Altered Expression of Regulators of the Cortical Chloride Transporters NKCC1 and KCC2 in Schizophrenia FREE

A core feature of schizophrenia is the impairment of certain cognitive functions, such as working memory,¹ that are dependent on the circuitry of the dorsolateral prefrontal cortex (DLPFC).² These cognitive deficits in schizophrenia are associated with altered neural network synchrony^3- 4 that may be attributable, at least in part, to functional abnormalities in γ-aminobutyric acid (GABA) neurotransmission in the DLPFC.⁵

Disturbed GABA neurotransmission in schizophrenia might arise from alterations in 1 or more of the following factors: (1) the strength of GABA neurotransmission, which depends on the amount of GABA available at the synapse; (2) the kinetics of GABA neurotransmission, which is determined, in part, by the subunit composition of GABA_A receptors; and (3) the nature of GABA neurotransmission, which can be hyperpolarizing, depolarizing, or shunting depending on the flow of chloride ions when GABA_A receptors are activated.⁶ Presynaptic and postsynaptic markers of the strength and kinetics of GABA neurotransmission are known to be altered in the DLPFC in subjects with schizophrenia,^7- 8 but measures of the nature of GABA neurotransmission have not been evaluated.

The flow of chloride ions through GABA_A receptors depends on intracellular levels of chloride, which are regulated by the relative activities of the sodium-potassium-chloride cotransporter 1 (NKCC1; SLC12A2), which mediates chloride uptake, and the potassium-chloride cotransporter 2 (KCC2; SLC12A5), which mediates chloride extrusion from the cell. In the central nervous system, NKCC1 is found in both neurons and glial cells, but KCC2 is strictly neuronal.^9- 11 The activities of both transporters are sensitive to intracellular chloride levels, and their activation depends on their phosphorylation (NKCC1) or dephosphorylation (KCC2) status.^12- 15 Oxidative stress response kinase (OXSR1; also known as OSR1) and Ste 20–related, proline-alanine–rich kinase (STK39; also known as SPAK) are highly expressed in the brain^16- 19 and bind to NKCC1 where they phosphorylate residues present in the N-terminal domain, resulting in an increase of NKCC1 activity.^17,20- 23 In addition, NKCC1 can be phosphorylated by the coexpression of STK39 and WNK4 (with no K [lysine] protein kinase 4) kinases in vitro.²⁴ Conversely, STK39-mediated phosphorylation has a dominant negative effect on KCC2 function.²⁴ Another WNK kinase, WNK1, activates both STK39 and OXSR1 via phosphorylation.^23,25- 27 Finally, a fifth kinase, WNK3, both activates NKCC1 and inhibits KCC2, regardless of cellular tonicity.^28- 29 Consequently, alterations in the relative expression levels of NKCC1 and KCC2, or of their regulatory kinases, in the DLPFC of subjects with schizophrenia could, by shifting intracellular chloride levels, alter the nature of GABA transmission and thereby contribute to impaired neural network synchrony and cognitive dysfunction in affected individuals.

To examine this possibility, we quantified messenger RNA (mRNA) expression levels of NKCC1 and KCC2 and their associated regulatory kinases STK39, OXSR1, WNK1, WNK3, and WNK4 in the DLPFC from subjects with schizophrenia and matched healthy comparison subjects. In subjects with schizophrenia, OXSR1 and WNK3 transcripts were substantially overexpressed, relative to comparison subjects. In contrast, NKCC1, KCC2, STK39, WNK1, and WNK4 transcript levels did not differ between subject groups. Thus, increased expression levels, and possibly increased kinase activities, of OXSR1 and WNK3 in schizophrenia would be predicted to both increase NKCC1 and decrease KCC2 activities, producing greater intracellular chloride concentration and thus altering the nature of GABA neurotransmission in the DLPFC.

Brain specimens (n = 84) were obtained during autopsies conducted at the Allegheny County Medical Examiner's Office (Pittsburgh, Pennsylvania) after consent was obtained from the next of kin. An independent committee of experienced research clinicians made consensus DSM-IV diagnoses for each subject on the basis of medical records and structured diagnostic interviews conducted with the decedent's family members, as described previously.³⁰ To control for experimental variance and to reduce biological variance between subject groups, each subject with schizophrenia (n = 42) was matched with 1 healthy comparison subject for sex and as closely as possible for age and postmortem interval (PMI); samples from a given subject pair were always processed together. The subject groups (Table; see eTable 1 for details on individual subjects) did not significantly differ in mean age, PMI, brain pH, RNA integrity number (RIN), or tissue storage time at −80°C (all t₈₂<1.67; all P > .10) or in race (χ² = 1.587; P = .21). All procedures were approved by the University of Pittsburgh's Committee for the Oversight of Research Involving the Dead and Institutional Review Board for Biomedical Research.

