GAD2 Alternative Transcripts in the Human Prefrontal Cortex, and in Schizophrenia and Affective Disorders

Genetic variation and early adverse environmental events work together to increase risk for schizophrenia. γ-aminobutyric acid (GABA), the major inhibitory neurotransmitter in adult mammalian brain, plays a major role in normal brain development, and has been strongly implicated in the pathobiology of schizophrenia. GABA synthesis is controlled by two glutamic acid decarboxylase (GAD) genes, GAD1 and GAD2, both of which produce a number of alternative transcripts. Genetic variants in the GAD1 gene are associated with increased risk for schizophrenia, and reduced expression of its major transcript in the human dorsolateral prefrontal cortex (DLPFC). No consistent changes in GAD2 expression have been found in brains from patients with schizophrenia. In this work, with the use of RNA sequencing and PCR technologies, we confirmed and tracked the expression of an alternative truncated transcript of GAD2 (ENST00000428517) in human control DLPFC homogenates across lifespan besides the well-known full length transcript of GAD2. In addition, using quantitative RT-PCR, expression of GAD2 full length and truncated transcripts were measured in the DLPFC of patients with schizophrenia, bipolar disorder and major depression. The expression of GAD2 full length transcript is decreased in the DLPFC of schizophrenia and bipolar disorder patients, while GAD2 truncated transcript is increased in bipolar disorder patients but decreased in schizophrenia patients. Moreover, the patients with schizophrenia with completed suicide or positive nicotine exposure showed significantly higher expression of GAD2 full length transcript. Alternative transcripts of GAD2 may be important in the growth and development of GABA-synthesizing neurons as well as abnormal GABA signaling in the DLPFC of patients with schizophrenia and affective disorders.


Introduction
State of Maryland, and the tissue was donated to the NIMH under the terms of a material transfer agreement. All samples were obtained with audio-taped informed consent from the legal next of kin to study brain tissue, as approved by the Institutional Review Board of the National Institutes of Health and University of Maryland, in lieu of a written consent form. Brains were hemisected, and cut into 1.0-1.5-cm-thick coronal slabs, flash frozen, and stored at -80°C. DLPFC gray matter was dissected using a dental drill. For CBDB cases, the DLPFC (Brodmann's areas 9 and 46) were dissected from middle frontal gyrus from coronal slab immediately anterior to the genu of the corpus callosum. For fetal cases, the PFC was obtained from the frontal cortex dissected at the dorsal convexity, midway between the frontal pole and anterior temporal pole. To measure the pH, pulverized cerebellum was used from each sample. To assess the expression of GAD2 alternative transcripts in the DLPFC, 176 schizophrenia patients, 61 bipolar disorder patients, 138 major depressive disorder (MDD), and 364 non-neurological and non-psychiatric controls were studied. A subset of non-psychiatric non-neurological controls, spanning from gestational weeks 14 to 20 and from birth up to 85 years of age, were used to measure expression of selected GAD2 transcripts in the prefrontal cortex (Table 1). Toxicological analysis was performed on all samples. Positive toxicology was exclusionary for control subjects but not for patients with psychiatric disorders. The majority of positive toxicology reports in psychiatric cases were due to the presence of psychotropic medications. In a subset of psychiatric cases, toxicology studies also revealed the presence of a variety of illicit substances. Nonpsychiatric controls were excluded if toxicology was positive for psychotropic medications or illicit substances. 43.6% of psychiatric patients (63.8% in MDD, 65.6% in bipolar disorder, and 20% in schizophrenia) are suicides. Nicotine exposure was positive in 41.8% patients with psychiatric disorders (43.4% in schizophrenia, 43.7% in bipolar disorder and 35.1% in MDD) and 24.6% controls.
Dissected tissue was pulverized and stored at -80°C; RNA extractions and reverse transcriptase reactions were performed as described previously [20]. Briefly, total RNA was extracted from 300mg of tissue with TRIZOL Reagent (Life Technologies, Grand Island, New York). The yield of total RNA was determined by measuring the absorbance at 260 nm. To generate a RNA Integrity Number (RIN), RNA quality was assessed by high-resolution capillary electrophoreses (Agilent Technologies, Palo Alto, California). Using SuperScript First-Strand Synthesis System for RT-PCR (Invitrogen, Carlsbad, CA, USA) to synthesize cDNA, 4μg of total RNA was reverse transcribed in a 50μl reaction system.

