Characterization of a Novel Mutation in SLC1A1 Associated with Schizophrenia

Schizophrenia is a chronic and debilitating disorder with a complex behavioral and cognitive phenotype thought to reflect a similarly complex etiopathogenesis involving interactions between various susceptibility genes and environmental factors [1]. Glutamatergic systems have been strongly implicated in the pathophysiology of schizophrenia [2]. Although initially focused on disturbances in neurotransmission at N-methyl-D-aspartate (NMDA)-type glutamate receptors, the ‘glutamate hypothesis' has since been expanded to include a potential role for dysfunction at other types of glutamate receptors, as well as other glutamate-related entities, such as transporters [3].

A growing amount of experimental, epidemiological and clinical evidence also implicates neuroinflammation as a possible key element in the pathophysiology of schizophrenia [4,5]. This emerging school of thought postulates that increased risk of psychosis occurs as a result of elevated proinflammatory molecule production following exposure to oxidative stress, infectious agents, toxins or even acute psychological stressors [6]. In support of this idea, several cytokines have been reported to be elevated in plasma from schizophrenic subjects [7,8], and neuroimaging studies have shown active inflammation in the brains of patients with psychosis [9]. Conversely, subjects with psychosis exhibit reduced symptoms when administered anti-inflammatory agents [10,11]. Unifying these findings, Steullet et al. [12] have recently proposed that a tightly-regulated balance exists between redox state, neuroinflammatory processes and glutamate systems, forming a ‘central hub' where imbalances can lead to the pathophysiological changes seen in the brains of subjects with schizophrenia.

We recently discovered a hemi-deletion at the SLC1A1 glutamate transporter gene, co-segregating with psychosis in multiple members of a 5-generation family from the Pacific island of Palau [13]. This finding led to the designation of a new susceptibility locus - SCZD18. SLC1A1 encodes the excitatory amino acid transporter 3 (EAAT3), a member of the neuronal high-affinity glutamate transporter family. Along with other EAATs, this protein helps terminate the postsynaptic action of glutamate and maintain extracellular glutamate concentrations below neurotoxic levels [14]. Consequently, EAAT3 plays a major role in regulating glutamate-mediated neuroplasticity [15]. In addition to these functions, EAAT3 is important for the synthesis of intracellular glutathione (GSH) and subsequent protection of neurons from oxidative stress [16,17]. Thus, if the hemi-deletion in SLC1A1 we identified acts as loss-of-function allele, it may increase the risk of psychosis through alterations in glutamate transport or impairment of cellular responses to oxidative stress or inflammation, suggesting that SLC1A1 may form a point of convergence for several lines of evidence in the field.

The specific hemi-deletion we found is an 84-kb copy number variation (CNV) in SLC1A1 that eliminates the entire promoter and first exon, including the first 59 amino acids of the EAAT3 protein, but preserves nearly all of intron 1 and the remaining structure of the SLC1A1 gene. Thus, the deletion impacts, at a minimum, the first transmembrane domain of the protein, which helps form the Na²⁺/dicarboxylate symporter domain that is critically involved in glutamate transport and also appears essential for trimerization of EAAT3 into functional complexes. Since our initial report, we have examined the available data from other groups that reported CNVs affecting SLC1A1 in subjects with schizophrenia and related psychiatric disorders. Stewart et al. [18 ]found a 135-kb hemi-deletion in the promoter and exon 1 of SLC1A1 in a single subject with both schizophrenia and epilepsy following a screen of 235 individuals with dual diagnoses of these disorders. Horiuchi et al. [19] found a 10-kb CNV in the first intron of SLC1A1 in a case-control study of 1,920 Japanese schizophrenic patients and 1,920 Japanese control subjects. Although the CNV failed to demonstrate association with schizophrenia in their sample, evidence for association with nearby single nucleotide polymorphisms (SNPs) was detected as was association of these SNPs with the expression level of SLC1A1 in postmortem samples of the prefrontal cortex. More recently, Priebe et al. [20] found an exonic 95,102-bp deletion in SLC1A1, spanning the first 5 exons in a single schizophrenic subject from a study of 1,637 cases and 1,627 controls. Costain et al. [21,22] reported the presence of a 487-kb deletion affecting all but the first exon of SLC1A1 in a single case (out of 459 unrelated adults with schizophrenia) and subsequently found the same CNV in a sibling of that proband, who exhibited mood and/or anxiety disorder, obsessive compulsive symptoms and learning deficits. Lastly, Rees et al. [23] reported exonic CNVs affecting SLC1A1 in a total of 10 schizophrenics (out of 21,450 cases) and 2 controls (out of 26,529) in different subpopulations of European Caucasians and African-American subjects, in a study that screened data from the International Schizophrenia Consortium (ISC). Taken together, the combined genetic evidence clearly suggests that rare CNVs of large effect involving SLC1A1 may help explain a small proportion of the incidence of schizophrenia across multiple populations. Moreover, it also appears that the greatest degree of overlap among the microdeletion cases occurs in the 5′ region of the gene. Each of these 5′ CNVs, including the one we observed in the Palau pedigree, could be predicted to result in haploinsufficiency or reduced SLC1A1 expression.

Complementing the human genetic data, there have also been consistent reports of decreased SLC1A1 expression in the striatum of idiopathic schizophrenic subjects compared to controls [24,25]. Moreover, earlier studies of the SLC1A1-knockout mouse also lend support to its importance in normal brain function. These mice exhibited cortical thinning, ventricular enlargement, reduced size of the CA1 field of the hippocampus and the corpus callosum, along with some behavioral changes and cognitive impairments [16,26,27]. In addition, neurons within the hippocampus and cerebral cortex of SLC1A1-knockout mice were found to have lower GSH content, increased oxidant levels and increased susceptibility to oxidant injury [16,26,28]. Interestingly, many of these changes were reversed by treating the knockout mice with N-acetylcysteine, a membrane-permeable precursor for cysteine which is required for GSH synthesis [16,26] and has recently been shown to improve psychotic symptoms in human subjects [12,29]. Importantly, however, although many of these features are consistent with findings in schizophrenia, the SLC1A1-knockout mouse studies to date have not specifically examined phenotypes considered the most relevant for models of this disorder. Moreover, no data have been published on the heterozygous mice which would more closely mimic the human subjects identified to date that have a single putative loss-of-function allele.

