Mice carrying a human GLUD2 gene recapitulate aspects of human transcriptome and metabolome development

Significance

A novel version of the glutamate dehydrogenase gene, GLUD2, evolved in the common ancestors of humans and apes. Based on sequence and expression pattern, GLUD2 has been suggested to play a role in glutamate metabolism in human and ape brains. We have generated transgenic mice carrying a human GLUD2 gene. Analysis of transcriptome and metabolome changes induced by GLUD2 in the cerebral cortex revealed no changes in glutamate concentration but instead changes to metabolic pathways centering on the TCA cycle during early postnatal development. These changes mirrored differences seen between human and macaque during cortex development, suggesting that GLUD2 may play a role during brain development in apes and humans, possibly by providing precursors for the biosynthesis of lipids.

Abstract

Whereas all mammals have one glutamate dehydrogenase gene (GLUD1), humans and apes carry an additional gene (GLUD2), which encodes an enzyme with distinct biochemical properties. We inserted a bacterial artificial chromosome containing the human GLUD2 gene into mice and analyzed the resulting changes in the transcriptome and metabolome during postnatal brain development. Effects were most pronounced early postnatally, and predominantly genes involved in neuronal development were affected. Remarkably, the effects in the transgenic mice partially parallel the transcriptome and metabolome differences seen between humans and macaques analyzed. Notably, the introduction of GLUD2 did not affect glutamate levels in mice, consistent with observations in the primates. Instead, the metabolic effects of GLUD2 center on the tricarboxylic acid cycle, suggesting that GLUD2 affects carbon flux during early brain development, possibly supporting lipid biosynthesis.

Glutamate dehydrogenase (GDH) is a metabolic enzyme catalyzing the conversion of glutamate to α-ketoglutarate and ammonia (1). Whereas the ammonia is metabolized via the urea cycle, the α-ketoglutarate enters the tricarboxylic acid (TCA) cycle in mitochondria. Besides its metabolic role, glutamate also functions as a major excitatory neurotransmitter (2, 3).

Whereas most organisms contain one copy of the GLUD gene encoding the GDH enzyme, humans and apes have two: GLUD1 and an additional gene, GLUD2, which originated by retroposition of the GLUD1 transcript after the split from apes and old world monkeys (4). Sequence analysis suggests that it is highly unlikely that GLUD2 would have contained an intact ORF and a K_a/K_s <1 throughout the evolution of apes without being functional (5). Moreover, positive selection has affected GLUD2 (4), including changes in amino acid residues that make it less sensitive to low pH and GTP inhibition, and resulting in a requirement for high ADP levels for allosteric activation (4, 6, 7). GLUD2 mRNA expression levels in tissues are lower than those of GLUD1, but similarly distributed across tissues. However, whereas the ancestral version of the GLUD enzyme occurs both in mitochondria and the cytoplasm, GLUD2 is specifically targeted to mitochondria (8, 9).

Changes in GLUD2 properties have been suggested to reflect functional adaptation to the metabolism of the neurotransmitter glutamate in the brain (4, 10), and the fact that GLUD2 has been positively selected and maintained during ape and human evolution suggests that it may have physiological effects important for the function of ape and human brains. However, the connection between the emergence of the GLUD2 gene in the ancestors of apes and humans and changes in brain function remains elusive. To date, the only direct insights into GLUD2 function come from a rare GLUD2 mutation linked to the onset of Parkinson’s disease (11) and from glioma cells carrying a mutated isocitrate dehydrogenase 1 gene (IDH1) where GLUD2 expression reverses the effects of the IDH1 mutation by reactivation of the metabolic flux from glucose and glutamine to lipids by way of the TCA cycle (12).

To investigate the physiological role the GLUD2 gene may play in human and ape brains, we generated mice transgenic for a genomic region containing human GLUD2. We compared effects on gene expression and metabolism during postnatal development of the frontal cortex of the brain in these mice and their wild-type littermates, with similar data obtained from humans and rhesus macaques, the closest relative of humans and apes, which lack the GLUD2 gene.

