Increases in [3H]Muscimol and [3H]Flumazenil Binding in the Dorsolateral Prefrontal Cortex in Schizophrenia Are Linked to α4 and γ2S mRNA Levels Respectively

Background GABAA receptors (GABAAR) are composed of several subunits that determine sensitivity to drugs, synaptic localisation and function. Recent studies suggest that agonists targeting selective GABAAR subunits may have therapeutic value against the cognitive impairments observed in schizophrenia. In this study, we determined whether GABAAR binding deficits exist in the dorsolateral prefrontal cortex (DLPFC) of people with schizophrenia and tested if changes in GABAAR binding are related to the changes in subunit mRNAs. The GABA orthosteric and the benzodiazepine allosteric binding sites were assessed autoradiographically using [3H]Muscimol and [3H]Flumazenil, respectively, in a large cohort of individuals with schizophrenia (n = 37) and their matched controls (n = 37). We measured, using qPCR, mRNA of β (β1, β2, β3), γ (γ1, γ2, γ2S for short and γ2L for long isoform, γ3) and δ subunits and used our previous measurements of GABAAR α subunit mRNAs in order to relate mRNAs and binding through correlation and regression analysis. Results Significant increases in both [3H]Muscimol (p = 0.016) and [3H]Flumazenil (p = 0.012) binding were found in the DLPFC of schizophrenia patients. Expression levels of mRNA subunits measured did not show any significant difference in schizophrenia compared to controls. Regression analysis revealed that in schizophrenia, the [3H]Muscimol binding variance was most related to α4 mRNA levels and the [3H]Flumazenil binding variance was most related to γ2S subunit mRNA levels. [3H]Muscimol and [3H]Flumazenil binding were not affected by the lifetime anti-psychotics dose (chlorpromazine equivalent). Conclusions We report parallel increases in orthosteric and allosteric GABAAR binding sites in the DLPFC in schizophrenia that may be related to a “shift” in subunit composition towards α4 and γ2S respectively, which may compromise normal GABAergic modulation and function. Our results may have implications for the development of treatment strategies that target specific GABAAR receptor subunits.


Introduction
The majority of inhibition in the cerebral cortex of the mammalian brain is mediated by the interaction of the neurotransmitter c-aminobutyric acid (GABA) with ionotropic GABA A receptors (GABA A R). These ligand-gated channels allow chloride ions to enter neurons, hyperpolarising their membrane and attenuating their activity. Dysfunctional cortical inhibition has been suggested as a mechanism through which symptoms of schizophrenia are mediated [1].
The functional GABA A R is composed of five subunits, typically two a's, two b's and a single c/d or e subunit, assembled into a pentamer. Nineteen subunits of the GABA A R can be grouped together in 8 families through sequence homology (a1-6, b1-3, c1-3, d, e, h, p, r1-3) [2]. The prevalent c2 subunit has a wellknown splice variant in which exon 10 (24 base pairs) is removed producing the c2 short isoform (c2S) [3,4]. Multiple binding sites are found on the GABA A R with an orthosteric binding site for GABA at the interface of the a and b subunits and an allosteric binding site for benzodiazepines at the junction of an a (a1, a2, a3 or a5) and a c subunit [2,[5][6][7][8][9]. The specific subunits present determine the pharmacological sensitivity, synaptic localisation and function of the GABA A R [10] and these subunits change during normal human cortical development as well as in disease states [5,[11][12][13][14].
Changes in the expression and function of GABA A R have been strongly implicated in schizophrenia. Binding studies targeting the orthosteric site of GABA A R have robustly showed an increase in [ 3 H]Muscimol binding particularly in the dorsolateral prefrontal cortex (DLPFC), and also in the caudate nucleus, posterior and anterior cingulate cortices, superior temporal gyrus, and hippocampal formation of post-mortem schizophrenic brains compared to controls [15][16][17][18][19][20][21][22]. This finding has been reinforced by the observation of increased GABA A R protein subunit a2 in the prefrontal cortex in schizophrenia [23]. However, studies on benzodiazepine binding (to the allosteric GABA A receptor site) in schizophrenia show conflicting results: Increased [ 3 H]Flunitrazepam binding has been reported across a variety of cortical areas including the medial frontal cortex, orbitofrontal cortex, temporal gyrus, and the parahippocampal cortex with the DLPFC yet to be assessed [24,25]. Other post-mortem [26] or imaging [27][28][29] studies have reported no changes or decreased [30] benzodiazepine binding in schizophrenia.
At the mRNA level, results have also been inconsistent with studies reporting an increase in a1 and a5 [31], a decrease in a1, a5 and b2 [12], a decrease in c2S [32] and our own group observing no diagnostic change in a1-4 and a decrease in a5 in people with schizophrenia [11]. Presumably, a substantial heterogeneity in the pathophysiology of schizophrenia coupled with small cohort sizes contributes to the lack of consistency amongst the above findings. Clarification on the alterations in the expression of GABA A R subunits is important as they may contribute to changes in both the orthosteric and allosteric binding sites in schizophrenia [5,33,34] affecting pharmacological receptor properties and response to treatment. However, these parameters have not been studied together in the same cohort.
Therefore, in the present study, we used a fairly large cohort of 37 schizophrenia cases and their matched controls to examine whether deficits in the GABA orthosteric and benzodiazepine allosteric binding site of the GABA A R exist in the DLPFC in schizophrenia. An important aspect of our study was to determine, in the same cohort, if mRNA expression of the subunits involved in the formation of these binding sites (a and b for GABA binding site and a and c or d for benzodiazepine binding site) were also different in schizophrenia and to determine how they may relate to changes in binding levels.