Given the well-established functional roles that SLC1A1/EAAT3 plays in neuronal function, there are compelling reasons to determine the potential importance of SLC1A1 CNVs in the risk for schizophrenia and psychosis. As a first step toward characterizing the biological consequences of the hemi-deletion we discovered, the present report examines four key questions: (1) In the absence of the native promoter and exon 1, can intron 1 function as a surrogate or alternative promoter for the remaining truncated portion of the gene? (2) Does a 5′-truncated version of SLC1A1 retain its normal cellular localization and glutamate transporter function? (3) Does the 5′-truncation of SLC1A1 affect the expression of any other glutamate-related genes that are strongly implicated in the pathophysiology of schizophrenia? (4) Does knockdown of SLC1A1 expression or overexpression of the 5′-truncated or wild-type (WT) protein impact the expression of other schizophrenia, neurotransmission and stress-related genes?

Methods of ascertainment and clinical assessment have been described in detail in prior publications [30]. All human subject protocols and procedures were approved by institutional review boards in the US and the Republic of Palau. All individuals in this study were members of the same pedigree reported in our previous study (pedigree 3,501) and provided written informed consent or assent to participate after receiving a full explanation of the study in both English and Palauan. A total of 21 family members including 12 subjects without the deletion (who were all unaffected) and 9 subjects with the deletion (5 of whom were affected) participated (online suppl. table S1; for all online suppl. material, see www.karger.com/doi/10.1159/000433599).

Dual Luciferase Assay

To determine if the sequence upstream of exon 2 was capable of promoting transcription of a 5′-truncated mRNA, we cloned 1 kb of sequence upstream of exon 1 or exon 2 into PGL4 Firefly luciferase vector (Promega) without any reporter gene promoter or enhancer elements and with minimal transcription factor (TF) binding sites. HEK293 cells were then co-transfected with plasmid containing either exon 1 or exon 2 promoter (upstream) insert as well as the Renilla luciferase vector as an internal control (at a ratio of 10:1 to reduce the trans effects between promoters on co-transfected plasmids and to provide low-level, constitutive expression of Renilla luciferase). A total of 10 wells of a 96-well plate for each promoter construct (20 wells in total) were used for these comparisons. After 24 h, a dual luciferase assay (Promega) was performed using a Synergy 2 plate reader (BioTek) to measure the activities of firefly and Renilla luciferases sequentially from each well. Comparisons of the relative levels of promoter activity were made using an unpaired Student's t test.

TF Profiling Array

To gain additional insight into the possible differences in regulation of the full-length WT SLC1A1 gene compared to the 5′-truncated construct lacking the native promoter and exon 1, we performed a competitive promoter binding assay (Promoter Binding TF Profiling Plate Array II; Signosis; Cat. No. FA-2002) according to the manufacturer's protocol. Briefly, 1 μM of the same sequences used in the dual luciferase assays (containing 1 kb of exon 1 or exon 2 promoter sequence) were hybridized in a solution containing a pre-made mixture of biotinylated TF probes, along with a total of 30 μg of nuclear extracts (as sources of unlabeled TFs) that we prepared from human neural stem cells (AppliedStemCell), or purchased for human liver and human heart (Active Motif). In this assay, if our SLC1A1 exon 1 or exon 2 promoter regions contain specific TF DNA binding sites, they will compete with the biotinylated TF probes provided in the kit resulting in reduced chemiluminescence at that location due to competitive inhibition. As a result, greater affinity of the SLC1A1 promoter region for a TF results in lower signal. The pattern of TF probe signal changes was measured using a Synergy2 plate reader, and relative comparisons were made between the native exon 1 promoter and putative exon 2 promoter constructs. Because these experiments were only performed once, we focused only on those TFs which showed changes of ±3-fold or greater (approximately the top 10% of the findings).

Confocal Microscopy

To examine whether full-length or 5′-truncated SLC1A1 proteins showed evidence of altered localization within cells, we prepared HEK293 cells for confocal imaging analysis following transfection with 1 μg of GFP-tagged full-length or 5′-truncated SLC1A1 construct. The full-length construct contained the sequence-verified WT human SLC1A1 ORF (1,575 bp) subcloned into pCMV6-AC-GFP vector (Origene; PS100040). The 5′-truncated construct used the same vector but contained a sequence-verified SLC1A1 ORF with the first 91 bp removed. The HEK293 cells receiving these constructs were transfected for 6 h using Lipofectamine 2000 (Invitrogen), maintained for 24 h, stained with the cell membrane dye CellMask Orange (Life Technologies) and fixed in 4% paraformaldehyde. Images were taken using a ×40 oil immersion lens on a Zeiss LSM 510 confocal microscope.