The right hemisphere of each brain was blocked coronally, immediately frozen, and stored at −80°C, as described previously.³¹ Area 9 of the DLPFC was identified cytoarchitectonically from Nissl-stained coronal sections³¹ spanning the rostrocaudal axis of the superior frontal sulcus. The cortical gray matter was dissected from cryostat sections (40 μm) in a manner that ensured limited white matter contamination and excellent RNA preservation, as described previously.³² Total RNA was isolated from tissue homogenates using the TRIzol protocol from Invitrogen (Invitrogen Corporation, Carlsbad, California) and further purified using the RNeasy kit (QIAGEN, Valencia, California). RNA integrity was assessed by measuring the RIN using the Agilent 2100 bioanalyzer (Agilent Technologies, Santa Clara, California).

Total RNA was converted to complementary DNA using the High-Capacity cDNA Archive Kit from Applied Biosystems (Foster City, California). For each reaction, we used 50 ng of total RNA from each subject. Priming was performed with random hexamers, according to the manufacturer's recommendations. The efficiency for each primer set (eTable 2) was between 92% and 100%, and the amplified product resulted in a specific single product in dissociation curve analysis. Given the high level of homology between certain domains of OXSR1 and STK39, the quantitative real-time polymerase chain reaction (qPCR) primer sets for these 2 transcripts were designed within a unique sequence in the 3′ untranslated region of each mRNA.

Samples from each matched pair of subjects with schizophrenia and comparison subjects were always assayed on the same plate. For each sample, amplified product differences for each transcript were measured with 4 replicates using SYBR Green chemistry-based detection.³³

β-Actin, cyclophilin A, and glyceraldehyde-3-phosphate dehydrogenase were used as endogenous reference genes. These 3 transcripts were selected based on their previously demonstrated stable expression across both subjects with schizophrenia and healthy comparison subjects.³² The qPCR reactions were carried out in an ABI Prism 7000 thermal cycler (Applied Biosystems) using the ABI Prism 7000 SDS software with the automatic baseline and threshold detection options selected. These data were exported to Microsoft Excel (Microsoft, Redmond, Washington) and delta cycle thresholds (dCTs) were calculated for each sample by using the geometric mean of the 3 endogenous reference genes as the normalization factor (ie, cycle threshold [CT] for each transcript in a sample minus the geometric mean of β-actin, cyclophilin A, and glyceraldehyde-3-phosphate dehydrogenase CTs for the same sample).³⁴ The expression level for each transcript was then calculated as the expression ratio value (where expression ratio = 2^−dCTs), and all results are reported as the expression ratio.

Two isoforms of the WNK3 gene are expressed in the brain: isoform 1, which is brain specific, and isoform 2, which is ubiquitous. Isoform 1 differs from isoform 2 only by the presence of an additional stretch of 47 amino acids in the protein.³⁵ Given the differential expression pattern of both isoforms, we designed 1 primer set detecting WNK3 isoform 1 (primer set WNK3-1) and 1 primer set detecting both isoforms 1 and 2 (primer set WNK3-1&2). WNK3-1 was used on the 42 pairs of subjects, and WNK3-1&2 was used only on an arbitrary subset of 12 pairs of subjects with schizophrenia and healthy comparison subjects (eTable 1A, pairs with *) that had been used in a previous real-time qPCR study.³²

Two-way cluster analyses were performed using delta-delta CTs (subject with schizophrenia dCT minus matched healthy comparison subject dCT) across all 42 subject pairs for all transcripts using Pearson mean analysis in Genes@Work (IBM, Armonk, New York).³⁶

The effect of long-term exposure to antipsychotic medication on the levels of transcripts showing altered expression in schizophrenia was examined using macaque monkeys exposed to haloperidol, olanzapine, and placebo, as described before^37- 38 (eAppendix). All housing and experimental procedures were conducted in accordance with National Institutes of Health guidelines and with approval of the University of Pittsburgh Institutional Animal Care and Use Committee.