Human post-mortem brain samples and ethics statement
All research was conducted in accordance with the latest version of the declaration of Helsinki and approved by the Human Research Ethics Committees at the University of Wollongong (#HE99/22) and at the University of New South Wales (#HREC0761) that follow the guidelines set out in the National Statement on Ethical Conduct in Research involving humans (http:/www.nhmrc.gov.au). Written consent for use of tissues in the study was obtained from next of kin. Characterization and tissue preparation for this Australian schizophrenia cohort has been described in detail previously [35]. Tissue samples and frozen cryostat sections were prepared from the large cohort of schizophrenia (n = 37) and controls (n = 37) cases matched for age, gender, pH, and post-mortem interval (see Table 1 for demographics information) [35].

Tissue dissection and section preparation
Tissue dissection has been described in detail previously [35]. Briefly, at autopsy, brain weight and volume were determined  [36]. The fresh tissue was cut into ,1 cm coronal slices and various anatomical areas were dissected for separate freezing. For the DLPFC (Brodmann's area 46, Figure 1A) dissections, frozen tissue was dissected on a dry ice platform using a dental drill (Cat# UP500-UG33, Brasseler, USA) for homogenates. DLPFC tissue (average weight of tissue ,0.5 g grey matter tissue from the crown of the middle frontal gyrus) was obtained from an adjacent coronal slab corresponding to the middle one-third (rostral caudally) found anterior to the genu of the corpus callosum. Coronal tissue sections of the DLPFC containing the superior and inferior frontal sulcus were cut (14 mm) on a cryostat, thaw mounted onto microscope slides and stored at 280uC until use.

