Loss of activation by GABA in vertebrate delta ionotropic glutamate receptors

Significance The neurotransmitters glutamate and GABA activate excitatory sodium ion influx and inhibitory chloride flux across neuronal membranes by binding and activating ionotropic glutamate receptors (iGluRs) and GABAA receptors, respectively, two superfamilies of ligand-gated ion channels. Curiously, mammalian receptors of the delta iGluR family mediate no sodium influx in response to neurotransmitter binding in most experimental settings, so we investigated delta iGluRs from numerous animal lineages, bioinformatically and experimentally, and found that numerous delta iGluRs are indeed ligand-gated channels. Surprisingly, most are activated by GABA, the classically inhibitory neurotransmitter. Our results identify several amino acid substitutions that occurred during evolution to make mammalian delta iGluRs inactive and reveal a potential excitatory signaling role for GABA in numerous invertebrates.


Significance
The neurotransmitters glutamate and GABA activate excitatory sodium ion influx and inhibitory chloride flux across neuronal membranes by binding and activating ionotropic glutamate receptors (iGluRs) and GABA A receptors, respectively, two superfamilies of ligand-gated ion channels.Curiously, mammalian receptors of the delta iGluR family mediate no sodium influx in response to neurotransmitter binding in most experimental settings, so we investigated delta iGluRs from numerous animal lineages, bioinformatically and experimentally, and found that numerous delta iGluRs are indeed ligand-gated channels.Surprisingly, most are activated by GABA, the classically inhibitory neurotransmitter.Our results identify several amino acid substitutions that occurred during evolution to make mammalian delta iGluRs inactive and reveal a potential excitatory signaling role for GABA in numerous invertebrates.
Rapid signals are conveyed between neurons of the central nervous system via chemical synapses, specialized directional interfaces between adjacent neurons (1).Most synapses in the mammalian brain are glutamatergic and excitatory, where glutamate released by the presynaptic cell binds to ionotropic glutamate receptors (iGluRs) on the postsynaptic cell (2,3).iGluRs are tetramers assembled by four homologous subunits, each of which has an extracellular N-terminal domain (NTD), extracellular ligand-binding domain (LBD), membrane domain, and intracellular C-terminal domain (CTD).The four mem brane domains together form a membrane-spanning, nonselective cation channel (3).The iGluR superfamily is broad, but two main families predominate in mammals and classical model organisms: AMPA receptors, along with close cousins kainate (KA) receptors, are rapidly activated by glutamate and depolarize cells quickly (4); NMDA receptors are slower acting, more calcium-permeable, and require the binding of glutamate and an ambient coagonist, D-Serine or glycine (5).
Additionally, there is a relatively mysterious "delta" family of iGluRs encoded by two genes, GluD1 and GluD2 in rodents and human (6)(7)(8).Like most iGluRs, delta iGluRs are expressed in excitatory synapses, their up-or downregulation leads to developmental disorders or neural malfunction, and when expressed heterologously in mammalian cells or frog oocytes, GluD1 and GluD2 subunits form homotetrameric channels (9).In stark contrast to other iGluRs, however, under numerous native and heterologous experimental conditions, no current through GluD1 or GluD2 iGluRs is activated by neurotransmitter binding (7,(10)(11)(12).This is despite two major lines of evidence that the ligands D-serine and glycine bind to the canonical iGluR LBD of GluD2 and induce local conformational change.X-ray structures of the excised GluD2 LBD show D-serine binding and LBD closure around the ligand (13), reflecting the first step of channel activation in most iGluRs (3).Second, mutant GluD2 channels carrying the "lurcher" (Lc) A654T substitution in the channel pore conduct current constitutively, and the binding of D-serine or glycine inactivates this current, indicative of ligand-induced conformational change in both the LBD and in the channel pore (12,13).
The absence of typical ligand-gated currents in heterologously expressed delta iGluRs is reflected in relatively unique biological function.Intracellular signaling pathways are suggested to activate delta iGluRs based on mouse cerebellum and midbrain recordings (10,14,15), and a principally structural and developmental role is served by delta iGluRs in the hippocampus and cerebellum, where they interact with pre-and intrasynaptic proteins via their large extracellular NTD (16)(17)(18).NTD dynamics may in turn relate to the absence of ligand-induced gating in het erologously expressed receptors.The compaction of GluD2 extra cellular domains via coexpression with synaptic proteins in densely cultured mammalian cells or via introduced cysteine-linked NTDs leads to small ligand-gated currents in heterologously expressed GluD2 iGluRs (19).Furthermore, delta iGluRs differ from AMPA and NMDA iGluRs in that cryoelectron microscopy structures of full-length GluD1 and GluD2 iGluRs show a "nonswapped" archi tecture, where the back-to-back NTD dimers of two adjacent subunits sit atop back-to-back LBD dimers of the same two sub units (20,21), contrasting the "domain-swapped" architecture of other iGluRs (3).Whether nonswapped NTD structure underlies delta iGluR inactivity is doubtful, however, as plant iGluRs are nonswapped yet capable of ligand-gated currents (22).
