Introduction

The cysteine-loop ligand-gated ion channels (cys-loop LGICs) constitute a superfamily of pentameric channels that are defined by a fixed-length cysteine-loop in the long, extracellular N-terminal domain, and by four membrane spanning regions (TM1–TM4). These cys-loop receptors, which mediate rapid chemical synaptic transmission (Hille 2001), include both cationic and anionic channels and are found in both vertebrates and invertebrates. Much has been learned about the ancestry of the cys-loop receptors (Le Novère and Changeux 1995; Ortells and Lunt 1995; Vassilatis et al. 1997; Xue 1998; Tasneem et al. 2004; Dent 2006) with their antecedents seen in prokaryote LGIC-like proteins. The prokaryote ancestors, however, lack the cysteines that define the cys-loop superfamily, suggesting that these receptors appeared after the prokaryote–eukaryote split (Tasneem et al. 2004).

The nicotinic acetylcholine (ACh) receptor is the prototype of the cationic cys-loop receptors (Karlin 2002), although in invertebrates both anionic and cationic forms of this receptor exist (van Nierop et al. 2005). The GABA-A receptor, on the other hand, is the prototype of anionic cys-loop receptors. Its invertebrate orthologs, however, include both cationic and anionic forms (Beg and Jorgensen 2003; Gisselmann et al. 2004; Dent 2006). Finally, there exists a subfamily of presumably only anionic receptors that are characterized by a second cys-loop typically occurring about 45 residues downstream from the signature cys-loop in the ligand-binding domain and are labeled here as 2-cys-loop receptors. The prototype of the 2-cys-loop receptor is the vertebrate glycine receptor (see Lynch 2004; Cascio 2004; Connolly and Wafford 2004; Lester et al. 2004), which is the only such receptor found in vertebrates. The second cys-loop has been shown to be critical for agonist binding in the glycine receptor (Rajendra et al. 1995; Grudzinska et al. 2005).

Subfamilies of 2-cys-loop receptors are found in both protostomes and deuterostomes. A recent phylogenetic analysis revealed a 2-cys-loop receptor in the urochordate Ciona intestinalis which joined the clade of vertebrate glycine receptors (Dent 2006). Until now, no other 2-cys-loop receptors have been identified in deuterostomes. In contrast, four distinct types of 2-cys-loop receptor have been described in protostomes, but so far none have been shown to be glycine-gated (Cully et al. 1996, Gisselmann et al. 2002, Zheng et al. 2002, Schnizler et al. 2005). It is only in ecdysozoa (nematodes and arthropods), however, that examples of 2-cys-loop receptors have been identified outside of the chordate subphylum. Consequently, we wanted to determine whether 2-cys-loop receptors from lophotrocozoa, the other superphylum of protostoma, have a similar spectrum of ligand sensitivities to that of ecdysozoa. With this in mind, we first cloned two 2-cys-loop receptors from Aplysia californica, a molluscan representative of the lophotrocozoa superphylum. Then, to better understand the evolution of the 2-cys-loop receptors, we searched Genbank for such receptors in a mammal, an echinoderm, a nematode, and in insects. We also searched the recently cloned genomes (DOE Joint Genome Institute) of a urochordate, a cephalochordate, an annelid, and another mollusk for the same receptor type. We then subjected all these sequences to a phylogenetic analysis, first alone, and then with 1-cys-loop GABA-A receptors from the same phyla and from a cnidarian. Finally, we profited from a recent model of the glycine binding pocket (Grudzinska et al. 2005) that we modified in light of the findings of Pless et al. (2008) to predict which of the 2-cys-loop receptors might be expected to bind glycine.

Methods

Cloning

Degenerate PCR primers were designed on the basis of an alignment of eight insect and nematode glutamate-gated Cl channels. Two partial sequences were obtained using the forward primer 5′-TGGRTGYC(Inosine)GAY(Inosine)B(Inosine)TTYTTYHV(Inosine)AAYGA-3′ which targeted the region beginning at the so-called loop A of the glycine receptor (Galzi et al. 1990; Vafa et al. 1999 and Fig. 1 below) and the reverse primer 5′-RTCNADCCARAANSW(Inosine)ACCCA-3′ which targeted the end of first transmembrane-spanning region. For each of these partial cDNA sequences (one of 456 base pairs, the other of 480 base pairs) the 3′ and 5′ cDNA ends were obtained using the SMART RACE cDNA amplification kit from Clontech. Once the complete coding region was obtained, gene-specific primers and the Eppendorf TripleMaster PCR System were used for the final series of PCR reactions. Each of the full-length cDNAs retrieved by these procedures was then subcloned into pcDNA3.1/V5-His (TOPO Cloning, Invitrogen). Miniprep DNA was screened by restriction digestion and verified by commercial sequencing. The GenBank accession numbers for GluClAc1 and GluClAc2, respectively, are GQ148562 and GQ148563.

Fig. 1
figure 1

Alignment of GluClAc1 and GluClAc2 with the glycine α1 receptor of Rattus norvegicus. Loop A is the first of four loops proposed to be involved in ligand binding in the ACh receptor (Galzi et al. 1990; Vafa et al. 1999) and was the region used here for making the 5′ degenerate primer for cloning the two Aplysia subunits (see “Methods”). Cys-loop #1 corresponds to the defining loop of the superfamily of cys-loop LGICs, whereas cys-loop #2 corresponds to the defining loop of the 2-cys-loop receptor subfamily. TM1, TM2, TM3 and TM4 indicate the four membrane spanning regions. The broken lines refer to the two intracellular loops M1–M2 and M3–M4, and the extracellular loop M2–M3, all three of which refer to a sequence of residues connecting the corresponding membrane spanning regions. At the junction of the M1–M2 intracellular loop and the transmembrane region TM2, three residues, PAR, have been underlined. These residues in the glycine α1 receptor subunit were labeled 250, 251 and 252, respectively, by Grenningloh et al. (1990a), because they defined as the first residue of the glycine α subunit the residue which follows the putative signal peptide cleavage site. Their numbering system has been used for all mutations performed on the glycine receptor. The numbering of the PAR residue positions have alternatively been referred to as −2, −1, and 0 (Sunesen et al. 2006), referring to the end of the M1–M2 loop. The nine residues in the α1 glycine sequence that have been bordered by a rectangle are the nine critical residues of the glycine binding motif (Grudzinska et al. 2005). These residues will be discussed near the end of the article, and are indicated here for later reference

Full size image

Culture and Transfection

Chinese Hamster Ovary (CHO-K1) cells were obtained from the American Type Tissue Culture Collection (ATCC, Molsheim, France) and maintained as previously described (Medina et al. 2000). Two days before transfection, cells were plated on coverslips (12 mm in diameter), which were placed inside 35-mm cell culture dishes with 2 ml of medium. On the day of, but prior to, transfection the old medium was replaced by 1 ml of fresh medium. CHO cells were transfected with approximately 1 μg/1 μl DNA using the Lipofectamine-reagent + transfection protocol (Life Technology, USA). To facilitate the identification of GluR-expressing cells, a Green Fluorescent Protein (GFP) cDNA was cotransfected (0.5 μg/μl). Three hours after the initial exposure of the cells to the cDNAs, a fresh cDNA-containing solution replaced the old one. Electrophysiological recordings were performed from fluorescent cells 24–48 h after transfection.