Electrophysiological Recordings and Western Immunoblot

To examine whether the 5′-truncated SLC1A1 protein can still function as a glutamate transporter, we performed in vitro transcription (using the T7 MaxiScript kit, Ambion) to generate capped mRNAs using the full-length and 5′-truncated cDNA constructs described above, except for the absence of the GFP tags. The RNAs were coded in a blind fashion and submitted to Ecocyte Biosciences for electrophysiological recording. Using a Robocyte automated injection system, oocytes isolated from adult Xenopus laevis were injected with one of the following: (1) water, (2) full-length SLC1A1 mRNA (200 ng/μl), (3) 5′-truncated SLC1A1 mRNA (200 ng/μl), or (4) a 50-50 mixture of full-length/5′-truncated SLC1A1 mRNA (100 ng/μl each). A total of 9-21 oocytes were used for each treatment. Glutamate transporter activity was measured after 3-5 days of incubation in Barth's solution. For these assays, oocytes were clamped to a holding potential of -60 mV and 2 mM glutamate or frog Ringer's solution (as a control) was bath applied at 3-4 ml/min. Glutamate-induced inward currents were recorded at a sampling rate of 1,000 Hz at room temperature. Two-electrode voltage-clamp recordings were performed at a holding potential of -60 mV or with 20-mV step voltage increments from -120 to 80 mV. Drug-induced current amplitudes were calculated with respect to the holding current. Results are given as means ± SEM. At the conclusion of the recording, the oocytes from each preparation were rapidly frozen and returned to our lab for confirmation of protein expression using Western immunoblot. Total protein was extracted from the cell pellets using standard conditions and transferred to PVDF membranes after separation using SDS-PAGE on two 10% gels. The EAAT3 protein was detected using rabbit antibody (Sigma-Aldrich) at 1:500 dilution. GAPDH was used as a loading control, and mouse monoclonal antibody to this protein was used at 1:1,000 dilution. The antibodies were visualized by chemiluminescence after incubation with HRP-conjugated antibody to rabbit IgG (Promega). Optical densities of the protein bands were measured using the ImageJ software. The protein band densities for EAAT3 were normalized to the densities of the GAPDH bands from the same samples in the second immunoblot.

Measurement of GABA and Glutamate Gene Expression by PCR Array on Human Peripheral Blood RNA

To examine the consequences of the hemi-deletion on gene expression, we compared mRNA levels in CNV carriers and noncarriers, as well as affected and unaffected subjects, in peripheral blood leukocyte (PBL) RNA obtained from subjects in the K3501 pedigree. For these studies, peripheral venous blood was collected into PaxGene RNA tubes (Qiagen) and shipped by courier to the US at ambient temperature. Total RNA was purified according to the manufacturer's protocol using the PAXgene Blood RNA Kit (Qiagen). Purified RNA samples were subjected to quality control using the Agilent Bioanalyzer RNA 6000 Nano Kit.

Because of the typically low level of expression of neurotransmission genes in leukocytes, we compared the expression level of 84 genes of interest (online suppl. table S2) using a human GABA and glutamate RT² Profiler PCR Array (Qiagen; PAHS-152ZG-4), which uses gene-specific reverse transcription to enhance the detectability of such transcripts during PCR. A total of 19 RNA samples were used for this assay, including 5 affected deletion carriers, 3 unaffected deletion carriers and 11 noncarrier controls from the K3501 pedigree. The total RNA was prepared according to the manufacturer's protocol. Thermal cycling was performed using a LightCycler 480 (Roche), and data were normalized according to the ΔΔC_t method, with the arithmetic mean of two reference genes used for standardization (HPRT1, RPLP). Resulting data were log2 scaled, and comparisons were made between deletion carriers (n = 8) and noncarriers (n = 11) in the pedigree and between deletion carriers with psychosis (n = 5) and deletion carriers without psychosis (n = 3) in the pedigree using an unpaired Student's t test without correction for multiple testing. Results were visualized in a volcano plot.

Examination of the Effects of Knockdown of SLC1A1 or Overexpression of WT and Truncated SLC1A1

Because of our interest in knowing how the expression level and the presence of a truncated SLC1A1 transcript might affect genes involved in brain function, we next examined the effects of shRNA-mediated knockdown of SLC1A1 or overexpression of full-length and 5′-truncated SLC1A1 in HEK293 cells. Four replicate wells in a 24-well plate were used for each experimental condition, each containing 150,000 cells at 70% confluence. For the overexpression studies, we transfected HEK293 cells with 1 μg of the same constructs used in the confocal imaging analysis using Lipofectamine 2000, as well as an empty cDNA vector as a negative control. For the knockdown studies, identical preparations of HEK293 cells were transfected using Lipofectamine 2000 with 0.7 μg of a cocktail of three validated human MISSION SLC1A1 shRNAs (Sigma-Aldrich; product No. SHCLNG-NM_004170) or 0.7 μg of pLKO1-puro nontarget shRNA (Sigma-Aldrich) or blank shRNA as a negative control. All cells were maintained for 48 h after treatment. For qualitative confirmation of the overexpression and knockdown of SLC1A1 in these cells, we also co-transfected an additional set of identically treated HEK293 samples with a mixture of both GFP-tagged full-length SLC1A1 cDNA and the shRNA cocktail or control construct and measured the SLC1A1 mRNA levels and fluorescence of these wells after 48 h.

After 48 h of treatment, the RNA from each well was isolated using the RNeasy Mini Kit (Qiagen). We first performed quantitative confirmation of overexpression and knockdown of SLC1A1 using real-time quantitative RT-PCR in triplicate reactions with a custom-designed primer pair that amplified a 102-bp sequence in exon 10 of the gene (left primer: ATTCGTGTTACCCGTTGGTG; right primer: CCCAAGTCCAGGTCATTCAA). Thermal cycling was performed using 5 μl of cDNA from each sample, with 0.5 μl of each forward and reverse primer in a standard qPCR reaction in 20-μl volumes using LightCycler 480 SybrGreen I Master Mix (Roche) and the following conditions: 95°C for 5 min, then 40 cycles at 95°C for 10 s, 59°C for 15 s and 72°C for 10 s. Quantification was performed using the ΔΔC_t method, with RPLP1 used as the reference gene. Resulting data were log2 scaled and comparisons made between full-length and 5′-truncated cDNA treated cells and their control, as well as between the shRNA treated cells and their control using unpaired Student's t tests.