Delta iGluR function is thus relatively mysterious.This impairs our understanding of the biophysical underpinnings of excitatory signaling, precludes the study of potential delta iGluR pharmacology, and makes biological inferences about the presence of delta iGluR genes in different animals difficult.We therefore sought to establish a molecular and functional signature of the delta iGluR family by investigating beyond the mammalian orthologues, using phyloge netics, electrophysiology, and mutagenesis.This uncovered surpris ingly active ligand-gated delta iGluRs in numerous invertebrates, uncovered pharmacological similarities between delta iGluRs and their AMPA receptor cousins, and traced the inactivity of vertebrate delta iGluRs to a distinct part of the iGluR gating machinery.

Results
GABA-gated Channels throughout the Delta iGluR Family.In questioning the divergence of relatively inactive mammalian delta iGluRs from their active, ligand-gated iGluR cousins, we sought a more definitive view of the delta iGluR family and its phylogenetic and functional relation to other iGluRs.We first generated a maximum likelihood phylogeny of iGluR genes from a broad selection of diverse animals (Fig. 1A and SI Appendix, Fig. S1).Mammalian GluD1 and GluD2 genes are found in a branch we will refer to as the delta family (green in Fig. 1A), whose closest relatives in terms of other previously characterized genes are AMPA/KA receptors (dark pink in Fig. 1A).Delta and AMPA/ KA iGluRs are thus closely related, and together with several uncharacterized paraphyletic relatives make up the AMPA/KA/ delta/phi (AKDF) branch that was proposed by others (23,24).The delta iGluR family comprises genes only from animals of the bilaterian lineage, i.e., xenacoelomorphs, a distinct group of simple marine worms lacking a circulatory system (25); protostomes such as molluscs; and deuterostomes such as vertebrates, hemichordates (e.g., acorn worms), and echinoderms (e.g., starfish, SI Appendix, Fig. S1).According to our maximum likelihood phylogeny, protostome delta iGluRs are the earliest branching within the delta family ( X in Fig. 1A), consistent with a previously published Bayesian phylogeny (23).
Apart from inactive wild-type (WT) and constitutively active, D-serine-inhibited, mutant mammalian delta iGluRs, the func tional signature of the delta iGluR family is unknown.We there fore expressed putative delta iGluRs from diverse bilaterian animals in frog oocytes and characterized their function with two-electrode voltage clamp.Genes included "DioGluD" from the acoel Diopisthoporus longitubus, (a xenacoelomorph); "CraGluD" from the oyster Crassostrea gigas (a protostome); "AcaGluD" from the crown-of-thorns-starfish Acanthaster planci (an echinoderm) and "SacGluD" from the acorn worm Saccoglossus kowalevskii (a hemi chordate; both invertebrate deuterostomes); and "DanGluD2A" from the zebrafish Danio rerio (a vertebrate deuterostome).Similar to RatGluD1 and RatGluD2, WT DanGluD2A receptors showed no response to the four ligands tested, but lurcher-mutant (Lc) DanGluD2A Lc receptors showed constitutive currents that were inactivated by D-serine and glycine (Fig. 1B).However, in contrast to RatGluD2 Lc , which were only sensitive to glycine and D-serine, DanGluD2A Lc receptors and RatGluD1 AC [the Lc-like A654C mutation (26)] were also sensitive to GABA, which caused 58 ± 8% (n = 4) or 86 ± 5% (n = 10) inactivation relative to D-serine at these two receptors (Fig. 1 B and C).This shows that numerous vertebrate delta iGluRs are inactive, but suggests that some of them bind the classical inhibitory neurotransmitter GABA, as indicated by a recent study of RatGluD1 (27).
In stark contrast to WT delta iGluRs of vertebrate deuterostomes, WT delta iGluRs of other bilaterians showed robust ligand-gated currents, and remarkably the most effective agonist was GABA (Fig. 1 D and E).At invertebrate deuterostome (starfish) AcaGluD, the GABA EC 50 of 13 ± 3 μM (n = 5) was much lower than that of glutamate (7.8 ± 3 mM, n = 3).Fellow invertebrate deuterostome (acorn worm) SacGluD showed smaller currents, making potency difficult to measure, although GABA was more potent than other potential ligands (SI Appendix, Fig. S2A).This was more evident in lurcher-mutant SacGluD Lc , which showed relatively small consti tutive current and inward GABA-gated currents of very high potency (EC 50 = 84 ± 14 nM, n = 4, SI Appendix, Fig. S2 B-D).At xenacoelmorph (acoel) DioGluD iGluRs, GABA was in fact the only ligand that elicited currents, with an EC 50 of 180 ± 20 μM (Fig. 1E).Finally, we tried measuring the activity of the earliest branching delta iGluR in our tree, protostome (oyster) CraGluD ( X in Fig. 1A).Unfortunately, we could not establish the function of this receptor, as we detected no ligand-gated currents in oocytes injected with WT or Lc mutant CraGluD mRNA, probably due to low surface expression, as assayed by oocyte immunolabeling (SI Appendix, Fig. S2 E and F).
