

Many of the glutamate receptor subunits were found to be expressed in the command interneurons that control worm locomotion (Fig.5). For example, the neuron AVA expresses both the non-NMDA subtypes GLR-1, GLR-2, GLR-4, and GLR-5 and the NMDA subtypes NMR-1 and NMR-2. Other command interneurons expressed various subsets of these subunits (Table 2). Several patterns emerged from the expression data. First, the NMDA-type receptor subunits NMR-1 and NMR-2 were always coexpressed and were found in a relatively small number of neurons that included four of the five command interneurons. Second, many neurons expressed multiple non-NMDA receptor subunits in various combinations. For example, AVB expressed GLR-1 and GLR-5; AVD expressed GLR-1, GLR-2, and GLR-5; RIM expressed GLR-1, GLR-4, and GLR-5; and RIF expressed GLR-4 and GLR-5. Presumably, these neurons could create a variety of mature heteromeric receptors from these subunits. Third, some neurons expressed only one subunit. For example, RIB expressed only GLR-4, LUA expressed GLR-5, and AIA expressed GLR-2. Whether these subunits can form functional homomeric receptors or whether other subunits were expressed at levels too low to detect remains to be determined.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 5.

Four non-NMDA and two NMDA ionotropic glutamate receptor subunits are expressed in the command interneurons of the locomotory control circuit. Confocal micrographs of transgenic worms expressing GFP fusions to glr-1, glr-2,glr-4, glr-5, nmr-1, andnmr-2 are shown. Only the anterior head and posterior tail regions of the worm are shown (two left columns). A schematic reconstruction of the neuronal cell bodies in the head and tail of the worm are also shown [two right columns; adapted from White et al. (1986)]. Cells filled in grayand black represent neurons that express the respective subunit. Each of these GFP fusions is expressed in at least one of the five pairs of command interneurons, AVA, AVB, AVD, AVE, and PVC (black cells). Neurons that express at least one of the six receptor subunits have been labeled in the bottom panel. The command interneurons of the locomotory control circuit are filled in black. Scale bars, 5 μm.

Although most glutamate receptor subtypes are expressed in multiple neurons, two non-NMDA subunits, GLR-3 and GLR-6, were exclusively expressed in the thermosensory interneuron RIA (Fig.6). This neuron also expressed the GLR-2 receptor subunit (Fig. 5). RIA is considered the major integrative thermosensory interneuron because it receives input from the interneurons AIY and AIZ. In addition, as revealed by laser ablation studies, it is required for worms to thermotax in defined temperature gradients (Mori and Ohshima, 1995).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 6.

Two ionotropic glutamate receptor subunits are exclusively coexpressed in the thermoregulatory interneuron RIA (arrows). A, B, Confocal micrographs (anterior region only) of transgenic worms expressingglr-3:: GFP (A) andglr-6:: GFP (B) are shown. C, Expression of both fusion proteins is detected in a single neuron, the thermoregulatory interneuron RIA (arrow). Scale bars, 5 μm.

Two genes, glr-7 and glr-8, were primarily expressed in the pharyngeal nervous system

The pharynx of C. elegans functions to ingest and process food. It contains 58 cells, including 20 neurons that comprise the pharyngeal nervous system. This collection of neurons is almost entirely separate and independent from the rest of the C. elegans nervous system (Albertson and Thomson, 1976). The function of some of these neurons is known to modulate pharyngeal pumping (Raizen and Avery, 1994). Two subunits, GLR-7 and GLR-8, were expressed in the pharyngeal nervous system (Fig.7). GLR-8 is expressed in all GLR-7-expressing neurons but is also found uniquely expressed in other pharyngeal and extrapharyngeal neurons (Fig. 7, Table 2). Both GLR-7 and GLR-8 are outliers in the sequence alignment, and GLR-8 cannot be obviously categorized into NMDA or non-NMDA classes (Fig. 3). However, coexpression of GLR-7 and GLR-8 in specific neurons suggests that they may encode subunits of the same subtype and combine to form glutamate-gated receptors.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 7.

