Circuitry-based gene expression profiles in GABA cells of the trisynaptic pathway in schizophrenics versus bipolars

Abstract

Significant reductions in GABAergic cell numbers and/or activity have been demonstrated in the hippocampus of subjects with schizophrenia and bipolar disorder. To understand how different subpopulations of interneurons are regulated, laser microdissection and gene expression profiling have been used to “deconstruct” the trisynaptic pathway, so that subtypes of GABA cells could be defined by their location in various layers of CA3/2 and CA1. The results suggest that the cellular endophenotypes for SZ and BD may be determined by multiple factors that include unique susceptibility genes for the respective disorders and altered integration among hippocampal GABA cells with extrinsic and intrinsic afferent fiber systems. The extensive and intricate data that has come from this study has provided insights into how a complex circuit, like the trisynaptic pathway, may be regulated in human hippocampus in both health and disease.

Gene expression plays a central role in the regulation of neural circuitry involved in cognitive behavior. Identifying molecular mechanisms within neurons of complex circuits presents one of the foremost challenges to understanding the human brain. In the past 20 years, postmortem studies of schizophrenia (SZ) and bipolar disorder (BD) have provided evidence for a dysfunction of GABAergic neurons in frontal cortices and hippocampus (1). It is well known that GABAergic interneurons provide potent inhibitory modulation of principle neurons (2) and are critical for the regulation of feed-forward inhibition (3) and oscillatory rhythms (4, 5). A network of genes involved in the regulation of glutamate decarboxylase 67 (GAD₆₇), a key marker for the GABA cell phenotype (6), shows changes in expression in SZ that are different from those seen in BD, suggesting that there may be unique molecular endophenotypes for each disorder. To learn more about the molecular regulation of hippocampal GABA cells in SZ and BD, a combination of laser microdissection (LMD) and gene expression profiling has been used to “deconstruct” the trisynaptic pathway into subtypes of GABA neurons defined by their location and connectivity. Several clusters of genes have been examined across a broad array of cellular functions that include transduction, signaling, metabolism, translation, transcription and cell cycle regulation. These clusters have been separately analyzed in various layers and sectors with a preponderance of GABA cells. To our knowledge, this is the first demonstration that the regulation of gene expression in GABA cells varies not only according to diagnosis, but also to location within a complex circuit.

Results

Samples of the stratum oriens (SO), pyramidale (SP), and radiatum (SR) of sectors CA3/2 and CA1 (Fig. S1) were obtained for the normal control (CON), SZ and BD groups (Table S1). The SO and SR samples were collected by LMD from a location where there were very few glial cells. Consistent with conventional microscopic criteria (7), the interneurons in these layers showed a distinctive cresyl violet staining of their cytoplasm. Glial cells, however, typically showed no such staining, appearing as isolated nuclei suspended within the neuropil. This fundamental difference between the two cell types suggests that the concentrations of RNA in interneurons of SO and SR is much higher than in glial cells. Additionally, preliminary assessments, using in situ hybridization demonstrated that antisense RNA associated with a broad array of genes (i.e., GAD₆₇, GAD₆₅, GluR5, GluR6, KCNJ3, KCJN6, HCN3, HCN4, HDAC1, DAXX, PAX5, Runx2) was found over neuronal cell bodies, whereas no grains were observed over glial cell profiles.

As shown in Table S1, the BDs showed different ratios for laterality (R/L) and gender (M/F) when compared with the CON and SZ groups that were well-matched with one another. An earlier study from our group did not find that laterality or gender influenced the nature of gene expression changes (8). Additionally, the expression of mRNA for BDNF and trkB also shows no relationship to these two variables (9). Although an influence of these potential confounds on gene expression cannot be definitively ruled out, it does seems unlikely that these variables could account for the unusually robust differences noted in the BD group.

The quality of amplified antisense RNA used was evaluated using 18S/28S ratios, 3′/5′ ratios and the percentage of present calls generated with dChip 1.0 (see ref. 6). The percentage of present calls showed good correspondence between the SZ and BD groups. When the changes in expression between the groups were compared, the number of genes satisfying the inclusionary criterion of P ≤ 0.05 (CON vs. SZ or BD) for the post hoc analysis of GenMapp biopathways and/or clusters resulted in large numbers being detected in SO of both CA1 and CA3/2 of the BD and SZ groups (Table 1). In the SR and SP of both sectors, very few genes satisfied the inclusionary criterion in both groups, particularly the BDs. The number of genes with P ≤ 0.05 level did not covary with the percentage of present calls (Table 1). For example, in the SP of CA3/2 of BDs and SZs the number of genes at the P ≤ 0.05 level was 181 and 244, respectively, whereas the percentage of present calls in the same samples of all three groups ranged from 30–36%.

