Three Distinct Glutamate Decarboxylase Genes in Vertebrates

Gamma-aminobutyric acid (GABA) is a widely conserved signaling molecule that in animals has been adapted as a neurotransmitter. GABA is synthesized from the amino acid glutamate by the action of glutamate decarboxylases (GADs). Two vertebrate genes, GAD1 and GAD2, encode distinct GAD proteins: GAD67 and GAD65, respectively. We have identified a third vertebrate GAD gene, GAD3. This gene is conserved in fishes as well as tetrapods. We analyzed protein sequence, gene structure, synteny, and phylogenetics to identify GAD3 as a homolog of GAD1 and GAD2. Interestingly, we found that GAD3 was lost in the hominid lineage. Because of the importance of GABA as a neurotransmitter, GAD3 may play important roles in vertebrate nervous systems.

Scientific RepoRts | 6:30507 | DOI: 10.1038/srep30507 We used recently generated genomic resources to ask whether gad3 is present and expressed in species other than the goldfish and abyssal grenadier. Our results revealed a surprisingly broad conservation of GAD3 in mammals, reptiles, birds, and amphibians, as well as fishes.

Methods
Throughout this paper, we use standard gene nomenclature. For fishes, gene symbols are lowercase and italicized and protein symbols are capitalized. For other vertebrates, human conventions are used: gene symbols in all capitals and italicized, protein symbols in all capitals.
In addition to the species listed in Table 1 Model Testing. MEGA 6 software was used to compare 24 DNA evolution models for the aligned GAD CDS sequences 38 . A generalized time-reversible plus gamma (GTR + G + I) model had the lowest BIC score (Bayesian Information Criterion) and AICc value (Akaike Information Criterion, corrected), so it was used for subsequent phylogenetic analyses. In this model, non-uniformity of evolutionary rates among sites is modeled by estimating a discrete Gamma distribution (+ G) of rates and by assuming that certain sites are evolutionarily invariable (+ I).
Diagnostics of the MCMC sampling were carried out using Tracer v1.6 (http://tree.bio.ed.ac.uk/software/ tracer/). The effective sample size (ESS) for each parameter was > 300 for each run, allowing adequate sampling of the Markov chain.
Orangutan was used as a reference species for cross-species synteny and protein similarity to highlight conservation of synteny and GAD3 protein sequence in vertebrates despite the absence of GAD3 in some hominids. For comparing primate synteny, simiiformes (last common ancestor of simians) was used as the reference taxon. Synteny data, protein similarity, and species images were downloaded from Genomicus.

Results
We found previously uncharacterized GAD3 genes in many vertebrate genomes, including diverse fishes and tetrapods. A teleost gad3 transcript was found in a transcriptome library generated from testis tissue from Astatotilapia burtoni: (> comp56037_c0_seq1_indA_testis). Zebrafish gad3 has been previously referred to with the identifier zgc:163121. Interestingly, GAD3 had already been annotated in the Xenopus tropicalis genome as GAD1.
2. It appears, however, that it has not yet been studied in Xenopus.
Sequence Similarity. Gad3 predicted protein sequence from spotted gar (Lepisosteus oculatus) is more similar to Gad1 (Pairwise Identity: 60.5%) than to Gad2 (Pairwise Identity: 53.9%) (Fig. 1). Gad1 and Gad2 share 67.1% pairwise identity. The N-terminal domain, which is quite variable between Gad1 and Gad2, is truncated and highly divergent in Gad3. The N-terminal 92 amino acids (aa) of Gad1 align to the N-terminal 84 aa of Gad2. Gad3 has 42 aa aligning in this range, only 27 of which align to Gad1 and Gad2 sequence (with a 15 aa gap). These 27 aa of Gad3 have: 23.1% pairwise identity with Gad1, 11.5% pairwise identity with Gad2.

