AMPA receptor auxiliary subunits emerged during early vertebrate royalsocietypublishing.org/journal/rsob evolution by neo/subfunctionalization of unrelated proteins David Ramos-Vicente1,2 and Àlex Bayés1,2 Research 1 Molecular Physiology of the Synapse Laboratory, Biomedical Research Institute Sant Pau, Barcelona, Spain 2 Universitat Autònoma de Barcelona, Barcelona, Spain Cite this article: Ramos-Vicente D, Bayés À. 2020 AMPA receptor auxiliary subunits DR, 0000-0002-2730-0850; ÀB, 0000-0002-5265-6306 emerged during early vertebrate evolution by In mammalian synapses, the function of ionotropic glutamate receptors is neo/subfunctionalization of unrelated proteins. critically modulated by auxiliary subunits. Most of these specifically regulate Open Biol. 10: 200234. the synaptic localization and electrophysiological properties of AMPA-type http://dx.doi.org/10.1098/rsob.200234 glutamate receptors (AMPARs). Here, we comprehensively investigated the animal evolution of the protein families that contain AMPAR auxiliary sub- units (ARASs). We observed that, on average, vertebrates have four times more ARASs than other animal species. We also demonstrated that ARASs Received: 29 July 2020 belong to four unrelated protein families: CACNG-GSG1, cornichon, shisa Accepted: 3 October 2020 and Dispanin C. Our study demonstrates that, despite the ancient origin of these four protein families, the majority of ARASs emerged during vertebrate evolution by independent but convergent processes of neo/ subfunctionalization that resulted in the multiple ARASs found in present vertebrate genomes. Importantly, although AMPARs appeared and diversi- Subject Area: fied in the ancestor of bilateral animals, the ARAS expansion did not occur neuroscience/molecular biology/bioinformatics until much later, in early vertebrate evolution. We propose that the surge in ARASs and consequent increase in AMPAR functionalities, contributed to Keywords: the increased complexity of vertebrate brains and cognitive functions. AMPA receptor auxiliary subunit, TARP, cornichon, shisa, Dispanin C, phylogeny 1. Introduction Ionotropic glutamate receptors are key to the physiology of the nervous system, Author for correspondence: as they mediate fast excitatory neurotransmission [1]. Previously, we reported Àlex Bayés that the evolution of the proteins that form these tetrameric receptors has e-mail: abayesp@santpau.cat been much more sophisticated than previously acknowledged [2]. We found that beyond the well-known AMPA, Kainate, NMDA and Delta classes, there are four other classes exclusive to invertebrate species. Because of this, species with simple nervous systems, such as the sea anemone N. vectensis, have a simi- lar number of ionotropic glutamate receptor subunits to animals with complex brains, such as mammals [2]. We argued that the high diversity of neuronal types [3] or the wide array of glutamate receptor functionalities [4] found in vertebrates is unlikely to be the result of an increased repertoire of genes coding for them. Notably, the subcellular traffic and function of ionotropic glutamate receptors is controlled by their auxiliary subunits [5]. These proteins add a new layer of complexity to glutamatergic transmission and might have contributed to an expanded functionality of these receptors in animals with complex brains. In mammals, sixteen proteins have been identified as ionotro- pic glutamate receptor auxiliary subunits. Of these, fourteen modulate AMPA- type glutamate receptors (AMPARs) [6], and they are referred to as AMPA receptor auxiliary subunits (ARASs). Only one auxiliary subunit has been Electronic supplementary material is available reported for NMDA receptors, Neto1 [7], which also regulates Kainate recep- online at https://doi.org/10.6084/m9.figshare. tors, as does Neto2 [8,9]. Currently known mammalian ARASs belong to five c.5179219. © 2020 The Authors. Published by the Royal Society under the terms of the Creative Commons Attribution License http://creativecommons.org/licenses/by/4.0/, which permits unrestricted use, provided the original author and source are credited. protein families: CACNG and GSG1 (both within the CACNG and GSG1 proteins belong, present homologues in the 2 superfamily of claudins), cornichon, shisa and Dispanin C. basal metazoan phylum of porifera [54,55], placing its origin royalsocietypublishing.org/journal/rsob Genes belonging to the same family are defined as paralo- prior to the divergence of these organisms. While CACNG gues; these originate from gene or genome duplication events homologues have been reported in different bilaterian species, [10]. In vertebrates, many paralogues resulted from the two including vertebrates and the fruit fly or C. elegans [17,19], it is rounds of whole-genome duplication (2R) occurred at the still unknown when GSG1s appeared during evolution. In base of this lineage, approximately 400 Ma [11]. After addition, some shisa homologues have been described in por- duplication, the two new genes can experience processes of ifera and Dispanin C homologues have been identified in neofunctionalization or subfunctionalization. In the first scen- brown algae [56,57]. Here, we present a comprehensive study ario, one of the new genes retains all functions performed by of the animal evolution of all protein families that include the ancestral gene, while the second acquires new ones AMPAR auxiliary subunits. Our work shows that cornichons, [12,13]. Alternatively, genes subfunctionalize when the mul- present in the ancestor of all metazoans, would be the most tiple functions carried out by the ancestral gene are realized ancient of all ARAS. TARPs, which appeared together with separately by descendent paralogues [13,14]. Establishing if AMPARs in the ancestor of bilaterians, also function as Open Biol. 10: 200234 paralogues have undergone a process of neo or subfunctionali- ARASs in invertebrate organisms. Nevertheless, most zation requires a precise understanding of all functions carried ARASs, including shisas, GSG1L and SynDIG1, would have out by the ancestral gene and the descending paralogues. When been recruited to modulate AMPAR function early in ver- this knowledge is not available, it is not possible to determine tebrate evolution, suggesting that during this period there which one of these two processes took place [15]; in these was an evolutionary pressure that favoured an expansion of occasions, the term neo/subfunctionalization can be used [13]. the functionality of AMPARs. Known invertebrate ARAS belong to the CACNG or cornichon families. These have been described in C. elegans [16,17] and the fruit fly [18,19]. The first ARAS to be reported was CACNG2, also called stargazin, as it is mutated in 2. Materials and methods stargazer mice [20]. Although this protein is phylogenetically related to CACNG1, an auxiliary subunit of voltage- 2.1. Identification of genes coding for AMPA receptor dependent calcium channels (VDCC), it specifically interacts auxiliary subunits in metazoan genomes with AMPAR [21,22]. Subsequently, six other CACNGs have Phylogenetic analysis was performed as described previously been discovered in mammals (CACNG3 to CACNG8 [23,24]). [2]. We searched sequences in 31 species belonging to different Apart from CACNG6, which also acts as a VDCC auxiliary metazoan phyla: Porifera, Ctenophora, Placozoa, Cnidaria, subunit [25,26], all other CACNGs function as AMPA recep- Lophotrochozoa, Ecdysozoa, Hemichordata, Chordata and tor auxiliary subunits [27–29]; these are usually referred to as Vertebrata. The same species were used to construct all phylo- TARPs (transmembrane AMPA receptor regulatory proteins) genies. We used slowly evolving species whenever possible. [27]. Phylogenetic analysis of mammalian TARPs identifies All sequences were retrieved from public databases. that they are more related to each other