The identification and functional implications of human-specific "fixed" amino acid substitutions in the glutamate receptor family

Background The glutamate receptors (GluRs) play a vital role in the mediation of excitatory synaptic transmission in the central nervous system. To clarify the evolutionary dynamics and mechanisms of the GluR genes in the lineage leading to humans, we determined the complete sequences of the coding regions and splice sites of 26 chimpanzee GluR genes. Results We found that all of the reading frames and splice sites of these genes reported in humans were completely conserved in chimpanzees, suggesting that there were no gross structural changes in humans after their divergence from the human-chimpanzee common ancestor. We observed low KA/KS ratios in both humans and chimpanzees, and we found no evidence of accelerated evolution. We identified 30 human-specific "fixed" amino acid substitutions in the GluR genes by analyzing 80 human samples of seven different populations worldwide. Grantham's distance analysis showed that GRIN2C and GRIN3A are the most and the second most diverged GluR genes between humans and chimpanzees. However, most of the substitutions are non-radical and are not clustered in any particular region. Protein motif analysis assigned 11 out of these 30 substitutions to functional regions. Two out of these 11 substitutions, D71G in GRIN3A and R727H in GRIN3B, caused differences in the functional assignments of these genes between humans and other apes. Conclusion We conclude that the GluR genes did not undergo drastic changes such as accelerated evolution in the human lineage after the divergence of chimpanzees. However, there remains a possibility that two human-specific "fixed" amino acid substitutions, D71G in GRIN3A and R727H in GRIN3B, are related to human-specific brain function.

remains a possibility that two human-specific "fixed" amino acid substitutions, D71G in GRIN3A and R727H in GRIN3B, are related to human-specific brain function.

Background
Glutamate is the most abundant fast-excitatory neurotransmitter in the central nervous system (CNS) and glutamate receptors (GluRs) play a vital role in the mediation of excitatory synaptic transmission. Because of their roles in neurotransmission and synaptic plasticity, GluRs are thought to be key molecules in cognitive functions such as learning and memory (reviewed in [1][2][3]). Based on their structural and functional characteristics, GluRs are classified into two major groups: ionotropic glutamate receptors (iGluRs) and metabotropic glutamate receptors (mGluRs) (Reviewed in [3]). Vertebrate iGluRs are pharmacologically classified into four subgroups by their ligand selectivity: NMDA, AMPA, kainate, and delta. While iGluRs directly regulate the ion flux across the cell membrane as ion channels, mGluRs are involved in a variety of intracellular signaling pathways by activating phospholipase C and/or suppressing adenlylate cyclase and subsequently mediating excitatory neurotransmission and synaptic plasticity by affecting iGluR activities.
Recent studies have reported that the genetic variations in GluRs are associated with multiple neurobehavioral phenotypes in humans including addictions, anxiety/dysphoria disorders, schizophrenia, and epilepsy [e.g., [4][5][6][7][8][9][10][11][12]]. These observations suggest that genetic variations in GluRs cause brain dysfunctions in humans. A study of the evolutionary genetic changes in the GluR genes would provide us with insights into the molecular basis of human-specific brain and nervous system functions.
The evolutionary changes that occurred in the GluR genes in the lineage leading to humans are poorly understood, mainly because of the limited availability of relevant information in public databases. To overcome this problem, we determined the complete coding sequences of 26 GluR genes in chimpanzees (Pan troglodytes) and conducted a comparative genomic analysis of all GluR genes. We examined whether positive selection plays a role in the evolution of the GluR gene family after the divergence of humans and chimpanzees and investigated human-specific "fixed" nonsynonymous substitutions in the GluR genes that might be associated with human-specific brain function.

The pairwise nucleotide divergence
We calculated the pairwise nucleotide divergence in total and synonymous sites between humans and chimpanzees for all of the genes encoding NMDA, AMPA, kainate, delta, and metabotropic glutamate receptors ( Table 1). The divergence for the entire set of GluR genes both at total and synonymous sites was 0.00461 and 0.01257, respectively, which is significantly lower than the genome-wide average values, which are 0.0059 and 0.0177 (more than 2 SD below the mean in a normal distribution, [14]). Indeed, except for the NMDA type, the divergence for each type was significantly lower than the genome-wide average at total and synonymous sites (more than 2 SD below the mean).
The AMPA type genes showed the lowest divergence of the GluR types at both total and synonymous sites (0.00242 ± 0.00053 and 0.00792 ± 0.00167 at total and synonymous sites, respectively). The divergence of the four individual AMPA genes ranged from 0.00075 to 0.00303 at total sites and from 0.00056 to 0.00103 at synonymous sites. Therefore, the AMPA type genes showed lower divergence than the other GluR types. The divergence for the NMDA type genes was the highest of all of the GluR types at both total and synonymous sites (0.00594 ± 0.00045 and 0.01477 ± 0.00141). This observed higher divergence can be attributed to two NMDA type GluR genes, GRIN3A and GRIN3B. The GRIN3B and GRIN3A genes are the most and the second most diverged GluR genes between humans and chimpanzees at both types of site (0.01351 and 0.00965 at total sites and 0.03023 and 0.02844 at synonymous sites). When we excluded GRIN3A and GRIN3B from this analysis, the NMDA type genes were found to have similar divergences to the other gene types at total and synonymous sites (0.00425 ± 0.00040 and 0.01154 ± 0.00135).