In vitro autoradiography
[ 3 H]Muscimol autoradiography was carried out based on the method described previously [17,21,30] with minor modifications. All sections were processed simultaneously to minimize experimental variance. On the day of the experiment, sections were preincubated three times for 5 min at 4uC in 50 mM Tris citrate buffer (pH 7.0). Sections were then incubated for 60 min at RT in the same buffer with the addition of 20 nM [ 3 H]Muscimol (specific activity 35.6 Ci/mmol, Perkin Elmer, USA). Non-specific binding was determined by incubating adjacent sections in the same solution with the addition of 100 mM GABA (Sigma). The concentration of [ 3 H]Muscimol was measured in 10 ml aliquots taken from the incubation mixture. After the incubation, sections were washed two times for 1 minute in cold (4uC) 50 mM Tris citrate buffer (pH 7.0). Sections were then dipped briefly in cold distilled water and then air dried.
[ 3 H]Flumazenil (Ro-15-1788) autoradiography was carried out based on the method described in Hand et al., 1997 [37] with minor modifications. On the day of the experiment, sections were pre-incubated for 20 min at room temperature in 50 mM Tris (pH 7.4). Sections were then incubated for 80 min at room temperature in the same buffer with the addition of 5 nM [ 3 H]Flumazenil (specific activity 83.4 Ci/mmol, Perkin Elmer, USA). Non-specific binding was determined by incubating adjacent sections in the same solution but in the presence of 10 mM flumazenil (Sigma). The concentration of [ 3 H]Flumazenil was measured in 10 ml aliquots taken from the incubation mixture. After the incubation, sections were washed two times for 1 minute in cold 50 mM Tris buffer (pH 7.4). Sections were then dipped briefly in cold distilled water and then dried.
Following the assays, dried sections were apposed to Kodak Biomax MR films together with autoradiographic standards (

Quantitative analysis of autoradiographic images
All films were analyzed by using a computer-assisted image analysis system, Multi-Analyst, connected to a GS-690 Imaging Densitometer (Bio-Rad, USA). Quantification of receptor binding in DLPFC was performed by measuring the average optical density in adjacent brain sections. Both ligands presented a similar distribution binding pattern across the grey matter of the DLPFC with ''superficial layers'' (I, II and III) having higher binding compared to ''deep layers'' (IV, V and VI) ( Figure 1B). Nonspecific binding was found to be negligible (,5%) for both [   2.5. Total RNA isolation and RNA quality assessment Total RNA was extracted from ,300 mg of frozen tissue per subject using Trizol (Invitrogen, Carlsbad, California) according to the manufacturer's instructions (Kozlovsky et al., 2004). The quality of extracted total RNA was determined using the Agilent Bioanalyzer 2100 (Agilent Technologies, Palo Alto, California). A volume containing 100-200 ng RNA was applied to an RNA 6000 Nano LabChip, without heating before loading. The RNA integrity number (RIN) was used as an indicator of RNA quality, ranging from 1 (lowest quality) to 10 (highest quality). The cDNA was synthesized in three reactions of 3 mg of total RNA in a 26.25 ml reaction using the Superscript First-Strand Synthesis Kit (Invitrogen) according to the manufacturer's protocol.

Statistical analysis
Statistical analyses were conducted using PASW Statistics 18.0 (SPSS, IBM) and GraphPad Prism (CA, USA) statistical packages. The data were normally distributed. All cases were included in the analyses. Mean values for binding and mRNA expression are reported 6 standard error mean (SEM). Student's t-tests were used to compare the mean brain pH, age at death, PMI, freezer storage time, brain weight, brain volume and RIN between the schizophrenia and control groups. We tested the continuous variables (brain pH, age at death, PMI, freezer storage time, brain weight, brain volume, RIN, age of illness onset, illness duration, and estimated lifetime exposure to antipsychotics) for significant Pearson's correlations with binding and mRNA subunits expression. Non-continuous variables such as gender (male/female), hemisphere (right/left), cause of death (suicide/other), daily alcohol intake (none 0, low: 1, moderate: 2, and high: 3), and tobacco smoking (moderate: 1, and heavy: 2) were used as grouping variables with t-tests or one-way ANOVA to evaluate their effects on binding and mRNA expression.
GABA A R bindings between diagnostic groups (schizophrenia and control) were analysed using two-way analysis of covariance (ANCOVA), with diagnosis (schizophrenia and control) and layers (superficial and deep) as independent variables and co-varying for continuous variables that were significantly correlated with binding. GABA A R subunit mRNA expression levels between diagnostic groups (schizophrenia and control) were analysed using ANCOVA, co-varying for continuous variables that were significantly correlated with mRNA subunit expression.
To assess the relative implication of GABA A R mRNA subunits on binding measures in patients with schizophrenia and controls, we used forward regression analysis with all a and b mRNA subunits as predictors of the GABA orthosteric binding site targeted with [ 3 H]Muscimol, and all a, c and d mRNA subunits as predicators of the benzodiazepine allosteric binding site, in both layers of the DLPFC. The criterion probability of F for predicators to enter the regression model was set at #0.05.