Results

A Mouse Model of GLUD2.

We constructed transgenic mice carrying a 176-kb-long human genomic region containing the GLUD2 gene, as well as 43 kb of upstream and 131 kb of downstream DNA sequences (SI Appendix, Fig. S1). To account for effects potentially caused by stochastic insertion of GLUD2 sequence into the mouse genome, we constructed two independent transgenic lines (a and b). Anatomical, neurophysiological, and behavioral analyses of adult mice did not reveal any overt effects of the GLUD2 genotype in the two lines (13).

We assessed the effects of GLUD2 on gene expression and metabolite concentrations in the developing frontal cortex of hemizygous (GLUD2⁺) mice from both lines as well as control littermates, using RNA sequencing (RNA-seq) and capillary electrophoresis coupled with mass spectrometry (CE-MS). RNA-seq data were collected from 30 GLUD2⁺ and 29 control individuals. Metabolite concentrations were measured in 56 GLUD2⁺ and 81 control individuals. The mice varied in age from 3 d to 18 mo, with ages and sexes matched between transgenic and control groups (Fig. 1 A and B and SI Appendix, Table S1).

Fig. 1.

Sample information. (A) Schematic representation of age, sex, and transgenic line information for mouse and primate samples used for transcriptome (RNA-seq) measurements. (B) Schematic representation of age, sex, and transgenic line information for mouse and primate samples used for metabolome (CE-MS) measurements. Each symbol represents an individual sample. The colors indicate genotype/species information: orange, transgenic mice; green, control mice; red, humans; blue, rhesus macaques. Lighter shades of color indicate younger ages. The bars show the corresponding age intervals: d, days; y, years. The symbols indicate mice lines: circles, line a; triangles, line b. Filled symbols indicate males; empty symbols indicate females.

To assess whether molecular changes induced by GLUD2 in the mice might recapitulate differences between hominoids (apes and humans) and other primates, we measured metabolite concentrations in prefrontal cortex of 35 humans and 31 rhesus macaques using CE-MS. For both species, samples covered the respective lifespans approximately corresponding to the one sampled in the mice: from 16 d postnatal to 90 y in humans, and from 18.4 wk postconception to 21 y in macaques (Fig. 1B and SI Appendix, Table S2). We furthermore analyzed RNA-seq data previously collected in the prefrontal cortex of 38 humans and 40 rhesus macaques covering the same age span (14) (Fig. 1A and SI Appendix, Table S2). Between these two datasets, 13 human and 20 macaque individuals overlapped for both the RNA-seq and CE-MS data (SI Appendix, Table S2).

GLUD2 Affects Gene Expression During Early Postnatal Development.

Based on 613 million, 100 nt-long sequence reads collected from the mouse polyA-plus RNA (SI Appendix, Table S3), we detected expression of 18,610 protein-coding and noncoding genes. Global expression variation among the samples showed a clear effect of age when analyzed by principal component analysis (PCA), whereas neither genotype nor sex affected gene expression substantially (Fig. 2A and SI Appendix, Fig. S2). Consistently, as many as 15,663 of the 18,610 detected genes, including the GLUD2 transgene, showed significant age-dependent expression changes in the mouse cortex [F-test, Benjamini–Hochberg (BH) corrected P < 0.05, false discovery rate (FDR) < 0.01].

Fig. 2.

Effect of the GLUD2 genotype on gene expression in the mouse model. (A) The first two principal components (PCs) of the principal component analysis, based on the expression of 18,610 genes detected in the mouse brain. Each circle represents an individual. The colors indicate genotype information: orange, transgenic mice; green, control mice. Symbols filled by lighter shades of color indicate younger ages. (B) The mean normalized gene expression divergence between transgenic and control mice, based on the 13 age-dependent genes with expression affected by the GLUD2 genotype (red curve). The colored area shows variation of the divergence estimates obtained by bootstrapping the 13 genes 1,000 times. (C) The mean normalized gene expression divergence between transgenic and control mice, based on the remaining 15,650 age-dependent genes (light-red curve). The colored area shows variation of the divergence estimates obtained by subsampling 13 genes 1,000 times.