Table 1.

Numbers of genes showing significant changes and percentages of present calls

Group	Sector	Layer	Number of genes (P ≤ 0.05)	Percentage of present calls
BD	CA1	Oriens	4,062.0	42.3
SZ			1,028.0	33.0
CON				30.5
BD	CA3/2	Oriens	2,979.0	36.4
SZ			1,875.0	27.3
CON				32.4
BD	CA1	Radiatum	173.0	34.5
SZ			539.0	33.1
CON				35.0
BD	CA3/2	Radiatum	339.0	27.8
SZ			15.0	21.9
CON				22.6
BD	CA1	Pyramidale	220.0	35.7
SZ			133.0	30.5
CON				30.3
BD	CA3/2	Pyramidale	181.0	30.0
SZ			244.0	36.2
CON				34.5

The number of genes with P ≤ 0.05 refers to comparison of bipolars (BDs), and schizophrenics (SZs) with normal controls (CONs) using a 2-way ANOVA. The percent of present calls were obtained with dCHIP 1.3.

QRT-PCR Validation.

As shown in Fig. S2 and predicted from the microarray data, the KCNJ3 gene (G protein dependent inwardly rectifying potassium channel; GIRK) (10) and the hyperpolarization-activated Ih (HCN3 and 4) (11) channels showed significantly increased expression in SZs, but decreased expression in BDs. The GRIA1 (ionotropic AMPA1 receptor) gene, that is thought to be a susceptibility gene for schizophrenia (12), showed significant negative fold changes in this group, but not in BDs. These various changes all occurred in the direction observed with microarray-based gene expression profiling.

Analyses of Functional Biopathways and Clusters.

The composite probability, P_c, was calculated (13) for 48 different functional gene clusters, including those associated with synaptic function, signaling, metabolism, transcription/translation and other miscellaneous categories (Table S2). For each biopathway/cluster in SO of CA3/2 for the CONs vs. BDs, the number of genes meeting the P ≤ 0.05 criterion (N_i), the total number of genes in the cluster (N_t) and the overall percentage change were calculated for each layer and sector of both groups. The most significant P_c values were observed in the SO of CA3/2 and CA1 of the BD and SZ groups (Table S2). In the SP and SR of both sectors, very few biopathways/clusters had a P_c ≤ 10⁻¹⁰. Some examples of gene clusters that showed very prominent changes in SO of CA3/2 and CA1 in BDs are neurogenesis (P_c = 10⁻¹⁰⁹ and 10⁻¹¹⁷, respectively), cell cycle regulation (P_c = 10⁻⁸¹ and 10⁻⁹², respectively) and synaptic transmission (P_c = 10⁻⁸⁷ and 10⁻⁹¹, respectively). In the SZs, similar clusters showed robust changes, except that the magnitude of the P_c values was not quite as robust as those seen in BDs, generally because the number of genes with P ≤ 0.05 was smaller (Table S2). The P_c values for neurogenesis (10⁻⁷⁷ and 10⁻³⁶, respectively), cell cycle regulation (10⁻⁶⁷ and 10⁻²², respectively) and for synaptic transmission (10⁻⁵⁴ and 10⁻²², respectively) generally showed a smaller negative exponent. The genes associated with the same biopathway/cluster of the same layer/sector in the SZs vs. BDs were typically quite different in terms of the specific genes and their direction of change (Fig. S3). For example, in the SO of CA3/2, the synaptic transmission and potassium ion transport clusters, respectively, contained many genes with predominantly decreased expression in BDs (blue). In SZs, however, the genes in the same clusters were both increased and decreased.

Mapping Genes Associated with Synaptic Transmission and Potassium Ion Transport to the Trisynaptic Pathway.