Inferring the putative functional properties of the ancestral receptor at the base of the delta branch is difficult without func tional data on early-branching protostome delta iGluRs.However, considering the presence of a) GABA sensitivity in both xenacoe lomorph, deuterostome invertebrate, and certain deuterostome vertebrate delta iGluRs, b) the absence of ligand-gated currents in WT vertebrate delta iGluRs, and c) the fact that AMPA/KA iGluR cousins are glutamate-gated channels, we suggest the fol lowing.When the first delta iGluR diverged from its AKDF ances tor, it was an active ligand-gated channel that quickly evolved selectivity for the neurotransmitter GABA.After chordates diverged from other invertebrate deuterostomes, ligand-induced channel gating was lost in delta iGluRs of the chordate or subse quent vertebrate lineage.And based on the presence of both GABA and glycine/D-serine sensitivity in vertebrate GluD1 and GluD2 receptors, we tentatively conclude that ligand selectivity changed from GABA to glycine and D-serine in early vertebrates or other chordates.
Computational Analysis of Ligand Binding.One potential explanation for the evolutionary scenario described above would involve substitutions in the ligand-binding residues of the vertebrate delta iGluR LBD, enabling selectivity for αamino acid D-serine (and glycine) instead of γamino acid GABA.We investigated this by computationally docking GABA and Dserine to a starfish AcaGluD AlphaFold structural model and the RatGluD2 X-ray structure (13).With an upper lobe-lower lobe separation similar to that of D-serine-bound RatGluD2 (ref.13 and SI Appendix, Fig. S3 A and D), our model likely represents an active ligand-bound conformation.As expected, based on experimental evidence for some sensitivity to GABA and D-serine (Fig. 1 C and D), both ligands docked to both receptors in the canonical binding site, but with more favorable energies for GABA than D-serine at AcaGluD and the converse at RatGluD2 (Fig. 2 A and B and SI Appendix, Fig. S3).
Aligning the ligand-binding residues shows only minor differ ences between GABA-selective AcaGluD and D-serine-(and glycine-) selective RatGluD2 (Fig. 2C).Indeed, the major bonding partner of the GABA γamine and D-serine αamine in both receptors is the carboxylate side chain of the conserved lower lobe D763/D742 residue (Fig. 2 A-C).Similarly, although the upper lobe E475/E450 carboxylate residue was recently shown to con tribute to GABA potency in RatGluD1 (27), it is conserved among GABA-selective invertebrate and D-serine-selective verte brate delta iGluRs (Fig. 2 A-C), and thus does not determine ligand selectivity.
We did notice two differences between the receptors, however.In the upper lobe, hydrophobic V548 of AcaGluD cannot form a polar interaction with the αamine of D-serine that is formed by the equivalent but polar T525 of RatGluD2 (Fig. 2 A-C).And in the lower lobe, invertebrate receptors have a polar S713 side chain where vertebrate receptors have a small, nonpolar A686 side chain (Fig. 2 A-C).However, when we tested the ligand selectivity profile of mutant V548T and S713A AcaGluD receptors, we saw no increase in D-serine activity relative to GABA (Fig. 2D), sug gesting that these differences do not determine ligand selectivity.This is reflected in the fact that invertebrate SacGluD, which has the upper lobe threonine residue like RatGluD2, shows high GABA selectivity (Fig. 2C and SI Appendix, Fig. S2D).Taken together, these computational and functional data show that most extant delta iGluRs either show GABA selectivity or retain some GABA sensitivity and that ligand-binding residues alone do not determine ligand selectivity in delta iGluRs.

Pharmacological and Biophysical Properties of Delta iGluRs
Reflect their Close Relationship to AMPA/KA Receptors.
Although phylogenetic relationships suggest that delta iGluRs are closely related to AMPA/KA receptors, previous pharmacological studies have emphasized similarities between delta iGluRs and NDMA receptors, such as glycine and D-serine binding, and channel block by pentamidine (19,28).We therefore sought a more extensive view of delta iGluR function, incorporating channel pore properties, competitive antagonist pharmacology, and modulation by extracellular calcium ions (Ca 2+ ), utilizing the crown-of-thorns starfish AcaGluD receptor because of its large, tractable ligand-gated currents in oocytes.