Electrophysiological Recording

Whole-cell recordings were conducted at room temperature (20–25°C) using an EPC-9 amplifier (HEKA Elektronik, Germany). Cells were continuously superfused through an independent tube with external solution containing (mM) NaCl 140, CaCl2 2, KCl 2.8, MgCl2 1, HEPES 20, glucose 10; pH 7.3; 320–330 mOsm. The patch pipette solution contained (mM) CsCl 140, MgCl2 2, MgATP 2, NaGTP 0.4, HEPES/CsOH 10, BAPTA/KOH 20; pH 7.3; 290 mOsm. In some experiments the CsCl in the pipette solution was reduced to 20 mM by replacing 120 CsCl by 120 K-gluconate. Pipettes were pulled from borosilicate glass capillaries (Harvard Apparatus Ltd, USA) and had resistances of 3–8 MOhms. A rapid perfusion system was used for obtaining concentration–response curves. Different concentrations of agonists were applied using a computer-driven fast exchange system (SF 77A Perfusion Fast-Step, Warner, USA). Three parallel square tubes for delivering the control and agonist-containing solutions were positioned 40–50 μm above the recorded cell. A 10–90% solution exchange was obtained in 1–2 ms, as evaluated using open electrode controls (1/10 NaCl). Cells with a low input resistance (<150 MOhms) and a rapid run-down (>30% with repetitive application) were excluded from analysis.

For quantitative estimations of agonist action, concentration–response relationships were fitted by the equation:

$$ {{I = I_{ \max } } \mathord{\left/ {\vphantom {{I = I_{ \max } } {\left( { 1 + \left( {{\text{EC}}_{ 50 } / \left[ C \right]} \right)^{n} } \right)}}} \right. \kern-\nulldelimiterspace} {\left( { 1 + \left( {{\text{EC}}_{ 50 ~ } / \left[ C \right]} \right)^{n} } \right)}} $$

where I is the current amplitude induced by the agonist at concentration [C], I max is the maximum response of the cell, n is the Hill coefficient and EC50 is the concentration for which a half-maximum response is induced. Concentration–response relationships were constructed using at least four points. The statistical differences between the EC50’s and Hill coefficients of the different receptors formed by transfections were evaluated using the one-way ANOVA test, the Kruskal–Wallis rank-sum test, or a paired-T test, as appropriate.

Glutamate, glycine, β-alanine, GABA, and ivermectin were all purchased from Sigma.

Phylogenetic Analysis

2-cys-loop channels were identified by BLAST analysis of non-redundant protein sequences from Genbank and from databases of filtered (F) and/or all gene (AG) protein models from a number of metazoan genomes available at Joint Genome Institute (JGI). The queries used were the truncated versions of the glycine α1 subunit from rat, the two glutamate-gated subunits cloned here from Aplysia californica, as well as histamine-gated and pH-sensitive (pHCl) receptor subunits from Drosophila melanogaster. The specific genomes targeted by the searches were Rattus norvegicus, Gallus gallus, Xenopus tropicalis, Morone americana, Ciona intestinalis, Branchiostoma floridae, Strongylocentrotus purpuratus, Capitella sp., Daphnia pulex, Lottia gigantea, and Nematostella vectensis.

Because vertebrate glycine receptors show such high sequence identity across species, only the four alpha sequences and the one beta sequence of Rattus norvegicus were necessary for representing the vertebrates. For the three other non-vertebrate deuterostome genomes that were searched, all 2-cys-loop receptors that were found were retained for the initial phylogenetic analysis: 12 sequences from the cephalochordate Branchiostoma floridae, 1 sequence from the urochordate Ciona intestinalis, and 2 sequences from the echinoderm Strongylocentrotus purpuratus. For the two lophotrocozoan genomes that were searched (Lottia gigantea, Capitella sp.), all non-redundant 2-cys-loop receptors that were found were likewise retained for the initial phylogenetic analysis (10 from Lottia gigantea, 9 from Capitella sp.). From the crustacean Daphnia duplex only a few sequences were selected to increase variety in the arthropod sequences which were already represented by insects in GenBank. A selection similar to that made for the vertebrate glycine receptor was made for glutamate-gated receptors from arthropods and nematodes, as well as for histamine-gated receptors from arthropods and the pHCl receptors from insects. The high interspecies similarity of the ecdysozoan receptors led to a choice of examples from Drosophila melanogaster, Tribolium castaneum and Apis mellifera for insects, Lepeophtheirus salmonis and Daphnia pulex for crustacea, and from C. elegans for nematodes. Preference was given to subunits for which the function had been confirmed by heterologous expression. It was only in C. elegans that, in addition to the 2-cys-loop receptors of known or presumed function, 2-cys-loop receptors with no known or predictable function were found. Four of six such sequences were retained for phylogenetic analysis, whereas the other two (NP_494579 and NP_001023062), both of which are likely to be cationic, were not retained (see “Results“, description of Fig. 1)

One genome from cnidaria (Nematostella vectensis; JGI) was queried as above, but no 2-cys-loop receptors could be found. Nevertheless, many 1-cys-loop sequences were identified that showed relatively strong bit scores and high % identities with both the glycine and the GABA-A receptors from rat. Three of them were retained for a phylogenetic analysis of 2-cys-loop and GABA-A 1-cys-loop receptors combined.

For the initial analyses, the selected 2-cys-loop receptors were aligned with ClustalX, and were truncated (but not edited), removing the poorly aligned and highly variable 5′ region preceding the 52nd residue of the α1 glycine receptor subunit—at which point a series of residues with high similarity over many of the 2-cys-loop receptors first appeared. At the 3′ end, the sequences were truncated following the third transmembrane region, because the subsequent M3–M4 intracellular loop and the last (4th) membrane spanning region varied markedly within and across species. Because some of the genomes used for this study are incomplete, a few sequences that were retained from the blast searches do not span the entire region defined by those two end points.