Both transgenic mouse lines showed the same developmental expression profile of GLUD2, which was distinct from the expression profile of GLUD1 (SI Appendix, Fig. S3). Notably, the GLUD2 expression trajectory in the developing mouse brain strongly resembled the trajectory of GLUD2 expression during prefrontal cortex development in humans (Pearson correlation, r = 0.86, P < 0.0001; SI Appendix, Fig. S4). We also detected expression of a long noncoding RNA (lncRNA) originating from the opposite strand of the transfected human genomic region upstream of the GLUD2 transcription start site (SI Appendix, Fig. S5). Expression profile of this lncRNA closely resembled the expression profile of GLUD2 (SI Appendix, Fig. S6), suggesting that it is expressed from the same bidirectional promoter as GLUD2 (SI Appendix, Table S4).

An effect of the GLUD2 genotype, albeit small, was detectable: using analysis of covariance (ANCOVA) with linear, quadratic, and cubic models, we identified 12 protein-coding genes and one lncRNA showing significant expression differences between the two transgenic lines and control littermates (F-test, BH corrected P < 0.05, permutations P < 0.05, FDR < 0.01, SI Appendix, Table S5). Strikingly, the effect of the GLUD2 genotype was observed only during early postnatal development and disappeared at approximately 2 wk of age (Fig. 2B). By contrast, other genes expressed in the mouse brain did not show any increased divergence between transgenic and control mice during early development (Fig. 2C).

Eleven of the 13 genes differentially expressed between transgenic and control mice fell into the same coexpressed module in an unsupervised hierarchical clustering analysis (Fig. 3A and SI Appendix, Table S5 and Fig. S7), an observation not expected by chance (permutations, P < 0.001) (SI Appendix, Fig. S8). The genes affected by GLUD2 are thus coexpressed during mouse development. The expression of these 11 genes decreased rapidly during early postnatal development in both transgenic and control mice, but in the transgenic mice, the decrease in expression was shifted to earlier stages of development (Fig. 3 B and C and SI Appendix, Fig. S9).

Fig. 3.

Patterns and functions of genes with expression affected by GLUD2 genotype. (A) Dendrogram based on unsupervised hierarchical clustering of 15,663 age-dependent genes. The colors represent coexpressed modules. The darker shades of color indicate larger modules. Dashed green rectangle shows module 1 containing 11 of the 13 genes with expression affected by the GLUD2 genotype. (B) Expression profiles of 3 of the 11 module 1 genes with expression affected by the GLUD2 genotype. Each point represents an individual sample. The colors indicate genotype information: orange, transgenic mice; green, control mice. The symbols indicate mice lines: circles, line a; triangles, line b. The lines show spline curves fitted to expression data points with four degrees of freedom. (C) Developmental time shift between transgenic mice and control mice, calculated based on expression of 11 module 1 genes with expression affected by the GLUD2 genotype (dark-red curve) and the remaining 15,652 age-dependent genes (light-red curve). The light-red–colored area shows variation of time-shift estimates for the remaining age-dependent genes, obtained by randomly subsampling 11 genes 1,000 times. The curves are obtained by aligning transgenic mice expression time series to that of control mice, using the modified dynamic time warping algorithm, showing the ages where transgenic mice expression levels correspond to those of control mice. During early stages of development, transgenic mice ages are mapped to older ages in control mice, indicating transgenic mice expression profiles shift to earlier developmental stages. (D) GO biological processes significantly enriched in 11 module-1 genes with expression affected by the GLUD2 genotype. Biological processes related to neural development are shown in bold.

Based on analysis of Gene Ontology (GO) terms (15), the 11 coexpressed genes were significantly enriched in several biological processes, many of them related to neural development and transcriptional regulation (hypergeometric test, FDR corrected P < 0.05) (Fig. 3D and SI Appendix, Table S6), a result robust to the use of different background gene distributions. The genes affected by GLUD2 may thus to some extent be functionally related.

GLUD2 Effects in the Mice Mirror Differences Between Primates.