Two ionotropic glutamate receptors are coexpressed in the pharyngeal nervous system. A, Confocal image of a transgenic worm expressing the glr-7:: GFPtransgene in the pharyngeal nervous system (head region only).B, Diagram of the pharynx and the neuronal cell bodies that express glr-7:: GFP. C, D, Confocal images of a transgenic worm expressingglr-8:: GFP in the pharyngeal nervous system (C, head region only) and in a single neuron in the tail (arrow; D). E, F, Neuronal cell bodies that express glr-8:: GFP. Neurons filled in black coexpressglr-7:: GFP andglr-8:: GFP. Neurons filled ingray only express glr-8:: GFP in either the pharyngeal nervous system (E; URB is the only neuron of this subset that is not considered part of the pharyngeal nervous system) or the tail (F). Scale bars, 5 μm.

Expression of all glutamate receptor subunits began during embryogenesis

Expression of the GFP reporter constructs could be observed at all larval stages and in the adult. The onset of expression was typically noted in late embryogenesis (Fig. 8).glr-5 had the earliest onset of expression in which GFP fluorescence could be observed at the twofold stage (∼450 min after fertilization). Within another 100 min of development (at the threefold stage), expression of glr-1, glr-2,glr-4, glr-7, glr-8, nmr-1, and nmr-2 was first observed. We could not detectglr-3 or glr-6 expression until near hatching. However, the GFP signal from the transgenic strains that expressed these constructs was rather weak, and the late onset of expression may simply reflect low expression levels.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 8.

Expression of ionotropic glutamate receptor subunits can be detected during embryogenesis. A–C, Image of a twofold stage embryo that expressedglr-5:: GFP (arrowheads). Expression of this fusion protein was detected the earliest of all of the GFP fusion transgenes. D–F, Image of a threefold stage embryo that expressed glr-1:: GFP. Most of the GFP fusions were first detected at this stage. G, Time course of embryogenesis (minutes) indicating the approximate time of the earliest detected expression of each of the receptor subunits.glr-3:: GFP andglr-6:: GFP expression was not detected during embryogenesis. A, D, Nomarski. B, E, FITC. C, F, Overlay. Scale bars, 5 μm.

Expression of GLR-4 and GLR-5 was dependent on the homeodomain protein UNC-42

The genetic control of glutamate receptor expression is not well understood. In a recent study, the homeodomain gene unc-42was found to be required for the development of neuronal identity and specifically required for the expression of GLR-1 in the AVA, AVD, and AVE interneurons and the RMD motor neurons (Baran et al., 1999). UNC-42 itself was expressed in many additional neurons that were not known to express glutamate receptors. To determine whether unc-42differentially regulates glutamate receptor expression, we examined the expression pattern of glutamate receptor subunits in anunc-42 mutant background. We confirmed thatunc-42 is required for GLR-1 expression in AVA, AVD, AVE, and RMD neurons (Fig. 9A). Of the other genes encoding glutamate receptor subunits, onlyglr-4:: GFP and glr-5:: GFP had altered GFP expression. GLR-4 and GLR-5 were normally expressed in many neurons; however, the expression of GLR-4 in AVA and RMD (Fig.9B) and the expression of GLR-5 in AVA, AVD, AVE, and RMD (Fig. 9C) were dependent on unc-42. Interestingly, expression of GLR-2 and the two NMDA subunits NMR-1 and NMR-2 was independent of unc-42. These subunits were also expressed in AVA, AVD, AVE (GLR-2, NMR-1, and NMR-2), and a subset of the RMD neurons (GLR-2). Thus, of the neurons that expressed glutamate receptor subunits and whose identity was controlled, in part, byunc-42, the expression of a subset of non-NMDA receptor subunits was dependent on unc-42. The glutamate receptor subunits GLR-3, GLR-6, GLR-7, and GLR-8 were expressed in neurons that did not express unc-42, and their expression was not found to be dependent on unc-42.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 9.