Two cationic 1-cys-loop receptors (ACh-a7 and 5HT-3 from rat) were included as the outgroup for rooted maximum likelihood phylogenetic analysis. An initial analysis was performed on the truncated ClustalX aligned, but unedited, sequences (see Supplement A) using the rapid maximum likelihood program Tree Puzzle 5.2 (Schmidt et al. 2002). Only sequences that were shown to join with others to form a clade were retained for further phylogenetic analysis. Of the 66 sequences, 7 failed to join a clade and will be referred to as loners. For the construction of PHYML trees, the truncated sequences were aligned using the Muscle alignment program (Edgar 2004), and were edited (Supplement B) before being introduced into the PHYML analysis program (Guindon and Gascuel 2003). SEQBOOT and CONSENSE were used for bootstrap analyses and the generation of consensus trees.

For the construction of PHYML trees, the truncated sequences were aligned using the Muscle alignment program (Edgar 2004) and were edited (Supplement B) before the choice of an amino acid replacement model was chosen by a ProtTest (Abascal et al. 2005) for evaluation of the alignments. For both the sets of sequences used (Figs. 4, 5), the optimal model recommended by the ProtTest and AIC evaluation was the JTT + I+G model, and the number of invariant sites and the alpha parameter for the gamma distribution were optimized in PhyML accordingly. A reexamination by ProtTest and AIC just prior to publication led to the recommendation of the more recent LG model, but the output trees obtained using that model were essentially identical to those obtained with JTT, so we retained our initial choice as described above.

Predictions Concerning the Possibility that Glycine Could Be a Ligand for a Given 2-cys-loop Receptor

We have used two means of predicting which ligand might or might not bind to the newly identified 2-cys-loop receptors. (1) Point mutations in the N-terminal domain of the glycine α1 subunit have revealed critical residues required for glycine binding (Grudzinska et al. 2005). Experimentally induced mutations shown to raise the EC50 of the α1 homomeric glycine receptor by 200-fold or more were labeled “fatal” mutations in any 2-cys-loop receptor in which such a mutation appeared naturally. Those receptors were predicted, without further evaluation, to be non-glycinergic. (2) In the cases for which the natural mutations seen in 2-cys-loop receptors had not been tested experimentally, we evaluated them in light of a modified version of the previously published model of the glycine binding pocket (Grudzinska et al. 2005) that takes into account recent experimental findings of Pless et al. (2008). Using this modified model (see Supplement D), we identified sequences that had other fatal or less than fatal mutations, and we have elaborated the justifications for our conclusions below and in Supplement E. For restrictions on these predictions see “Predictions concerning which of the 2-cys-loop subunits might bind glycine” in “Results”.

Results

The Two Subunits Cloned from Aplysia californica

Figure 1 shows the amino acid sequences of the two Aplysia subunits, labeled GluClAc1 and GluClAc2, which were cloned by degenerate PCR. The full protein sequences were used for searching a non-redundant protein database (BLAST, NCBI GenBank) for entries with which they showed strong similarities, measured in bit scores (see BLAST glossary, NCBI). The alpha subunits of the vertebrate glycine receptor gave the highest bit scores for each of the Aplysia sequences (309 for GluClAc1, and 338 for GluClAc2), and consequently these sequences have been aligned in Fig. 1 with the glycine α1 subunit (from Rattus norvegicus).

As can be seen in Fig. 1, the GluClAc1 and GluClAc2 subunits, like all other cys-loop receptors, have a large hydrophilic extracellular domain at the N-terminus, four hydrophobic transmembrane segments (TM1–TM4), and an extended cytoplasmic domain (M3–M4).

The comparison of the vertebrate and molluscan 2-cys-loop sequences shown in Fig. 1 reveals two significant differences. First, the M3–M4 intracellular loop is considerably longer in the vertebrate sequence than it is in either GluClAc1 or GluClAc2. An analysis of other known 2-cys-loop receptors from arthropods and nematodes, along with those discovered here by the screening of recently cloned genomes (a urochordate, a cephalochordate, a mollusk, an annelid, and a crustacean, see “Methods” and below) have shown that the shorter M3–M4 loop of the Aplysia receptors is typical of the 2-cys-loop receptors of the superphylum to which mollusks belong (lophotrocozoa). Whereas in sequences from mollusks and annelids that loop contained a mean number of residues of 47.6 (SD 11.4, n = 13) and 41.5 (SD 20, n = 8), respectively, the mean number of residues in that loop from other phyla that were studied here were significantly greater: 81.5 (SD 24.5, n = 10) for nematodes, 69.5 (SD 26.3, n = 14) for arthropods, and 75.4 (SD 22.4, n = 17) for chordates (one-way ANOVA, P = 0.0001). This characteristic is independent of the ligand that gates the 2-cys-loop receptors. Because this loop is known to be important for protein–protein interactions involved in trafficking, clustering, and related processes (see Lynch 2004; Kracun et al. 2008), this difference might limit the possibilities for intracellular modulation of the lophotrocozoan 2-cys-loop receptors.

The second difference observed in Fig. 1 between the vertebrate and molluscan sequences is that the two Aplysia sequences have more residues in the second cys-loop (12 and 11 for GluClAc1 and GluClAc2, respectively) than does the glycine α1 subunit (10 residues). The variability of the length of the 2nd cys-loop will be discussed further. That variability can be contrasted with the constant number of residues (13 between the cysteines) in the first cys-loop which is shared by all members of the superfamily of cationic and anionic cys-loop receptors (see cys-loop #1 in Fig. 1).

Both the glycine α1 subunit and the Aplysia sequences share the PAR signature at the end of the M1–M2 intracellular loop (residues underlined in Fig. 1). This signature is associated with most anionic cys-loop LGICs, and, specifically in the 2-cys-loop receptor type, appears to be found, in published subunits, only in those which have not been categorized or which have been labeled as alpha subunits. Consequently, the signature (PAR) has been used here to define the 2-cys-loop alpha subunits as opposed to so-called non-alpha subunits which have alternative residues in one or more of those three positions. These three residue positions have been referred to either as 250, 251, and 252 (see figure legend and Grenningloh et al. 1990a) or as positions −2, −1 and 0 of the M1–M2 intracellular loop (Sunesen et al. 2006).

In our search for presumably anionic 2-cys-loop receptors (see Supplement A), we found, and rejected for phylogenetic analysis, two C. elegans receptors (NP_494579 and NP_001023062) that we suspect to be cationic rather than anionic because the PAR signature in the two sequences was replaced by EGR and DVR, respectively. Wotring and Weiss (2008) have shown that a negatively charged residue at 250 (position −2) renders the resulting homomeric GABA-A ρ receptors permeable to cations. These two sequences are the only ones we found which, if shown to be truly cationic, would render invalid the generalization that the 2-cys-loop receptors are exclusively anionic.