To assess whether the effect of GLUD2 in the mice recapitulated differences between primate species that lack and that have GLUD2, we reanalyzed 579 million, 100-nt-long RNA-seq reads from the prefrontal cortex of rhesus macaques and humans. We detected the expression of 23,115 protein-coding and noncoding genes. PCA analysis of the expression of these genes demonstrated substantial species-dependent as well as age-dependent divergence (Fig. 4A).

Fig. 4.

Gene expression in primates. (A) The first and third PCs of the principal component analysis, based on the expression of 23,115 genes detected in human and rhesus macaque prefrontal cortex. Each circle represents an individual (red, humans; blue, macaques). Lighter shades of color indicate younger ages. (B) The mean normalized gene expression divergence between humans and macaques based on the 9 genes with expression affected by GLUD2 genotype in the mouse model (red curve). The colored area shows variation of the divergence estimates obtained by bootstrapping the 9 genes 1,000 times. “H-birth” x-axis marks represent human birth age. (C) The mean normalized gene expression divergence between humans and macaques based on the remaining 23,106 detected genes (light-red curve). The colored area shows variation of the divergence estimates obtained by subsampling 9 genes 1,000 times. (D) Expression profiles of E2F transcription factor 7 (E2F7), SOX11, and SOX4, which are among the 7 genes showing the similar expression and divergence pattern in primates and the transgenic mice. Each point represents an individual. The colors indicate species information: red, humans; blue, macaques. Filled symbols indicate postnatal ages; empty points, prenatal ages. The vertical dashed line shows the human birth age.

Among the 13 genes differentially expressed between transgenic and control mice, 9 were expressed in human and macaque prefrontal cortex (SI Appendix, Table S5). Remarkably, the expression of these 9 genes showed a large divergence between humans and macaques at the earliest stages of postnatal development after which the divergence rapidly decreased. Thus, the expression differences between humans and macaques resembled those seen between mice carrying a GLUD2 gene and control littermates (Fig. 4B). By contrast, other genes expressed in human and macaque prefrontal cortex showed no such divergence pattern (permutation P = 0.05, Fig. 4C).

Furthermore, seven of the nine genes showed a rapid expression level decrease during human and macaque early postnatal development, i.e., the same developmental expression trajectory as the genes affected by GLUD2 in the transgenic mice (Pearson correlation, r > 0.9, permutation P < 0.001, SI Appendix, Figs. S10 and S11). After adjusting for the differences in lifespan between humans and macaques, we found a shift in the timing of the expression to earlier developmental stages in humans relative to macaques (Fig. 4D and SI Appendix, Fig. S10), similar to what is seen in the GLUD2 mice. This was not caused by any general mismatching of human and macaque developmental timing, as other genes did not show such a coordinated shift to earlier developmental stages (Pearson correlation, r > 0.9, permutation P = 0.02, SI Appendix, Fig. S12). Thus, gene expression changes in GLUD2 mice indeed recapitulate gene expression differences between human and macaque expression profiles during early postnatal development.

Among the 7 genes showing similar expression and developmental shifts in the transgenic mice and in human cortex, 3 are transcription factors (TFs) implicated in neural development: SRY (sex determining region Y)-box 4 (SOX4), SRY (sex determining region Y)-box 11 (SOX11), and Distal-less homeobox 2 (DLX2) (16–22). A total of 15 known neural-related target genes of these TFs (16–29) were expressed in the mice and 14 in the primates (SI Appendix, Table S7). The expression divergence profiles of these target genes and the corresponding TFs between the transgenic and control mice and between humans and macaques at different ages were significantly correlated (Pearson correlation, r > 0.5; permutation P = 0.008 for mice and P = 0.001 for primates, SI Appendix, Fig. S13), suggesting that the effects of GLUD2 may go beyond the 13 genes showing differences between the transgenic and control mice.