Results
The mean age, pH, PMI, freezer storage time, brain weight, brain volume and RIN did not differ between the schizophrenia and control groups as shown by non-significant Student t-tests

Disease related effects on [ 3 H]Muscimol and [ 3 H]Flumazenil binding
Significant associations were found between [ 3 H]Muscimol binding and age and freezer storage time in both superficial and deep layers (age: 20.523,r,20.524, p,0.001, freezer storage: 0.403,r,0.425, p,0.001, see Table S1 for details and section 3.4). In the whole cohort, these associations accounted for less than 20% of the variance and were co-varied for by performing twoway ANCOVA. This analysis revealed  Table S1 for details and section 3.4), that accounted for less than 28% of the variance and were covaried for in the ANCOVA.

Disease related effects on GABA A R mRNA subunits expression (bs and cs)
One-way ANCOVA (controlling for demographic variables significantly correlated to GABA A R mRNA subunit expression) revealed no significant effect of diagnosis ( Figure 3) in any of the nine transcripts examined (b1-b3, c1, c2, c2S, c2L, c3 and d) (all F's,3.4, all p's.0.05).

Correlation and regression analysis
Pearson's correlation between [ 3 H]Muscimol binding and a (measured in our previous study [11]) and b mRNA subunits are presented in Table 2 with separate analyses in the whole cohort, in patients with schizophrenia only, and in controls only. Pearson's correlation between [ 3 H]Flumazenil binding and a, c and d mRNA subunits are presented in Table 3 with separate analyses in the whole cohort, in patients with schizophrenia only, and in controls only.
In patients with schizophrenia, we used forward regression analysis with all a and b mRNA subunits (constituting the GABA orthosteric binding site) as predictors for [ 3 H]Muscimol binding in both layers of the DLPFC. Interestingly, the a4 subunit mRNA was most strongly associated with binding and alone could account for 34% and 23% of the variance in [ 3 H]Muscimol binding in superficial and deep layers of the DLPFC respectively (see Figure 4 for correlation plot and Table 4 for regression results). In contrast, in the superficial layers of the control group, no predictors (mRNA subunits) passed the criterion probability of F to enter the regression model. In the deep layers however, b2 mRNA subunit contributed the most (14%) to [ 3 H]Muscimol variance in controls (Table 4).
In patients with schizophrenia, we used forward regression analysis with all a, c and d mRNA subunits (constituting the benzodiazepine allosteric binding site) as predictors of [ 3 H]Flumazenil binding in both layers of the DLPFC. Interestingly, the c2S subunit mRNA was most strongly associated with binding and alone could account for 31% and 13% of the variance in [ 3 H]Flumazenil binding in superficial and deep layers of the DLPFC respectively (see Figure 5 for correlation plot and Table 5 for regression results). In contrast, when applying the same regression method in the control group, several mRNA subunits contributed to [ 3 H]Flumazenil variance, including c1, c2 pan and c2L in the superficial layers and a1, a2, c2 pan and c2S in the deep layers of the DLPFC (Table 5).