To assess the distribution and relative levels of OXSR1 protein, we conducted immunocytochemical experiments using immersion-fixed cryostat sections (40 μm) of DLPFC area 9 from 6 white male human subjects (17-62 years of age; 3 subjects with a short PMI [5-8.2 hours] and 3 subjects with a long PMI [14.5-37 hours]). Tissue sections were processed for OXSR1 immunoreactivity using a 1:200 dilution of a goat anti-OXSR1 (SC49473; Santa Cruz Biotechnology, Santa Cruz, California) and the Vectastain ABC kit (Vector Laboratories, Burlingame, California), according to a previously described protocol.³⁹ The OXSR1 antibody was raised against the C-terminal portion (amino acids 370-420) of the human OXSR1 protein. Antibody specificity was demonstrated by the absence of labeling in the thymus⁴⁰ and by the absence of specific labeling when the primary antibody was omitted.

Western blots were conducted using brain tissue lysate prepared as previously described⁴¹ (eAppendix). The specificity of the anti-OXSR1 antibody was confirmed by detection of a band of identical molecular weight (58 kDa) compared with cloned human OXSR1 expressed in vitro and by lack of cross-reactivity with a cell lysate of in vitro–expressed STK39 (OriGene, Rockville, Maryland) (data not shown).

To assess the effect of PMI on OXSR1 immunoreactivity, 4 adjacent coronal tissue blocks (2-3 mm thick) were prepared from the DLPFC of the same adult macaque monkey, stored in room-temperature artificial cerebrospinal fluid for 0, 6, 12, or 24 hours, and then flash frozen.⁴² The OXSR1 protein content was reported as percentage relative to the 0-hour PMI.

To assess the laminar and cellular patterns of the expression differences in OXSR1 and WNK3-1&2, we used laser microdissection techniques to obtain mRNA samples from layer 3 tissue and from individually dissected layer 3 pyramidal cells from a subset of 10 subject pairs (eTable 1A and B, pairs with •). Using the results from qPCR experiments conducted on the entire area 9 gray matter for all 42 subject pairs, we selected the 10 pairs with mean expression differences between subjects with schizophrenia and control subjects that were closest to the mean of the entire group.

Cryostat sections (12 μm) were cut and thaw-mounted onto glass polyethylene naphthalate membrane slides (Leica Microsystems, Bannockburn, Illinois) that had been previously UV-treated at 254 nm for 30 minutes, dried, and stored at −80°C. On the day of the microdissection, slides were immersed in an ethanol–acetic acid fixation solution, stained with thionin, dehydrated through 100% ethanol, and air dried. Using a Leica microdissection system (LMD 6500), 2 independent samples of layer 3 pyramidal neurons (eFigure 3A) per subject from 2 different slides (×40 objective; power, 15; aperture, 9; speed, 12; balance, 14; and offset, 120) or 3 strips of tissue from layer 3, as previously described⁴³ (×5 objective; power, 43; aperture, 18; speed, 13; balance, 25; and offset, 60), were obtained. An average of 50 single pyramidal cells or 4.1 mm² of layer 3 tissue were collected per sample in 0.5-mL microtube caps (Ambion/Applied Biosystems) and lysed by vortexing for 30 seconds in 200 μL of RLT Buffer Plus (QIAGEN). The RNA was extracted and PCR analyses were conducted, and cell type specificity and absence of glial contamination of the single pyramidal cell samples were confirmed, as described in the eAppendix. We elected to measure both isoforms of WNK3 using the WNK3-1&2 primer set because of the lower expression of WNK3-1, although the expression difference between schizophrenia and comparison samples using WNK3-1&2 was only half of the expression difference obtained using WNK3-1.

The effect of diagnosis and the influence of potential confounding variables on the transcript expression ratios were assessed using 2 analyses of covariance (ANCOVA) models. In the first model, transcript expression ratio was used as the dependent variable; diagnostic group, as the main effect; and subject pair, as a blocking factor. The RIN, pH, and freezer storage time were entered as covariates because they may affect RNA integrity. Subject pairing may be considered an attempt to control for experimental variance by the parallel processing of tissue samples from each subject pair and not a true statistical pair design. Therefore, to validate the first model, a second unpaired ANCOVA model was performed, using diagnostic group as the main effect and with age, sex, PMI, RIN, pH, and freezer storage time as covariates. Because the 42 pairs of subjects were used in 2 successive qPCR runs of 24 and 18 pair cohorts (eTable 1A) and 18 (eTable 1B) that involved different lots of reagents, qPCR run was also entered as a covariate in both models. Tissue storage time did not have a significant effect in either model for any transcript and was excluded in the reported analyses. The reported P values for each model were corrected for multiple comparisons (n = 21) using the Bonferroni procedure.