After confirmation of the desired effect on SLC1A1 expression in the various cell treatments, the same HEK293 RNA samples were subjected to a more comprehensive expression profiling using a custom gene expression panel of 375 genes related to schizophrenia, neurotransmission and immune function (online suppl. table S3). These genes were selected based on their public annotations and relevance to neuropsychiatric disorders, as judged from the results of large genome-wide association and CNV studies, meta-analyses and expression studies.These genes were interrogated using custom-designed targeted RNA-Seq (TREx) assay (Illumina) according to the manufacturer's protocol with an Illumina MiSeq system. Only genes with at least 5 read counts in 5 or more of the samples were retained for analysis (n = 240). The resulting read counts for these genes were normalized within each sample to the geometric mean of two reference genes (PPIA and RPLP), subjected to cube-root transformation, then mean-centered and range-scaled (divided by the range of values for that gene) across the samples. To simplify the statistical comparisons, we combined the data for the 3 control groups (12 samples from 4 blank cDNA, 4 blank shRNA and 4 nontargeting shRNA preparations) into a single control group. A 1 × 4 ANOVA was then performed to compare the effects of knockdown of SLC1A1 or overexpression of the full-length or 5′-truncated cDNA on the expression of the 240 genes in the TREx assay. False-discovery rate (FDR) correction was applied to the main effect p values, with a threshold of q < 0.15. Post hoc contrasts were obtained from Tukey's honestly significant difference. Genes with significant changes were subjected to hierarchical cluster analysis to help discern patterns in the data. Due to the number of control samples (n = 12), replicates within each control group were averaged to single data point for clustering.

Using a dual luciferase assay, we compared the transcriptional activities of 1 kb upstream of exon 1 and exon 2 of SLC1A1. The alternative exon 2 promoter activity was significantly increased more than 7-fold (p < 5.7E-8) compared to the native promoter (fig. 1). These data strongly suggest that a 5′-truncated form of EAAT3 could be expressed as a result of an alternative start site introduced by the deletion of the native promoter and exon 1.

Based on the dual luciferase assay results, we next sought to determine possible TFs that might contribute to increased expression of a truncated protein. This was performed using the Promoter Binding TF Profiling Plate Array II. Our results indicated the presence of several putative TFs that showed increased binding affinity (NFAT, SATB1, Myb, HNF-1, all >3-fold increased), along with others that showed decreased binding affinity [vitamin D receptor (VDR), TFEB, SOX18, SOX9, TFE3 and USF-1, all >3-fold decreased] for the alternative promoter compared to the native promoter (fig. 2; table 1). These observations strongly suggest that the expression of the full-length and truncated isoforms is differentially regulated by a number of potentially important TFs.

Previous immunocytochemical studies have shown that EAAT3 can be localized both intracellularly and on the cell membrane of neurons [31,32]. We used confocal microscopy to determine if the full-length and truncated isoforms of the protein were differentially localized using C-terminal GFP-tagged constructs. High-resolution imaging demonstrated that most of the full-length EAAT3 was found on the cell membrane, whereas most of the truncated EAAT3 was found intracellularly (fig. 3). These differences were present despite generally similar levels of expression, although some cells expressing the truncated EAAT3 clearly had reduced fluorescence. Overall, our results strongly suggest different patterns of localization for these two proteins.

We next sought to determine if the 5′-truncated EAAT3 that would be expressed from an intron 1 promoter was functional as a glutamate transporter. For this purpose, glutamate-induced inward currents in X. laevis oocytes were measured after injections with water (as a negative injection control) or the mRNA encoding full-length or 5′-truncated SLC1A1. In the oocytes injected with full-length SLC1A1 mRNA, bath-applied glutamate (2 mM) induced a robust inward current characteristic of glutamate transport (fig. 4a, b). However, electrogenic glutamate transport was minimal in the oocytes injected with water or truncated SLC1A1 mRNA. These differences existed despite comparable overall expression levels of each protein, according to Western immunoblot (fig. 4d).

Based on the results from the initial oocyte recordings, in a separate experiment, we examined the potential for the 5′-truncated form of EAAT3 to possibly function in a dominant-negative manner. Electrophysiological recordings were repeated after injections in another set of oocytes (n = 9-21) with either the full-length SLC1A1 mRNA (200 ng/μl) or a 1:1 mixture of full-length and 5′-truncated mRNA (100 ng/μl each). Expression of the protein was confirmed by immunoblot (fig. 4d). Results indicated that injecting half as much full-length mRNA in the presence of an equal amount of 5′-truncated mRNA still enabled inward glutamate currents to be produced after bath application. Moreover, the amount recorded was approximately 60% of that recorded after injections of twice as much full-length mRNA alone (fig. 4c). These observations do not support a dominant-negative effect of the truncated mRNA or protein.

To examine the consequences of the hemi-deletion on gene expression, we compared the expression level of 84 genes of interest related to glutamate and GABA transmission in 8 subjects with the deletion and 11 noncarrier controls from the same pedigree using a commercial PCR array. Comparisons between these subject groups were made using an uncorrected t test and visualized in a volcano plot (fig. 5). Our results demonstrate nominally significant changes in expression of a small subset of genes in deletion carriers (fig. 5a). Interestingly, these genes encode well-characterized members of the AMPA, NMDA and metabotropic classes of glutamate receptors (GRIA1, GRIN2A and GRM8) as well as a neuronal high-affinity glutamate/aspartate transporter (SLC1A6).

In addition to comparing all deletion carriers versus noncarriers from the pedigree, we also performed an exploratory analysis using the same PCR array data to compare expression in the peripheral blood RNA of deletion carriers with psychosis (n = 5) to those without psychosis (n = 3). Although the results must be viewed as preliminary, they revealed nominally significant expression increases in a kainate receptor subunit (GRIK4) and decreases in a glial glutamate transporter (SLC1A3) as well as α-synuclein (SNCA) (fig. 5b), which is involved in the recycling and maintenance of presynaptic glutamate-containing vesicles.