To evaluate the potential significance of these various changes for the activity of the trisynaptic pathway, the gene expression profile changes for various functional categories were mapped onto schematic diagrams of GABA cells representing the various layers and sectors studied. To focus primarily on GABA cells, the layers reported herein are those in which interneurons are the predominant neuronal cell type and include the SO and SR of CA3/2 and CA1. SP is not reported below because it contains both pyramidal cells and GABA cells in a ratio of ≈10:1. If GAD₆₇ expression was significantly decreased, particularly if other signaling and/or metabolic pathways also showed an overall decrease in expression, a cell was given a blue background coloration (Fig. S4). If, however, GAD₆₇ expression was unchanged and/or other signaling and metabolic pathways were markedly up-regulated, then the cell was given a red background to indicate that its neuronal activity may be significantly increased. Because of the importance of synaptic transmission for the up- and down-stream regulation of the trisynaptic pathway, this functional cluster is given primary emphasis. Some of the genes related to potassium ion transport were differentially regulated in SZs and BDs and, because of their importance for understanding changes in synaptic mechanisms, are also included in the results described below:

SO of CA3/2.

Schizophrenia.

As shown in Fig. S4A for genes related to synaptic transmission, GAD₆₇ (GAD1) and GAD₆₅ (GAD2) expression showed significant decreases in expression in the SO of CA3/2. There was also an increase in the rho1 subunit of the GABA_A receptor (GABRR1). The rho1 subunit has been associated with GABA-to-GABA ionotropic inputs to cerebellar Purkinje cells (14) and this finding is consistent with a previously reported increase of specific GABA_A binding activity on interneurons in CA3/2 in SZs (15). The expression of another gene involved in the degradation of GABA, ALDH5A1 gene, i.e., succinate semialdehyde dehydrogenase (16), was also significantly decreased at this locus and could potentially be associated with a compensatory increase of GABA concentrations. Several other genes in the synaptic transmission cluster also showed significant changes in SZs (refer to Fig. S3A). Those genes that showed decreased expression included the GRIA1 [fold change (FC) = −1.77, P = 0.008], GRIA3 (FC = −1.47, P = 0.05), GRIK1 (the kainate receptor subunit, GluR5; FC = −1.35, P = 0.05) and the metabotropic glutamate receptor (mGluR) subunits GRM3 (FC = −1.91, P = 0.001) and GRM5 (FC = −2.02, P = 0.004). In contrast, some glutamate receptor subunits were up-regulated, including the GRIK2 (GluR6; FC = 1.5, P = 0.04), GRIK 3 (GluR7; FC = 1.6, P = 0.01) and GRIN2A (NR2A; FC = 1.46, P = 0.04). The synaptosomal 25-kDa protein (SNAP25) showed a very robust decrease in expression (FC = −2.24, P = 0.006). For the voltage-gated potassium ion transport cluster (Fig. S3B), the G protein dependent, inwardly rectifying potassium (KCNJ3; FC = 1.51 P = 0.015) and hyperpolarization-activated cationic (HCN3; FC = 1.68; P = 0.001) channels were also significantly up-regulated.

Bipolar disorder.

As shown in Fig. S3A and S4C, BDs, like the SZs, also showed a significant decrease in the expression of GAD₆₇ and GAD₆₅ in SO of CA3/2. For the neurotransmitter receptors, glutamate and dopamine subtypes generally showed changes. Although the expression of the GRIK1 kainate receptor subunit (GluR5) was also significantly decreased, there were no other changes in AMPA or NMDA receptors. In contrast, the expression of muscarinic (CHRM5; FC = −1.88, P = 0.02) and nicotinic cholinergic receptor polypeptides alpha 3 (CHRNA3; FC = −1.34, P = 0.03), alpha 7 (CHRNA7; FC = −1.58, P = 0.04), delta (CHRND; FC = −1.23, P = 0.04) and epsilon (CHRNE; FC = −1.51, P = 0.03) were all decreased in SO of CA3/2 in BDs. As in SZs, mRNA for the synaptic protein SNAP25 (FC = −2.98, P = 0.002) was also decreased; in BDs expression of synapsin II (FC = −1.56, P = 0.03) was also reduced. With one exception, the expression of all genes associated with potassium ion transport was decreased in the BDs, including KCNJ6 or GIRK (FC = −2.31, P = 0.003) and HCN4 or Ih (FC = −1.38, P = 0.04) (see Fig. S3B).

SO of CA1.

Schizophrenia.