As the effects of GLUD2 in the mice were strongest at the earliest stages of postnatal development, it is possible that GLUD2 plays a role mainly in prenatal development. To test if this may be the case in humans, we analyzed the expression of the 13 genes affected by GLUD2 expression in the mice using RNA-seq data from prenatal prefrontal cortex development of humans (30) and macaques. Of the 13 genes, 9 had detectable expression in our human and macaque data, as well as in public fetal human cortex data. Strikingly, expression of the 9 genes was substantially more divergent between humans and macaques before than after birth (Fig. 5 A and B and SI Appendix, Fig. S14), whereas other expressed genes showed no obvious increase in expression differences before birth (Fig. 5C).

Fig. 5.

Prenatal gene expression. (A) Prenatal expression profiles of E2F7, SOX11, and SOX4 genes, which are among the 7 genes showing similar expression and divergence patterns in primates and the transgenic mice. Each point represents an individual. The colors indicate species information: red, humans; blue, macaques. Filled symbols indicate postnatal ages; empty points, prenatal ages (17 in humans and 4 in macaques). The vertical dashed line shows the human birth age; pcw, weeks postconception. (B) The mean normalized gene expression divergence between humans and macaques based on the 9 genes with expression affected by the GLUD2 genotype, calculated using a published fetal human dataset (red curve). The colored area shows variation of the divergence estimates obtained by bootstrapping the 9 genes 1,000 times. (C) The mean normalized gene expression divergence between humans and macaques, based on the remaining 15,897 detected genes in the public fetal human and our macaque time series data (light-red curve). The colored area shows variation of the divergence estimates obtained by subsampling 9 genes 1,000 times.

Brain Metabolism in GLUD2 Mice and Primates.

We assessed the effects of GLUD2 on metabolism in the brains of the GLUD2 mice using CE-MS. For comparisons, we similarly analyzed human and macaque brains. We detected and quantified 110 and 88 metabolites in the mice and in humans and macaques, respectively (SI Appendix, Tables S8 and S9). PCA analysis based on concentration levels of these metabolites revealed a substantial effect of age in both mice and primates (Fig. 6 A and B).

Fig. 6.

Metabolite differences and glutamate concentration among genotypes/species. (A and B) The first two PCs of the principal component analysis, based on the concentrations of 110 metabolites detected in mouse frontal cortex and 88 metabolites detected in human and macaque prefrontal cortex. Each circle represents an individual. The colors indicate genotype/species information: orange, transgenic mice; green, control mice; red, humans; blue, rhesus macaques. Lighter shades of color indicate younger ages. (C) Glutamate concentration profiles measured in mice using CE-MS and in primates using CE-MS and GC-MS. The colors indicate genotype/species information as in A and B; purple, chimpanzees. Lines show spline curves fitted to concentration data points with four degrees of freedom.

We first focused on the concentration levels of glutamate, the direct substrate of the GDH2 enzyme encoded by GLUD2. We detected no effect of GLUD2 on glutamate concentration in the mouse cortex (Fig. 6C). By contrast, glutamate concentrations in the prefrontal cortex of the humans are substantially lower than in the macaques (Fig. 6C). Previous work using gas chromatography coupled with mass spectrometry (GC-MS) has shown that compared with humans, chimpanzees as well as macaques have higher glutamate levels in the brain (31), something that reexamination of the published glutamate data confirms (Fig. 6C). Because both humans and chimpanzees carry GLUD2 genes, a mechanism other than the mere presence of GLUD2 must be responsible for the lower glutamate concentrations in the human brain.

Despite the absence of a GLUD2 effect on glutamate levels in the mice, we detected that the overall differences between the metabolomes of the transgenic and the control mice were about threefold larger shortly after birth relative to 2 mo after birth (SI Appendix, Fig. S15). To test if this may be at least partially connected to gene expression differences, we compared the concentration profiles of 24 metabolites associated with the metabolic pathways where the 13 genes differentially expressed between transgenic and control mice are located (SI Appendix, Table S8). Concentration differences of the 24 metabolites were substantially larger than the differences for the 86 remaining compounds analyzed (Fig. 7A). Similar results were obtained using the Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway annotation and the Small Molecule Pathway Database (32) to identify metabolites and pathways associated with the 13 genes (SI Appendix, Fig. S16).