Heterologous Expression of GluClAc1 and GluClAc2

Formation of Homomeric Glutamate-Gated Receptors

Each of the two subunits from Aplysia californica was shown to be capable of forming a homomeric receptor that responds to both glutamate and β-alanine, as does the receptor formed by double-transfection. Figure 2 shows three concentration–response curves obtained with both glutamate and β-alanine from three different cells transfected, respectively, with either (A) GluClAc1, (B) GluClAc2 or with (C) GluClAc1 + GluClAc2. Figure 2d shows the mean EC50’s obtained for the response to glutamate from many cells for each of the three transfection types. The mean EC50 for GluClAc1 (EC50 = 499 μM, SD 150, n = 7) differs significantly (P < 0.01) from both that of the homomeric receptor formed from GluClAc2 (EC50 = 196 μM, SD 132, n = 6) and from that of the receptor created by the double transfection (EC50 = 210 μM, SD 120, n = 17). On the other hand, no statistical difference could be seen between the mean EC50’s of the GluClAc2 homomer and the receptor formed by transfection with both subunits, so it is impossible to conclude that the latter (GluClAc1 + GluClAc2) is truly heteromeric. No significant differences were found for the mean Hill coefficients for the responses of the three receptor types to glutamate: 0.83 (SD 0.1) for GluClAc1, 1.0 (SD 0.2) for GluClAc2, and 0.9 (SD 0.05) for the double transfection.

Fig. 2
figure 2

Functional expression of GluClAc1 and GluClAc2 subunits. Whole cell recordings from individual CHO cells transiently transfected with a GluClAc1 cDNA, b GluClAc2 cDNA, or c GluClAc1 + GluClAc2 cDNAs. In a–c top: superimposed traces of ionic currents induced by a rapid application of 100 μM or 10 mM glutamate (left traces, filled circles) and 100 μM or 10 mM β-alanine (right traces, open circles). The duration of applications are indicated by bars above traces. V h = −30 mV. External and pipette solutions contained 150 mM Cl. a–c bottom: concentration–response relationships for whole cell currents induced either by glutamate (filled circles) or by β-alanine (open circles). All recordings in the top and bottom of each section (a–c) are from the same cell. a The EC50’s for glutamate- and β-alanine-induced currents were 323 μM and 1561 μM, respectively; the corresponding Hill coefficients, 0.7 and 1.2. b The EC50’s for glutamate- and β-alanine-induced currents were 86 μM and 311 μM, respectively; the corresponding Hill coefficients, 0.95 and 0.92. c The EC50’s for glutamate- and β-alanine-induced currents were 245 μM and 1150 μM, respectively; the corresponding Hill coefficients, 1 and 1.6. d Mean EC50 and standard deviation of the mean for cells transfected with GluClAc1 = 499 μM, SD 150 (n = 7), GluClAc2 = 196 μM, SD (n = 6) and GluClAc1 + GluClAc2 = 210 μM, SD 120 (n = 17). * P < 0.0001, one-way ANOVA

Full size image

Response of GluClAc1 and GluClAc2 to β-Alanine

In all three types of transfection of the Aplysia subunits, the response to β-alanine consistently yielded an EC50 considerably higher than that seen for glutamate (see Fig. 2 a–c). For the five cells exposed to the two agonists (one cell each for the two homomers, and three cells having received the double transfection), the mean EC50 for β-alanine was 6-fold higher than that for glutamate (SD 3.1, n = 5) with a paired T test rejecting the null hypothesis at P = 0.03. The mean Hill coefficients obtained from the same concentration–response curves did not differ significantly.

This differential sensitivity to glutamate and β-alanine was also seen in the native receptors (Kehoe and Vulfius 2000), as was a similar differential sensitivity to glycine and β-alanine in the vertebrate glycine receptor (Schmieden et al. 1993, Schmieden and Betz 1995). β-alanine has not been tested on either the nematode or the arthropod heterologously expressed glutamate-gated receptors.

The Receptors Gated by Glutamate Are Cl Channels

As can be seen in Fig. 3, lowering the internal Cl concentration from 150 to 24 mM caused a close to Nernstian shift in the reversal potential of the response (predicted 45 mV shift; obtained, 37 mV). Three additional experiments in which low-Cl pipette solutions were used produced similar shifts in ECl compared to the values obtained in 14 I–V curves made from cells which were patched with high-Cl pipette solutions. The fact that the same cell cannot be used for both control and low-Cl experiments, and the fact that one would need a full intracellular exchange to obtain the predicted shift might contribute to the less than full Nernstian shift.

Fig. 3
figure 3

Glutamate-gated channel exhibits Cl selectivity. a, b superimposed traces of whole-cell currents induced by a rapid application of 1 mM glutamate at different holding potentials (V h in mV are shown on the left of traces). CHO cells transiently transfected with GluClAc1 + GluClAc2 cDNAs. External solution contained 152 mM Cl. a pipette solutions contained 24 mM Cl, b pipette solutions contained 150 mM Cl, c I–V relations for glutamate-induced currents recorded in whole-cell from two different CHO cells transfected with GluClAc1 + GluClAc2 cDNAs. Recording pipette solutions contained 24 mM Cl (circles) and 150 mM Cl (squares). Reversal potentials were −42 and −6 mV, respectively, close to the E Cl values predicted by the Nernst equation

Full size image

Evaluation of the Responses of the Expressed Receptors to GABA, Ivermectin, and Glycine

GABA

Both the homomeric receptors and the receptor resulting from a double transfection respond to high concentrations of GABA (not shown), as do the native receptors (Oyama et al. 1990; Kehoe and Vulfius 2000). Concentration–response curves were not constructed, but for the values compared (300 μM glutamate and 10 mM GABA), the sensitivity to glutamate was over 600-fold greater than that for GABA. It was also confirmed that, as in the native receptors (Oyama et al. 1990; Kehoe and Vulfius 2000) 100 μM GABA (which itself elicited no measurable response) reduced the amplitude of the response to glutamate (300 μM) by 20–60%. The effect of GABA on the glutamate response was tested exclusively on the receptors resulting from a double transfection.

Some, but not all, homomeric glycine receptors also respond to high concentrations of GABA (de Saint Jan et al. 2001), and GABA has recently been shown to modulate glycinergic transmission at a mammalian auditory synapse (Lu et al. 2008). On the other hand, 1 mM GABA (the highest concentration tested) failed to activate the glutamate-gated Cl channels from the ecdysozoan glutamate receptors (Vassilatis et al. 1997; Cully et al. 1996; Gisselmann et al. 2002; Schnizler et al. 2005; Horoszok et al. 2001; Eguchi et al. 2006).

Ivermectin

The glutamate-gated receptors of nematodes and insects have been shown to be the target of ivermectins—insecticides and antiparasitics isolated from soil-dwelling Streptomyces—and of their semi-synthetic equivalent, ivermectin (see Cully et al. 1996; Zheng et al. 2002; Schnizler et al. 2005; Dent et al. 2000; Yates et al. 2003). As was found to be the case for the native glutamate-gated chloride channels (Kehoe and Vulfius, unpublished observations), the heterologously expressed glutamate-gated Cl receptors described here responded to 5 μM ivermectin with an irreversible increase in Cl conductance (tested only in the doubly transfected cells). Vertebrate glycine receptors have also been shown to respond to ivermectin with an irreversible increase in Cl conductance (Shan et al. 2001).