We confirmed that in HEK293 cells treated with SLC1A1 shRNA or cDNA-containing plasmids, the expression level of this gene significantly decreases or increases, respectively. Specifically, when normalized to the SLC1A1 expression level in the nontarget shRNA-treated HEK293 cells, the cocktail of three validated human MISSION SLC1A1 shRNAs (Sigma-Aldrich) had knocked down the SLC1A1 mRNA expression by more than 37% (fig. 6a). Conversely, compared to untreated cells, the HEK293 cells transfected with full-length SLC1A1 plasmids had significantly increased SLC1A1 mRNA expression according to qRT-PCR (>400-fold) (fig. 6b). This increase was significantly attenuated when the full-length cDNA was co-transfected with SLC1A1 shRNA, although the levels were still nearly 100-fold above baseline (fig. 6b). These findings were strongly supported at the protein level based on the trends in the fluorescent signal intensity observed in the same cell cultures (data not shown). Notably, the specificity of the overexpression of full-length versus 5′-truncated SLC1A1 was also strongly confirmed, since we observed no increase in SLC1A1 exon 1 reads in the HEK293 cells receiving the 5′-truncated construct, compared with the marked increase in exon 10 reads in the same samples.

Examination of the more global effects of altered SLC1A1 mRNA expression on genes relevant to schizophrenia and brain function was enabled by targeted RNA-sequencing expression (TREx) analysis. A total of 36 unique genes demonstrated significant changes in expression (FDR q < 0.15) out of the total 375 genes included in the assay and the 240 genes subjected to statistical testing (table 2). Not surprisingly, the most significant changes were seen in cells that received the SLC1A1 overexpression constructs. Note that data for two different regions of SLC1A1 are shown. The first region includes a sequence in exon 10 common to all known SLC1A1 isoforms. It demonstrated a clear pattern of increase in the 8 samples overexpressing either 5′-truncated (TR) or full-length (WT) cDNA (fig. 7). The second region is in exon 1, which also showed a significant increase in expression, but only in the full-length samples compared to all other samples (fig. 7; table 2). These results were not unexpected given the construct design and the qRT-PCR results.

In addition to SLC1A1 itself, there were several other significant changes seen as a result of either knockdown or overexpression of either truncated or full-length SLC1A1. Rather than describing each of these observations individually, we will merely point out a few general patterns of change and highlight some of the genes. Overall, most of the significantly changed genes showed expression changes that mirrored those seen for exon 10 of SLC1A1, with increases following transfection with either truncated or full-length cDNA (e.g. SLC1A3, S100B, GRIA2, GFAP, ARRB1, AKT3, SNAP25, PPP1R1B, GRIA1, GRIN1 and SLC1A2) (fig. 7a). However, some genes showed patterns of decreased expression in the shRNA samples and increased expression in the full-length samples (e.g. GABRA4, GABRG2, GRM5, NDE1, GRM3 and CTNNA3). Of particular note were genes that appeared to be somewhat selectively decreased in the shRNA samples and increased in the truncated cDNA samples (e.g. DLX1, TSPO, SORBS3, NTRK3, PDE4A and PLCB1) or somewhat selectively decreased in the truncated samples and increased in the full-length samples (e.g. CYP2D6, NAT1). Representative expression profiles for some of these genes make these patterns easier to discern (fig. 7b).

The present study has four major findings. First, the region of chromosome 9 upstream of the second exon of SLC1A1 appears to have the ability to drive expression of a truncated form of SLC1A1, potentially through activation by TFs that are distinct from those which act on the native promoter. Second, the expression of a 5′-truncated isoform of SLC1A1 lacking exon 1 reduced the membrane localization of the protein and led to its inability to transport glutamate. Third, in PBLs from human subjects who inherited the exon 1 and native promoter deletion, there appears to be compensatory or downstream effects that lead to significant changes in a set of genes related to glutamate and GABA neurotransmission. Fourth, knockdown or overexpression of either full-length or 5′-truncated SLC1A1 mimics some of the changes seen in the human PBLs as well as some of the more general changes in gene expression previously reported in schizophrenia. We now discuss some of the implications of these findings.

As a first step toward characterizing the potential consequences of our newly discovered hemi-deletion, we used a dual luciferase assay and showed that the intron 1 sequence upstream of exon 2 can promote expression at an even higher level than the native promoter. This sequence would replace the native promoter in the hemi-deleted allele in subjects from the K3501 pedigree and function as a ‘surrogate' or alternative promoter for the expression of a truncated protein at one of three possible start codons at the beginning of exon 2. Although there was some indication from public databases that this region of SLC1A1 could function as a ‘weak promoter' in normal contexts, the robust and highly significant result we obtained was completely unexpected.

Because different TFs act on promoters to drive or reduce gene expression in different cellular and physiological conditions, we also performed a screen to identify TFs which would have different actions on the two promoters. Our results show that the two promoters are acted upon differently by a relatively small subset of TFs, including some with considerable relevance for brain function. The well-characterized VDR was one such TF and showed more than 7-fold less binding affinity for the alternative promoter region. VDR is expressed widely in many tissues but in the brain shows particular enrichment in the hypothalamus and in the dopaminergic neurons within the substantia nigra [33]. The VDR acts in combination with other entities, such as retinoid X receptors, to form transcription complexes that induce or repress gene expression [34]. Interestingly, low vitamin D status has been linked to a range of psychiatric conditions [35], with a recent meta-analysis demonstrating a strong association between vitamin D deficiency and schizophrenia [36]. Moreover, Graham et al. [37] reported a direct association between lower vitamin D levels and increased severity of negative symptoms and cognitive deficits in schizophrenic subjects, and McGrath et al. [38] reported that vitamin D supplementation during the 1st year of life is associated with reduced risk of schizophrenia in males. Complementing these findings, there is also a well-established vitamin D deprivation model of schizophrenia in rodents [39,40]. Interestingly, Simmons et al. [41] recently found that the expression of SLC1A1 was increased more than 7-fold by the VDR ligand, 1,25-dihydroxyvitamin D₃ (1,25D), in hTERT-HME cells. Taken together, these results suggest that the psychiatric disturbances associated with vitamin D deficiency might be due, in part, to reduced SLC1A1 expression. Furthermore, the loss of the VDR binding site in subjects with a hemi-deleted SLC1A1 allele provides a possible mechanistic explanation for the observed findings.