Effect of continuous and non-continuous variables
In the whole cohort and in each diagnostic group, age at death was negatively correlated with  Table S1, along with significant correlations between GABA A R mRNA subunit expression and continuous variables.
The t-tests for gender showed no significant male/female variation in binding measures (data not shown). Concerning mRNA subunits, only a4 showed a significant increase in males compared to females (+15.2%, t(71) = 2.003, p = 0.049). All the other subunits did not show a statistically significant change according to gender (data not shown).
The hemisphere analysed, right or left, did not have any significant effect on binding or mRNA subunits measures (data not shown). Cases who died by suicide in the schizophrenia group were characterised by an increase in

Disease-related effects
We report increases of both the GABA and benzodiazepine binding sites of the GABA A R in the DLPFC in schizophrenia that are linked to a4 and c2S mRNA subunits respectively. Our findings are in line with studies reporting increases in [ 3 H]Muscimol binding (GABA binding site) in the DLPFC [18] and other cortical regions [16,21]. The increase we observed in the benzodiazepine binding site in the DLPFC in schizophrenia confirms previous studies which report increases across a variety of cortical areas including the medial frontal cortex, orbitofrontal cortex, temporal gyrus, and the parahippocampal cortex [24,25] but contradicts other studies that found no changes [26] or decreased binding in individuals with schizophrenia [30].
It is known that many demographic and peri-mortem factors including medication before death can influence binding and/or mRNA expression in post-mortem studies. Although in the present study several associations between these factors and binding/ mRNA expression were identified, the sizes of these associations in the whole cohort were small and taken into account in the ANCOVA analysis. Moreover, no significant correlations between [ 3 H]Muscimol and [ 3 H]Flumazenil binding and antipsychotic medication (lifetime chlorpromazine) were found in the schizophrenia group. This suggests that the observed increases in GABA A R binding are not secondary to medication or demographic and/or peri-mortem factors.
We found significant correlations between [ 3 H]Muscimol and [ 3 H]Flumazenil binding in the whole cohort and in the schizophrenia group, but not in the control group. While the association between [ 3 H]Flumazenil and [ 3 H]Muscimol binding was statistically significant only in the patient group it cannot be discounted that a similar, though weaker, association was present in controls. This suggests that in general there is a positive association between the orthosteric GABA and allosteric benzodiazepine binding site but this is stronger in the disease state. The increases in [ 3 H]Muscimol and [ 3 H]Flumazenil binding we observed in the present study were not accompanied by cohortwide changes in any of GABA A R subunit mRNA measured here. Despite this lack of overall change in primary transcript levels in schizophrenia, regression analysis revealed that, in the schizophrenia group, the a4 subunit mRNA levels contributed the most to the [ 3 H]Muscimol binding variance in both layers of the DLPFC. In contrast in the control group, no mRNA reached a significant level of correlation with binding to enter the regression model in the superficial layers, whereas in the deep layers, b2 mRNA was the one mainly contributing to [ 3 H]Muscimol binding variance.
The c2 subunit is an important functional determinant of GABA A receptors and is essential for formation of high-affinity benzodiazepine binding sites [39][40][41]. Using transgenic mice, Baer et al (2000) demonstrated that expression of either the long or the short c2 splice variant resulted in mice that had indistinguishable [ 3 H]Flumazenil binding, c2 protein levels and phenotypes [42]. It is interesting that mRNA levels encoding the c2S subunit was the   These observations suggest that the relationships between binding and mRNA expression at the benzodiazepine binding site are probably indirect and quite varied amongst the control population.
Overall the results of our regression analysis suggests that there may be a ''shift'' in the subunit composition of GABA A R towards inclusion of a4 when studying the GABA orthosteric binding site and c2S when analysing the benzodiazepine allosteric binding site in schizophrenia.