Fig. 7.

Metabolome analyses of transgenic mice and primates. (A) The mean normalized metabolite concentration divergence between transgenic and control mice, based on the 24 metabolites sharing the same KEGG pathway as the 13 genes with expression affected by the GLUD2 genotype (dark-blue curve) and the remaining 86 detected metabolites (light-blue curve). The colored area shows variation of the divergence estimates of the 86 metabolites obtained by bootstrapping 1,000 times. (B) The mean normalized metabolite concentration divergence between humans and macaques based on 11 of the 24 metabolites linked to the GLUD2 genotype effect in mice (dark-blue curve) and the remaining 56 detected metabolites (light-blue curve). The colored area shows variation of the divergence estimates of the 56 metabolites obtained by bootstrapping 1,000 times.

The 13 genes and 24 metabolites showing increased divergence between the transgenic and control mice converged within seven KEGG pathways (SI Appendix, Table S10). For three of these pathways, the HIF-1 signaling pathway, the pentose phosphate pathway, and carbon metabolism, metabolite concentration differences were significantly greater than expected by chance (permutations, P < 0.05) (Fig. 8A). The three pathways are closely related as the “carbon metabolism” pathway includes the pentose phosphate pathway and the TCA cycle, which is in turn linked to the HIF-1 signaling pathway (Fig. 8B).

Fig. 8.

Pathway analysis of the GLUD2 genotype effect on mouse metabolome. (A) Metabolite concentration divergence between transgenic and control mice, based on metabolites in a pathway (red circles). The boxplots show metabolite concentration divergence between transgenic and control mice, calculated by sampling 1,000 times the same number of metabolites as detected in a given pathway from the bulk of remaining detected metabolites. Each pathway contains at least two detected metabolites and at least one gene differentially expressed between transgenic and control mice. The filled circles show the pathways with significantly greater metabolic divergence than expected by chance. (B) The schematic representation of the three KEGG pathways showing significantly greater metabolite concentration divergence between transgenic and control mice: HIF-1 signaling pathway, pentose phosphate pathway, and carbon metabolism. Detected metabolites are shown in blue and genes with expression affected by the GLUD2 genotype are shown in red. Dashed pink rectangle delineates the pathway affected by GLUD2 overexpression in IDH1-mutant glioma cells.

We next analyzed metabolic differences between human and macaque brains. As the human samples were frozen after substantial postmortem delay (PMD), we first analyzed the effects of PMD by comparing three macaque samples that were intentionally collected with substantial PMD with macaque samples that were not (SI Appendix, Table S11). Of the 88 metabolites detected in the macaque and human brains, 21 were affected by PMD in at least one of the three samples analyzed (SI Appendix, Fig. S17 and Table S9) and were therefore excluded from further analyses.

For the remaining 67 metabolites, the differences between humans and macaques explained 26% of the total variation in metabolite concentrations. By comparison, the GLUD2 genotype explains 2% of the metabolic variation in the mice (SI Appendix, Fig. S18). Despite the much greater overall metabolic differences between macaques and humans than between the transgenic and control mice, it is notable that in both cases, the greatest metabolic differences are seen in early development (SI Appendix, Figs. S15 and S19).

Of the 24 metabolites linked to the genes differentially expressed between GLUD2 and control mice, 11 were among the 67 metabolites that were detected and unaffected by PMD in humans and rhesus macaques (SI Appendix, Table S9). As in the mice, these 11 metabolites diverged more between humans and macaques than the other detected metabolites in early development (Fig. 7B). However, this effect was less pronounced in development and also apparent at later age, possibly due to fewer metabolites detected in the primates than in the mice, as well as differences in lifespan. Still, for 8 of the 11 metabolites detected in both primates and mice, the direction of concentration change coincided between humans and transgenic mice (permutation, P = 0.08) (SI Appendix, Fig. S20). Thus, GLUD2 transgenic mice recapitulate some of the metabolic differences seen between human and macaque brains.