RNA Sequencing
To identify the library of transcripts derived from the GAD2 gene, RNA sequencing was performed on pooled commercial fetal and adult whole human brain poly A + RNAs (Clontech, Lot#7110099, Lot#110415). mRNA molecules were fragmented into small pieces using divalent cations under elevated temperatures. Reverse transcriptase and random hexamers were used to convert the cleaved RNA fragments into first strand cDNA. Second strand cDNA was synthesized using DNA Polymerase I and RNaseH. These cDNA fragments were then subjected to an end repair process using T4 DNA polymerase, T4 polynucleotide kinase (PNK) and Klenow DNA polymerase, the addition of a single 'A' base using Klenow exo (3' to 5' exo minus), and then ligation of the Illumina P adapters using T4 DNA Ligase. An index (up to 12) was inserted into Illumina adapters so that multiple samples could be sequenced in one lane of an 8-lane flow cell if necessary. These products were then purified and enriched by PCR to create the final cDNA library for high throughput DNA sequencing using HiSeq 2000 (Illumina, Inc.). To analyze the RNASeq data, sequencing reads were mapped against the whole reference genome by using TopHat software. The Cufflinks software was used to assemble transcripts and estimate their abundance.

Rapid amplification of cDNA (RACE) ends and end-to-end PCR
To identify the 3' ends of GAD2 transcript in human brain, we performed 3' RACE, using fetal and adult brain poly A+ RNA with GAD2 gene specific sense primers binding at exon 1. We used the SMART RACE cDNA Amplification Kit (Clontech) and Advantage 2 PCR Kit (Clonetech) for these assays. Commercial human fetal and adult brain poly A+ RNAs (Clontech) were reverse-transcribed to cDNA by MMLV reverse transcriptase (Clontech) according to the manufacturer's protocol.  Table 2). mRNA expression levels of two GAD2 alternative transcripts (GAD2 full length transcript and truncated transcript) were normalized to the geometric means of three constitutively-expressed genes: β-actin (ACTB), β2-microglobulin (B2M) and β-glucuronidase (GUSB). The geometric means of the three housekeeping genes did not show diagnostic group difference. Genotypes were generated as described previously by Straub et al [21].

Statistical analysis
Statistical analyses were performed using R package (Version3.2.2). The expression data was log2 transformed to make the data more symmetrical for ANCOVA/linear modeling. Levene's test showed that the assumption of homogeneity of variance was met. The lifespan curve was generated using LOESS fit (local polynomial regression fitting) by using an R package with default parameters. ANCOVA/linear model was used for all comparisons. Multiple regression analyses were carried out to evaluate the contributions of age, pH, postmortem interval (PMI), race, sex and RNA integrity number (RIN), ethanol exposure and nicotine exposure on mRNA expression. Multiple regression analyses were also performed to assess the effect of completed suicide, psychotropic medications (toxicology screen results about antidepressants, antipsychotics, anticonvulsants, benzodiazepines, and opiates), lifetime neuroleptic exposure, average daily neuroleptic dose, final neuroleptic dose, and illicit substances on mRNA expression in the patients with psychiatric disorders. Estimated of lifetime neuroleptic exposure, average daily dose and final neuroleptic dose were all converted to chlorpromazine equivalents for statistical comparisons. Comparisons between diagnostic groups were made by ANCOVA for mRNA expression with diagnosis as an independent variable. Covariates were chosen for each ANCOVA from multiple regression analyses. Comparisons between patients and controls following overall ANCOVA were conducted by post-hoc Tukey HSD tests. Comparisons within the different diagnostic groups for completed suicide, nicotine exposure, ethanol exposure and psychotropic medications were conducted by ANCOVA, followed by Bonferroni correction.