Glycine

In spite of the sensitivity of the Aplysia glutamate receptors to one of the agonists of the glycine receptor (β-alanine), no response to glycine itself (1–10 mM) could be seen in either of the two homomeric receptors formed from GluClAc1 or GluClAc2 or in the receptor resulting from double transfection. This echoes the finding from the native glutamate-gated Cl channels (Kehoe and Vulfius, unpublished data) which even 10 mM glycine failed to activate. Electrogenic responses to glycine have been observed in Aplysia neurons (Kehoe 1976), but they do not reflect the activation of a Cl channel.

Glycine, tested only at 1 mM, likewise failed to elicit a response from any of the ecdysozoan 2-cys-loop receptors (Cully et al. 1996; Vassilatis et al. 1997; Dent et al. 1997; Zhen et al. 2002; Schnizler et al. 2005; Eguchi et al. 2006).

The above findings reveal much in common between the vertebrate glycine-gated and the Aplysia glutamate-gated chloride channels. In spite of their independent ligands, it was seen that the two receptor types have similar sensitivities to a number of other compounds. In the section below we attempt to establish the phylogenetic relationship between the vertebrate glycine receptors, the glutamate-gated Aplysia channels, and many other receptors of the same type taken from a broad range of both deuterostome and protostome phyla.

The Analysis of 66 2-cys-loop Receptors has Revealed Many Novel LGIC Clades

Combining the known 2-cys-loop receptors selected from GenBank, the sequences found in the five genomes from JGI (Branchiostoma floridae, Strongylocentrotus purpuratus, Capitella sp., Lottia gigantea, Daphnia pulex and Nematostella vectensis) and the two Aplysia sequences described here gave a total of 66 2-cys-loop sequences for our first analysis (see “Methods” and Supplement A for procedures used in the selection, truncating, and editing of sequences).

The ClustalX alignment of the truncated, unedited 66 2-cys-loop sequences and two 1-cys-loop cationic channels (NP_0773705 and NP_036964; used as an outgroup) is presented in Supplement A. The alignment reveals that 59 of the 2-cys-loop sequences contain the so-called alpha PAR signature at positions 250, 251 and 252, whereas 7 sequences have an alternative, non-alpha signature at those same positions. In Supplements A and B, the alpha signatures have been presented in red font, whereas the 7 non-alpha signatures (including two labeled beta; see Grenningloh et al. 1990b and Cully et al. 1994, and one labeled zeta, see Hutton et al. 1993) have been underlined.

The above aligned sequences were then introduced into the Tree puzzle Maximum Likelihood program (Schmidt et al. 2002). This program yielded a rooted, bootstrapped tree (not shown), and revealed the formation of clades of sequences. A few sequences, however, failed to form a clade with other sequences. That was the case for the only 2-cys-loop sequences from Strongylocentrotus purpuratus, as well as for 1 sequence from Branchiostoma floridae, 3 sequences from Lottia gigantea, and 1 from Capitella sp. These 7 loners (see “Methods“) were not retained for further phylogenetic analysis.

The remaining 59 2-cys-loop sequences plus the 2 outgroup cationic LGICs were then subjected, in the form of an edited Muscle alignment (see Sup. B), to a maximum likelihood phylogenetic analysis using PHYML (Guindon and Gascuel, 2003). The resulting dendogram is presented in Fig. 4, where names of the subunits which have been heterologously expressed and for which a ligand had been verified are underlined. If the ligand has been identified for one member of a family, all the members of the family were assumed to contribute to the same receptor type, as indicated by the label juxtaposing the relevant clade.

Fig. 4
figure 4

Dendogram of 62 2-cys-loop channel subunits. Maximum likelihood phylogenetic tree derived from truncated, aligned (Muscle), and edited amino acid sequences obtained from GenBank and four species-specific databases from JGI (see text for details and Supplement B for alignment). Sequences which have been expressed and the function confirmed are underlined. Clade names given in bold represent families for which at least 1 member has been expressed and the ligand successfully determined. Sequence names written in italics are non-alpha sequences. Sequence labels end in abbreviations indicating the name of the species: Cs Capitella sp., Lg Lottia gigantea, Ac Aplysia californica, Lym.s Lymnaea stagnalis, Ce C. elegans, Dp Daphnia pulex, Bf Branchiostoma floridae, Rn Rattus norvegicus, Ci Ciona intestinalis, Dm D. melanogaster, Tc Tribolium castaneum, Am Apis mellifera, Ls L. salmonis. The “Cs, Dp, Lg, and Bf” proteins, taken from the relevant JGI databases, are indicated as “F” or “AG” referring to protein sequences which were taken from either “filtered gene models” or “all gene” models, respectively. All the other sequences were taken from GenBank. The bootstrap values written in bold and in larger font are those referred to in the text

Full size image

The most striking revelation of the PHYML analysis of the 59 2-cys-loop receptors described here (see Fig. 4) was the independence of the ecdysozoan and lophotrocozoan clades of glutamate-gated receptors. GluClAc1 and GluClAc2 form a well-supported clade with subunits from the mollusk Lottia gigantea (labeled Mollusca Glutamate with a bootstrap of 93, Fig. 4), forming a less well-supported clade when joined either by the annelid sequences from Capitella sp. (bootstrap 75) or by both the annelid sequences and the molluscan non-alpha sequences to yield a clade with a bootstrap of 64 of Lophotrocozoa putative Glutamate receptors. There is no association between these lophotrocozoan known or presumed glutamate subunits and those of the ecdysozoa, which form an independent well-supported ecdysozoa glutamate clade (bootstrap 90).

Several other features of the 2-cys-loop subunit phylogeny shown in Fig. 4 are noteworthy. It expands the diversity of subunits to include an annelid-specific clade and a lophotrocozoan-specific clade of 2-cys-loop receptors of unknown ligand sensitivity. There is no evidence in Fig. 4 for a molluscan ortholog to the arthropod histamine receptors.

In previous studies, the only non-vertebrate chordate 2-cys-loop channel that was shown to belong to the glycine receptor clade (Dent, 2006) was from Ciona intestinalis. Here we expand this clade (Chordata Glycine clade, bootstrap 97) to include three subunits from the primitive chordate Branchiostoma floridae which appear to be orthologs of the vertebrate glycine receptors. In addition, there are two clades of Branchiostoma subunits that are more divergent, and of indeterminate ligand sensitivity (one, a 2-sequence clade, another, a 6-sequence clade, both with bootstraps of 100 and of unknown function).