To establish the channel pore properties of AcaGluD receptors we measured current-voltage (IV) relationships and channel block by pentamidine.The IV relationship at AcaGluD iGluRs is inwardly rectifying, with small outward currents only appearing at potentials more positive than 50 mV, both in the absence and presence of divalent cations (Fig. 3A).This suggests that AcaGluD receptors are blocked by intracellular polyamines and not by extra cellular Mg 2+ ions, which is qualitatively similar to AMPA recep tors (29) and chimeric receptors carrying KA receptor LBDs and RatGluD2 channel domains (30).Running voltage ramps from −80 to 60 mV during GABA-gated currents in regular extracellular solution yielded a reversal potential of −16 ± 2 mV (n = 4), as expected for a nonselective cation-permeable channel in these conditions (31).This indicates that AcaGluD is a mixed cation channel like most iGluRs (3) and thus an excitatory GABA receptor.
Pentamidine is a diarylamidine compound previously shown to block NMDA receptors in the low micromolar range and con stitutively active mutant RatGluD2 Lc channels in the high micro molar range (28,32).We observed that pentamidine (100 µM) blocked 68 ± 5% (n = 5) of the GABA-gated current and 78 ± 2% (n = 7) of the glutamate-gated current through AcaGluD iGluRs (Fig. 3B).As the pentamidine sensitivity of AMPA recep tors has been explored relatively little, we tested this ourselves and observed that pentamidine (100 µM) elicited 37 ± 3% (n = 5) inhibition of glutamate-gated current through RatGluA2 AMPA receptors (SI Appendix, Fig. S4A).Combined with earlier studies, our results show that AMPA receptors and delta iGluRs share moderate pentamidine sensitivity, in contrast to NMDA receptors, which have higher pentamidine sensitivity.

The NTD Does not Determine the Ligand-gated Activity of
Starfish Delta iGluRs.Our data suggest that after diverging from other AKDF iGluRs, early delta iGluRs were active, homotetrameric ligand-gated channels, even in the absence of accessory synaptic proteins.And while this ligand-gated activity is conserved in several extant delta iGluRs, delta iGluRs in the lineage to vertebrates lost ligand-gated channel activity.We questioned the molecular basis for this loss, hoping to establish a molecular blueprint for the mysterious function of mammalian delta iGluRs.
In most animal iGluR subunits, the large clamshell-shaped NTD contributes greatly to tetramer assembly but relatively mod erately to ligand-gated channel activity (3).However, NTDs are indirectly implicated in the ligand-gated (in)activity of mouse/ human delta iGluRs, as the experimental separation of NTDs from each other, or from the LBDs below, impairs channel func tion (17,19).We therefore questioned whether substantial diver gence in the NTDs-19% amino acid sequence identity between AcaGluD and RatGluD2 NTDs cf 38% identity between AcaGluD and RatGluD2 LBD and channel domains-determines the absence of ligand-gated currents in vertebrate delta iGluRs.To answer this, we measured ligand-gated currents in two chimeric AcaGluD constructs in which the NTD was replaced with that of RatGluD2: one retaining the NTD-LBD linker of AcaGluD, "AcaGluD RatNTD "; and the other with the NTD-LBD linker of RatGluD2, "AcaGluD RatNTDlink " (Fig. 4 A and B and SI Appendix, Supporting text).Neither chimeric receptor showed drastic differ ences from WT AcaGluD, with each showing large GABA-gated currents (Fig. 4 C and D).However, we did notice a small increase in the relative efficacy of D-serine at AcaGluD RatNTDlink (20 ± 3%, n = 3) compared to WT and AcaGluD RatNTD (12 ± 3%, n = 7, and 5 ± 1%, n = 3, respectively).While confirming that the NTDs and their link to the LBD make some sort of contribution to delta iGluR function (17), these results show that great divergence in the NTDs is unlikely to have determined the stark loss of function in vertebrate delta iGluRs relative to other deuterostome delta iGluRs.