UNC-42 is required for expression of GLR-1, GLR-4, and GLR-5. A–C, Confocal images ofunc-42(e270) transgenic worms that expressedglr-1:: GFP (A),glr-4:: GFP (B), andglr-5:: GFP (C).D–F, Diagram of neuronal cell bodies that expressglr-1 (D), glr-4(E), and glr-5(F). Cell bodies filled in grayand black represent those that expressed the respective subunits in wild-type worms. Cell bodies filled in grayrepresent those that no longer express the subunits inunc-42 mutants. unc-42 is required forglr-1 expression in the command interneurons AVA, AVD, and AVE and in the RMD motor neurons (Baran et al., 1999). It is also required for glr-4 and glr-5 expression in these same cells. Scale bars, 5 μm.

unc-42 regulates axon outgrowth in a subset of the command interneurons

The command interneurons have been divided into two interconnected subclasses, those that primarily regulate backward movement (AVA, AVD, and AVE) and those that primarily regulate forward movement (AVB and PVC) (Chalfie et al., 1985) (Fig.10H). Whether moving forward or backward, unc-42 mutants move slowly and haltingly and often display kinking behavior. They also have abnormalities in the escape response to tactile stimulation (Baran et al., 1999) (Table3). Rather than moving backward in response to anterior stimulation (Chalfie et al., 1985), unc-42 mutants either fail to respond or move forward (Table 4). The backing response to tactile stimulation of the nose is also disrupted in unc-42 mutants (Baran et al., 1999) (Table 4). The normal response to posterior tactile stimulation is forward movement, and this response, although diminished and uncoordinated, still occurs in unc-42mutants. How can these various defects be explained?

• Download figure
• Open in new tab
• Download powerpoint

Fig. 10.

unc-42 disrupts the axon outgrowth of AVA, AVD, and AVE. A–D, Confocal images of transgenic unc-42(e419) mutants (A, B) and unc-42(e270) mutants (C, D) that expressed the nmr-1:: GFP transgene. Abnormal processes can be observed extending anterior to the nerve ring (arrows) or along dorsal or lateral paths (arrowheads). E, Confocal image of a transgenic unc-42(e270) mutant expressing bothnmr-1:: GFP and a wild-typeunc-42 genomic clone. Axon outgrowth is indistinguishable from that of wild-type worms. F, Wild-type worm that expressed nmr-1:: GFP. All images show the head region only, with anterior on the left and more posterior regions on the right of each panel. See Figure 5 for the identity of neurons that expressnmr-1:: GFP. G, Processes of the ventral cord in an nmr-1:: GFP; unc-42(e270)mutant. Only four processes were visible in the ventral cord (three of these indicated by arrows), whereas nine are expected in a wild-type worm. Migration of a subset of the processes terminated prematurely in the ventral cord (arrowhead). H, I, Schematic diagram of the locomotory control circuit in wild-type (H) and unc-42mutants (I). The circuits have been divided into forward (black) and backward (gray) components. Gap junctions and chemical synapses between sensory neurons (triangles), command interneurons (hexagons), and motor neurons (circles) are represented by lines andarrows, respectively. In wild-type worms (H) the DA motor neurons that direct backward movement receive synaptic input from the backward command interneurons (AVA, AVD, and AVE). In unc-42mutants (I) these connections are disrupted, and the worms have difficulty moving backward. Note that both the forward and backward command interneurons receive synaptic input from sensory neurons (AVM and ASH) that normally drive backward movement in response to anterior tactile stimulation. Scale bars, 5 μm.

Because nmr-1:: GFP expression was not dependent onunc-42, we were able to use this transgene to study the morphology of the command interneurons in unc-42 mutants. Transgenic unc-42 mutants showed severe axon outgrowth defects in the backward command interneurons AVA, AVD, and AVE (Fig. 10). Normally, the processes of these neurons enter the nerve ring and then travel posteriorly in the ventral cord (Fig.10F) (White et al., 1986). However, inunc-42 mutants, the processes of these neurons extended anteriorly, laterally, or dorsally, and in many of these worms multiple displaced processes were evident (Table 4; Fig.10A–D). Although it was difficult to identify the origin of each aberrant process, the processes of AVD were disrupted in most worms. Furthermore, because the migration of some processes terminated prematurely, fewer than the expected number of processes were observed in the ventral cord (Fig. 10G). This suggests that the aberrant processes reflect misdirection of the primary wild-type process rather than additional ectopic projections. In contrast to the processes of AVD, the processes of PVC, a forward command interneuron that does not express UNC-42, never appeared disrupted (data not shown). The unc-42 axon outgrowth defects were completely rescued in transgenic worms that expressed a wild-type genomic clone encoding UNC-42 (Table 4; Fig.10E). These axonal defects may depend solely on the lack of UNC-42 expression in the affected neurons, or it may be dependent on expression of UNC-42 in neighboring neurons or cells. The rescue of the axon outgrowth defects observed in AVA, AVD, and AVE was accompanied by rescue of the movement and touch-response abnormalities.unc-42 mutants that expressed the wild-typeunc-42 transgene had normal forward and backward movement, and touch sensitivity and escape responses to nose touch, anterior touch, and posterior touch were comparable with that of wild-type worms (Table 4).