The expression of GAD 65 showed a significant decrease (6) (Figs. S4B and S5). Other genes related to synaptic transmission, however, did not show prominent changes in this group (Fig. S5 Left). Additionally, mRNA for the NR1 subunit (GRIN1) of the NMDA receptor was significantly decreased, as were those for two subunits of the nicotinic cholinergic receptor (CHRND and CHRNA2), GRM4 or the mGluR receptor and norepinephrine transporter (SLC6A2). Unlike SO in sector CA3/2, potassium ion transport in the KCNJ and HCN classes did not show significant changes in SZs at this locus. However, other genes in this cluster, including several voltage-gated potassium channels (e.g., KCNS3, KCND1, KCNE1, KCNA2 and KCNAB3) showed significant decreases in expression (Fig. S6).

Bipolar disorder.

GAD₆₇ and GAD₆₅ showed no expression changes (6), although the overall pattern of expression among the various functional clusters of genes was otherwise uniquely up-regulated in this layer/sector (Fig. S5 Right). The GRIA 1,2 and 3 receptor genes (AMPA1,2 and 3) all showed 2-fold increases of expression in SO of CA1; the FC for GRIA 1 was 2.10 (P = 0.04). Other genes showing an up-regulation at this locus included the glutamate (SLC1A3) and GABA (SLC6A1) transporters and glycine receptor β subunit. Genes that showed decreased expression notably included several nicotinic cholinergic receptor subunits (CHRNA4, CHRNB3, CHRNA2, CHRNA1, and CHRNG). Additionally, the GluR5 subunit of the kainate receptor (GRIK 1), the glycine α2 (GLRA2), mGLUR4, NR1 NMDA subunit, dopamine D1 (DRD1), and D3 (DRD3) receptors and norepinephrine transporter (SLC6A2 and -3) all showed decreased expression. For potassium ion transport in the SO of CA1 in BDs (Fig. S6), KCNJ6 and HCN4 showed no changes, except for one inwardly rectifying channel (KCNJ15) that was significantly decreased. Several other voltage-gated potassium channels showed increased (KCNA5 and KCNA3) or decreased (KCNS3, KCND1, KCNAB1 and -3, KCNE1, and KCNA2) expression.

Many genes associated with metabolic pathways were strikingly up-regulated in the SO of CA1 of BDs (Fig. S4D). As shown in Fig. S7, notable examples included glycolysis and gluconeogenesis, Krebs tricarboxylic acid cycle and the electron transport chain where most genes showed robust increases in expression. Genes associated with translation, ribosomal regulation, t-RNA synthesis, proteosomal degradation, TGF-β and Wnt signaling and cell cycle regulation also showed a preponderance of up-regulated genes in the SO of CA1 in BDs (data not shown).

SR of CA3/2 and CA1.

Schizophrenia.

The expression of GAD₆₇ was significantly decreased in SZ of CA3/2 (6); however, none of the functional gene clusters shown showed any significant expression changes in this group (Table S2). Very few functional gene clusters attained significance in the SR; however, the P_c for calcium channel regulation, voltage-gated ion channels, electron transport and neurogenesis exceeded 10^−10. Within the synaptic transmission cluster (data not shown), a small number of genes, including SNAP25, the potassium voltage-gated KCNQ2 channel, 4-aminobutyrate aminotransferase (ABAT) and dynamin I were all up-regulated.

Bipolar disorder.

GAD₆₇ showed no change in expression in SR of CA3/2 of BDs. In CA1, none of the functional gene clusters at this locus in BDs showed significant differences (Table S2).

Psychotropic Medications.