Vertebrate-like Mutations in the Lower Lobe of the LBD Abolish
the Activity of Starfish Delta iGluRs.These results suggest that the relative inactivity of vertebrate delta iGluRs such as RatGluD2 derives from amino acid sequence divergence in the remainder of the receptor, i.e., the LBD, transmembrane channel domain (TMD), and the intracellular CTD.We therefore aligned amino acid sequences of delta iGluR subunits, and from ~410 positions comprising the LBD and TMD, we identified 41 at which the amino acid residue shares biophysical properties in AcaGluD, SacGluD, and DioGluD and differs in inactive DanGluD2A, RatGluD1, and RatGluD2 (SI Appendix, Fig. S5).As the long CTD is implicated little in channel gating and more in expression and trafficking of iGluRs (3), we excluded it from this analysis.Taking AcaGluD again as our active experimental model, we substituted each of these 41 amino acid residues with the equivalent amino acid from inactive RatGluD2, generating 39 mutant AcaGluD receptors, which we tested for responses to GABA, glutamate, glycine, and D-serine.(Two of the mutants incorporated substitutions of two adjacent residues for efficiency.) Thirty-five of the 41 substitutions had little or no effect on AcaGluD activity, with the respective mutants showing large responses to GABA and smaller responses to the other agonists (Fig. 5 A and B).In contrast, six substitutions substantially altered function, with the respective mutants showing either no currents greater than uninjected oocytes (F640Y, S713A, R721K, P741N, and D825P) or a substantial change in relative agonist efficacies combined with decreased current amplitude (G567S; Fig. 5 A and  B).The loss of currents via F640Y, S713A, and R721K substitu tions derives primarily from altered channel function and not decreased oocyte surface expression, as the latter was similar to WT, whereas for P741N and D825P, surface expression was evi dent but greatly reduced compared to WT (Fig. 5C).An addi tional four mutants, Q772D, K782N, M805Q, and V806R, showed milder but noticeable effects, decreasing maximum GABA-gated current amplitude to less than 1 μA (cf.8.2 ± 1.1 μA at WT, n = 4) and retaining the typical agonist selectivity profile (Fig. 5B).
Questioning how these 10 residues contribute to receptor func tion, we considered their putative location in the receptor tertiary structure.For the LBD, we used the AlphaFold structural model, and for the TMD, we examined the position of homologous res idues in published rat delta and AMPA receptor structures (13,20,21,38).From the 10 crucial AcaGluD delta iGluR residues, F640 is in the second helix of the TMD, abutting the channel pore (SI Appendix, Fig. S6A), whereas the other nine residues are all in the LBD, remarkably with seven specifically in the lower lobe of the LBD clamshell (Fig. 5D).Ligand-induced activation of iGluRs involves the upward/outward movement of the lower lobe toward the upper lobe, closing the clamshell around the ligand and pulling the pore-lining third TMD helix outward to open the channel (38,39).Eight of the nine residues are thus in a crucial part of the receptor, close to the putative ligand binding site in the cleft of the LBD clamshell but probably located and/or oriented away from the putative ligand (Fig. 5D and SI Appendix, Fig. S6B).Indeed, G567, R721, D825, and K782 of the lower lobe and M805 and V806 of the upper lobe are all on the "rear" of the LBD, interfacing with the rear of an adjacent LBD within one of two LBD dimers in the homotetramer (Fig. 5D).P741 is in the outer lip of the lower lobe, and S713 is more central (Fig. 5D), although its hydroxyl moiety is unlikely to interact directly with GABA or other ligands (Fig. 2).

Vertebrate-like Mutations in the Lower Lobe of the LBD Induce an
Inactive Channel State.In AMPA/KA iGluRs, interactions between interfacing LBDs control entry into or recovery from agonistinduced desensitization (40)(41)(42)(43).Given the relation of delta and AMPA/KA iGluRs, and the position of loss-of-function AcaGluD substitutions S713A, R721K, P741N, and D825P in or near the interface of LBDs, we hypothesized that the loss-of-function may derive from altered LBD dynamics and increased desensitization compared to WT AcaGluD.We therefore tested responses of mutant AcaGluD iGluRs after treatment with concanavalin A, a plant lectin that prevents desensitization and thus uncovers otherwise small responses in some iGluRs (44,45).Concanavalin A treatment caused a remarkable (>>100-fold) increase in F640Y, S713A, and R721K AcaGluD iGluR activity, with GABA-gated currents now resembling those through WT channels (Fig. 5E).
The same restoration of function was not observed with P741N and D825P mutants, although a relatively small fourfold increase in current amplitude was observed in D825P channels, similar to WT channels (Fig. 5E).Thus, active delta iGluRs are rendered inactive by three vertebrate delta iGluR-like mutations that bias the receptor toward desensitization or another inactive state, and two additional vertebrate delta iGluR-like mutations primarily reduce surface expression of active delta iGluRs.
Considering that such desensitization or inactivation may underlie the absence of currents in WT vertebrate delta iGluRs, we tested the effects of D-serine and glycine on RatGluD1 and RatGluD2 after concanavalin A treatment, but we observed no difference to inactive, untreated receptors (SI Appendix, Fig. S4E).As RatGluD1 and RatGluD2 contain only three predicted N-linked glycosylation sites per subunit compared to eight in AcaGluD (seven in the S713A mutant), however, it could be that concanavalin A is incapable of modulating the vertebrate receptors to rescue them from such a state, precluding conclusions along these lines.