than to auxiliary sub- A. digitifera and P. flava sequences were obtained from the units of VDCCs [25,27], being divided in two types [27,30,31] Marine Genomics Unit [58,59]. A complete list of species that differentially modulate AMPAR [28,29]. The most included in the analysis and the corresponding database recently identified ARAS is GSG1L [32,33]. In vertebrate gen- where sequence search was done can be found in electronic omes, GSG1L has two paralogues: GSG1, which interacts supplementary material, table S1. with the polymerase TPAP, and GSG1L2 [32,34], of unknown Sequences from protein families of interest were identified biological function. These proteins also belong to the super- by homology-based searches with reciprocal identifications. family of claudins, like CACNGs. Despite its structural Mouse proteins were used as search queries, when different iso- similarity with TARPs, GSG1L modulates AMPAR in a differ- forms were present we used the longest amino acid sequence. ent way, downregulating its traffic to the plasma membrane We searched for homologues using the BLASTP and TBLASTN and accelerating deactivation and desensitization [35,36]. The tools [60] with default parameters. Subject sequences with an remaining ARASs have been identified in three other protein E-value below 0.05 were selected as candidate homologues. families: cornichon, shisa and Dispanin C. In all cases, these These sequences were re-blasted against mammal proteins in families contain ARAS and proteins with different biological the NCBI database of ‘non-redundant protein sequences’ functions. Among cornichons, CNIH2 and CNIH3 modulate using the BLASTP and BLASTX tools, respectively. Identified AMPAR function, but CNIH1 and CNIH4 regulate the traffic homologues which have a protein length that is less than 50% of TGFα and GPCRs, respectively [37–41]. Four members of of the query protein were discarded. the shisa family, Shisa6 to Shisa9 (originally referred to as CKAMPs [42–46]), interact with AMPAR, yet Shisa2 and Shisa3 are involved in the traffic of FGF [47] and WNT [48] receptors, respectively, and Shisa5 participates in the p53/ 2.2. Phylogenetic analyses TP53-dependent apoptosis pathway [49]. Finally, SynDIG1, The CACNG-GSG1 tree was constructed using a total of from the Dispanin C family, has also been identified as an 114 sequences, from which fifteen are used as the outgroup ARAS [50,51], this protein has two paralogues in vertebrates, (electronic supplementary material, file S1). The cornichon SynDIG1L and TMEM91, with poorly understood functions. tree includes 70 sequences, of which five are cornichons from The evolutionary origin of ARASs is well established for Arabidopsis thaliana used as the outgroup (electronic supplemen- cornichons, which have been identified in the ancestor of tary material, file S2). The shisa tree includes 98 sequences, eight eukaryotes and are present in a large range of species, includ- of which are used as the outgroup (electronic supplementary ing plants and yeasts [52,53]. The claudin superfamily, to which material, file S3). Finally, the Dispanin C tree includes 41 sequences, four of which were used as the outgroup (electronic all bilaterians investigated. The other families were lost in cer- 3 supplementary material, file S4). tain lineages. For instance: (i) the GSG1 subfamily is only royalsocietypublishing.org/journal/rsob Protein sequences were aligned with the MUSCLE algor- present in deuterostomes; (ii) molluscs would have lost the ithm [61], included in the software package MEGAX [62] with protostome-specific subfamily; and (iii) the VDCC-subunit default parameters. ProtTest v3.4.2 was used to establish the subfamily was independently lost by cephalochordates, best evolutionary model [63]. Trees were constructed using echinodermates and ecdysozoans. MrBayes v3.2.7 [64] for Bayesian inference and IQ-TREE [65] Among the phyla studied, vertebrates present the highest for maximum-likelihood analysis. For Bayesian inference number of CACNG-GSG1s (11 proteins; table 1), of which 7 phylogenies were run for 10 000 000 generations. Markov are ARAS. Although Capitella teleta, an annelid, has eight mem- chain Monte Carlo (MCMC) was used to approximate the bers in this family, of which six would be TARPs according to posterior probability of the Bayesian trees. Bayesian analyses the phylogenies, most invertebrates investigated generally included two independent MCMC runs, each using four paral- have fewer CACNG-GSG1s. All mammalian proteins within lel chains composed of three heated and one cold chain. the branch of TARPs are known AMPAR regulatory proteins Twenty-five per cent of initial trees were discarded as burn-in. [27,28,71]. Due to their position in the trees we propose that Open Biol. 10: 200234 Convergence was assessed when the potential scale reduction invertebrate orthologues to vertebrate TARPs and their bilater- factor (PSRF) value was between 1.002 and 1.000. In ian ancestor would interact and regulate AMPARs. However, maximum-likelihood analysis the starting tree was estimated with the exception of cephalochordate sequences, invertebrate using a neighbour-joining method and branch support TARPs are more divergent than their vertebrate counterparts, was obtained after 1000 iterations of ultrafast bootstrapping as indicated by their longer branches. It is thus plausible that [66]. Gene/protein names were given based on their position the ability to interact with AMPARs has been altered in this in the tree. Phylogenetic trees were rendered using FigTree species. Functionally, vertebrate TARPs are classified into (http://tree.bio.ed.ac.uk/software/figtree/). Phylogenetic Type I and Type II [27,28,71]. This classification is mirrored calculations were performed at the CIPRES science gateway [67]. by the phylogenies, as vertebrate sequences form two mono- phyletic groups, one for each type. While invertebrate proteins cannot be unambiguously related to Type I or Type 2.3. Protein nomenclature II TARPs, as the statistics metrics generated by the Bayesian Proteins from non-vertebrate species were systematically (figure 1) and maximum-likelihood (electronic supplementary named following this nomenclature: (i) the name of the subfam- material, figure S1) phylogenies are not high enough, the trees ily which they belong, or family if the sequence is not assigned suggest that the majority of them belong to Type II and that to a subfamily; (ii) a Greek letter to identify non-vertebrate Type I would have been lost in non-vertebrates. Among paralogues, if any; and (iii) a three-letter species code. GSG1s, vertebrate GSG1Ls are the only ones known to act as ARASs [32,33]. Yet, our tree indicates that all GSG1 paralogues arouse from recent duplications at the base of the vertebrate 2.4. Prediction of PDZ motifs classes phylum, suggesting that the emergence of ARAS in this The software Eukaryotic Linear Motif (ELM) [68] was used family occurred by neo/subfunctionalization early in to identify if the C-terminus of proteins presented a PDZ vertebrate evolution. binding motif and, if so, to what class it belonged. 