Discussion

GLUD2 originated as an evolutionary novel version of the glutamate dehydrogenase gene as a result of the retroposition of a GLUD1 transcript in the common ancestors of humans and apes. We investigated the function of GLUD2 by inserting the human GLUD2 gene and surrounding sequences carrying putative regulatory elements into the mouse genome. We isolated two independent transgenic lines to exclude artifacts caused by the insertion of the human gene in the mouse genome. When analyzed in the frontal cortex, both lines displayed similar developmental GLUD2 expression trajectories, closely resembling GLUD2 expression in human prefrontal cortex during development.

In both transgenic lines, GLUD2 affects the expression of 13 genes, 11 of which show similar ontogenetic expression profiles. These genes include several TFs, among which DLX2, SOX4, and SOX11 play important roles in neuronal differentiation and neurogenesis (16, 18, 19, 25). Some of the previously described targets of these TFs show changes in expression in the transgenic mice that are correlated with expression changes of the corresponding TFs. Nine primate orthologs of the 13 mouse genes affected by GLUD2, including DLX2, SOX4, and SOX11, differ in expression between human and macaque ontogenesis in ways that mirror the changes seen in the mice. Similarly, primate orthologs of target genes of DLX2, SOX4, and SOX11 show expression differences correlating with those of these TFs. These results illustrate that the introduction of GLUD2 into the mouse genome induces effects paralleling evolutionary differences between primate species that carry a GLUD2 gene and those that do not. This adds to a mounting amount of evidence suggesting that human-specific variants of genes such as FOXP2, ASPM, EDAR, SRGAP2, and CMAH can be fruitfully studied in mouse models (33).

The gene expression changes induced by GLUD2 in the mice and the differences between humans and macaques were restricted to early development and were not observed past the first 2 wk postpartum in the mice or the first 2 y of life in humans. This suggests that GLUD2 mainly functions during brain growth and early development, a notion that is in apparent contradiction to the idea that GLUD2 has a major role in the metabolism of the neurotransmitter glutamate.

The metabolic data lend support to the notion that GLUD2 mainly influences aspects of metabolism different from glutamate recycling. We detect no effects on glutamate concentrations in the mouse frontal cortex and our previous results show that glutamate concentrations do not differ between macaques and chimpanzees throughout prefrontal cortex developmental and adult stages, even though chimpanzees carry a GLUD2 gene. Instead, metabolic differences between GLUD2⁺ and control mice center on the metabolic pathways surrounding the TCA cycle. In agreement with this, it has been reported that overexpression of human GLUD2, but not GLUD1, in mutant murine glioma progenitor cells results in shunting of carbon into lipid biosynthesis via the TCA cycle (12). Our data further show that metabolic differences between GLUD2⁺ and control mice, as well as between humans and macaques, are particularly prevalent at early developmental stages characterized by rapid brain growth. Given that lipids comprise more than 50% of the brain’s dry weight (34), we speculate that GLUD2 may support the rapid growth of the large ape and human brains by enhancing lipid biosynthesis.

Materials and Methods

A detailed description of materials and methods is provided in SI Appendix. Briefly, a human bacterial artificial chromosome containing the GLUD2 gene (RP11-610G22) was linearized by NotI and injected into the male pronucleus of C57BL/6 mice to construct GLUD2^+/− transgenic mice. Transgenic animals were identified by PCRs targeting the 5′-end, the coding region, and the 3′-end of the gene. For metabolome analysis, CE-MS measurements were conducted in frontal cortex samples of 81 control and 56 transgenic mice and prefrontal cortex samples of 35 humans and 34 rhesus macaques. Transcriptome analysis was conducted in the frontal cortex of 29 control and 30 transgenic mice used for metabolite profiling on the Illumina platform, as well as in prefrontal cortex of 38 humans and 40 rhesus macaques (14). Written consent for the use of human tissues for research was obtained from all donors or their next of kin. Use of human autopsy tissue is considered nonhuman subject research and is institutional review board exempt under NIH guidelines. Biomedical Research Ethics Committee of Shanghai Institutes for Biological Sciences completed the review of the use and care of the animals in the research project (approval ID: ER-SIBS-260802P).