Alternative transcripts of GAD2 in human brain
An analysis of pooled RNA from normal fetal or adult brains (acquired commercially) by RNA sequencing failed to identify any novel splicing junctions of GAD2 in the human DLPFC, but confirmed the production of a previously identified truncated transcript (ENST00000428517) composed of several 5' exons of GAD2 (Fig 1).
To validate the findings from the RNA-Seq data, we carried out 3'RACE and observed an alternative termination point with a poly A+ tail in intron 3. By end-to end PCR, we confirmed the well-known full length transcript encoding for the 65kDa GAD protein and the previously identified truncated transcript, ENST00000428517, consisting of the first four exons of GAD2 (Fig 2A). RNA sequencing data suggested that the expression of the GAD2 truncated transcript is about 78% of the expression of the full length GAD2 in the fetal brain, about 12% in the childhood (0~10 years old) and 4% after 10 years old [22]. The expression levels of the full length and truncated GAD2 transcripts in 20 pooled human tissues (acquired commercially from Clontech, Human Total RNA Master Panel II, Lot#1403130A) showed that both GAD2 transcripts are primarily expressed in brain ( Fig 2B). The truncated transcript shares the same exon1 and translational start site with the full length transcript, but lacks an in-frame stop codon.

Lifespan expression pattern of GAD2 alternative transcripts
In a cohort of 326 non-neurologic non-psychiatric control subjects ranging from the second trimester in the fetus through 85 years of age, prefrontal cortical samples were studied to determine the expression pattern of the two transcripts across the human lifespan. Unlike the lifespan pattern of the GAD2 full length transcript (Fig 3A), whose expression was low in the fetus and then gradually rose to its adult peak at about 20 years of age, the GAD2 truncated transcript, ENST00000428517, showed a mildly inverted expression pattern. ENST00000428517 expression peaked during fetal period, and then gradually decreased after birth (Fig 3B).
Expression of GAD2 splice variants RNA in schizophrenia, bipolar disorder and major depression  Using pooled whole brain, known and putative novel exons for GAD2 in both adult and fetal human brain were identified using RNA-Seq. The y-axis is the coverage information, while the x-axis shows the position of the reads. The red arc-shapes between exons represented the splicing junctions. The thickness of the red arc-shapes represented the abundance of the splicing events. We observed the continuously splicing events across all the known junction of the GAD2 gene in adult brain but not in fetal brain. These results suggested there isn't a novel splicing junction for GAD2 in the human DLPFC, and potential truncated transcripts composed by several 5' exons. doi:10.1371/journal.pone.0148558.g001 Expression of GAD2 full length transcript with nicotine exposure and suicide The toxicology screen provided the positive or negative results for a large number of substances including nicotine, ethanol, opiates, anticonvulsants, antidepressants, and antipsychotics. The clinical and medical examiner records provided basic demographic information plus lifetime neuroleptic exposure, average daily dose, final neuroleptic dose, and suicide. The multiple regression analyses revealed that the expression of GAD2 full length transcript in DLPFC was significantly correlated with completed suicide (p = 9.61E-09) and nicotine exposure (p = 2.59E-02) in the psychiatric patients taken as a whole across diagnoses. The subjects with   Fig 5B) showed significantly higher expression of GAD2 full length transcript compared with natural deaths or nicotine free patients with schizophrenia. As there was a significant correlation between sex and the expression of GAD2 full length transcript, we ran further analyses to determine if sex might affect the expression of GAD2 full length transcript in the DLPFC within each diagnostic group. The female control subjects had significantly lower expression of GAD2 full length transcript compared to male controls (p = 1.73E-03). This was  also seen independently in schizophrenia (P = 1.19E-02) and MDD (1.20E-03) patients ( Fig  5C). There was no significant effect of psychotropic medications on the expression of GAD2 transcripts in the DLPFC.