3.2. Sequence analysis reveals common motifs in 3. Results CACNG-GSG1s 3.1. TARPs are ancient AMPAR auxiliary subunits We next constructed multiple sequence alignments to investi- gate primary sequence features among CACNG-GSG1s widespread in bilaterians (figure 2a; electronic supplementary material, figure S3). We Among protein families including AMPA receptor auxiliary found a highly conserved motif of 7 residues in all subunits, CACNGs and GSG1s belong to the superfamily of CACNG-SGS1s (consensus sequence: Y(174)SYGWSF, resi- claudins [32,69,70]. We thus incorporated GSG1s into the phy- due numbering corresponds to human CACNG2). The last logeny of CACNGs to investigate if they are part of the same five residues of this motif present the highest conservation, family within the claudin superfamily. We found that GSG1 S179 being the most conserved position. We also investigated proteins confidently fell within the ingroup (figure 1; electronic which proteins from this family would present PDZ binding supplementary material, figure S1). Thus CACNGs and GSG1s motifs, as mammalian TARPs [5] present Class 1 PDZ motifs. are more evolutionarily related to each other than they are to These short C-terminal sequences contribute to the anchoring the rest of the claudin superfamily, belonging to the same of TARPs at the postsynaptic membrane [72,73]. All TARP family, which we have named CACNG-GSG1. We only subfamily sequences except two, TARP_Pfl and TARPα_Pca, found CACNG-GSG1 sequences in bilaterian species are predicted to have a PDZ binding motif; most of these (table 1), indicating that this family appeared in a common being Class 1 motifs. TARP_Spu, TARP_Apl, TARPβ_Lgi ancestor of bilaterians. Phylogenetic analysis identified four and TRAPβ_Aca are predicted to have changed this into subfamilies among CACNG-GSG1s: TARPs, GSG1s, VDCCs Class 2 motifs. Similarly, most GSG1 subfamily proteins and a protostome-specific subfamily. As these are widely rep- also have a PDZ binding motif, except GSG1s from hemichor- resented among bilaterian phyla, we propose that the ancestral dates (figure 2b). These motifs are present in GSG1 proteins CACNG-GSG1 experienced three duplication events before the that act as ARAS but also in those with other functions. Inter- split of bilaterians (electronic supplementary material, figure estingly, this subfamily presents three types of PDZ motifs. S2A). Nevertheless, only the TARP subfamily is conserved in Vertebrate GSG1s are predicted to have Class 3 motifs, CACNG2_Hsa 62 CACNG2_Mmu 100 4 CACNG2_Gga Chordata vertebrata 100 CACNG2_Xtr 100 Cephalochordata royalsocietypublishing.org/journal/rsob CACNG2a_Dre 100 CACNG2b_Dre Hemichordata CACNG3_Hsa 98 100 Echinodermata 99CACNG3_Mmu 99CACNG3_Gga Lophotrochozoa 98 CACNG3b_Dre Ecdysozoa CACNG3a_Dre Type I TARPs 100 CACNG4_Hsa 100 100CACNG4_Mmu CACNG4_Gga 100 CACNG4a_Dre 100 CACNG4b_Dre 97 CACNG8_Hsa 100 CACNG8_Mmu 100 CACNG8a_Dre 100 CACNG8b_Dre CACNG7_Hsa 88 CACNG7_Mmu 100 CACNG7_Gga TARPs subfamily 100 CACNG7a_Dre 100 CACNG7b_Dre 70 CACNG5_Hsa Type II TARPs 69 CACNG5_Mmu CACNG5_Gga 100 CACNG5a_Dre Open Biol. 10: 200234 CACNG5b_Dre TARPa_Lgi 87 TARPa_Cgi 70 93 TARPb_Lgi 99 TARPa_Aca 91 TARPb_Aca TARPa_Cte 51 94 100 TARP b_Cte TARP g_Cte 75 TARP b_Cgi 81 TARPd_Cte 99 58 TARPe_Cte TARPa_Pca 100 TARPb_Pca TARPa_Bfl 97 100 75 TARPb_Bfl 100 TARPa_Bbe 87 TARP_Bla TARP_Spu 100 TARP_Apl 81 TARPg_Lgi 97 94 TARPg_Aca 98 TARPg_Cgi 58 TARPz_Cte 81 TARP_Ame 83 100 TARP_Sma TARPa_Sko TARP_Pfl TARPg_Bfl 100 100 TARPd_Bfl 79 TARPb_Bbe TARPb_Sko GSG1l_Mmu 66 GSG1L_Gga GSG1L_Xtr AMPAR 62 86 GSG1L_Hsa interactors GSG1L_Dre 54 GSG1_Hsa 100 100 GSG1_Mmu TPAP location 100 GSG1_Gga GSG1_Xtr to ER 100 100 GSG1L2_Hsa GSG1l2_Mmu GSG1 GSG1L2a_Dre 100 60 100 GSG1L2b_Dre GSG1L2_Gga GSG1L2_Xtr subfamily GSG1a_Bfl 79 GSG1a_Bla 100 100 GSG1_Bbe 91 GSG1b_Bfl 100 GSG1b_Bla 77 GSG1_Sko 100 60 GSG1_Pfl GSG1_Apl CACNG1_Hsa 65 100 100 CACNG1_Mmu CACNG1_Gga 100 100 68 CACNG1b_Dre CACNG1a_Dre CACNG6 VDCC-subunit 100 100 100 100 CACNG6_Mmu CACNG6a_Dre CACNG6b_Dre subfamily VDCC_Pfl 96 VDCC_Cte 100 100 ProtCACNG_Ame VDCC_Aca ProtCACNG_Cte ProtCACNG 99 subfamily ProtCACNG_Pca PMP22_Hsa 100 97 PMP22b_Dre PMP22a_Dre 100 EMP3_Hsa 100 98 EMP3b_Dre EMP2_Hsa 99 94 EMP2_Dre 100 EMP1_Dre EMP1_Hsa LIM2.4 77 61 100 LIM2.5 LIM2_Hsa 98 LIM2.1_Dre 100 98 LIM2.3_Dre LIM2.2_Dre 2.0 Figure 1. Bayesian inference phylogenetic tree of the CACNG-GSG1 family. The subfamilies in which the CACNG-GSG1 protein family is divided are highlighted by dashed line boxes. The name of each subfamily is presented at the right of the corresponding box. Posterior probabilities are shown at tree nodes and protein names at the end of branches. Tree branches are coloured based on phylum, as indicated in the legend. The closest relatives to vertebrate CACNG in the claudin superfamily were used as outgroup. The Type I and Type II TARPs are indicated. If known, the function of vertebrate GSG1 subfamily sequences was also included. Scale bar denotes number of amino acid substitutions per site. The amino acid substitution model used was Vt + G + F, the analysis ran for 10 000 000 generations, the final standard deviation was 0.012330 and the final potential scale reduction factor (PSRF) was 1.000. Table 1. Number of proteins found in the four families known to include AMPA receptor auxiliary subunits (ARASs). The number of ARASs is given in brackets. 5 Proteins from Homo sapiens are experimentally confirmed as ARAS. Invertebrate proteins are defined as candidate ARASs based exclusively in the phylogenies. royalsocietypublishing.org/journal/rsob phylum species CACNG-GSG1 CNIH Shisa Dispanin C Metazoa Bilateria Vertebrata H. sapiens 11 (7) 4 (2) 10 (4) 3 (1) Chordata C. intestinalis 0 2 (1) 0 0 B. lanceolatum 3 (1) 3 (1) 6 (0) 1 (1) B. belcheri 3 (2) 2 (1) 6 (0) 7 (1) B. floridae 6 (4) 3 (2) 8 (0) 3 (0) Hemichordata S. kowalevskii 3 (2) 2 (1) 16 (0) 1 (1) P. flava 3 (1) 1 (0) 4 (0) 0 Open Biol. 10: 200234 Echinodermata A. planci 2 (1) 1 (1) 2 (0) 0 S. purpuratus 1 (1) 1 (1) 1 (0) 2 (0) Mollusca L. gigantea 3 (3) 2 (1) 0 2 (0) C. gigas 3 (3) 3 (1) 0 1 (1) A. californica 4 (3) 3 (2) 1 (0) 0 Annellida C. teleta 8 (6) 2 (1) 2 (0) 1 (0) Arthropoda A. mellifera 2 (1) 1 (1) 0 0 S. maritima 1 (1) 1 (0) 1 (0) 0 Priapullida P. caudatus 3 (2) 2 (1) 0 0 non-Bilateria Cnidaria N. vectensis 0 2 (1) 0 0 A. digitifera 0 2 (1) 0 0 O. faveolata 0 2 (1) 0 0 H. magnipapillata 0 1 (1) 0 1 (0) Placozoa T. adhaerens 0 2 (1) 0 0 Porifera O. carmela 0 2 (1) 0 2 (1) S. cilliatum 0 1 (0) 0 0 L. complicata 0 2 (1) 0 1 (0) A. queenslandica 0 2 (1) 0 0 Ctenophora M. leidyi 0 1 (1) 0 1 (0) although their cephalochordate orthologues would have Class subfamilies are highly conserved, having experienced few dupli- 1 motifs. Mammalian GSG1L2 would also present Class 1 cation or deletion events throughout metazoan evolution. motifs although Danio rerio orthologues present a Class 3 Nevertheless, while the CNIH1/2/3 subfamily is found in all motif. Finally, mammalian GSG1Ls, which are the only mem- metazoan phyla studied, ctenophores and echinoderms appar- bers of the family known to interact with AMPARs, present a ently lost the CNIH4 subfamily. From all species studied, PDZ motif class termed Trp-1 [74]. This motif results from an vertebrates have the highest number of cornichon proteins, pre- insertion of 7 amino acids exclusive to the mammalian lineage. senting four. Among invertebrates, B. lanceolatum, B. floridae, Other vertebrate species present a Class 3 PDZ motif (electronic C. gigas and A. californica are the species with more cornichons, supplementary material, figure S4). with three coding genes in their genomes. The vertebrate CNIH4 protein interacts with GPCRs, promoting their traffic to the cell surface [41]. The high conser- 3.3. A process of subfunctionalization would have led vation of this subfamily allows us to hypothesize that all its members will act in GPCR trafficking to the cell surface, like to vertebrate cornichon ARAS their vertebrate counterparts. Instead, vertebrate members of We next investigated the phylogeny of cornichon proteins. the CNIH1/2/3 subfamily have different functions. While Noticeably, we found cornichon sequences in all species CNIH1 is involved in the maturation of TGFα [18], both investigated, covering all metazoan phyla (table 1). Our phylo- CNIH2 and CNIH3 function as ARASs [33]. These three genetic analysis indicates that this family is divided into two genes appeared by duplication in the vertebrate ancestor; subfamilies: CNIH1/2/3 and CNIH4 (figure 3; electronic sup- thus, the phylogeny alone does not allow us to