Neuronal expression of glutamate receptors inC. elegans

In the vertebrate nervous system, 18 different receptor subunits, as well as alternatively spliced or RNA-edited variants, are differentially expressed throughout the CNS (Petralia and Wenthold, 1996; Hollmann, 1999). Individual subunits may be widely distributed (NR1, NR2A, and GluR2), have a more limited tissue distribution (GluR4 and NR2C), have lower levels of expression (kainate receptors), or have a distribution that is limited to only a few cell types (delta2 receptor). Individual neurons often express many different receptor subunits, including AMPA, kainate, and NMDA subtypes that may be intermingled at synapses (Takumi et al., 1999a). In addition, in some neurons subunits are found to be differentially localized to distinct synapses (Landsend et al., 1997; Rubio and Wenthold, 1997).

In C. elegans, our cloning and expression studies revealed a surprising number of putative glutamate receptor subunits for an organism with only 302 neurons. By comparing the amino acid sequence of these subunits with known vertebrate receptor subunits, we have assigned subunits to NMDA and non-NMDA classes. However, whether receptors formed from these subunits are characteristically activated by the selective pharmacological agents AMPA, kainate, and NMDA remains to be determined. The receptor subunits were predominantly expressed in interneurons, and most neurons expressed a number of receptor subunits. Subunits closely related by sequence were likely to be coexpressed in the same neurons, e.g., GLR-1 and GLR-2 or NMR-1 and NMR-2. However, some neurons expressed only one subunit, which might therefore form functional homomeric receptors. Alternatively, we have not identified all of the ionotropic glutamate receptor subunits in C. elegans (see Materials and Methods for additional candidate glutamate receptors). In contrast, other neurons, such as AVA, expressed six receptor subunits. Interestingly, unlike vertebrate neurons, most C. elegans neurons do not express NMDA receptors. Consequently, the majority of neurons that express non-NMDA receptor subtypes do not express either of the NMDA subtypes NMR-1 or NMR-2. Unlike the case for vertebrate glutamate receptor subunits, we found no evidence of alternative splicing or RNA editing of any subunits. In Drosophila, neuromuscular transmission is glutamatergic; however, in C. elegans, glutamate receptor expression in muscle cells was not observed.

The subunit composition of vertebrate glutamatergic synapses changes with activity. Whether the initial formation of glutamatergic synapses is also dependent on activity is unclear. One model suggests that glutamatergic synapses initially contain only NMDA receptors and that non-NMDA receptors are later recruited to the synapse in an NMDA receptor and activity-dependent process. Thus, synaptic plasticity is achieved by modulating the relative level of non-NMDA receptor expression in response to patterns of synaptic activity. In this view, the expression of glutamate receptors at synapses is a dynamic process and is partly dependent on the activation of NMDA receptors. Thus, in hippocampal CA1 pyramidal neurons, certain synapses, termed “silent synapses,” contain only NMDA receptors (Liao et al., 1995). During development or with activity that leads to the activation of NMDA receptors, these synapses acquire AMPA receptors and become more active (Petralia et al., 1999a; Shi et al., 1999; Takumi et al., 1999b). However, there is also strong evidence that AMPA-mediated synaptic neurotransmission develops in the absence of NMDA receptor activity (O'Brien et al., 1997; Rao and Craig, 1997;Gomperts et al., 2000). In C. elegans, synapse formation in most neurons must be independent of NMDA receptor activity, because we have shown that most neurons that express non-NMDA receptors do not express NMDA receptors. Thus, for these neurons, synapse formation and function are independent of NMDA receptor activity. Whether the command interneurons that coexpress NMDA and non-NMDA receptor subunits show activity dependence of synapse formation remains to be determined.