Comparison of the Phylogenic Profile of the GABA-A Receptor Family and that of 2-cys-loop Receptor Families

To better specify when the 2-cys-loop clades diverged, it is helpful to compare them with a subunit clade represented in all species examined (see also Xue 1998 and Dent 2006). GABA-A receptors (either known or predicted) are found in all the phyla studied here. For the phylogenetic analysis shown in Fig. 5, we eliminated from the 59 2-cys-loop sequences used in Fig. 4 the sequences from Branchiostoma and Ciona, 2 somewhat redundant sequences from both the Arthropod Histamine clade and the Lophotrocozoa putative Glutamate clade, as well as all 16 sequences making up the three clades of receptors of unknown function (see Fig. 4). To the remaining 30 2-cys-loop sequences we added the 2 outgroup sequences, 6 GABA-A subunits from Rattus norvegicus (αγβρτ) and 17 homologs of these sequences from C. elegans, Drosophila melanogaster, Aplysia californica, Lottia gigantea, Lymnaea stagnalis and 3 GABA-A-type receptors from the cnidarian, Nematostella vectensis, yielding a total of 58 sequences represented in Fig. 5.

Fig. 5
figure 5

Dendogram of 30 2-cys-loop channel subunits and 26 1-cys-loop GABA-A or presumed GABA-A subunits. Maximum likelihood phylogenetic tree derived from truncated, aligned (Muscle), and edited amino acid sequences obtained from GenBank and four species-specific databases from JGI (see text for details and Supplement C for alignment). Sequence labels end in abbreviations indicating the name of the species: Cs Capitella sp., Lg Lottia gigantea, Ac Aplysia californica, Lym.s Lymnaea stagnalis (in Lophotrocozoa), Ce C. elegans, Dp Daphnia pulex, Rn Rattus norvegicus, Dm D. melanogaster, Tc Tribolium castaneum, Am Apis mellifera, Ls L. salmonis (in Insecta). The “Cs, Dp, and Lg” amino acid sequences, taken from the relevant JGI databases, are indicated as “F” or “AG” referring to protein sequences which were taken from either “filtered gene models” or from “all gene” models, respectively. All the other sequences were taken from GenBank. The bootstrap values written in bold and in larger font are those referred to in the text

Full size image

The dendogram in Fig. 5 confirms the independence of the lophotrocozoan and ecdysozoan clades of glutamate-gated channels (bootstrap support 74 and 84, respectively). It also shows that all the putative and known protostome and deuterostome GABA-A subunits form a clade (bootstrap 93). Moreover, subunits from the cnidarian N. vectensis also join the GABA-A receptor clade.

Figure 5 also reveals an interesting pattern of divergence among the protostome GABA-A-type receptors. To the ρ-, β- and τ-type GABA-A receptor clade we added lophotrocozoan orthologs to the previously described ecdysozoan GABA receptors (one each from nematodes and arthropods, see bootstrap 98), whereas for the α–γ-type GABA-A receptor clade (bootstrap 100) we were able to identify lophotrochozoan orthologs for each of the two insect receptors (CAA55144, known as GRD, and NP_573090), as well as for the nematode receptor (NP_499662).

Interestingly, of the invertebrate homologs of the anionic α- and γ-type GABA-A vertebrate receptors, 9 of the 11 appear to be cationic and form a distinct clade with a bootstrap of 86. Among them is the Drosophila channel CAA55144 (GRD), which has been shown to form a heteromultimeric cationic receptor with the beta subunit NP_727916, LCCH3 (Gisselmann et al. 2004). As in the GRD channel, in the other 8 uncharacterized subunit sequences, an aspartate replaces the alanine at position −1 (e.g. underlined SDR, ADR, or EDR signatures in alignment 3S, Supplement C). It has been shown that mutating the alanine in position −1 to a negatively charged amino acid yields cationic receptors in a glycine receptor subunit (Keramidas et al. 2000), in a glutamate-gated Cl subunit (Sunesen et al. 2006), in a GABA-A β subunit (Jensen et al. 2005), and in the C. elegans MOD-1 Cl channel (Menard et al. 2005).

Predictions Concerning Which of the 2-cys-loop Subunits Might Bind Glycine

The phylogenetic analysis presented in Fig. 4 provides a categorization of the members of the 2-cys-loop receptor family which suggests, among other things, that glycine receptors are not exclusively vertebrate, but, rather, appear to be exclusively chordate. Furthermore, the analysis suggests that in at least one chordate, both glycinergic and non-glycinergic 2-cys-loop receptors coexist. Recently proposed hypotheses concerning the binding motif of the vertebrate glycine receptor (Grudzinska et al. 2005; Pless et al. 2008) offer a second approach that we exploit below for predicting glycine receptivity of the 66 2-cys-loop receptors.

The model of Grudzinska et al. (2005) incorporates nine critical residues and predicts their interactions both with glycine and with the other residues of the binding motif. These nine residues have been enclosed in a rectangle in the α1 glycine receptor sequence (a1glyrat) in the alignment in Fig. 1, and are also indicated in bold blue font in the α1 glycine subunit NP_037265 in the alignment of the 66 2cys-loop receptors in Supplement A.

A 2-dimensional representation of the model of Grudzinska et al. (2005) can be found in Supplement D (Fig. 1SA). Recently Pless et al. (2008) experimentally tested certain of the hypotheses expressed in the Grudzinska model. In Fig. 1SB of Supplement D we present a modification of the Grudzinska model that takes into account the findings of Pless et al. (2008). Using this modified model and the results from experimental mutations described and/or referred to in Grudzinska et al. (2005; see also Supplement F), we predict whether or not a naturally occurring mutation could be expected to disrupt glycine binding (see Supplements E and F). When such a disruption was predicted to lead to an increase in the EC50 of the heterologously expressed monomeric α-subunit of 200-fold or more, the mutation was labeled fatal, and the receptor was predicted to be non-glycinergic.

This evaluation does not permit us to conclude that a receptor predicted to bind glycine would forcibly yield a functional glycine receptor, because binding to the critical residues is only one component of many required for forming an active glycine receptor (see Grudzinska et al. 2005; Grenningloh et al. 1990b; Cully et al. 1994; Etter et al. 1996 and Supplement E for further discussion).

Furthermore, our predictions are also subject to the assumption that there is only one glycine binding motif and that the model of Grudzinska et al. (2005), revised in light of the findings of Pless et al. (2008), is correct. Finally, because many of the 2-cys-loop sequences were predicted by a bioinformatic approach, we cannot completely exclude that errors in individual residues could possibly perturb the analysis.

Supplements E and F contain the analysis of individual mutations in light of the revised model (Supplement D, Fig. 1SB). Figure 6 is presented to facilitate the overview of the conclusions from our analysis. It is a copy of Fig. 4 in which predictions have been indicated concerning the likelihood that a given 2-cys-loop sequence would or would not be able to bind glycine effectively.