To evaluate the effects of medications, the GenMapp clusters/pathways related to synaptic transmission were analyzed according to low versus high dose treatment with antipsychotic medications (APDs). These evaluations were performed for all sectors and layers reported in this study; however, only those for sector CA3/2 are described below (Fig. S8). The expression profiles on high dose neuroleptics (i.e., those receiving 500 CPZ mg-equivalents/day or higher; 4 cases; average = 775 mg) were compared with those receiving a low dose (i.e., <500 CPZ mg-equivalents per day; 3 cases; average = 312 mg-equivalents per day). The number of genes showing significant changes in the high dose group was 37.5% lower (n = 35) when compared with the low dose group (n = 56 highlighted in red and blue). A similar pattern was observed in other functional clusters, such as those related to cell cycle regulation and neurogenesis (data not shown) where the differences were 36 and 38%, respectively, suggesting that APDs may have a generalized tendency to suppress the expression of genes representing a wide variety of cell functions that were both increased or decreased. Some genes showed a significant change in the low dose group, but not the high dose group (n = 40; 58%). A smaller proportion of genes showed significant changes only in the high dose group (n = 16; 23%). For example, SNAP-25 mRNA was significantly decreased in SZs receiving high dose APDs, but not in those receiving low dose treatment. In contrast, one rodent study reported that chronic APD administration results in an increase of SNAP-25 expression in the CA3 region (17). Other genes, however, may not be influenced by APD exposure, because similar changes were observed in both the low and high dose groups (n = 12; 17%). These genes included, but were not limited to GAD 65, GAD 67, GRM3, GRM5 and SLC1A1 (a neuronal high affinity glutamate transporter).

Only 3 subjects (two SZs and one BD) were treated with APDs alone, but no other mood stabilizing agents. All of the remaining 11 SZ and BD cases received APDs in combination with lithium carbonate and/or other mood stabilizing agents, suggesting that the differences in gene expression vs. CONs observed in the respective groups were probably not related to medication effects.

Discussion

The findings reported herein demonstrate that GABAergic interneurons at specific loci along the trisynaptic pathway show unique expression profiles that vary according to layer, sector and diagnosis (Fig. 1A and 1B). As reported in ref. 6, both the SZs and BDs showed significant decreases in GAD₆₇ expression in layers containing a preponderance of GABA cells, particularly those located in sector CA3/2. Although the number of cases in each group was small, the fold changes and probabilities that were detected were quite robust, indicating that the variance in the data for each group was low. As shown in Fig. 1, changes in expression profiles noted in these cells of SZs and BDs may potentially limit or enhance their ability to provide inhibitory modulation to pyramidal neurons. However, substantial differences in expression were also found for many functional gene clusters (Fig. 1), suggesting that a reduced expression of GAD₆₇ in SO of CA3/2 occurs within the context of complex molecular changes that are unique to the two disorders. Additionally, the findings presented here suggest that GABA cell regulation may be related to a complex interaction of layer and sector with psychiatric diagnosis and exposure to psychotropic medications.

Fig. 1.

Schematic diagrams of the trisynaptic pathway showing pyramidal neurons (triangular) and GABA cells (square) in stratum oriens (SO), stratum pyramidal (SP), and stratum radiatum (SR) of sectors CA3/2 and CA1. The perforant path projection from the entorhinal cortex project to the granule cells and these in turn send mossy fiber projections that synapse on the apical dendrites of pyramidal neurons in sector CA3/2. The latter cell sends projects axons into the SO where they travel as Schaffer collaterals that eventually form excitatory synapse with the apical dendrites of pyramidal neurons in sector CA1. Collateral branches of pyramidal neurons in CA3/2 and CA1 form synapses with GABAergic interneurons on the SO. The arrows at the bottom of each diagram show the direction of feed-forward excitation along the trisynaptic pathway and suggest that this may be increased in schizophrenics because of diminished GABAergic tone in both CA3/2 and CA1. In BDs, however, feed-forward excitation may be attenuated at the level of CA1 because of the heightened activity of GABA cells with in SO at this locus.

Medication Effects.

Psychotropic effects may potentially increase the intersubject variability of gene expression profiling within the BD and/or SZ groups. We noted prominent qualitative differences between these two groups, but it is difficult to specifically relate them to one or another class of drugs to which they were exposed (refer to Fig. S9). For example, APDs and mood stabilizers did not appear to influence GAD₆₇ expression. In an earlier study, GAD₆₅-containing terminals showed APD dose-related increases, suggesting that these drugs may contribute to compensatory sprouting of GABAergic terminals, particularly in the SO of CA3/2 (18). Consistent with this, rats receiving chronic administrations of haloperidol showed marked increases in the number of GABA-containing terminals forming axosomatic contacts with pyramidal neurons in the medial prefrontal cortex (19). Moreover, in the current report, mRNA for GAD₆₅ showed no significant changes in either SZs or BDs, suggesting that APD exposure might have increased its expression to normal levels. Most of the SZs and BDs included in this study were treated with both APDs and mood stabilizers, making the interpretation of the complex expression patterns difficult. Gene expression profiles in the two groups were fundamentally different, despite overlapping treatments, suggesting that most of these changes are not related to psychotropic treatment alone. Finally, the decrease in the expression of various nicotinic receptor subunits raises the possibility that smoking cigarettes might have contributed to these changes in SZs (20) and BDs (21). Because BDs do not smoke as heavily as SZs (clinical observation), they might be expected to show less striking changes in nicotinic receptor expression; however, this group showed more subunits with decreased expression when compared with SZs, suggesting that nicotine ingestion alone does not account for the expression changes noted for these receptors.