hypothesize plementary material, figure S5). Furthermore, it also shows about the function of the ancestral gene. Yet both the fruit fly that the ancestor of all metazoans already presented the two and C. elegans present a single cornichon orthologue each genes that later gave rise to these two subfamilies (electronic sup- [16,18]. Interestingly, the fly protein has been found to partici- plementary material, figure S2B). Interestingly, both cornichon pate in the maturation of TGFα [18], as mammalian CNIH1s, (a) TM3 extracellular TM4 (b) 6 royalsocietypublishing.org/journal/rsob TARPs subunits VDCC- Open Biol. 10: 200234 GSG1s CACNG prot- Figure 2. Multiple sequence alignment of the TM3, TM4 and PDZ binding motif of CACNG-GSG1 proteins. The sequence alignment includes sequences from the four subfamilies in the CACNG-GSG1 protein family. From vertebrate species, only Homo sapiens and Danio rerio sequences were included. In the TARPs subfamily, only representative sequences from invertebrate species are shown. A complete alignment of TARPs sequences can be seen in electronic supplementary material, figure S3. The name of the subfamilies is indicated at the left of sequence names. (a) Alignment of the TM3 and TM4 of CACNG-GSG1s proteins, which are implicated in the interaction with AMPAR. The extension of the transmembrane segments and the extracellular loop is marked on the top of the alignment. Residues involved in the highly conserved YSYGWSF motif are highlighted by a red box. The conservation of each position of the alignment is represented by an intensity gradient of the background, higher conservation corresponding to more intense blue, and by a bar chart at the bottom. Also the quality chart (a measure of the probability of seeing mutation in an alignment position), the consensus sequence and the occupancy chart are shown in the bottom. Residue numbers shown on top indicate the start and finish of each transmembrane helix. Protein numbering corresponds to human CACNG2 sequence. (b) The last eight residues of proteins containing a PDZ binding motif are shown. If no PDZ binding motif is found it is labelled as No PDZ. The class of each PDZ binding motif is also displayed. The figure was prepared with Jalview v2.11.0. while the C. elegans orthologue, cni-1, acts as an ARAS. To the supplementary material, figure S2C). Based on the proteins best of our knowledge the role of the fly protein as ARAS has they contain, we have termed them Shisa1/L1 and ShisaL2 not been investigated, nor has the role of cni-1 been studied (figure 4; electronic supplementary material, figure S6). The in the context of TFGα maturation. Nevertheless, we propose Shisa1/L1 subfamily is better conserved, as it has only been that the invertebrate orthologue of CNIH1, 2 and 3 performs lost in echinoderms; instead, the ShisaL2 is only found in both functions and that in the vertebrate lineage this gene deuterostome species, having been lost in the ancestor of pro- was duplicated and subfunctionalized so that CNIH1 retained tostomes. While the Shisa1/L1 greatly expanded in the role as a factor for TFGα maturation, and the ancestor of vertebrates, the ShisaL2 expanded in cephalochordates and CNIH2 and 3 retained the ARAS function. hemichordates. Protostome species with shisas show a low number of sequences compared to chordates and S. kowalevs- kii, having 1 or 2 genes per species. The ML phylogeny shows 3.4. The phylogeny of the shisa family shows how the Shisa1/L1 subfamily might be further divided in two classes: one comprising vertebrate Shisa4 and Shisa5 independent expansions in deuterostome species with invertebrate sequences and a second including ver- Shisa proteins were only found among bilaterian species, tebrate shisas: L1, 2, 3 and 6 to 9 (electronic supplementary although not in all of them. Species such as Ciona intestinalis, material, figure S6). Nevertheless, the tree constructed with an urochordate, or L. gigantea, a mollusc, have lost this family the BI method presents a different topology for the Shisa1/ (table 1). The phylogenies revealed that this family is orga- L1 subfamily (figure 4); this does not allow us to fully con- nized in two subfamilies. The ancestral shisa gene would clude that there would be two classes within the Shisa1/L1 have appeared in bilaterians and duplicated before their subfamily, one being specific to vertebrates, as we found for diversification to generate these two subfamilies (electronic Type I TARPs. CNIH1_Hsa 100 7 Chordata vertebrata CNIH-1_Mmu 54 CNIH1_Gga TGF a royalsocietypublishing.org/journal/rsob Cephalochordata 97 CNIH1_Xtr maturation Hemichordata CNIH1_Dre Echinodermata CNIH2_Hsa Lophotrochozoa 100 100 CNIH-2_Mmu 100 Ecdysozoa CNIH2_Xtr 96 Cnidaria CNIH2_Dre AMPAR Placozoa 100 CNIH3_Hsa Porifera 100 CNIH-3_Mmu interactors 65 100 Ctenophora CNIH3_Gga 98 CNIH3_Xtr CNIH1/2/3 subfamily 65 CNIH3_Dre CNIH1/2/3_Bla 50 CNIH1/2/3 a_Bfl 100 CNIH1/2/3 b_Bfl CNIH1/2/3_Bbe Open Biol. 10: 200234 CNIH1/2/3_Cin 63 CNIH1/2/3_Sko 65 97 CNIH1/2/3_Apl 100 CNIH1/2/3_Spu CNIH1/2/3_Cte CNIH1/2/3a_Aca 100 81 52 CNIH1/2/3b_Aca 63 CNIH1/2/3_Lgi 96 52 CNIH1/2/3_Cgi CNIH1/2/3_Sma CNIH1/2/3_Pca CNIH1/2/3_Ofa 100 82 CNIH1/2/3_Adi 53 CNIH1/2/3_Nve CNIH1/2/3_Tad CNIH1/2/3_Hma 52 CNIH1/2/3_Lco CNIH1/2/3_Oca 96 CNIH1/2/3_Aqu CNIH1/2/3_Mle CNIH4_Hsa 97 CINH-4_Mmu 85 100 CNIH4_Gga GPCR 100 CNIH4_Xtr trafficking CNIH4_Dre 100 CNIH4a_Bla 100 CNIH4b_Bla 100 CNIH4_Bfl 100 83 CNIH4_Bbe CNIH4 subfamily CNIH4_Cte 100 CNIH4_Tad 100 CNIH4_Pfl 100 62 61 CNIH4_Sko CNIH4_Ofa 100 100 CNIH4_Adi CNIH4_Nve 96 CNIH4_Pca CNIH4_Lgi 100 CNIH4_Aca 100 CNIH4a_Cgi 100 81 CNIH4b_Cgi 61 CNIH4_Cin CNIH4_Ame 100 CNIH4_Sci 100 99 CNIH4_Lco CNIH4_Oca CNIH4_Aqu CNIH2_Ath 96 CNIH4_Ath 100 100 CNIH3_Ath 100 CNIH5_Ath CNIH1_Ath 0.5 Figure 3. Bayesian inference phylogenetic tree of the cornichon family. The two subfamilies in which the cornichon family is divided are highlighted by dashed line boxes. The name of each subfamily is presented at the right of the box. Posterior probabilities are shown at tree nodes and protein names at the end of each branch. Tree branches are coloured based on phylum, as indicated in the legend. Cornichon proteins from Arabidopsis thaliana were used as outgroup. When known the function of vertebrate sequences is indicated. Scale bar denotes number of amino acid substitutions per site. The amino acid substitution model used was Lg + G, the analysis ran for 10 000 000 generations, the final standard deviation was 0.004018 and the final potential scale reduction factor (PSRF) was 1.000. Shisa9_Hsa 100 99 Shisa9_Gga 8 Chordata vertebrata 100 Shisa9_Xtr Shisa9a_Dre royalsocietypublishing.org/journal/rsob Cephalochordata 100 Shisa9b_Dre Hemichordata 100 Shisa8_Gga 96 Echinodermata 85 85 Shisa8_Xtr Shisa8b_Dre AMPAR Lophotrochozoa Ecdysozoa 100 99 Shisa8_Hsa Shisa7_Hsa interactors Shisa7_Xtr 100 Shisa7a_Dre Shisa1/L1 subfamily 100 76 Shisa7b_Dre Shisa6_Hsa 100 Shisa6_Dre 99 Shisa2_Gga 98 Shisa2_Xtr 100 100 Shisa2_Hsa 60 Shisa2b_Dre Shisa1_Xtr FGF and WNT 67 Shisa3_Hsa 96 100 Shisa3_Xtr receptors 100 100 Shisa3_Gga Shisa3_Dre Shisa2a_Dre Open Biol. 10: 200234 Shisa1/L1 a_Sko 100 53 Shisa1/L1 b_Sko Shisa1/L1_Aca Shisa1/L1_Bbe 100 51 Shisa1/L1_Bfl 57 Shisa1/L1a_Cte 79 Shisa1/L1b_Cte Shisa1/L1_Sma 60 Shisa4_Hsa 97 Shisa4_Gga 100 Shisa4_Xtr 100 Shisa4_Dre 89 Shisa5_Hsa 100 100 Shisa5_Gga Shisa5_Dre p53-induced apoptosis ShisaL1a_Dre 99 ShisaL1b_Dre 100 ShisaL1_Hsa ShisaL1_Gga ShisaL2a_Sko 85 100 ShisaL2b_Sko 100 ShisaL2g_Sko ShisaL2a_Pfl 82 ShisaL2b_Pfl 100 100 ShisaL2g_Pfl 63 ShisaL2d_Pfl ShisaL2a_Bla 99 ShisaL2b_Bla 94 ShisaL2A_Hsa 92 98 ShisaL2A_Gga 91 ShisaL2A_Xtr 100 ShisaL2A_Dre ShisaL2B_Hsa 100 ShisaL2 subfamily ShisaL2B_Xtr ShisaL2 a_Bfl 100 100 ShisaL2b_Bfl ShisaL2a_Bbe 100 100 ShisaL2g_Bla ShisaL2g_Bfl 90 ShisaL2b_Bbe ShisaL2g_Bbe 100 74 99 ShisaL2d_Bfl ShisaL2d_Sko 82 ShisaL2a_Apl 100 