Lineage-specific K A /K S ratios and selection tests
Intrigued by the possible functional implications of the GluR genes in the evolutionary process of the human lineage, we examined whether positive selection acted on the GluR genes in humans and chimpanzees. At first, using macaque sequences as an outgroup, we calculated the human and chimpanzee lineage-specific nonsynonymous and synonymous substitution rates (K A and K S , respectively) and their ratios (K A /K S , Table 2). Notably, the K A / K S ratios for the GluR genes were less than one in both the human and chimpanzee lineages, although we could not calculate the ratios for 11 human and 11 chimpanzee GluR genes due to the absence of nonsynonymous substitutions. Excluding the genes with no substitutions, the K A / K S ratios ranged from 0.0004 to 0.417 and 0.0005 to 0.242 in human and chimpanzee lineages, respectively. The K A /K S ratios for 24 human GluR genes and all of the chimpanzee GluR genes are smaller than the species-specific genome-wide mean values (0.259 and 0.245 in human and chimpanzee, respectively [15]). These results indicate that the functional constraint on the GluR is relatively strong.
Two maximum likelihood ratio tests (implemented in PAML [16]) were employed to examine the K A /K S for the GluR genes in the human lineage and to evaluate the accelerated selection for the GluR genes in humans and chimpanzees. First, we compared the K A /K S ratios in the human lineage vs. background (chimpanzee and macaque) lineages. We observed that the human-specific K A /K S ratio is significantly different from the background ratio in GRM7 according to test B outlined in [17] (Additional file 1). Taking the K A /K S value into account, this result implies that the human K A /K S ratio is significantly lower than that of the background in GRM7. We then applied the improved branch-site model [18] to examine whether positive selection acted on the GluR genes in humans and chimpanzees, but could not detect any significant accelerated selection in either human or chimpanzee lineages (Additional file 2).

Identification of human-specific "fixed" nonsynonymous mutations
To evaluate the functional changes of the GluR genes in the human lineage, we searched for human-specific "fixed" nonsynonymous substitutions. First, we carried out a pairwise comparison between the human and chimpanzee orthologs of 26 GluR genes. We found a total of 80 nonsynonymous substitutions including four indels (insertion/deletion) ( Table 3). Out of the 80 substitutions, two substitutions were excluded from further analysis due to discrepancies among the UCSC reference and our chimpanzee sequences; these two substitutions are possibly polymorphic within chimpanzees. Second, we sequenced the remaining 78 substitution/indel sites in five additional apes: the bonobo, gorilla, orangutan, siamang, and crab-eating macaque. We regarded human alleles that were not shared with any of the great apes to be "human-specific". We pooled substitutions that were found specifically either in chimpanzees or in bonobos as "chimpanzee-specific" substitutions, because these mutations must have occurred in the chimpanzee and bonobo lineages after the divergence of the human lineage. Out of the 78 human-chimpanzee substitution sites, we identified 37 human-specific (35 substitutions and 2 indels) and 31 chimpanzee-specific substitutions/indels (29 sub- stitutions and 2 indels). The remaining 10 are recurrent substitutions in the primate lineages. To determine whether these substitutions/indels are "fixed" or "polymorphic" in human populations, we sequenced 80 human samples representing seven different populations around the world for the 37 human-specific substitution/ indel sites. Table 3 shows the 30 "fixed" (28 substitutions and 2 indels) and 7 "polymorphic" substitutions/indels that we confirmed in the human populations. The 30 human-specific "fixed" substitutions are potentially responsible for human-specific functions.