Functional Implications of Gene Expression Changes.

An important factor that probably influenced gene expression changes is the unique connectivity found within each layer and sector of the trisynaptic pathway (Fig. 1). The mossy fiber system provides afferent inputs to pyramidal neurons and GABA cells in the SR of CA3 (22); kainate (23) and AMPA receptors both contribute to feed-forward inhibition in this circuit (24–26). No changes in the expression of any of the glutamate receptors were observed in the SR of CA3/2 in SZs or BDs. Pyramidal neurons receiving mossy fiber inputs provide excitatory recurrent collaterals to GABAergic cells; these are mediated by AMPA receptors and generate NMDA-mediated LTP responses (27). In SR of CA3/2, AMPA and NMDA receptor subunits showed no expression changes in either group, suggesting that mossy fiber activity may not be altered in SZ or BD.

In addition to SR, GABAergic neurons provide inhibitory modulation to pyramidal neurons in the SO of CA3/2. In SZs, complex changes in the expression of AMPA, NMDA and kainate receptor subunits were observed in this layer/sector; this occurred to a lesser degree in BDs. With GABAergic activity decreased at this locus, these differences in gene expression profiles for glutamate receptors suggest that the integration of glutamatergic inputs with interneurons is fundamentally different in the two disorders.

Neurons of CA3/2 also receive inputs from subcortical regions, such as the hypothalamus, basal forebrain and basolateral amygdala (2). Cholinergic (28) and GABAergic (29) inputs to GABA cells in SO of CA3/2 originate in the septal nuclei, whereas glutamatergic projections (30) originate in the basolateral amygdala (BLa) (31). Experimental stimulation of the BLa is associated with a reduction in the number of GABA cells in sector CA3/2, a pattern remarkably similar to that seen in postmortem studies of SZ (1). Accordingly, GABA cell dysfunction inferred from our gene expression profiling studies in the SO of CA3/2 in SZs and BDs could involve glutamatergic inputs from the BLa (32). Kainate receptors may potentially mediate some of the amygdalar effects on GABA cells and regulate the expression of GAD₆₇ in interneurons of this layer/sector (6). The up-regulation of kainate receptor subunits channels in SO of CA3/2 of SZs could be part of a larger mechanism that promotes neuronal excitability along GABA cell dendrites (32) and increases their firing rate (33, 34). This effect could be potentiated through an activation of Ih channels (35) that are associated with an attenuation of after-hyperpolarizing currents (34, 36) associated with the activation of GABA cells by pyramidal neurons via their recurrent collaterals. The up-regulation of the HCN4 gene is consistent with such a mechanism operating in SZ.

Kainate receptors are also critical for the generation of gamma oscillations (37). Interestingly, Ih channels are involved in pacemaker activity (35) and have the potential to reset the phase currents that comprise oscillatory rhythms (38). Together with kainate receptors, Ih channels expressed by “horizontal,” fast-spiking interneurons located in the region of the oriens-alveus (39) may contribute to the generation of gamma (40) and theta (41, 42) oscillations that influence long-range synchronization (43), involving the induction of GABA currents in pyramidal neurons (44). Oscillatory rhythms are thought to be important in establishing the cognitive relevance of hippocampal output (45). Given the gene expression findings reported here for GABA cells in SO of CA3/2, it is not surprising that abnormalities in gamma oscillations have been reported in schizophrenia (46).