Discussion
Our results confirmed that GAD2 produces a truncated transcript in the human brain, in addition to the well-known full length transcript that encodes the 65kDa GAD enzyme. Characterization of the expression of these transcripts across the lifespan in the human DLPFC suggests that they are developmentally regulated. Finally, the expression of both GAD2 transcripts differed significantly between patients with psychiatric disorders compared to non-psychiatric controls.

Alternative splicing of GAD2
This study is the first to characterize alternative splicing of GAD2 in the human brain. We confirmed the presence of the full length transcript responsible of GAD2, and identified its developmental expression pattern in the prefrontal cortex. The truncated GAD2 transcript consisting of the first four exons of GAD2, ENST00000428517 lacks an in-frame stop codon. While the sequence for ENST00000428517 previously has been deposited into the UCSC Genome Browser, nothing has been published regarding its expression in human brain. Lacking of an in-frame stop codon, the truncated transcript is prone to nonstop decay, a cellular mRNA surveillance mechanism to prevent mRNA without a stop codon from translation [23,24]. As alternative splicing can lead to increased functional diversity in mRNA transcripts [25,26], this truncated GAD2 splice variant probably works as on-off regulation to adjust the level of the enzymatic protein produced by full length GAD2 transcript, which has a well established role in GABA synthesis. This truncated transcript might be critical to maintain a low level of full length GAD2 expression during fetal brain development since its lifespan expression pattern is the inverse of the full length GAD2 transcript. Additional experiments will be needed in order to define the function of this truncated transcript

GAD2 splice variants expression is developmentally regulated
Studies have shown that multiple transcripts derived from genes have expression patterns that differ across the lifespan [27,28]. Chan and colleagues detected the full length GAD2 mRNA in the human fetal frontal pole at 12 gestational weeks, with expression being highest at the beginning of the second trimester and decreasing rapidly and becoming undetectable by gestational week 19 [29]. However, in this study we show that the expression of the full length GAD2 mRNA increased with age, and had a particularly rapid rise during the first two decades of life, which mirrors (albeit on a different time scale) with expression studies in the rodent [30,31]. The lifetime expression trajectory of the ENST00000428517 is highest during fetal periods and gradually decreases after birth. Since the truncated transcript will be degraded quickly as it is an mRNA lacking a stop codon, its expression more likely acts as a component of the regulation of full length GAD2 expression levels during fetal brain development. During early CNS development, GABA levels are tightly controlled, as this neurotransmitter plays a major role in the regulation of cell proliferation, differentiation, and migration [32][33][34]. The role of GABA in early brain development may be mediated in part by the ENST00000428517 transcript since it is highly expressed during in the second trimester fetus, at a time before the rise in levels of synaptosomal full length GAD2.