ShisaL2b_Apl ShisaL2e_Bfl 99 ShisaL2z_Bfl 100 ShisaL2d_Bla 100 ShisaL2d_Bbe ShisaL2e_Bla ShisaL2z_Bla 91 ShisaL2h_Bfl 65 100 ShisaL2e_Bbe ShisaL2_Spu ShisaL2e_Sko 100 100 ShisaL2z_Sko ShisaL2h_Sko 100 ShisaL2q_Sko 83 100 ShisaL2 _Sko 64 ShisaL2 _Sko ShisaL2l_Sko 82 100 ShisaL2m_Sko 75 ShisaL2 _Sko ShisaL2x_Sko 98 ShisaL2e_Pfl VOPP1_Hsa 100 100 VOPP1_Dre VOPP1b_Dre WBP1La_Dre 94 100 100 WBP1L_Hsa 97 WBP1L1b_Dre 100 WBP1_Hsa WBP1b_Dre 0.9 Figure 4. Bayesian inference phylogenetic tree of the shisa family. The two subfamilies in which the shisa family is divided are highlighted by dashed line boxes. The name of each subfamily is presented at the right of the box. Posterior probabilities are shown at tree nodes and protein names at the end of each branch. Tree branches are coloured based on phylum, as indicated in the legend. The closely related vertebrate proteins VOPP and WBP1 were used as outgroup. The function of vertebrate sequences, if known, is indicated. Scale bar denotes number of amino acid substitutions per site. The amino acid substitution model used was Vt + I + G + F, the analysis ran for 10 000 000 generations, the final standard deviation was 0.005589 and the final potential scale reduction factor (PSRF) was 1.000. Vertebrate proteins from the Shisa1/L1 perform very differ- the complexity of their nervous system [2]. Here, we present 9 ent biological functions. Shisa5 is involved in p53-induced how the neo/subfunctionalization of unrelated proteins into royalsocietypublishing.org/journal/rsob apoptosis, Shisa2 in FGF receptor traffic, Shisa3 in WNT recep- ARASs has resulted in a different evolutionary strategy to tor function and Shisas 6 to 9 are ARASs. Due to its position in increase the synaptic proteome in vertebrates. both phylogenies and its grouping with invertebrate shisas, we Mammalian ARASs were thought to belong to five protein propose that Shisa5 would be phylogenetically closer to the families: CACNG, GSG1, cornichon, sisha and Dispanin ancestral shisa than the other paralogues and that the ancestral C. CACNGs and GSG1s are known to belong to the superfam- gene would also be involved in apoptosis. In this scenario the ily of claudins [85], a large group of proteins that presents over emergence of ARAS in this subfamily would be the result of 40 members in mammals and that includes three families, one neofunctionalization events occurring during vertebrate evol- of them being that of CACNGs [86]. Nevertheless, the exact ution in the branch of Shisa6 to Shisa9. position of GSG1s within claudins was unknown. Our phylo- genetic analysis indicates that GSG1s actually belong to the CACNG family of claudins. We thus refer to this family as 3.5. The Dispanin C family is poorly conserved among CACNG-GSG1. Therefore, ARASs are organized into four evo- Open Biol. 10: 200234 metazoans lutionarily unrelated protein families. In contrast with what we reported for the family of ionotropic glutamate receptors, SynDIG1, a member of the Dispanin C family, has recently which diversified into 12 phylogenetic groups, including four been reported as an AMPAR auxiliary subunit [75]. Our phy- subfamilies and ten classes [2], the animal evolution of the logenies indicate that members of this family can be traced to families containing ARAS has not been particularly complex. the early-diverging phylum of Ctenophores (figure 5; elec- cornichon, sisha and Dispanin C families can be divided into tronic supplementary material, figure S7). Yet, this family has just two phylogenetic groups (subfamilies) and the CACNG- been lost in multiple lineages and species during metazoan GSG1 family in four. Nevertheless, all these families have evolution (table 1). Only half of the 26 species investigated pre- increased their members along animal evolution, and ver- sent at least one member of the Dispanin C family. Our trees tebrates have especially increased their number of ARAS. indicate that in metazoans this family is also divided into This is particularly true for the CACNG-GSG1 and shisa two subfamilies, which we have named Dispanin C1 and Dis- families, which include 11 and 10 proteins, respectively, in panin C2. The L. gigantea DispaninC1/2_Lgi does not belong to the vertebrate species investigated, of which 7 and 4 are any subfamily, since we didn’t find other proteins in the same ARASs. Altogether vertebrates generally present 14 ARASs, tree branch we did not define a new subfamily for it. The cte- while, based on our study, invertebrate bilaterals would have nophore sequence DispaninC_Mle is the first to diverge, not fewer, between 1 and 7, and non-bilaterals even fewer, 1 or belonging to any subfamily; thus we propose that the ancestor 2. Sequences from invertebrate bilaterals only fall confidently of all metazoans had a single gene coding for Dispanin C and in two subfamilies with experimentally identified ARAS: that after the split of ctenophores it duplicated, giving rise to the of TARPs within the CACNG-GSG1 family, and that of two subfamilies (electronic supplementary material, figure CNIH1/2/3 among cornichons. Furthermore, non-bilaterals S2D). Virtually nothing is known about the two vertebrate would only present proteins phylogenetically related to the paralogues of SynDIG1, SynDIG1L and TMEM91, although ARAS subfamily of CNIH1/2/3 cornichons. Additionally, a SynDIG1L has been found downregulated in mouse models number of invertebrates, including the basal sponge O. carmela, of Huntington’s disease [57]. Due to the lack of functional infor- could have retained one Dispanin C1 phylogenetically close to mation in other vertebrate or invertebrate Dispanin Cs it not the mammalian ARAS SynDIG1, although the phylogenies are possible to establish when SynDIG1 function as ARAS arose not fully conclusive in this regard. The loss of the shisa and Dis- during evolution. panin C families in multiple invertebrate species, and even in entire phyla, suggests a less relevant role of these proteins in invertebrates as compared with vertebrates, which present 4. Discussion high conservation levels in both families. Furthermore, our Evolutionary studies have demonstrated that the synaptic pro- analysis suggests that ARAS from the shisa family are only teome has importantly increased in vertebrates [76], expanding present in vertebrates, postulating them as an innovation of the molecular tools available to synaptic physiology [77–79] this lineage. and multiplying synaptic molecular types [80,81]. The two As ARASs belong to four unrelated protein families that rounds of whole-genome duplication (2R) that occurred at include proteins with other functions, we aimed at using the base of the vertebrate lineage [82] were the major driving our phylogenetic study to propose when in evolution these force behind this expansion, as genes expressed at the synapse proteins acquired their function as AMPAR modulators, were retained at high frequencies after these duplication events although functional information would be required to com- [83]. The result of this increased complexity in the vertebrate pletely establish their role. Our data indicate that all synaptic proteome has been associated with the higher cogni- proteins identified in the TARP subfamily might function as tive functions found in mammals [84]. Nevertheless, there are ARASs, which would mean that the TARP subfamily exceptions to this general model of synaptic proteome evol- would be the only one in which their proteins are solely dedi- ution. This is the case of glutamate receptors, key nervous cated to modulating AMPAR function. On the other hand, we system proteins driving the excitatory synaptic transmission. found a possible example of vertebrate neofunctionalization These proteins have undergone a highly sophisticated evol- of ARASs in the subfamily of GSG1Ls, as its vertebrate para- utionary pattern in animals, with many