Functional and therapeutic implications
Our results have several functional and therapeutic implications. The increases in both [ 3 H]Muscimol and [ 3 H]Flumazenil binding together with the putative ''shift'' in the subunit composition support the theory of altered GABAergic neurotransmission in the DLPFC. The increases in binding observed in the present study and others [18,20,23,27,28] could be due to compensatory adaptations for reduced GABA synthesis [11,[43][44][45][46][47] and reuptake [48] at parvalbumin-positive chandelier neurons that synapse on the axon initial segment of pyramidal cells [26,53,54]. However, compensation at GABA A R level has been suggested to be insufficient to restore the synchronized oscillatory activity of cortical pyramidal neurons, in the gamma band range, necessary for normal cognitive functioning [49]. If a dysfunctional GABA system in schizophrenia does lead to cognitive impairments, therapeutic strategies to correct or modulate disrupted GABAergic pathways could be used to treat cognitive symptoms of schizophrenia, regardless of whether these GABAergic deficits are primary or compensatory [50].
Our results point to therapeutic potential of drugs that target a4 or c2S subunits in the treatment of symptoms of schizophrenia. GABA A R that include a4 subunits are generally insensitive to diazepam, the prototypical benzodiazepine [51]. If, as we propose, schizophrenia patients undergo a ''shift'' in their subunit composition of GABA A R towards a4 subunits, this could account for the 50% failure of response to benzodiazepine prescribed as ''add-on'' therapy in schizophrenia [52,53]. Moreover, GABA A R that include a4 subunits, assembled with a c or d subunit [50], are located at extrasynaptic sites and are responsible for tonic inhibition. Tonic inhibition is defined as the constant activation of extrasynaptic receptors that reduces the probability of generating an action potential [54]. A ''shift'' in the subunit composition of GABA A R towards a4 subunits in our schizophrenia cohort might therefore reflect an abnormal shift towards extrasynaptic GABA A R mediated tonic inhibition, that could compromise normal GABAergic modulation of cortical excitability, contributing to schizophrenia symptoms.
GABA A R containing the c2 subunit predominantly mediate phasic inhibition and have a synaptic localisation [50]. Phasic inhibition is defined as the rapid and synchronous opening of synaptic receptors that results in an inhibitory postsynaptic potential [54]. The long isoform of c2 subunit (c2L) has been shown to preferentially accumulate at synapses through the protein kinase C (PKC) phosphorylation of Ser343. Conversely, the c2S isoform lacking Ser343 [7,8] might not be able to ''transfer'' to synapses and rather localise extrasynaptically. In our schizophrenia cohort, a putative ''shift'' in the subunit composition of GABA A R towards c2S isoforms possibly reflects a similar ''shift'' in the ratio of extrasynaptic/synaptic GABA A R as we suggested for a4 subunits. Alternatively, an increase in the c2S isoform may be linked to down-stream changes in PKC phosphorylation signalling associated with abnormal modulation and function of the GABA A R in schizophrenia.
Therefore, increased contributions of both a4 and c2S subunits to the formation of the orthosteric and allosteric GABA A R binding sites in schizophrenia, may result in an increase of GABA A Rs located extrasynaptically, and thus affect tonic inhibition of cortical neurons. Although the implication of synaptic GABA A R (controlling phasic inhibition) in the generation of rhythmic activity in neuronal networks (thus cognitive processes) is established [54], the role of extrasynaptic GABA A R (responsible for tonic inhibition) is still unclear. Mutant mice with a loss of tonic inhibition in the hippocampus exhibited an increase in the power of gamma band oscillations [55] and increased gamma band power at baseline (non-evoked) has been reported for people with schizophrenia [56]. It is possible that a dysfunction in tonic inhibition affects gamma band oscillations resulting in cognitive impairments in schizophrenia. Interestingly, neurosteroids acting on extrasynaptic GABA A R have been found to improve cognition in schizophrenia patients, suggesting that extrasynaptic receptors could be involved in cognitive impairments in schizophrenia [57,58]. These studies, together with ours, reinforce the interest in developing therapeutics that target extrasynaptic GABA A Rs for the treatment of cognitive symptoms in schizophrenia.

Conclusions
Our study showed parallel up-regulation of the GABA orthosteric and benzodiazepine allosteric GABA A R binding sites in the DLPFC of a large cohort of schizophrenia patients and controls. We reported associations between a4/c2S mRNA subunit and binding that may suggest a shift in GABA A R subunit composition in schizophrenia affecting receptor localization and function and may have implications for the development of novel treatment strategies. Figure S1 Design for the custom GABA A R c2 long and short variants TaqMan assay. Probes (in grey) span exonexon junctions in order to eliminate genomic signals. One set of primers was used to amplify transcripts both including and excluding exon 10 (in green). (TIF)