The expression level and membrane localization of EAAT3 promote its ability to act as a glutamate transporter, which helps maintain the strength and dynamics of normal glutamatergic transmission. This transport activity limits the diffusion of glutamate molecules in one synapse to its neighboring synapses, a phenomenon known as ‘spillover' [42]. Despite the fact that EAAT3 must be expressed at the cell membrane to function as a transporter, a large proportion of the protein is known to be localized in intracellular pools [31,32,43]. In fact, some estimates have indicated that up to 80% of EAAT3 can be located intracellularly under baseline conditions [44,45]. However, it is also thought that when cells are activated, various intracellular messengers (including PKC and PDGF) promote rapid mobilization, cell surface expression and activity of EAAT3 [44,45,46,47,48]. Moreover, Levenson et al. [49] suggested that translocation of the EAAT3 from the cytosol to the plasma membrane is required for long-term potentiation in area CA1 of the hippocampus. Our physiological and confocal imaging results indicate that even if subjects with the hemi-deletion express a 5′-truncated form of EAAT3, this protein lacks normal cell surface expression and the ability to transport glutamate. Importantly, however, the deficit created by the expression of 5′-truncated EAAT3 does not appear to act in a dominant-negative manner, further suggesting that some residual EAAT3 function should still be present in these subjects, though they may clearly be at an increased risk for psychosis. Support for the nondominant nature of the truncated EAAT3 is also seen in previous studies. By co-expressing WT EAAT3 subunits with different mutant subunits [50,51], different groups have shown that EAAT3 assembles as trimers but individual subunits in the assembly have distinct glutamate binding sites and transport pathways which function independently of each other. These studies help explain our observation that coexpression of full-length WT and 5′-truncated EAAT3 generated approximately half as much glutamate transport as when only WT mRNA was injected.

To probe for potential compensatory changes occurring in the cells of subjects with the hemi-deletion, we examined the expression of genes involved in glutamate and GABA function using a commercial PCR array. Our results showed that a small subset of these genes demonstrated altered (increased) expression as a consequence of the deletion, including GRIA1, GRIN2A and GRM8, as well as a neuronal high-affinity glutamate/aspartate transporter, SLC1A6. Moreover, in subjects with the deletion who exhibited psychosis, there was an apparent increase in the expression of GRIK4 and decreased expression of SLC1A3 and SNCA compared to subjects with the deletion who did not exhibit psychosis. We now discuss the potential significance of these findings.

Deletion Carriers versus Noncarriers

Three of the genes with increased expression in the peripheral blood of SLC1A1 deletion carriers encode glutamate receptor subunits, specifically the α₂-subunit of the NMDA receptor (GRIN2A), the ionotropic AMPA 1 receptor (GRIA1) and the metabotropic glutamate receptor 8 (GRM8). Current evidence indicates that in schizophrenia, the levels of glutamate receptors such as these can be increased, decreased, or unchanged, depending on the brain region being studied [52] with variability likely arising from different drug histories, comorbid diseases and underlying clinical heterogeneity [53]. Despite these limitations, there are considerable data available that strengthen the potential impact of our observations regarding these genes.

GRIN2A. Glutamatergic theories of schizophrenia were originally based heavily on the observation that NMDA receptor antagonists such as phencyclidine and ketamine can induce schizophrenia-like symptoms in normal humans and worsen psychotic symptoms in persons with schizophrenia [54,55]. Since then, converging evidence suggests hypofunction of NMDA receptors in schizophrenia and decreased expression of GRIN2A in the dorsolateral prefrontal cortex of subjects with schizophrenia [2,56,57]. Moreover, auto-antibodies to NMDA receptors have been reported to cause schizophrenia-like symptoms in previously normal subjects [58,59], and the presence of such antibodies has been described in up to 10% of patients with an initial clinical diagnosis of schizophrenia [5,60]. Drugs that potentiate or modulate NMDA receptors have also been studied as potential therapeutic targets for schizophrenia [61]. In our subjects with the deletion, the increased expression we detected in the peripheral blood could be interpreted as a compensatory (defensive) mechanism to prevent or diminish psychosis subsequent to the hemi-deletion of SLC1A1 and resultant disruption of glutamate transmission. However, it is also important to point out that the measures we obtained in the blood may not reflect the state of what is occurring in the brain.

GRIA1. The AMPA1 receptor subunit is a good candidate gene for susceptibility to schizophrenia since it maps to 5q33, a region consistently implicated in several independent genome-wide linkage scans and meta-analyses [62,63,64,65], and also because its expression has been found to be decreased in the brains of some schizophrenia patients [66]. Furthermore, polymorphisms in GRIA1 have been associated with schizophrenia and its clinical symptoms in different populations [66,67]. Moreover, positive allosteric modulators of the AMPA receptor have become of increasingly high interest as potential new treatments for schizophrenia [68]. Based on these data, the increased expression we detected in the blood could likewise represent a possible compensatory response to glutamate transporter hypofunction in subjects with the hemi-deletion.

GRM8. Group III mGluRs are generally thought to be expressed presynaptically and to serve as autoreceptors that modulate glutamate transmission [69,70]. Although genetic variations of GRM8 in schizophrenia and the antipsychotic effects of mGluR8 agonists have been inconsistently reported in the literature [71,72,73,74], there are conclusive reports regarding its role in mediating excitatory transmission in response to environmental stressors. In fact, the mGlu8 receptor has been specifically implicated in anxiety and stress-related behaviors, and its activation has been shown to produce strong anxiolytic effects [75,76]. Thus, the observed increase in its expression might represent another adaptive response in SLC1A1 deletion carriers, thereby helping them to deal better with stressful stimuli or events.