Fig. 6
figure 6

A repeat of Fig. 4 containing indications concerning which of the 62 2-cys-loop sequences are predicted to bind glycine. Figure 6 is a copy of Fig. 4 in which predictions have been indicated for each of the 2-cys-loop sequences concerning the likelihood that they would be able bind glycine effectively (see text and Supplements E and F). The five sequences in the Chordate Glycine clade which are neither struck through nor bordered by a rectangle are known to be glycine receptors. The 40 sequences that have been struck through have been found to contain an experimentally tested fatal mutation (see Supplements E and F). The 14 sequences bordered by a rectangle are those that failed to contain an experimentally demonstrated fatal mutation, so have been subjected to an analysis in light of the glycine binding pocket (see Supplements D–F). For the sequences for which the analysis led to the prediction of a fatal mutation, their names have been struck through. For those found to contain mutations which would reduce glycine binding, but not sufficiently so to be considered fatal, one asterisk follows the name. For sequences for which all the binding residues are identical to or presumably isofunctional to those of the glycine α subunit, two asterisks have been attached to the sequence name. The two sequence names that have been underlined represent the cationic receptors serving as Outgroup. As for Fig. 4, sequence labels end in abbreviations indicating the name of the species: Cs Capitella sp, Lg Lottia gigantea, Ac Aplysia californica, Lym.s Lymnaea stagnalis (Lophotrocozoa), Ce C. elegans, Dp Daphnia pulex, Bf Branchiostoma floridae, Rn Rattus norvegicus, Ci Ciona intestinalis, Dm D. melanogaster, Tc Tribolium castaneum, Am Apis mellifera, Ls L. salmonis (Insecta). The “Cs, Dp, Lg, and Bf” proteins, taken from the relevant JGI databases, are indicated as “F” or “AG” referring to protein sequences which were taken from either “filtered gene models” or “all gene models”, respectively. All the other sequences were taken from GenBank

Full size image

In Fig. 6, the names of the five vertebrate sequences known to be glycine receptors are left as is. For the 40 2-cys-loop sequences containing an experimentally demonstrated fatal mutation, the names of the sequences have been struck through. For the 14 receptors which contain no experimentally tested fatal mutation, and which hence were evaluated in light of the glycine binding motif, the names of the sequences are surrounded by a rectangular border.

These latter 14 sequences were evaluated, mutation by mutation, at the nine critical residue positions (see detailed analysis in Supplements E and F). Among the 14 sequences are 3 for which all the nine residues are identical or isofunctional to those of the glycine receptor. They join the Chordata Glycine clade (bootstrap 97) and their names have been followed by two asterisks. Of the 14 sequences 6 were predicted to contain a mutation (or collection of mutations) categorized as fatal, so their names have been struck through, but retain their black border.

In the five remaining evaluated sequences mutations were expected to increase the EC50, though not sufficiently to be considered a fatal mutation. To identify such sequences, one asterisk was placed after the sequence name. Two of the sequences so classified are from Branchiostoma and form together a 2-sequence clade distinct from the Chordata Glycine clade. The three protostome sequences predicted to weakly bind glycine (1 asterisk) are each members of a clade in which all other members contain fatal mutations.

Of the seven “loner” sequences (none of which was included in the phylogenetic analysis), the 2 sea urchin receptors and 1 sequence from Lottia gigantea contained experimentally demonstrated fatal mutations. The remaining four loners were examined in light of the binding pocket model: 1 of the 4 (from Lottia gigantea) was predicted to contain a fatal mutation, 2 others (1 from Lottia gigantea and 1 from Branchiostoma floridae) were predicted to have mutations that would be expected to reduce the sensitivity to glycine, and the final loner (from Capitella sp.) was found to have identical or isofunctional residues at all the nine critical sites of the glycine binding motif. For details of these analyses see Supplement E and F.

The predictions concerning the ability to bind glycine of all the receptors subjected to analysis were identical whether we used the original or the revised model (Supplement D) for making the predictions.

Having analyzed the 59 2-cys-loop sequences of Fig. 4 in light of the binding pocket motif, we briefly return at this point to examine some structural aspects of the sequences belonging to the different clades formed by the phylogenetic analysis. The importance of the 2nd cys-loop in the binding pocket motif of the glycine receptor is evident, because 4 of the 9 critical residues are located between the cysteines of that loop. Although the amino acids composing that loop in the non-glycinergic receptor types are clearly different from those of the glycine receptor, it is interesting to note that a high identity of amino acids at many of the 10 residue positions between the cysteines is found not only in the sequences in the Chordata glycine clade (5 identical plus 1 isofunctional out of the 10–11 residues between the cysteines), but also in those of the ecdysozoa glutamate clade (6/10), of the Arthropoda histamine clade (7/10), and of the Insecta pHCl clade (6/10–11). In the lophotrocozoan clades, in contrast, almost no amino acids are identical at a given position in all members of a clade: 1/10–12 for the Mollusca Glutamate clade, 0/10–14 for the Putative Lophotrocozoa Glutamate clade, 1/10–12 for the Annelida Unknown function clade, and 2/10–11 for the Lophotrocozoa Unknown function clade. Two other clades were found to have low identity of amino acids at the same residue positions in the 2nd cys-loop: the Branchiostoma Unknown function clade (1/10) and the C. elegans Unknown function clade (2/10–19). The relevant alignments, the number of members of a clade, and the clades’ bootstrap values are given in Supplement G. It can be seen that, in general, the clades with the highest identity of amino acids in the 2nd cys-loop across members of the clade are also those that have either 10 or 10–11 residues between the cysteines, whereas in the clades with low identity, the length of the second cys-loop is usually much more variable.

Discussion

The expression studies of the two 2-cys-loop receptors cloned here from Aplysia have confirmed what seemed apparent from previous electrophysiological and pharmacological studies of the glutamate responses in Aplysia ganglia (e.g., Kehoe and Vulfius, 2000)—that the mollusc, like the insects and nematodes (e.g., Cully et al. 1996; Cully et al. 1994), express glutamate-gated chloride channels. The Aplysia receptors expressed in CHO cells, like the native receptors, were shown to respond to glutamate, β-alanine, and ivermectin. They also, like the native receptors and many glycine receptors, were found to share a weak sensitivity to GABA. It was, however, when these identified molluscan glutamate receptors were subjected to phylogenetic and sequence analysis in company of the other 2-cys-loop receptors that the originality of these lophotrocozoan 2-cys-loop receptors was revealed.