Phylogeny. Pyridoxal 5′-phosphate (PLP)-dependent decarboxylase genes include Glutamate
Decarboxylase-Like 1 (GADL1), Cysteine Sulfinic Acid Decarboxylase (CSAD), and histidine decarboxylase (HDC), in addition to GADs. Therefore, we tested the phylogenetic relationship of GAD3 to other genes in this group, using HDC as an outgroup for GAD, GADL1, and CSAD genes 42 . A neighbor-joining tree of aligned predicted amino acid sequences from the spotted gar (Lepisoteus oculatus) place gad3 most closely related to the gad1/gad2 clade (Fig. 2). A phylogenetic tree of vertebrate GAD1, GAD2, and GAD3 nucleotide coding sequences was generated using MrBayes (Fig. 3). In insects, GAD1 is the single homolog of vertebrate GAD genes (insect GAD2 is homologous to vertebrate CSAD and GADL1). Therefore we chose Drosophila melanogaster GAD1 as the outgroup for the vertebrate and deuterostome GAD genes.
Exon-intron Structure. Spotted gar gad1 and gad2 each have 16 exons (Fig. 4). Spotted gar gad3 has 17 exons. While both gad1 and gad2 have coding sequence beginning in exon 1, gad3 coding sequence (CDS) begins in the second exon (exon 2). The predicted coding sequence of gad3 has a gap (does not align) with the 5′ CDS sequence found in gad1 and gad2 exon 1, exon 2, and part of exon 3. The 3′ portion of the gad3 CDS is included on exon 17, while gad1 and gad2 stop codons are found in exon 16.
Aside from these differences, gad3 exon structure is largely similar to gad1 and gad2. All of the exon junctions from exon 3 to exon 16 are in identical locations for all three gad genes. In our alignment of the three gad genes, the only gaps introduced in gad3 are located in exon 2 and exon 17.
Synteny. In spotted gar, gad1 and gad2 are located adjacent to the myosin genes myo3b and myo3a, respectively. Similarly, in humans GAD1 is located near MYO3B on chromosome 2, and GAD2 is located adjacent to MYO3A on chromosome 10. On the other hand, gad3 is not located near a myosin gene in the genome of spotted gar. The genes located adjacent to spotted gar gad3 are mc4r and cdh20. This syntenic block of genes is conserved  GAD3 Conservation and Gene Loss. GAD3 predicted protein sequence is highly conserved in diverse vertebrate genomes (Fig. 6). Yet primates appear to have experienced varying degrees of gene loss at the GAD3 locus (Fig. 7). We identified a human transcript (ENST00000592477) with homology to GAD3 (Table 1), derived from a pseudogene located in the human genome in the conserved GAD3 syntenic position between MC4R and CDH20 (Fig. 7). Similarly, gorilla (Gorilla gorilla) GAD3 is annotated as a pseudogene in Ensembl (ENSGGOG00000027455). Although macaque (Macaca mulatta) and orangutan (Pongo pygmaeus) predicted GAD3 protein sequences share relatively high pairwise identity (84.5%), the GAD3 genes in these two species appear to have large insertions (or deletions) in their predicted coding sequence.