What are the functions of C. elegans glutamate receptor subunits? How might they contribute to neuronal processing in the worm? Hints are provided by their respective neuronal expression patterns. Thus, GLR-3 and GLR-6 were expressed in an interneuron required for thermotaxis and may be required for normal thermotaxis behavior; GLR-7 and GLR-8 were expressed in the pharyngeal nervous system and may participate in pharyngeal function; and GLR-1, GLR-2, GLR-4, GLR-5, NMR-1, and NMR-2 were expressed in command interneurons and may contribute to locomotory control. The large number of subunits in some neurons suggests that considerable receptor diversity may exist. However, which subunits form functional glutamate-gated channels remains to be determined.

How many neurons are glutamatergic?

The widespread expression of glutamate receptors in interneurons and motor neurons suggests that many sensory neurons and interneurons are glutamatergic. A recent study identified EAT-4 as a putative Na-P_i cotransporter in C. elegans that is required for vesicular uptake of glutamate (Lee et al., 1999). Furthermore, the mammalian homolog of EAT-4 has now been identified and functions to transport glutamate into synaptic vesicles (Bellocchio et al., 2000; Takamori et al., 2000). The EAT-4 cotransporter is expressed in at least 15 different classes of neurons in C. elegans. On the basis of the expression patterns of EAT-4 (Lee et al., 1999), the various glutamate receptors described here, and the serial electron microscopic analysis of the C. elegans nervous system (White et al., 1986), one can begin to predict which synaptic inputs are glutamatergic. For example, AVA expresses a variety of glutamate receptors and receives synaptic inputs from a large number of neurons. These inputs make contact in different regions of the AVA process. However, only a small subset of these input neurons express EAT-4 (Lee et al., 1999). Thus, the potentially glutamatergic neurons include ASH, AUA, and FLP in the nerve ring, PVD and PLM in the ventral cord, and LUA in the preanal ganglion. These predictions can be directly tested. For example, by the use of GFP-tagged presynaptic and postsynaptic markers, GLR-1 has been shown recently to colocalize with a subset of ASH presynaptic specializations (Rongo et al., 1998). Interestingly, the number of GLR-1 synapses (320) reported by Rongo et al. differs markedly from that predicted by counting the synaptic inputs provided by EAT-4-expressing neurons (∼130). This discrepancy can be explained if low, undetected levels of EAT-4 were expressed in additional neurons. Alternatively, other unidentified transporters may participate in vesicular uptake of glutamate.

Expression of glutamate receptor subunits is differentially regulated by UNC-42

Mutations that affect the connections among the locomotory command interneurons should disrupt locomotion. For example, vab-8mutants are uncoordinated and have particular difficulty moving backward. In these mutants the outgrowth of most posteriorly directed axons, including the axons of a subset of the command interneurons, is disrupted (Wightman et al., 1996). Interestingly, unc-42mutants are also severely uncoordinated and do not move forward or backward in a normal manner. They are sensitive to body wall touch, but rather than moving backward in response to anterior tactile stimulation, they usually move forward. We have shown thatunc-42, required for the expression of GLR-1 in the backward command interneurons (AVA, AVD, and AVE) (Baran et al., 1999), is also required for GLR-4 and GLR-5 expression in these neurons. Defective expression of GLR-1 in unc-42 mutants does not cause the uncoordinated locomotion because glr-1 mutants move in a coordinated manner. Because unc-42 mutants are also defective in the expression of GLR-4 and GLR-5, it may be that compound loss of all three subunits leads to the severe locomotory defects observed in unc-42 mutants. Loss of single subunits, such as GLR-1, may be compensated by the presence of the remaining subunits. However, because unc-42 mutants show multiple defects, the phenotype is unlikely to be caused by loss of glutamate receptors alone.