lineage-specific logue GSG1 is involved in trafficking of TPAP. Identifying the gains, losses and expansions of entire classes of these receptors. function of non-vertebrate GSG1s would be required to estab- This process has resulted in a similar number of glutamate lish more conclusively this event of neofunctionalization. receptors being present in most animal species, regardless of Vertebrates have two cornichons acting as ARASs (CNIH2 Chordata vertebrata 10 Cephalochordata Hemichordata SynDIG1_Gga royalsocietypublishing.org/journal/rsob 92 Echinodermata 63 SynDIG1_Xtr Lophotrochozoa SynDIG1_Dre AMPAR 100 Cnidaria SynDIG1_Hsa interactors Porifera 51 SynDIG1_Mmu Ctenophora 100 SynDIG1L_Hsa 100 Dispanin C1 82 SynDIG1l_Mmu subfamily 96 SynDIG1L_Gga 100 SynDIG1L_Dre TMEM91_Hsa 100 TMEM91_Mmu 97 TMEM91_Xtr 54 TMEM91_Dre 95 DispaninC1_Cgi 98 Open Biol. 10: 200234 86 DispaninC1_Oca DispaninC1_Sko DispaninC1_Bbe 100 DispaninC1_Bla DispaninC2a_Bbe 100 Dispanin C2 subfamily 100 DispaninC2a_Bfl DispaninC2b_Bbe 98 DispaninC2g_Bbe 97 100 100 DispaninC2b_Bfl 56 DispaninC2d_Bbe DispaninC2_Lgi 94 DispaninC2a_Spu 87 DispaninC2_Hma 51 DispaninC2e_Bbe 76 100 DispaninC2z_Bbe DispaninC2g_Bfl 83 XM_011453764.2_Cgi DispaninC2_Cte 78 100 DispaninC2_Lco 67 DispaninC2b_Spu DispaninC2_Oca DispaninC1/2_Lgi DispaninC_Mle PRRT2_Hsa 100 91 PRRT2_Mmu 100 PRRT2_Acar PRRT2_Dre 0.9 Figure 5. Bayesian inference phylogenetic tree of the Dispanin C family. The two subfamilies in which the Dispanin C family is divided are highlighted by dashed line boxes. The name of each subfamily is presented at the right of the box. Posterior probabilities are shown at tree nodes and protein names at the end of each branch. Tree branches are coloured based on phylum, as indicated in the legend. The vertebrate proteins PRRT2 from the Dispanin B family were used as outgroup. The function of vertebrate sequences, if known, is also indicated. Scale bar denotes number of amino acid substitutions per site. The amino acid substitution model used was Vt + I + G + F, the analysis ran for 10 000 000 generations, the final standard deviation was 0.003312 and the final potential scale reduction factor (PSRF) was 1.000. and 3), emerging from a vertebrate-specific duplication. Their we found in the CACNG-GSG1 and cornichon subfamilies closest paralogue is CNIH1, which is involved in TGFα matu- containing ARAS, the Shisa1/L1 subfamily includes very ration. Interestingly, the fly sole orthologue to CNIH1, 2 and 3 few sequences from non-vertebrates. Thereby, TARPs and is known to participate in the maturation of TGFα [18], while CNIH1/2/3 subfamilies present 31 and 23 sequences from the unique orthologue of these proteins in C. elegans has a invertebrate species, respectively, while the Shisa1/L1 sub- well-proven role as an ARAS [16]. We thus propose that family includes only eight. Importantly, all these eight invertebrate orthologues of CNIH1/2/3 perform both func- sequences fall in the branch of Shisa5, which is not an tions and that vertebrate paralogues underwent a process of ARAS. Thus, due to the topology of the phylogenetic trees subfunctionalization by which CNIH1 retained the TGFα we propose that Shisa6 to 9 would have experienced a pro- maturation function and the ancestor of CNIH2 and 3 cess of neo/subfunctionalization early in the vertebrate retained the ARAS role. Within the shisa family all known lineage to become ARASs. Finally, the phylogenetic evolution ARASs (Shisa6 to 9) are in the Shisa1/L1 subfamily, yet this of the Dispanin C family would also suggest a neo/subfunc- phylogenetic group also contains proteins performing other tionalization process in the vertebrate lineage resulting in functions, as Shisa5 participates in p53-induced apoptosis members of the C1 subfamily becoming ARASs, although, [49] and Shisa2 and 3, the closest paralogues to Shisas6–9 as before, research on vertebrate and invertebrate orthologues with known function, are involved in FGF [47] and WNT of SynDIG1 will be required to fully validate this hypothesis. [48] receptor trafficking, respectively. In opposition to what The fact that large-scale proteomics experiments of synaptic preparations [77–79] and repositories of synaptic proteins [87] vertebrate evolution, which suggests that these proteins 11 do not identify SynDIG1L or TMEM91 suggests that they are might have importantly contributed to the development of royalsocietypublishing.org/journal/rsob unlikely to function as ARASs, thus supporting the hypoth- the complex nervous systems found in these animals. The esis of a neo/subfunctionalization process of ARASs in this parallel recruitment of unrelated proteins to perform synaptic family. functions represents another strategy by which evolution has In summary, this study reveals that the large set of ARASs favoured an increased complexity in the synaptic proteome. found in vertebrates is absent from other species. Interest- ingly, the surge in ARASs happened much later than the Data accessibility. This article has no additional data. emergence of AMPARs [2]. The class of AMPA receptors Authors’ contributions. D.R.-V. performed the analysis. D.R.-V. and À.B. appears and diversifies in the ancestor of bilateral species, designed the study analysed the data and wrote the manuscript. around 800 Ma, while the increase in ARAS occurs early in À.B. allocated funding to the project. vertebrate evolution, approximately 400 Ma [11]. Our analy- Competing interests. The authors declare no conflict of interest. sis suggests that this increase is due to an expansion of Funding. D.R.-V. and À.B.’s financial support was provided by: BFU2012-34398, BFU2015-69717-P and RTI2018-097037-B-100 proteins belonging to the TARP subfamily, the neofunctiona- (MINECO), Career Integration Grant (ref. 304111), Ramón y Cajal Fel- Open Biol. 10: 200234 lization of a group of Shisa1/L1 subfamily proteins and lowship (RYC-2011-08391p), IEDI-2017-00822 and AGAUR (SGR14- parallel processes of neo/subfunctionalization occurring in 297 and 2017 SGR 1776). the cornichon and Dispanin C families. All these resulted in Acknowledgements. Our gratitude to Dr David Soto (Universitat de the recruitment of a large number of ARASs early in Barcelona) who provided useful advice on this research. References 1. Sobolevsky AI, Rosconi MP, Gouaux E. 2009 X-ray Genet. Dev. 9, 630–636. (doi:10.1016/s0959- 19. Walker CS, Brockie PJ, Madsen DM, Francis MM, structure, symmetry and mechanism of an AMPA- 437x(99)00029-5) Zheng Y, Koduri S, Mellem JE, Strutz-Seebohm N, subtype glutamate receptor. Nature 462, 745–756. 11. Kumar S, Stecher G, Suleski M, Hedges SB. 2017 Maricq AV. 2006 Reconstitution of invertebrate (doi:10.1038/nature08624) TimeTree: a resource for timelines, timetrees, and glutamate receptor function depends on stargazin- 2. Ramos-Vicente D et al. 2018 Metazoan evolution of divergence times. Mol. Biol. Evol. 34, 1812–1819. like proteins. Proc. Natl Acad. Sci. USA 103, 10 glutamate receptors reveals unreported (doi:10.1093/molbev/msx116) 781–10 786. (doi:10.1073/pnas.0604482103) phylogenetic groups and divergent lineage-specific 12. Hopkins DH, Rane RV, Younus F, Coppin CW, Pandey 20. Letts VA et al. 1998 The mouse stargazer gene events. Elife 7, e35774. (doi:10.7554/eLife.35774) G, Jackson CJ, Oakeshott JG. 2019 The molecular encodes a neuronal Ca2+-channel gamma subunit. 3. Hodge RD et al. 2019 Conserved cell types with basis for the neofunctionalization of the juvenile Nat. Genet. 19, 340–347. (doi:10.1038/1228) divergent features in human versus mouse hormone esterase duplication in Drosophila. Insect 21. Chen L, Chetkovich DM, Petralia RS, Sweeney NT, cortex. Nature 573, 61–68. (doi:10.1038/s41586- Biochem. Mol. Biol. 106, 10–18. (doi:10.1016/j. Kawasaki Y, Wenthold RJ, Bredt DS, Nicoll RA. 2000 019-1506-7) ibmb.2019.01.001) Stargazin regulates synaptic targeting of AMPA 4. Traynelis SF et al. 2010 Glutamate receptor ion 13. He X, Zhang J. 2005 Rapid subfunctionalization receptors by two distinct mechanisms. Nature 408, channels: structure, regulation, and function. accompanied by prolonged and substantial 936–943. (doi:10.1038/35050030) Pharmacol. Rev. 62, 405–496. (doi:10.1124/pr. neofunctionalization in duplicate gene evolution. 