Discussion
We identified a novel glutamate decarboxylase homolog, GAD3, found in many vertebrate genomes. We provide phylogenetic and intron/exon structural evidence that GAD3 is an ancient paralog of GAD1 and GAD2. The conserved chromosomal synteny of GAD3 in vertebrates supports an ancient origin for this gene. Surprisingly, GAD3 was lost in the hominid lineage. Taken together, the phylogenetic analyses, comparisons of gene structure, and synteny data suggest that GAD3 arose via gene duplication of a protovertebrate GAD homolog, likely before the duplication of another paralog which gave rise to GAD1 and GAD2. GAD3 Evolution. Although our data do not rule out the possibility of a local duplication that gave rise to GAD3, they are consistent with an origin of GAD3 in an early vertebrate via whole-genome duplication. Whole-genome duplication is thought to have played a major role in early vertebrate evolution [43][44][45] . Following genome duplication, these duplicated gene pairs (ohnologs) experienced a range of outcomes including non-functionalization, sub-functionalization, and neo-functionalization 46,47 . Sub-functionalization may happen via protein changes 48 or via regulatory element loss 49 in which ancestral expression domains are differentially lost in different genes 50 . For example, recent evidence indicates that duplication of a corticotropin-releasing hormone (CRH) gene in an early vertebrate led to a broadly expressed CRH1 and a CRH2 with expression restricted to a single hindbrain nucleus 51 . Like our recent analyses of CRH genes, the discovery of GAD3 as a conserved vertebrate gene relied on freely available genomic resources, pointing to the likelihood that many gene families have unannotated homologs remaining to be found in sequenced genomes 51,52 . GAD3 Function. GAD3 is phylogenetically closer to GAD1 and GAD2 than to GADL1, but nonetheless it is possible that its enzymatic functions differ from those of GAD1 and GAD2. The absence of much of the N-terminal region, which regulates intracellular localization of GAD1 and GAD2 proteins, suggests that GAD3 protein may have different localization 5 .
Little is known regarding the function of GADL1 enzyme, though polymorphisms are linked to differential response to lithium treatment for bipolar disorder 53 . Mammalian GADL1 does not appear to have glutamate decarboxylase activity, despite its name. Instead, it catalyzes the decarboxylation of aspartate, cysteine sulfinic acid, and cysteic acid to β -alanine, hypotaurine, and taurine, respectively 14 . Recently, GADL1 and CSAD were found to have preference for cysteine sulfinic acid as a substrate 42 . Future studies of GAD3 biochemical substrates will be necessary to address the possibility of substrates other than glutamate.
Intriguingly, zebrafish gad3 (referred to as zgc163121) mRNA expression was significantly downregulated by treatment with dexamethasone, a glucocorticoid agonist, in 25hpf larval zebrafish, as measured by microarray and qPCR 54 . In the deep-sea fish in which gad3 was first described, the armed grenadier, Coryphaenoides (Nematonurus) armatus, gad2 mRNA levels were found to be expressed in the brain in a sexually dimorphic manner, i.e. higher in male hypothalamus than in female, but no differences were found in gad3 levels 35 . In the goldfish, however, gad3 mRNA levels in the telencephalon were highest in sexually mature fish of both sexes during the breeding period 55 . Since the specific role of gad3 is unknown in any taxa, the full range of factors that regulate gad3 expression in the brain, and potentially elsewhere, awaits further investigation.

Loss in hominids.
The loss of GAD3 in both chimpanzees and humans appears to have been preceded by changes to GAD3 sequences in other hominids. Predicted gorilla, orangutan, and gibbon GAD3 transcripts appear to be truncated relative to fish gad3 sequence, but it may be that not all the exons in these sequences are fully annotated in Ensembl. Glutamate metabolic pathways appear to have been under positive selection in hominids, as seen for example in the origin of glutamate dehydrogenase 2 (GLUD2) by retroposition of GLUD1 56 .
Gene losses have played major roles in human evolution 57 . For example, loss of L-gulonolactone oxidase (GULO) makes humans and other Haplorhini susceptible to scurvy, a vitamin C deficiency. Despite conferring this disadvantage, GULO gene loss has occurred in multiple mammalian lineages, including guinea pigs 58 and some bats 59 . Hominids are not the only lineage that has lost GAD3. Rodent genomes, including mice, rats, and squirrels also appear to be missing GAD3 homologs. The absence of GAD3 in both humans and mice likely explains why this gene was not discovered sooner, since many investigators choose to focus on these two species.  Human and chimpanzee genomes have no functional gene at the location corresponding to GAD3 in other primates. Humans have a pseudogene at this location. A thick blue line between two genes indicates a "gap", i.e. a gene lost relative to the ancestral Simiiformes genome. A thin blue line between two genes indicates a "break" in the continuity of the alignment, i.e. a gene added or lost relative to the ancestral Simiiformes genome. Figure modified from Genomicus AlignView 41 , which relies on data and images from Ensembl 34 .