Starfish-like Mutations Partly Reawaken Inactive Rat Delta
iGluRs.If the 10 mutations discussed above drove the loss of function in vertebrate delta iGluRs, one might expect that the latter could be "reawakened" via active invertebrate delta iGluRlike mutations here.We therefore engineered mutant RatGluD2 iGluRs to contain at these positions the equivalent residues from starfish delta iGluRs.These mutants were RatGluD2 5x , containing Y613F, A686S, K694R, N720P, and P806D substitutions (pink in Fig. 6A), and RatGluD2 9x , additionally containing N763K, D753Q, Q786M, and R787V (cyan in Fig. 6A).We also generated A654T Lc-mutant versions, RatGluD2 Lc-5x and RatGluD2 Lc-9x , hypothesizing that constitutively active Lc-versions could offer tangible insight on ligand sensitivity in case the former mutants remained inactive.
RatGluD2 5x and RatGluD2 9x mutants showed no responses to glycine, D-serine, GABA, or glutamate (Fig. 6B), indicating that these nine mutations alone are not capable of reawakening ligand-gated currents in vertebrate delta iGluRs.On the artificial Lc-mutant background, however, tangible effects of the starfish GluD-like mutations were observed.Whereas RatGluD2 Lc-5x behaved much like the typical Lc-mutant RatGluD2 Lc , RatGluD2 Lc-9x  receptors were inactive at rest and conducted inward currents in response to glycine binding (Fig. 6 C and D).These glycine-gated currents were inhibited by the pore blocker pentamidine (Fig. 6D).Finally, we generated RatGluD2 Lc-10x , additionally containing S644G (light green in Fig. 6A), the converse of the G567S substi tution that altered ligand selectivity in starfish AcaGluD.Ligand selectivity was slightly altered by the addition of the S644G muta tion, as RatGluD2 10x receptors showed inward currents in response to both glycine and D-serine (Fig. 6 E and F).Thus, the reintro duction of key residues from active delta iGluRs into inactive ver tebrate delta iGluRs is not enough to reawaken the latter, but on an artificial Lc-mutant background, it awakens the gating machinery, leading to RatGluD2 channels that are inactive at rest and activated by glycine and D-serine binding.

Discussion
The relative inactivity of mammalian delta iGluRs compared to other iGluRs has intrigued biophysicists, impaired pharmacolog ical developments, and confounded our inferences from genome sequencing data.We therefore investigated the evolution and bio physical properties of the delta iGluR family, uncovering GABAgated delta iGluRs in numerous invertebrates, dissecting their relationships with other iGluRs, and identifying a putative molec ular basis for the loss of activity in vertebrate delta iGluRs.
Divergence of Delta Receptors from Other iGluRs.Our phyloge netic analysis finds delta iGluR genes only in bilaterians, consistent with a previous, detailed phylogenetic study (23).But as our study also included xenacoelomorphs, a group of bilaterians that likely branched before the divergence of various nephrozoans (i.e., protostomes and deuterostomes) (46), we now suggest that delta iGluRs emerged from the duplication of an AKDF gene in early bilaterian animals, shortly after they split from cnidarians (e.g., sea anemones and jellyfish).
Delta iGluRs form a family within the AKDF branch of the iGluR superfamily, a branch that includes the AMPA/KA family of fast-activating and deactivating glutamate-gated channels and numerous other uncharacterized genes.We cannot infer the func tional nature of the very first delta iGluR, because we were unable to measure the activity of early branching delta iGluRs found in protostomes.Nor do we know the functional nature of the gene from which the first delta iGluR emerged, as much of the sister group to delta iGluRs, the combined [(AMPA/KA)+(Phi)+(uncharacterized AKDF)] branch, is uncharacterized, and branch support toward the base of the AKDF branch is relatively weak (SI Appendix, Fig. S1 and ref. 23).However, the presence of GABA sensitivity in xenacoe lomorph, invertebrate deuterostome, and even certain vertebrate delta iGluRs, suggests that early delta iGluRs were GABA-sensitive.And the ligand-gated channel activity of numerous delta iGluRs, together with that of AMPA/KA receptors, suggests that mutations in delta iGluR genes either chordates or vertebrates led to the inactivity of extant vertebrate delta iGluRs.