22. Hashimoto K, Fukaya M, Qiao X, Sakimura K, 109.002451) Genetics 169, 1157–1164. (doi:10.1534/genetics. Watanabe M, Kano M. 1999 Impairment of AMPA 5. Jackson AC, Nicoll RA. 2011 The expanding social 104.037051) receptor function in cerebellar granule cells of ataxic network of ionotropic glutamate receptors: TARPs 14. Lambert MJ, Cochran WO, Wilde BM, Olsen KG, mutant mouse stargazer. J. Neurosci. 19, and other transmembrane auxiliary subunits. Cooper CD. 2015 Evidence for widespread 6027–6036. (doi:10.1523/JNEUROSCI.19-14- Neuron 70, 178–199. (doi:10.1016/j.neuron. subfunctionalization of splice forms in vertebrate 06027.1999) 2011.04.007) genomes. Genome Res. 25, 624–632. (doi:10.1101/ 23. Klugbauer N, Dai S, Specht V, Lacinová L, Marais E, 6. Haering SC, Tapken D, Pahl S, Hollmann M. 2014 gr.184473.114) Bohn G, Hofmann F. 2000 A family of gamma-like Auxiliary subunits: shepherding AMPA receptors to 15. Innan H, Kondrashov F. 2010 The evolution of gene calcium channel subunits. FEBS Lett. 470, 189–197. the plasma membrane. Membranes (Basel) 4, duplications: classifying and distinguishing between (doi:10.1016/s0014-5793(00)01306-5) 469–490. (doi:10.3390/membranes4030469) models. Nat. Rev. Genet. 11, 97–108. (doi:10.1038/ 24. Burgess DL, Gefrides LA, Foreman PJ, Noebels JL. 7. Ng D et al. 2009 Neto1 is a novel CUB-domain nrg2689) 2001 A cluster of three novel Ca2+ channel gamma NMDA receptor-interacting protein required for 16. Brockie PJ et al. 2013 Cornichons control ER export subunit genes on chromosome 19q13.4: evolution synaptic plasticity and learning. PLoS Biol. 7, e41. of AMPA receptors to regulate synaptic excitability. and expression profile of the gamma subunit gene (doi:10.1371/journal.pbio.1000041) Neuron 80, 129–142. (doi:10.1016/j.neuron.2013. family. Genomics 71, 339–350. (doi:10.1006/geno. 8. Zhang W et al. 2009 A transmembrane accessory 07.028) 2000.6440) subunit that modulates kainate-type glutamate 17. Wang R, Walker CS, Brockie PJ, Francis MM, Mellem 25. Chu PJ, Robertson HM, Best PM. 2001 Calcium receptors. Neuron 61, 385–396. (doi:10.1016/j. JE, Madsen DM, Maricq AV. 2008 Evolutionary channel gamma subunits provide insights into the neuron.2008.12.014) conserved role for TARPs in the gating of glutamate evolution of this gene family. Gene 280, 37–48. 9. Tang M, Pelkey KA, Ng D, Ivakine E, McBain CJ, receptors and tuning of synaptic function. Neuron (doi:10.1016/s0378-1119(01)00738-7) Salter MW, McInnes RR. 2011 Neto1 is an auxiliary 59, 997–1008. (doi:10.1016/j.neuron.2008.07.023) 26. Hansen JP, Chen R-S, Larsen JK, Chu P-J, Janes DM, subunit of native synaptic kainate receptors. 18. Bökel C, Dass S, Wilsch-Bräuninger M, Roth S. 2006 Weis KE, Best PM. 2004 Calcium channel gamma6 J. Neurosci. 31, 10 009–10 018. (doi:10.1523/ Drosophila Cornichon acts as cargo receptor for subunits are unique modulators of low voltage- JNEUROSCI.6617-10.2011) ER export of the TGFalpha-like growth factor activated (Cav3.1) calcium current. J. Mol. Cell. 10. Gogarten JP, Olendzenski L. 1999 Orthologs, Gurken. Development 133, 459–470. (doi:10.1242/ Cardiol. 37, 1147–1158. (doi:10.1016/j.yjmcc.2004. paralogs and genome comparisons. Curr. Opin. dev.02219) 08.005) 27. Tomita S, Chen L, Kawasaki Y, Petralia RS, Wenthold receptors. J. Neurosci. 32, 9796–9804. (doi:10.1523/ 53. Powers J, Barlowe C. 1998 Transport of axl2p 12 RJ, Nicoll RA, Bredt DS. 2003 Functional studies and JNEUROSCI.0345-12.2012) depends on erv14p, an ER-vesicle protein related to royalsocietypublishing.org/journal/rsob distribution define a family of transmembrane 40. Castro CP, Piscopo D, Nakagawa T, Derynck R. 2007 the Drosophila cornichon gene product. J. Cell Biol. AMPA receptor regulatory proteins. J. Cell Biol. 161, Cornichon regulates transport and secretion of 142, 1209–1222. (doi:10.1083/jcb.142.5.1209) 805–816. (doi:10.1083/jcb.200212116) TGFalpha-related proteins in metazoan cells. J. Cell. 54. Fahey B, Degnan BM. 2010 Origin of animal 28. Kato AS et al. 2007 New transmembrane AMPA Sci. 120, 2454–2466. (doi:10.1242/jcs.004200) epithelia: insights from the sponge genome. Evol. receptor regulatory protein isoform, gamma-7, 41. Sauvageau E, Rochdi MD, Oueslati M, Hamdan FF, Dev. 12, 601–617. (doi:10.1111/j.1525-142X.2010. differentially regulates AMPA receptors. J. Neurosci. Percherancier Y, Simpson JC, Pepperkok R, Bouvier 00445.x) 27, 4969–4977. (doi:10.1523/JNEUROSCI.5561- M. 2014 CNIH4 interacts with newly synthesized 55. Leys SP, Riesgo A. 2012 Epithelia, an evolutionary 06.2007) GPCR and controls their export from the novelty of metazoans. J. Exp. Zool. B Mol. Dev. Evol. 29. Soto D, Coombs ID, Renzi M, Zonouzi M, Farrant M, endoplasmic reticulum. Traffic 15, 383–400. 318, 438–447. (doi:10.1002/jez.b.21442) Cull-Candy SG. 2009 Selective regulation of long- (doi:10.1111/tra.12148) 56. Pei J, Grishin NV. 2012 Unexpected diversity in form calcium-permeable AMPA receptors by an 42. von Engelhardt J, Mack V, Sprengel R, Kavenstock N, Shisa-like proteins suggests the importance of their atypical TARP, gamma-5. Nat. Neurosci. 12, Li KW, Stern-Bach Y, Smit AB, Seeburg PH, Monyer roles as transmembrane adaptors. Cell. Signal. 24, Open Biol. 10: 200234 277–285. (doi:10.1038/nn.2266) H. 2010 CKAMP44: a brain-specific protein 758–769. (doi:10.1016/j.cellsig.2011.11.011) 30. Deng F, Price MG, Davis CF, Mori M, Burgess DL. attenuating short-term synaptic plasticity in the 57. Sällman AM, Bringeland N, Fredriksson R, Schiöth 2006 Stargazin and other transmembrane AMPA dentate gyrus. Science 327, 1518–1522. (doi:10. HB. 2012 The dispanins: a novel gene family of receptor regulating proteins interact with synaptic 1126/science.1184178) ancient origin that contains 14 human members. scaffolding protein MAGI-2 in brain. J. Neurosci. 26, 43. Farrow P, Khodosevich K, Sapir Y, Schulmann A, PLoS ONE 7, e31961. (doi:10.1371/journal.pone. 7875–7884. (doi:10.1523/JNEUROSCI.1851-06.2006) Aslam M, Stern-Bach Y, Monyer H, von Engelhardt J. 0031961) 31. Tomita S. 2010 Regulation of ionotropic glutamate 2015 Auxiliary subunits of the CKAMP family 58. Simakov O et al. 2015 Hemichordate genomes and receptors by their auxiliary subunits. Physiology differentially modulate AMPA receptor properties. deuterostome origins. Nature 527, 459–465. (Bethesda) 25, 41–49. (doi:10.1152/physiol. Elife 4, e09693. (doi:10.7554/eLife.09693) (doi:10.1038/nature16150) 00033.2009) 44. Khodosevich K, Jacobi E, Farrow P, Schulmann A, 59. Mao Y, Economo EP, Satoh N. 2018 The roles of 32. Shanks NF et al. 2012 Differences in AMPA Rusu A, Zhang L, Sprengel R, Monyer H, von introgression and climate change in the rise to and kainate receptor interactomes facilitate Engelhardt J. 2014 Coexpressed auxiliary subunits dominance of Acropora corals. Curr. Biol. 28, identification of AMPA receptor auxiliary subunit exhibit distinct modulatory profiles on AMPA 3373–3382. (doi:10.1016/j.cub.2018.08.061) GSG1L. Cell Rep 1, 590–598. (doi:10.1016/j.celrep. receptor function. Neuron 83, 601–615. (doi:10. 60. Altschul SF, Gish W, Miller W, Myers EW, Lipman DJ. 2012.05.004) 1016/j.neuron.2014.07.004) 1990 Basic local alignment search tool. J. Mol. 33. Schwenk J et al. 2012 High-resolution proteomics 45. Klaassen RV et al. 2016 Shisa6 traps AMPA receptors Biol. 215, 403–410. (doi:10.1016/S0022- unravel architecture and molecular diversity of at postsynaptic sites and prevents their 2836(05)80360-2) native AMPA receptor complexes. Neuron 74, desensitization during synaptic activity. Nat. 61. Edgar RC. 2004 MUSCLE: a multiple sequence 621–633. (doi:10.1016/j.neuron.2012.03.034) Commun. 