SLC1A6. Similar to what has been suggested regarding GRIN2A, GRIA1 and GRM8, we also believe that the increased SLC1A6 expression may represent a potential compensatory response to the hemi-deletion of SLC1A1. SLC1A6 encodes EAAT4, a neuronal high-affinity glutamate/aspartate transporter, and its expression is highly localized, mostly expressed in the Purkinje cells of the cerebellum [77]. Both EAAT3 and EAAT4 are required to form trimeric complexes to function as glutamate transporters [78]. However, Nothmann et al. [79] suggested that unlike other EAATs, EAAT3 and EAAT4 can co-assemble into stable heterotrimers in which individual subunits are functionally independent. Moreover, they found that the particular composition of these trimeric glutamate transporters may affect cellular localization. When expressed alone in canine kidney cells, EAAT3 homotrimers were inserted exclusively in the apical membrane. However, when EAAT4 was co-expressed with EAAT3, some of the EAAT3 was trafficked to the basolateral surface and associated with EAAT4 in heterotrimers. Thus, increased EAAT4 in SLC1A1 hemi-deletion carriers might indicate an effort by the cells to replace the nonfunctional (5′-truncated) EAAT3 transporters with EAAT4 in order to keep the number of excitatory glutamate transporter oligomers constant. However, it is quite likely that such a response could also lead to altered localization and functional properties.

Psychotics versus Nonpsychotics within Deletion Carriers

Although the number of subjects with the SLC1A1 hemi-deletion in the K3501 pedigree was relatively small, we attempted to gain additional insight into the relevance of the deletion for psychosis by comparing the deletion carriers with and without psychosis in an exploratory fashion. This analysis revealed decreased expression of two glutamate-related genes and increased expression of another.

SLC1A3. Among the hemi-deletion carrier subjects, those who manifested psychosis showed a >3-fold decreased expression of SLC1A3 when compared to the nonpsychotic carriers. This gene encodes EAAT1, a glial-enriched high-affinity glutamate/aspartate transporter, which is the major glutamate transporter in the cerebellum [80], although it is also found at high levels throughout the brain. It is tempting to speculate that along with the hemi-deletion of SLC1A1, the lower expression of SLC1A3 in our psychotic deletion carriers might be involved in the primary pathogenesis of schizophrenia symptoms. Evidence for this speculation stems from studies of the EAAT1-knockout mouse, which shows cerebellar dysfunction as well as behavioral abnormalities that model the positive, negative and cognitive/attentional disturbances seen in the disorder [81,82,83]. Moreover, reversal of some of these symptoms was seen with either haloperidol or mGlu2/3 agonist administration [82].

SNCA. Alterations in SNCA sequence and copy number have been very well-described in Parkinson's disease and α-synucleinopathies [84,85,86,87,88]. With regard to the present study, it is particularly interesting to note the common occurrence of psychosis and hallucinations in subjects with amplifications of SNCA [89,90,91,92], as well as the finding that such subjects often display reductions in SNCA mRNA expression. Given these findings, the 2.2-fold decreased expression in SNCA that we detected in deletion carriers with psychosis might represent a protective mechanism to attenuate the pathological condition.

GRIK4. The kainate receptor subunit KA1 plays a key role in regulating glutamate transmission in several brain areas implicated in psychosis as well as cognitive and memory functions, including the amygdala, hippocampal formation and entorhinal cortex [93]. Moreover, genome-wide association studies support the role of GRIK4 as a risk factor for schizophrenia as well as bipolar disorder [94,95]. A deletion variant within GRIK4 that caused increases in its mRNA expression and protein abundance has been found to be associated with a reduced risk of bipolar disorder [95,96]. Complementing these data, GRIK4-knockout mice demonstrate impairments in pre-pulse inhibition, spatial memory acquisition and recall [97]. Thus, it is possible that the increased expression of KA1 in our psychotic deletion carrier subjects compared to nonpsychotic carrier subjects is a compensatory response to reduce psychosis as well as mood disorders.

Our expression data from the HEK293 cells overexpressing the full-length or 5′-truncated forms of SLC1A1 confirmed the highly specific increases in each of these constructs and confirmed the significant reductions in SLC1A1 seen following treatment with shRNA to model a state of haploinsufficiency. Interestingly, with the artificially increased or decreased expression of SLC1A1, we observed changes in several other genes altered in deletion carriers and also reported to be changed in other studies of subjects with schizophrenia. Space limitations do not allow a full description of all of these expression changes. Nonetheless, we highlight some of the more salient ones. For example, with overexpression of both full-length or 5′-truncated isoforms of SLC1A1, the expression level of the two glial members of this glutamate transporter family SLC1A2 and SLC1A3 increased as well. These genes encode EAAT2 and EAAT1, respectively, both of which are mainly expressed in the astrocytes in all parts of the brain, with EAAT1 enriched in the cerebellum and EAAT2 enriched in the hippocampus [80,98]. Depending on the brain region examined, the expression level of these genes has been reported as either decreased or increased in schizophrenia. Smith et al. [99] detected significantly higher levels of EAAT1 and EAAT2 mRNA expression in the thalamus of subjects with schizophrenia, while another group found decreased expression of EAAT1 and EAAT2 proteins in the superior temporal gyrus and decreased EAAT2 protein in the hippocampus in schizophrenia [100].

SLC1A3

The function of EAAT1 and the implications of its reduced expression have been discussed earlier. Similar to what is seen in psychotic deletion carriers, the reduced expression of this gene in cells with the knockdown of SLC1A1 and its increased expression in cells overexpressing SLC1A1 implicates a strong association between the expression level of these two transporters, one responsible for glutamate uptake on neurons and the other on glial cells. Moreover, since EAAT1-knockout mice exhibit symptoms similar to the positive symptoms of schizophrenia [82], such a strong positive association could exacerbate the psychiatric/metabolic consequences of the low levels of EAAT3.