In BD, a prominent decreased in the expression of nicotinic cholinergic receptor subunits was the most prominent change observed in SO of CA3/2 and CA1. These findings are consistent with an earlier postmortem study reporting that this receptor activity was reduced in subjects with psychotic disorders (47). A decrease of excitatory cholinergic activity impinging on interneurons could contribute to GABAergic dysfunction in SO of CA3/2 and CA1 (48). In BDs, GABA cells within the SO of CA3/2 might be particularly compromised in their ability to provide inhibitory modulation, because the kainate receptor subunits and Ih channels also show significantly decreased expression, changes that could promote a diminished firing of GABA cells (33, 34). In CA1 of BDs, the expression of kainate receptor subunits is also decreased; however, dramatic increases in the expression of genes associated with a broad range of metabolic signaling, transcriptional and translational clusters are present at this locus in BDs. Together with a normal expression of GAD₆₇ the GABA cells found therein may be hyperactive and capable of exerting higher than normal levels of inhibitory modulation on the pyramidal neurons in SO of CA1. This locus represents the final common pathway for the output of excitatory activity from the trisynaptic pathway to its many termination sites and its modulation by GABA cells has important implications for functional integration within the hippocampus.

The model shown in Fig. 1 A and B postulates that the overall flow of excitatory activity along the trisynaptic pathway may be increased in SZ and help explain the paradoxical finding of an increased basal cerebral blood flow that was recorded in the hippocampus of SZs, using positron emission tomography (49). In CA1 of BDs, however, information processing by the trisynaptic pathway may be fundamentally different from that seen in SZ.

Conclusions

In summary, the current study describes the results of a cross-sectional analysis of gene expression profiling at key sites along the trisynaptic pathway that almost exclusively contain GABAergic interneurons. These results suggest that the cellular endophenotypes for SZ and BD may be determined by multiple factors that include unique susceptibility genes for the respective disorders and altered integration among hippocampal neurons with extrinsic and intrinsic afferent fiber systems. The extensive and intricate data that has come from this study has provided novel insights into how a complex circuit, like the trisynaptic pathway may be regulated in human hippocampus in both health and disease.

Methods

All methodological details have been described (6) and careful assessments of RNA quality and reproducibility have been reported (50). A brief description appears below:

Subjects.

As shown in Table S1, the cohort used in these studies consisted of 7 CONs, 7 SZs and 7 BDs matched for age, postmortem interval, pH and 18S/28S ratios and the method for establishing retrospective diagnoses is described elsewhere (6).

Tissue and RNA Preparation.

Frozen tissue sections were cut (8 μm) on a Microm HM 560 CryoStar cryostat, mounted on LEICA Frame Slides (PET-membrane 1.4 μm) and fixed in Streck Tissue Fixative (STF; Streck Laboratories). As shown in Fig. S1, the sections were lightly stained with 0.1% cresyl violet to visualize the cytoarchitectonic details of the hippocampus and to identify the stratum oriens (SO), stratum pyramidale (SP) and stratum radiatum (SR) within sectors CA3/2 and CA1. The frame slides were mounted on a LEICA AS LMD apparatus and a pulsed UV laser was used to “cut” samples (2 mm × 1 mm) of the various layers that fell into a lysis/denaturing solution. Total RNA was extracted using the Rneasy Micro Kit (Qiagen) and T7-based linear amplification was performed by MessageAmp II aRNA amplification kit (Ambion). Subsequently, target labeling was performed with the MessageAMP Biotin Enhanced Kit (Ambion) and generated aRNA with the incorporation of biotinylated nucleotides necessary for signal detection on the Affymetrix Genechip system. Yields of aRNA were determined using a μQuant Microplate Spectrophotometer (Bio-Tek Instruments) and the quality of aRNA was determined using a Bioanalyzer 2100 (Agilent Technologies).

Gene Expression Profiling.

Biotinylated target RNA from each sample was individually hybridized to the U133A array (Affymetrix) as described in ref. 6.

GenMapp Biopathway and Cluster Analyses.

To identify biologically relevant clusters of interrelated genes, GenMapp algorithms (www.genmapp.org) were used to relate the ANOVAs generated with dChip to several different biochemical pathways and/or biologically related clusters of genes. The α-level of significance for individual genes was set at P ≤ 0.05. To assess the changes in gene expression for each GenMapp biopathway or cluster, an ad hoc metric, composite probability, P_c, based on a combination of probability theory and two separate corrections for multiple comparisons was used (13).

Quantitative RT-PCR.

Five genes were used to validate the microarray findings with qRT-PCR and were processed as previously described (6), and the data are shown in Fig. S2.