7, 10682. (doi:10.1038/ncomms10682) alignment method with reduced time and space 34. Choi H-S, Lee S-H, Kim H, Lee Y. 2008 Germ cell- 46. von Engelhardt J. 2019 AMPA receptor auxiliary complexity. BMC Bioinformatics 5, 113. (doi:10. specific gene 1 targets testis-specific poly(A) proteins of the CKAMP family. Int. J. Mol. Sci. 20, 1186/1471-2105-5-113) polymerase to the endoplasmic reticulum through 1460. (doi:10.3390/ijms20061460) 62. Kumar S, Stecher G, Li M, Knyaz C, Tamura K. 2018 protein-protein interactions. FEBS Lett. 582, 47. Hedge TA, Mason I. 2008 Expression of Shisa2, a MEGA X: molecular evolutionary genetics analysis 1203–1209. (doi:10.1016/j.febslet.2008.01.065) modulator of both Wnt and Fgf signaling, in the across computing platforms. Mol. Biol. Evol. 35, 35. Gu X, Mao X, Lussier MP, Hutchison MA, Zhou L, chick embryo. Int. J. Dev. Biol. 52, 81–85. (doi:10. 1547–1549. (doi:10.1093/molbev/msy096) Hamra FK, Roche KW, Lu W. 2016 GSG1L suppresses 1387/ijdb.072355th) 63. Darriba D, Taboada GL, Doallo R, Posada D. 2011 AMPA receptor-mediated synaptic transmission and 48. Chen C-C et al. 2014 Shisa3 is associated with ProtTest 3: fast selection of best-fit models of uniquely modulates AMPA receptor kinetics in prolonged survival through promoting β-catenin protein evolution. Bioinformatics 27, 1164–1165. hippocampal neurons. Nat Commun 7, 10873. degradation in lung cancer. Am. J. Respir. Crit. Care (doi:10.1093/bioinformatics/btr088) (doi:10.1038/ncomms10873) Med. 190, 433–444. (doi:10.1164/rccm.201312- 64. Ronquist F et al. 2012 MrBayes 3.2: efficient 36. Mao X, Gu X, Lu W. 2017 GSG1L regulates the 2256OC) Bayesian phylogenetic inference and model choice strength of AMPA receptor-mediated synaptic 49. Bourdon J-C, Renzing J, Robertson PL, Fernandes across a large model space. Syst. Biol. 61, 539–542. transmission but not AMPA receptor kinetics in KN, Lane DP. 2002 Scotin, a novel p53-inducible (doi:10.1093/sysbio/sys029) hippocampal dentate granule neurons. proapoptotic protein located in the ER and the 65. Nguyen L-T, Schmidt HA, von Haeseler A, Minh BQ. J. Neurophysiol. 117, 28–35. (doi:10.1152/jn. nuclear membrane. J. Cell Biol. 158, 235–246. 2015 IQ-TREE: a fast and effective stochastic 00307.2016) (doi:10.1083/jcb.200203006) algorithm for estimating maximum-likelihood 37. Shi Y, Suh YH, Milstein AD, Isozaki K, Schmid SM, 50. Kalashnikova E, Lorca RA, Kaur I, Barisone GA, Li B, phylogenies. Mol. Biol. Evol. 32, 268–274. (doi:10. Roche KW, Nicoll RA. 2010 Functional comparison of Ishimaru T, Trimmer JS, Mohapatra DP, Díaz E. 2010 1093/molbev/msu300) the effects of TARPs and cornichons on AMPA SynDIG1: an activity-regulated, AMPA- receptor- 66. Hoang DT, Chernomor O, von Haeseler A, Minh BQ, receptor trafficking and gating. Proc. Natl Acad. Sci. interacting transmembrane protein that regulates Vinh LS. 2018 UFBoot2: improving the ultrafast USA 107, 16 315–16 319. (doi:10.1073/pnas. excitatory synapse development. Neuron 65, 80–93. bootstrap approximation. Mol. Biol. Evol. 35, 1011706107) (doi:10.1016/j.neuron.2009.12.021) 518–522. (doi:10.1093/molbev/msx281) 38. Kato AS et al. 2010 Hippocampal AMPA receptor 51. Díaz E. 2010 SynDIG1 regulation of synaptic AMPA 67. Miller MA, Pfeiffer W, Schwartz T. 2010 Creating the gating controlled by both TARP and cornichon receptor targeting. Commun. Integr. Biol. 3, CIPRES Science Gateway for inference of large proteins. Neuron 68, 1082–1096. (doi:10.1016/j. 347–349. (doi:10.4161/cib.3.4.11765) phylogenetic trees. In 2010 Gateway Computing neuron.2010.11.026) 52. Wudick MM et al. 2018 CORNICHON sorting and Environments Workshop (GCE), pp. 1–8. 39. Coombs ID, Soto D, Zonouzi M, Renzi M, Shelley C, regulation of GLR channels underlie pollen tube 68. Kumar M et al. 2020 ELM—the eukaryotic linear Farrant M, Cull-Candy SG. 2012 Cornichons modify Ca2+ homeostasis. Science 360, 533–536. (doi:10. motif resource in 2020. Nucleic Acids Res. 48, channel properties of recombinant and glial AMPA 1126/science.aar6464) D296–D306. (doi:10.1093/nar/gkz1030) 69. Price MG, Davis CF, Deng F, Burgess DL. 2005 The 75. Díaz E et al. 2002 Molecular analysis of gene synapses across the mouse life span. Science 369, 13 alpha-amino-3-hydroxyl-5-methyl-4- expression in the developing pontocerebellar 270–275. (doi:10.1126/science.aba3163) royalsocietypublishing.org/journal/rsob isoxazolepropionate receptor trafficking regulator projection system. Neuron 36, 417–434. (doi:10. 82. Kasahara M. 2007 The 2R hypothesis: an update. ‘stargazin’ is related to the claudin family of 1016/s0896-6273(02)01016-4) Curr. Opin. Immunol. 19, 547–552. (doi:10.1016/j. proteins by Its ability to mediate cell–cell adhesion. 76. Emes RD, Grant SGN. 2012 Evolution of synapse coi.2007.07.009) J. Biol. Chem. 280, 19 711–19 720. (doi:10.1074/ complexity and diversity. Annu. Rev. Neurosci. 35, 83. Bayés À, Collins MO, Reig-Viader R, Gou G, Goulding jbc.M500623200) 111–131. (doi:10.1146/annurev-neuro-062111- D, Izquierdo A, Choudhary JS, Emes RD, Grant SGN. 70. Adato A et al. 2002 USH3A transcripts encode 150433) 2017 Evolution of complexity in the zebrafish clarin-1, a four-transmembrane-domain protein 77. Bayés A, Collins MO, Croning MDR, van de synapse proteome. Nat Commun. 8, 14613. (doi:10. with a possible role in sensory synapses. Lagemaat LN, Choudhary JS, Grant SGN. 2012 1038/ncomms14613) Eur. J. Hum. Genet. 10, 339–350. (doi:10.1038/sj. Comparative study of human and mouse 84. Nithianantharajah J et al. 2013 Synaptic scaffold ejhg.5200831) postsynaptic proteomes finds high compositional evolution generated components of vertebrate 71. Kato AS, Siuda ER, Nisenbaum ES, Bredt DS. 2008 conservation and abundance differences for key cognitive complexity. Nat. Neurosci. 16, 16–24. AMPA receptor subunit-specific regulation by a synaptic proteins. PLoS ONE 7, e46683. (doi:10. (doi:10.1038/nn.3276) Open Biol. 10: 200234 distinct family of type II TARPs. Neuron 59, 1371/journal.pone.0046683) 85. Chen S, Gouaux E. 2019 Structure and mechanism 986–996. (doi:10.1016/j.neuron.2008.07.034) 78. Emes RD et al. 2008 Evolutionary expansion and of AMPA receptor–auxiliary protein complexes. Curr. 72. Kim E, Sheng M. 2004 PDZ domain proteins of anatomical specialization of synapse proteome Opin. Struct. Biol. 54, 104–111. (doi:10.1016/j.sbi. synapses. Nat. Rev. Neurosci. 5, 771–781. (doi:10. complexity. Nat. Neurosci. 11, 799–806. (doi:10. 2019.01.011) 1038/nrn1517) 1038/nn.2135) 86. Maher GJ, Hilton EN, Urquhart JE, Davidson AE, 73. Dakoji S, Tomita S, Karimzadegan S, Nicoll RA, Bredt 79. Bayés A, van de Lagemaat LN, Collins MO, Croning Spencer HL, Black GC, Manson FD. 2011 The DS. 2003 Interaction of transmembrane AMPA MDR, Whittle IR, Choudhary JS, Grant SGN. 2011 cataract-associated protein TMEM114, and receptor regulatory proteins with multiple Characterization of the proteome, diseases and TMEM235, are glycosylated transmembrane proteins membrane associated guanylate kinases. evolution of the human postsynaptic density. Nat. that are distinct from claudin family members. FEBS Neuropharmacology 45, 849–856. (doi:10.1016/ Neurosci. 14, 19–21. (doi:10.1038/nn.2719) Lett. 585, 2187–2192. (doi:10.1016/j.febslet.2011. s0028-3908(03)00267-3) 80. Zhu F et al. 2018 Architecture of the mouse brain 05.060) 74. Ernst A, Appleton BA, Ivarsson Y, Zhang Y, Gfeller D, synaptome. Neuron 99, 781–799. (doi:10.1016/j. 87. Koopmans F et al. 2019 SynGO: An evidence-based, Wiesmann C, Sidhu SS. 2014 A structural portrait of neuron.2018.07.007) expert-curated knowledge base for the synapse. the PDZ domain family. J. Mol. Biol. 426, 81. Cizeron M, Qiu Z, Koniaris B, Gokhale R, Komiyama Neuron 103, 217–234. (doi:10.1016/j.neuron.2019. 3509–3519. (doi:10.1016/j.jmb.2014.08.012) NH, Fransén E, Grant SGN. 2020 A brainwide atlas of 05.002)