SLC1A2

Most of the glutamate uptake in the brain and importantly about 95% of such an uptake in young adult forebrain tissue is accomplished by EAAT2 [101,102]. It has been shown that the deletion of the EAAT2 causes a synaptic glutamate spillover and has dramatic developmental consequences. Mutations of this gene have been associated with schizophrenia, alcoholism, smoking behavior and bipolar disorder [102]. The higher expression of this gene with the knockdown of SLC1A1 can represent a compensatory mechanism to clear the excess glutamate from the synapses and prevent the glutamate spillover.

SLC7A11

Another glutamate transporter with increased expression in response to SLC1A1 knockdown or overexpression was SLC7A11, which encodes xCT, the light chain subunit of the glutamate/cystine exchange protein system x_c^- (SXC). The main function of the xCT is to provide substrate for the synthesis of GSH in glial cells, a function which is performed by EAAT3 in neurons [103,104]. GSH is the most abundant antioxidant in the central nervous system, protecting it against oxidative stress. First postulated in the 1930s by Roy Hoskins, the significance of oxidative stress in the pathophysiology of schizophrenia is now well established [12,105]. It has also been suggested that cells are able to increase SXC expression and activity when increased intracellular redox control is required [106]. Thus, the higher expression of SLC7A11 after knockdown of SLC1A1 might represent a mechanism to maintain GSH concentrations at normal levels and hence protect the cells against oxidative stress.

Another group of genes with significant expression changes encode different subunits of ionotropic (GRIN1, GRIA1 and GRIA2) or metabotropic (GRM3) glutamate receptors, all having been implicated in the pathophysiology of schizophrenia. The importance of functional interaction between glutamate transporters and receptors for normal glutamatergic tone and higher brain functions cannot be overstated. By rapid uptake of glutamate from the synaptic space, glutamate transporters not only supply the neurotransmitter to the nerve terminals and decrease spillover of glutamate and putative excitotoxic processes, they also help regulate the duration and amplitude of excitatory postsynaptic potentials, limit receptor desensitization and control synaptic plasticity [107]. As an example of such an intricate interplay, Cao et al. [108] have suggested that EAAT3 is responsible for regulating the plasma membrane trafficking of GluR1 (GRIA1) in the hippocampus. Since rapid trafficking of AMPA receptors is known to be a fundamental process in learning and memory, EAAT3 plays a critical role in the biochemical cascade of such cognitive functions. Thus, disruption of its normal interplay with glutamate receptors can be one of the mechanisms by which decreased or increased expression of SLC1A1 can increase the risk of psychosis.

Finally, we observed changes in the expression of several GABA-related genes in the HEK293 studies (e.g. GABBR1, GABRA4, GABRG2 and DLX1). These changes generally followed the direction of change that was introduced to SLC1A1 (i.e. reduced in the SLC1A1 knockdown samples, increased in the SLC1A1 overexpression samples). These observations lend further support to the coupling of GABA and glutamate systems and underpin previous reports that imbalances in excitatory/inhibitory signaling have potential profound consequences for normal cognition and higher brain functions. In fact, there is abundant evidence that abnormal GABAergic activity contributes to the pathology of schizophrenia [109]. As a suggested mechanism, reduced NMDA receptor activity on inhibitory GABAergic interneurons leads to disinhibition of the glutamatergic input from the basolateral nucleus of the amygdala to the hippocampus, increasing the synaptic activity of glutamate, which in turn leads to excitotoxicity [2,110,111]. Thus, reducing glutamate neurotransmission through a modulatory mechanism such as agonism of GRM2 and GRM3, which are inhibitory presynaptic glutamate autoreceptors, can be a potential therapeutic approach that might correct for the decreased GABAergic activation. On the other hand, the DLX family of homeobox TFs are necessary for GABAergic neuron differentiation, migration and survival [112,113,114,115]. The changes in DLX1 expression that we observed in HEK293 cells following alteration of SLC1A1 expression suggest that a hypofunctional SLC1A1 allele might affect GABAergic neuron proliferation and development. Interestingly, mice lacking the DLX1 gene have been shown to exhibit postnatal subtype-specific loss of GABA interneurons and specific behavioral abnormalities consistent with models of schizophrenia [116], further supporting the involvement of GABAergic systems in the pathophysiology of this disorder.

Although CNVs in SLC1A1 may be rare, they appear to be of large effect size. Our recently discovered hemi-deletion affecting SLC1A1 maintains an alternative promoter region which enables the expression of a 5′-truncated form of EAAT3. However, the protein that might be expressed from such a locus lacks normal cell surface expression and the ability to transport glutamate, although it does not appear to act in a dominant-negative manner or affect expression from the other intact allele. Interestingly, however, the presence of the hemi-deleted allele does alter the expression of many genes related to GABA and glutamate signaling, potentially in a compensatory manner. Cell culture studies following knockdown of SLC1A1 or overexpression of either a 5-truncated form of EAAT3 or WT EAAT3 produced changes in some of the same genes altered in the human subjects with the hemi-deletion and also revealed several strong biological relationships between many genes implicated in the pathophysiology of schizophrenia. Overall, the present study adds to the growing literature strongly implicating glutamatergic dysfunction in schizophrenia and may help explain some of the phenotypes reported with the knockout of this gene. Future studies will be needed to formally address the impact of the specific hemi-deletion we identified on redox state, neuroinflammation and glutamatergic signaling in the intact brain as well as its behavioral relevance. The data we have presented suggest that SCZD18 may represent one of many glutamatergic pathways that could lead to psychosis either directly or in combination with other genetic and environmental factors. Clearly, this report represents merely a first step toward characterizing the full biological consequences of alterations in the SLC1A1 gene.

This study was supported by NIH grants MH080373 and MH096257 and a Brain and Behavior Research Foundation NARSAD Award. We are also grateful to the Palauan community members who participated in this research as well as K. Gentile, B. Devlin, K. Roeder and N. Melhem for their assistance with the CNV mapping.

All human subject protocols and procedures were approved by institutional review boards in the US and the Republic of Palau. All individuals provided informed consent or assent after receiving a full explanation of the study in both English and Palauan.

The authors have no conflicts of interest to report.