Biophysical Mechanisms.We identified several mutations that occurred in delta iGluRs of the vertebrate lineage that cause reduced cell surface expression and/or induce inactive, potentially desensitized states.Notably, several of these residues are in the mid-to-rear of the LBD (Fig. 5D), a part of the receptor implicated in AMPA/ KA iGluR desensitization (40)(41)(42)(43), the phenomenon of upperchannel collapse when ligand-bound LBDs separate from each other, loosening the tension between LBDs and channel-forming helices (3).Another vertebrate delta iGluR mutation we identified, G567S in the "hinge" between upper and lower lobes at the rear of the LBD, reduced the GABA selectivity of the invertebrate delta iGluR.How this mutation alters ligand selectivity is not addressed by our experiments, and our dockings to static LBDs do not address, e.g., flexibility of the LBD as determined by residues like G567 in the hinge region.However, previous experimental and computational work on RatGluD2 suggests that D-serine affinity is substantially altered by mutations in the hinge that alter its flexibility (47).That ligand-selectivity may be allosterically controlled is also reflected in the altered glycine/D-serine preference shown by a chimera with an NTD-LBD linker substitution.
Attempting to "reverse" the loss-of-function mutations in ver tebrate delta iGluRs and "reawaken" RatGluD2 iGluRs, we found that even when combined, nine reverse substitutions did not reawaken ligand-gated activity.This suggests that additional muta tions have accumulated in vertebrate delta iGluRs, and receptor reawakening is now contingent upon these additional residues.Foreseeably, some combination of up to 41 of the potential reverse mutations would achieve this, despite the fact that many of these forward mutations were not noticeably detrimental to starfish delta iGluR function.Reverse mutations in RatGluD2 iGluRs on the Lc-mutant background, however, indeed led to ligand-gated channels, either glycine-gated or glycine-and D-serine-gated, depending on the combination of mutations.This confirms pre vious work showing that delta iGluR LBDs are capable of sub stantial ligand-induced conformational change (13,48) and delta iGluR channels are capable of gating (12,30), but moreover, it shows that only a few amino acid residues prevent the coupling of these two processes in vertebrate delta iGluRs.
Our results also highlight the fact that the effects of upper-M3 mutations are difficult to predict, as they vary depending on amino acids in the LBD.Four of the above reverse mutations in the LBD caused RatGluD2 Lc-mutant channels to be closed at rest.This iden tifies candidate LBD determinants of the differing effects of M3 mutations in different iGluR families (11,26,49,50).Indeed, despite high conservation of upper-M3 amino acid sequence in various delta iGluRs (SI Appendix, Fig. S5), we show that the Lc-mutation has very different effects in RatGluD2 and SacGluD, as it converts the latter from a generally inactive channel to one that is largely inactive at rest and potently and efficaciously activated by GABA.
Our experiments suggest that mutations in the NTD played relatively little role in the loss of ligand-gated currents in vertebrate delta iGluRs.However, changes in both the NTDs and LBDs may affect delta iGluR activity via a remarkably similar mecha nism.Borrowing from broadly accepted mechanisms of AMPA receptor activation and desensitization (4), delta iGluR activity may depend on tight interfaces at the rear of adjacent LBDs so that clamshell closure at the front of the LBD pulls the lower lobe upward, pulling the channel open via the LBD-channel domain linker.Thus, the inter-LBD interface can be stabilized and channel activity enhanced by either introduced disulfides linking NTDs or LBDs (19,37); pre-and intersynaptic proteins that lock delta iGluR NTDs together to keep the receptor taut (17,19); or amino acid identity in and around the rear of the LBD (present study).
Pharmacological Avenues.No selective pharmacological modulators of delta iGluRs are known.Whether such modulators would be directly relevant to pharmacotherapy is unclear, as the links between low GluD1 expression and schizophrenia and GluD2 overactivity and cerebellar ataxia may pertain to early development (51)(52)(53) and thus be inaccessible to pharmacotherapy.But delta iGluR modulators would drastically improve physiological experiments aiming to dissect the function of delta iGluRs in neural circuitry in vivo/ex vivo.
Small currents through reawakened vertebrate delta iGluRs and certainly large currents through active invertebrate delta iGluRs offer a tangible experimental system for testing the effects of poten tial drug molecules on delta iGluR function.Such use of inverte brate receptors would rely on their pharmacological profile matching that of vertebrate GluD1 and GluD2 iGluRs, but our study shows that sensitivity to certain competitive antagonists and pore blockers seems similar in vertebrate and invertebrate delta iGluRs.Our study also suggests that care must be taken when dissecting synaptic iGluR composition with classical AMPA/KA receptor modulators, as some of these also affect delta iGluRs.
Excitatory GABA Receptors.The most surprising finding in our study was that many delta iGluRs are activated by the transmitter GABA.GABA mediates most of the rapid inhibitory signals in mammalian central synapses via its activation of type-A GABA receptors (GABA A receptors) of the pentameric (or "Cys-loop") ligand-gated ion channel superfamily (54).The presence of GABAgated receptors throughout the delta iGluR branch suggests that delta iGluRs became GABA-sensitive early after their emergence, resulting in excitatory GABA receptors that have been inherited by numerous bilaterian animals, including acoels and other xenacoelmorphs, and invertebrate deuterostomes such as starfish and acorn worms.Future studies are needed to determine whether GABAergic neurons synapse onto GABA-gated delta receptors and whether such excitatory GABA receptors also occur in protostomes and early-branching chordates.