Schizophrenia and GAD2
Recent studies indicate that genes involved in the etiology of schizophrenia may also be involved in the pathology of bipolar disorder and major depression [28,[35][36][37]. Accordingly we examined the expression of GAD2 splice variants across these disorders. The full length GAD2 mRNA was significantly decreased in expression in the DLPFC of patients with schizophrenia and bipolar disorder compared with controls. While a decrease in full length GAD2 mRNA expression in bipolar disorder has been previously reported in the hippocampus, this same group did not see an effect in schizophrenia [38]. Another group has reported that GAD2 full length protein expression was decreased in the primary auditory cortex and cerebellum of schizophrenia patients [37,39] and cerebellum of bipolar disorder patients [37]. Our finding of decreased full length GAD2 mRNA potentially reflects an immature level of prefrontal development or regression after onset of illness, a supposition based on the developmental trajectory of GAD2 full length expression across normal development. Schizophrenia and bipolar disorder not only have some common clinical phenotypes, but appear to share some genetic risk factors [40,41]. It is possible that the decrease of full length GAD2 expression in both diagnostic groups might be involved in the molecular mechanism of the shared phenotypes. The truncated GAD2 transcript showed a significant decrease in expression level in patients with schizophrenia and an increase in patients with bipolar disorder, but no difference was detected in MDD patients. The expression level of ENST00000428517 suggests that this transcript behaves differently in the pathology associated with schizophrenia and bipolar disorder. The role of the truncated transcript ENST00000428517 is undefined, but it may be a component of the molecular dysfunction associated with these two psychiatric disorders. It must be highlighted that any differences in mRNA expression between diagnostic groups are difficult to interpret. It is very hard to disambiguate changes related to the pathophysiology of illness from those secondary to the epiphenomena of illness, such as medications, cigarette smoking, environmental deprivation, and/or the stress of a lifetime combating chronic disability. Regardless, there appears to be diagnostic variation for alternative GAD2 transcripts in schizophrenia and affective disorders.
Our toxicology screen results suggest, unsurprisingly, that the psychiatric patients in this study have been exposed to multiple substances since positive toxicology was not an exclusion criterion for the patients with a psychiatric disorder. We studied the possibility that exposure to either prescription medications, nicotine, or illicit substances may contribute to expression changes seen in patients. Chronic administration of psychotropic medications, including clozapine, fluoxetine, haloperidol, lithium, olanzapine and valproic acid reportedly alters the level of full length GAD2 and full length GAD1 in rat, measured by both real-time PCR and western blotting [42]. In particular, the mood stabilizer valproic acid stimulated the proliferation of the GABAergic neurons with an associated dramatic increase in levels of GABA and full length GAD1/2 [43]. In contrast to animal studies, our analyses didn't reveal positive correlations between the expression of GAD2 transcripts in the DLPFC and psychotropic medications in psychiatric patient cohorts.
The most common psychoactive substance on toxicological screening was nicotine, and this was not an exclusion criterion in any diagnostic group. The schizophrenia patients with nicotine exposure had increased expression of GAD2 full length transcript in the DLPFC comparing to nicotine-free patients with schizophrenia. There have been no previous studies on the impact of nicotine exposure on the expression of GAD2 in brain. Further animal studies are needed to confirm this finding and delineate the molecular mechanism. The correlation between GAD2 full length and nicotine is intriguing and suggests a previously unrecognized role for nicotine in cortical function in schizophrenia. This is especially important given the prevalence of cigarette smoking among schizophrenia patients [44]. 43.6% psychiatric patients involved in this study were completed suicide. More than half of patients with bipolar disorder or MDD were suicides. We observed the up-regulated expression of GAD2 full length transcript in patients with schizophrenia who are suicides. Previously, suicidal patients with mood disorders had an increase in the relative density of GAD-immunoreactive neuropil [45] and abnormalities in the glutamate-glutamine and GABA-glutamine cycles had significant impact on suicidal behavior [46]. Our finding suggests that up-regulated 65kDa GAD, the translated protein produced by the GAD2 full length transcript, might in part be responsible for the increased level of GAD in suicidal patients with mood disorders, and provides additional and potentially important information regarding the role of GABA on suicidal behavior. If confirmed, this offers a new insight into the neurobiology of suicide.
Full length GAD2 expression is present in most classes of GABA neurons and is particularly prominent in axon terminals [47]. The data in this study is derived from homogenized tissue, and contains GAD2 transcripts from a variety of cell types and axon terminals. It is possible that the expression changes of GAD2 gene are restricted to a particular class of neurons or projections, which requires another level of investigation, which is necessary to understand the expression changes of GAD2 gene at cellular level in brain.
In general, the findings presented here confirm that splice variants of GAD2 are expressed differently in the human DLPFC across normal brain development and that these two transcripts may be involved in the pathophysiology of schizophrenia and affective disorders. While our results did show that ENST00000428517 is a fetal predominant transcript, its role in development is unknown. In order to address this issue, it would be insightful to conduct behavioral tests on transgenic ENST00000428517 knockout, knockdown and knock-up mice, followed by molecular biological experiments to observe if there are any changes in brain morphology and neurochemistry. It is known that GAD2 knockout mice are susceptible to seizures, have impairments in plasticity-related tasks, such as ocular dominance plasticity during the critical period and induction of early LTD in the visual cortex [14,16,17,48]. The identification of a fetal predominant novel truncated transcript produced by GAD2 suggests that this gene may play a more complex role in brain development and function than is widely appreciated.