Functional implications of human-specific "fixed" nonsynonymous substitutions
To evaluate the functional significance of each amino acid substitution in the GluR genes, we calculated Grantham's distance [19], a measurement of the chemical drasticity of amino acid replacements. We examined the differences in amino acid substitution patterns between humans and chimpanzees using 35 human-specific and 29 chimpanzee-specific substitutions after excluding 4 indel sites. We classified amino acid substitutions into two groups: substitutions with Grantham's distances greater than 100 (the 51  GRID1  T295M  ACG-ATG  C  T  T  T  T  -T  T  Human  Fixed  52 GRID1 Abbreviations are: Hum -human, Chimp -chimpanzee, Bon -bonobo, Gor -gorilla, Ora -orangutan, Gib -gibbon, in -insertion, del --deletion, rs --RefSNP accession ID. 1 : This shows the lineage in which the substitution (insertion/deletion) occurred. In cases in which the substitution independently occurred at the same site in more than two lineages, the specificity is not assigned (N.A.). We pooled substitutions found specifically in chimpanzees or in bonobos as "chimpanzee-specific" substitutions. " ": The sequence was obtained from the UCSC genome database.
-: We failed to determine the sequence of the site, and the sequence was not available from the UCSC genome database. *: The other substitution was found in the codon. mean chemical distance from the three-property formula [19]) were classified as "radical" changes and substitutions with Grantham's distance less than 100 were classified as "non-radical" changes. We found four radical and 30 non-radical changes in the human lineage and four radical and 25 non-radical changes in the chimpanzee lineage, indicating that there are no significant differences in the amino acid substitution patterns between humans and chimpanzees (p = 1 in Fisher's exact test, 2 × 2, twotailed). We then summarized Grantham's distance for each GluR gene that contained one of the 28 human-specific "fixed" substitutions (Table 4). Among the 11 GluR genes that have human-specific "fixed" substitutions, Grantham's distance analysis showed that GRIN2C and GRIN3A are the most diverged in humans from the human-chimpanzee ancestor sequence (476 in GRIN2C and 438 in GRIN3A). Four out of the five "fixed" substitutions in GRIN2C and five out of the six "fixed" substitutions in GRIN3A are non-radical changes with Grantham's distances of less than 100. These substitutions are not located in any particular region, but instead are distributed throughout the entire gene regions (Table 4). These results imply that it is unlikely that the accumulation of non-radical substitutions in GRIN2C and GRIN3A caused their functional divergence from their ancestors.
Using the MEMSAT3 [20] and MyHits [21], we annotated the transmembrane and protein motif domains around the 30 "fixed" human-specific substitution/indel sites (Table 5). We found that ten substitutions are located within either transmembrane or protein motif domains: four in transmembrane domain sites, two in N-glycosylation sites, two in N-myristoylation sites, and two in phosphorylation sites. Out of these ten substitutions, there are two human-specific and "fixed" substitutions, D71H in GRIN3A and R27H in GRIN3B, which alter the functional assignments of their respective genes as determined by the aforementioned annotation software. D71G in GRIN3A V653I GTC-ATC Transmembrane abolishes an N-myristoylation site that is conserved in other apes and R727H in GRIN3B generates a novel phosphorylation site for protein kinase C in the human lineage. These substitutions may cause functional changes in human GluRs that contribute to human brain function.

Discussion
In this study, we examined the evolutionary changes of the glutamate receptor (GluR) genes in humans and chimpanzees. We found no gross differences in the coding regions or splice sites of the GluR genes between humans and chimpanzees. We also demonstrated that the average rate of protein evolution (i.e. the K A /K S ratio) is significantly lower in the GluR genes than the genome-wide average values for humans and chimpanzees. This pattern is consistent with previous genome-wide studies [15,22,23], indicating that strong purifying selection acts on brain-expressed genes including the GluR genes due to their strict functional constraint. There are no significant differences between humans and chimpanzees with regard to their K A /K S values ( Table 2) or substitution patterns (Table 4). These results imply that no gross functional changes occurred in either lineage after the humanchimpanzee divergence.
Dorus et al. [24] found an increase in the K A /K S ratio of genes involved in the nervous system of primates relative to rodents when housekeeping genes are treated as a control, leading to the conclusion that the primate nervous system genes have experienced accelerated evolution. Our results indicate similar K A /K S ratios between these primate nervous system genes and the GluR genes. Although our K A /K S ratio values for the GluR genes are higher than those of housekeeping genes as discussed in Dorus et al. [24], we conclude that the GluR genes have been subject to strong functional constraint rather than rapid positive selection detected as accelerated evolution for the following reasons: First, we could not detect any positive selection for the GluR genes using the statistical tests reported by Dorus et al. [24]. Although the statistical power of the tests may have been somewhat affected by the scarcity of substitutions, the improved branch-site test can detect single amino acid substitutions positively selected [18]. Second, there is no local accumulation of amino acid substitutions that could have caused the functional divergence of domains in the GluR genes. Grantham's distance analyses showed that GRIN2C and GRIN3A are the most and the second most diverged GluR genes between humans and chimpanzees. However, most of the substitutions in these genes are non-radical and are not clustered in any particular region. Third, we identified only two out of the 28 human-specific "fixed" substitutions in the assigned functional transmembrane region. This observation strongly supports severe functional constraint acting on the coding regions of the GluR genes, especially on the functionally important domains.
Niemann et al. [25] reported a common null allele of GRIN3B with no particular phenotype, indicating relaxed functional constraints on GRIN3B in the human lineage. However, we observed a low K A /K S ratio (0.2976) and a low Grantham's distance (98) for human GRIN3B. Gene loss or decay might still contribute to functional changes by modifying the genetic network. In fact, NR3B knockout mice have been reported to show highly increased social interaction with their cage mates in their home cage but moderately increased anxiety-like behavior and decreased social interaction in a novel environment [26]. The presence of NR3 in NMDA receptors has been shown to decrease Mg +2 sensitivity and Ca +2 permeability, reduce agonist-induced current responses, and give rise to a new class of excitatory glycine receptors [27]. These observations suggest that NR3 has a significant role in higher brain functions through tetrameric formation with other NR subunits. Two out of the 28 human-specific "fixed" substitutions, D71G in GRIN3A and R727H in GRIN3B, changed the functional assignments between humans and other apes, causing the loss of a myristoylation site and the gain of a phosphorylation site, respectively. Since myristoylation and phosphorylation are commonly involved in processes related to synaptic plasticity including longterm potentiation and long-term depression in glutamate receptors [28], these two substitutions possibly affect human-specific brain function by modulating NMDA receptor characteristics.