There are now numerous examples of ligand-gated ion channels that overturn the conventional view of glutamate as excitatory and GABA and glycine as inhibitory transmitters.Within the iGluR superfamily, there are excitatory glycine-gated NMDA receptors in mammals (55) and numerous glycine-gated and presumably excitatory iGluRs of the "epsilon" family in invertebrates (23,56).And in the pentameric ligand-gated ion channel superfamily, there are excitatory GABA receptors and inhibitory glutamate receptors (57,58).The fact that ligand sensitivity and ion permeability are so readily evolvable in different ligand-gated ion channel super families and in different animals means that burgeoning transcrip tomic studies must be cautious in assigning neuronal identity to different cells based on the presence of, e.g., iGluR genes.While the prediction of function from sequence will always be susceptible to unidentified switches in function, we foresee that systematic studies of the evolution of receptor families, such as ours, will improve future assessments of neuronal function based on the presence of different ligand-gated ion channel genes.
Sequences were aligned with MAFFT v7.450 (60) in Geneious Prime (Dotmatics).Sequences that were >95% identical to another and sequences that were excessively long or short were removed.The final alignment contained 246 sequences with 4692 columns (including gaps).From this alignment, we generated a maximum-likelihood phylogenetic tree using IQ-Tree (61) with a Q.pfam+G4 substitution model (62) (log-likelihood of model -410679) and both SH-aLRT (63) and ultrafast bootstrap (64) branch support.
Molecular Biology and Heterologous Expression.Delta iGluR sequences were commercially synthesized and subcloned (Genscript Biotech Netherlands) into a custom vector so that finally each coding sequence had silent mutations to remove internal restriction sites, a C-terminal glycine-serine linker and cMyc tag (except for Saccoglossus kowalevskii GluD) and was preceded by a loosely Kozak consensus sequence, flanked by Xenopus laevis β-globin 5′ and 3′ untranslated sequences, and followed by a poly(A) tail.DanGluD2 and CraGluD sequences were codon-optimized for Xenopus leavis in iCodon as we learnt of this software

Fig. 1 .
Fig. 1.GABA-gated channels in the delta iGluR family.(A) Maximum likelihood phylogeny of animal iGluR genes.Selected iGluR branches are colored and labeled.*, genes characterized in panels B-E.X, O , Crassostrea gigas GluD and Saccoglossus kowalevskii GluD, see SI Appendix, Fig. S2.SH-aLRT support for selected branches indicated.Detailed branch support and all gene names in expanded phylogeny, SI Appendix, Fig. S1.(B-E) Example two-electrode voltage clamp recordings of oocytes expressing indicated delta iGluR genes in response to different ligands (Left) and mean ± SEM (n = 4 to 6) normalized concentration-dependent responses (Right).Scale bars: x, 10 s; y, as indicated.(B and C) Wild-type channels were inactive; lurcher-mutant ( Lc ) DanGlu2A and RatGluD2 and A654C-mutant ( AC ) RatGluD1 channels were constitutively active.The dashed line indicates zero current; mean responses reflect ligand-induced inhibition of constitutive current normalized to maximum inhibition of constitutive current.(D and E) Wild-type channels were active; mean responses reflect ligand-induced current amplitude normalized to maximum ligand-induced current amplitude.Dan, Danio rerio.Rat, Rattus norvegicus.Aca, Acanthaster planci.Dio, Diopisthoporus longitibus.(F) % amino acid sequence identity of selected delta iGluRs.

Fig. 3 .
Fig. 3. Pore properties and pharmacology of AcaGluD receptors.(A) Normalized current (I/I −80 mV ) -voltage relationship of GABA-gated currents through AcaGluD-expressing oocytes.Data points mean ± SEM (n = 4), joined by straight lines.(B and C) Example recordings (scale bars x, 5 s; y, 1 μA) and summary data (columns, mean; circles, individual data points) of modulation of GABA-and glutamate-gated current by indicated drugs at AcaGluD-expressing oocytes.(D) Example responses to increasing concentrations of GABA in the absence or presence of extracellular Ca 2+ (Left) and normalized (to maximum current in respective condition, "I/I max ") responses to different agonists (Right, mean ± SEM, n = 5 to 7).(E) Fold enhancement ("I Ca2+ /I zeroCa2+ ") of GABA-gated current by 1.8 mM Ca 2+ at oocytes expressing WT or indicated mutant AcaGluD.Inset: example recording from oocyte expressing D558A mutant AcaGluD, scale bars in D and E: x, 10 s; y, 0.15 μA.