UNC-42 is required for axon outgrowth of the backward command interneurons

By using the nmr-1 promoter to express GFP in AVA, AVD, and AVE in unc-42 mutants, we have shown that the most likely cause of uncoordinated movement in unc-42 mutants was disruption of synaptic communication secondary to defective axon outgrowth of AVA, AVD, and AVE. Whether these axon outgrowth defects were secondary to defects in glutamate receptors, defects in the expression of additional gene products in interneurons, or defects in the target tissues remains to be determined. Our results suggest that the sensory defects (nose touch and body touch) may also be caused by disruption of the interneuronal circuitry. When these neurons were killed either by a laser (Chalfie et al., 1985) or by specific expression of a caspase (Zheng et al., 1999), the worms were touch insensitive and uncoordinated. The severe axon outgrowth defects observed in the backward command interneurons in unc-42mutants are predicted to be incompatible with normal neuronal signaling and therefore should phenocopy the sensory and motor defects observed with neuronal ablation. The requirement reported by Baran et al. (1999)for expression in the AB.p cell lineage suggests that UNC-42 function may be required in AVE(R) or that low levels of UNC-42 are expressed in RIM. Ablation of RIM suggests that it may also have a role in touch response and coordinating backward and forward locomotion (Zheng et al., 1999).

UNC-42 is also expressed in the ASH sensory neurons that are required for the nose touch response. However, the processes of these neurons are not disrupted in unc-42 mutants (Baran et al., 1999), suggesting that the nose touch defect observed in unc-42mutants is caused by a postsynaptic defect. We provide evidence of a postsynaptic mechanism by showing that processes of AVA, AVD, and AVE are disrupted in unc-42 mutants.

The forward movement of unc-42 mutants is disrupted, although the forward command interneurons do not express UNC-42 and their processes are normal in unc-42 mutants. These results support a model in which the function of the command interneurons is, in part, distributed (Zheng et al., 1999). unc-42 mutants move forward in response to posterior touch, and paradoxically, they also often move forward in response to anterior touch. This result can be explained by the known wiring of the body wall touch receptors (Fig.10H,I) (Chalfie et al., 1985; White et al., 1986). The anterior touch receptors make synaptic contacts with both the forward and backward command interneurons (Fig.10H). Presumably, only contacts to the backward circuit are disrupted in unc-42 mutants, leaving contacts to the forward circuit intact (Fig. 10I). Consequently, the worms move forward rather than backward in response to anterior touch. Our results show that the initiation of an appropriate forward or backward response is dependent on subsets of the command interneurons but that coordinated forward or backward movement requires an intact circuit that contains all of the command interneurons (Zheng et al., 1999).

Summary

In a small and relatively simple nervous system, such as that found in C. elegans, single neurons perform many tasks undertaken by ensembles of neurons in larger nervous systems. For example, the amphid sensory neurons that detect olfactory cues express numerous genes encoding odorant receptors, whereas in the vertebrate olfactory system each neuron is thought to express only one specific receptor (Chess et al., 1994; Troemel et al., 1995). A situation similar to that of amphid sensory neurons may exist for interneurons in the C. elegans nervous system. The small number of command interneurons receives a large number of synaptic inputs from sensory neurons and interneurons and integrates this information to direct locomotory output. These neurons also express a large assortment of glutamate receptor subunits. Presumably, this diversity of subunits is required in part for the integrative function of these interneurons. The expression pattern of glutamate receptor subunits suggests that they may have critical roles in the control of locomotory response, pharyngeal function, and sensory processing. How these subunits are partitioned within neurons and how they function to control neuronal function await more detailed cell biological and electrophysiological analyses.

Note added in proof. Several recent reviews have addressed the predicted glutamate receptors that were identified by theC. elegans genome sequencing consortium. These reviews have used different receptor nomenclatures. Littleton and Ganetsky (2000)used the names of the cosmids that contained the receptor sequences.Hollmann (1999) and Sprengel et al. (2001) assigned names somewhat different than those we use here. Our nomenclature follows that introduced by Maricq et al. (1995) and Hart et al. (1995) and is based on the sequences of the cloned cDNAs.