Conclusion
The results of our comparative genetic study enable us to speculate about the evolutionary changes affecting human-specific brain function that occurred in the GluR genes. We showed that strong purifying selection is the major evolutionary force in the GluR genes shared by humans and chimpanzees. We identified 30 human-specific "fixed" amino acid substitutions/indels including two amino acid substitutions that potentially alter the functional roles of their genes as candidate sites responsible for human-specific brain function. Our results are valuable for understanding the molecular basis of the brain and nervous system in humans and help us to clarify human GluR functions in in vitro and in vivo experiments.

Sequence data for human, chimpanzee and macaque
Using the longest isoform transcript as a reference (Additional file 3), we retrieved the coding sequences of 26 glutamate receptor (GluR) human and chimpanzee genes from the UCSC Genome Browser (hg18 and panTro2 for humans and chimpanzees, respectively) [12]. Since only partial genomic sequences were available for 21 chimpanzee genes, we determined the chimpanzee sequence for the 21 genes using either PCR-based direct sequencing or a BAC-based cloning-sequencing method (Additional file 3). These genomic GluR sequences were deposited in Gen-Bank (accession numbers: AB514205-AB514225). The genomic sequences of macaque homologs were also obtained from the UCSC Genome Browser (rheMac2).

PCR amplification and sequencing
The PCR primers were designed based on the alignments of human and chimpanzee GluR genes using Primer3 [30]. PCR amplification was carried out using 10 μl samples containing 1 μg of genomic DNA, PCR buffer, 2.5 mM dNTPs (Promega, Madison, WI), 0.7 units Taq DNA polymerase (Promega, Madison, WI), 25 mM MgCl 2 , and 10 μM forward and reverse primers. Primer pairs and PCR conditions are described in Additional file 5. After the PCR reaction, we treated the reaction mixtures with 1 U of Exonuclease I (New England Biolabs) and 0.1 U of SAP (Shrimp Alkaline Phosphatase; Roche Applied Science, Indianapolis, IN) to remove the primers.
All sequencing reactions were performed using 10 μl samples containing 1 μl of PCR product, 1.6 μM sequencing primer, and 0.25 μl of BigDye Terminator v1.1 or v.3.1 (Applied Biosystems, Foster City, CA). The conditions for the sequencing reaction were 96°C for 90 seconds, 50°C for 5 seconds, and 60°C for 60 seconds for 25 cycles. The sequencing products were purified by ethanol precipitation and then analyzed on an ABI 3100 or 3730 (Applied Biosystems). Mutation Surveyor v2.2 (SoftGenetics, LLC) was used to compile the electropherograms.

Data analysis
The pairwise nucleotide divergence was estimated by MEGA [31]. The Tamura and Nei model [32] and Jukes and Cantor model [33] were used for total and synonymous sites, respectively. Standard errors were computed by the bootstrap method (1000 replicates). The nonsynonymous and synonymous lineage-specific rates (K A and K S , respectively) were estimated by the modified Nei and Gojobori [34] and ML [35] methods as implemented in the codeml module of PAML [16].
The Grantham's distances for each substitution were obtained from Table in [19]. Since this distance method is only applicable to amino acid substitutions, we excluded insertion/deletion (indel) sites from this analysis. We used the mean chemical distance from the three-property formula [19], 100, for the cutoff value of two classes, radical and non-radical substitution.
Using PAML software, the likelihood ratio test was applied to examine the following two hypotheses: (1) the K A /K S ratio was significantly higher in the branch of interest than in the background branch (tests B and D in [17]) and (2) positive selection acted on the branch of interest (improved branch site model; test 2 in [18]). The tests were carried out by comparing the log-likelihood values between the null and alternate hypotheses. Bonferroni correction was applied to correct for multiple comparisons.