First, the lophotrocozoan clade of presumed glutamate receptors was found to be completely distinct and independent of the glutamate receptors from ecdysozoa. The degree of separation of the two glutamate clades is similar to the distance between clades representing channels of diverse ligand specificities, indicating that the mollusk and ecdysozoan glutamate-gated Cl channels have long evolved independently. This finding suggests either convergent evolution of glutamate binding in the two families or the existence of a common glutamate-sensitive ancestor in the deep roots of the metazoan lineage.

The phylogenetic analysis of the 2-cys-loop subunits seen in Fig. 4 also reveals the diversity of subunits with the appearance of an annelid-specific clade and a lophotrocozoan-specific clade of 2-cys-loop receptors of unknown ligand sensitivity. In spite of electrophysiological (Gruol and Weinreich 1979; McCaman and Weinreich 1985) and immunological (Evans et al. 1999) evidence strongly arguing for the presence of histamine-gated Cl channels in Aplysia, there is no evidence in Fig. 4 for a molluscan ortholog to the arthropod histamine receptors. This suggests that one of the lophotrocozoan clades of unknown function may represent independently evolved subunits of histamine receptors, paralleling the independence of the ecdysozoan and lophotrocozoan clades of glutamate 2-cys-loop receptors.

Our phylogenetic analysis of the 2-cys-loop receptors has also resulted in the expansion of the Chordata Glycine clade, shown now to include not only one sequence from a urochordate (Dent, 2006), but also, in addition, three subunits from a cephalochordate, Branchiostoma floridae. Furthermore, it has revealed the presence of two other Branchiostoma clades of indeterminant ligand sensitivity. The finding that all the Branchiostoma 2-cys-loop receptors share a monophyletic clade with the vertebrate 2-cys-loop receptors is consistent with their having diverged after the protostome–deuterostome split, although loss of the Branchiostoma-specific clades from other phyla cannot be ruled out.

The independence of the lophotrocozoan and ecdysozoan clades of glutamate-gated channels was further confirmed by the phylogenetic analysis (see Fig. 5) that included GABA receptors. That analysis also showed that all the putative and known protostome and deuterostome GABA-A subunits form a clade (Bootstrap 93, Fig. 5), thereby confirming that GABA and glutamate receptor clades diverged before the protostome–deuterostome split. Moreover, subunits from the cnidarian N. vectensis also join the GABA-A receptor clade, consistent with a model wherein the GABA-A subunits had already diverged from ancestors of the other subunit clades prior to the divergence of bilateria and cnidaria, a more ancient event. If we assume roughly equal rates of evolution for all subunit clades, a parsimonious interpretation of Fig. 5 leads to the conclusion that the various 2-cys-loop subunit clades represented here (including the glutamate clades) all have ancestors that likewise predate the divergence of cnidaria and bilateria. Although the rate of evolution certainly varies markedly among the species we examined, there is little evidence that the rate of variations significantly distort the phylogenetic tree. If high rates of evolution were distorting the phylogenetic analysis, then functionally homologous subunits from species with high evolutionary rates would be attracted to subunits from phylogenetically unrelated branches or form basal groups. Yet, although Drosophila and C. elegans exhibit relatively high rates of evolution, the glutamate-gated chloride channels of these species form a well-supported clade reflecting the evolutionary relationship of their respective phyla.

Similarly, nematode and insect GABA receptor subunits (CAA55144, NP_573090 and NP_499662) find apparent orthologs among the relatively slowly evolving lophotrocozoan GABA receptors instead of forming basal-branching, highly diverged clades, as might be expected from fast-evolving species. Although we cannot rule out branch attraction artifacts for clades that are present only in one phylum (e.g., the Arthropod Histamine clade and the Annelida unknown function clade of Fig. 4), the fact that homologous subunits from fast-evolving species form clades suggests that the deep branches represent ancient divergence events rather than rapid evolution. This conclusion is especially convincing for deep-branching lophotrochozoan clades because lophotrochozoa evolve relatively slowly (Moroz et al. 2006, Putnam et al. 2007).

Our analysis of the GABA-A receptors from protostomes also reveals interesting evidence for selective loss of some ecdysozoan α-type GABA-A receptors. The data from lophotrocozoa suggest that the three α-type receptors were present in the ancestor of ecdysozoa and lophotrochozoa: all three are still present in lophotrochozoa, one was lost from arthropods and the other two lost from nematodes. This analysis also reveals the probable difference in ion selectivity of the 1-cys-loop vertebrate GABA-A receptor (known to be anionic) and that of many of their protostome orthologs (predicted to be cationic).

Comparison of the Conclusions from the Phylogenetic Analysis and the Analysis in Light of the Glycine Binding Pocket Model

For deuterostoma, with one exception (sequence 234093BfF in the Chordate Glycine clade), the data from both the phylogenetic analysis and the analysis of the sequences in light of the binding pocket model yield similar conclusions: that glycine receptors are not exclusively vertebrate, and that, in the cephalochordate Branchiostoma at least, both glycinergic and non-glycinergic 2-cys-loop receptors coexist.

For protostoma, the situation is less clear. Examination of all of the ecdysozoan sequences confirmed the findings of the phylogenetic analysis because they all had experimentally tested fatal mutations. However, 9 out of the 22 lophotrocozoan sequences described in this paper (including three loners) did not contain an experimentally demonstrated fatal mutation, and 3 of them that were included in the phylogenetic analysis were predicted to bind glycine, although less effectively than do the vertebrate glycine receptors. Each of these three sequences (see Fig. 6) belongs to a distinct clade in which all the other members of the clade bear fatal mutations. The presence of sequences that can weakly bind glycine in clades of non-glycinergic receptors reflects somewhat contradictory conclusions from the two methods of analysis. However, it might suggest the possibility that, via such a subunit, glycine might play the role of modulator of non-glycinergic receptors, much as GABA does for both the glutamate-gated molluscan receptors and the glycine-gated vertebrate receptors (Kehoe and Vulfius 2000; de Saint Jan et al. 2001; Lu et al. 2008, respectively).

The lophotrocozoan loner sequence 165308CsF (Supplement F) that shares all the critical residues (or isofunctional equivalents) with the vertebrate glycine receptor demands an even closer look. If indeed the critical residues of the glycine binding pocket are sufficient for predicting sensitivity to glycine, that sequence would be the best candidate for a glycine receptor among protostomes.

No experimental mutations have been undertaken to establish binding motifs for either the ecdysozoan glutamate receptors or the lophotrocozoan glutamate receptors of which the first have been cloned and expressed here. However, it is evident from the analysis of the sequences in the two glutamate receptor clades that whereas many residues of the 2nd cys-loop of the ecdysozoan receptors could potentially be ligand-binding sites, as in the glycine receptor, that region in the lophotrocozoan glutamate receptor would be highly unlikely to serve that function due to the lack of amino acid identity in the 2nd cys-loop across the members of the clade—a characteristic found in all of the clades of lophotrocozoan receptors (see Supplement G).