The influences of potential confounding variables on the expression ratio values in subjects with schizophrenia were assessed with ANCOVA models using each confounding variable as the main effect and sex, age, PMI, pH, RIN, and run as covariates. Pearson correlation was used to assess the relationships of the expression ratios for all pairs of transcripts; reported P values were also corrected for multiple comparisons (n = 21) using the Bonferroni procedure.

A 1-way analysis of variance model with expression ratios as the dependent variable and treatment group as the main effect was used to compare transcript expression levels in the haloperidol-, olanzapine-, and placebo-exposed monkeys.

After corrections for multiple comparisons, neither the mean level of NKCC1 mRNA (Figure 1A) nor KCC2 mRNA (Figure 1B) was significantly altered in subjects with schizophrenia (paired ANCOVA, both F_1,38<3.8 and P > .43; unpaired ANCOVA, both F_1,75<4.1 and P > .33). For the chloride transporter–related kinases, group differences in gene expression levels were not significant for STK39, WNK1, or WNK4 (paired, all F_1,38<5.1 and all P > .20; unpaired, all F_1,75<4.11 and all P > .32). In contrast, the mean expression level of OXSR1 mRNA (Figure 1C) was significantly (paired, F_1,38 = 54.3 and P < .001; unpaired, F_1,75 = 38.1 and P < .001) 19.4% greater in the subjects with schizophrenia. In addition, mean WNK3-1 mRNA levels (Figure 1D) were significantly (paired, F_1,38 = 50.9 and P < .001; unpaired, F_1,75 = 45.6 and P < .001) 48.7% greater in the subjects with schizophrenia. Relative to their matched healthy comparison subjects, OXSR1 and WNK3-1 mRNA levels were higher in 41 and 37, respectively, of the 42 subjects with schizophrenia (Figure 1C and D). None of the covariates (age, sex, PMI, RIN, brain pH, or freezer storage time) were significant in any of the analyses for OXSR1 and WNK3-1 mRNA (paired, all F_1,38<2.69 and all P > .11; unpaired, all F_1,75<3.76 and all P > .06).

Because of the low expression level of the brain-specific WNK3-1 mRNA (primer set WNK3-1) in the DLPFC, we designed a new set of primers binding to both isoforms 1 and 2 of WNK3 in the human brain (primer set WNK3-1&2). This primer set produced lower CT values than the WNK3-1 primer set in the same subjects (n = 12 pairs), reflecting the higher level of expression for both WNK3 isoforms combined than for isoform 1 alone. Consistent with the findings for isoform 1 alone, the combined expression of both WNK3 isoforms was also higher (+27.0%; paired, F_1,8 = 3.97 and P = .08; unpaired, F_1,19 = 7.1 and P = .01) in the subjects with schizophrenia, and the expression levels from both primer sets were highly correlated (r = 0.561; P < .01) across all subjects.

The expression ratios for OXSR1, WNK3-1, and WNK3-1&2 did not differ (all F_2,15<0.92 and all P > .42) across the monkeys with long-term exposure to placebo, haloperidol, or olanzapine (Figure 2).

The mean expression ratio in the subjects with schizophrenia did not differ for OXSR1 (Figure 3A) (all F <1.941 and all P ≥ .17, uncorrected) or WNK3-1 (Figure 3B) (all F ≤3.99 and all P > .05, uncorrected) as a function of sex; diagnosis of schizoaffective disorder; suicide; antidepressant medication, benzodiazepines or sodium valproate, or antipsychotic medication use at the time of death; or diagnosis of substance abuse/dependence at time of death.

A 2-way cluster analysis performed on delta-delta CT values across the 42 pairs of matched subjects for all 7 transcripts with Pearson mean distance analysis resulted in the separation of the transcript expression changes in 2 different clusters (eFigure 1). OXSR1 clustered in a pathway with WNK1, WNK4, NKCC1, and STK39, whereas WNK3-1 and KCC2 clustered in a distinct pathway.

Pearson correlations were performed using the dCT values obtained for each transcript for each of the 84 subjects (Figure 4 and eTable 3); the reported P values were calculated using Bonferroni correction for multiple comparisons. Expression levels of OXSR1 were significantly positively correlated with those for WNK3-1 (r = 0.744; P < 10⁻⁷) and WNK1 (r = 0.478; P < 10⁻⁴), and WNK3-1 expression was significantly correlated with WNK1 (r = 0.346; P = .03). In contrast, WNK3-1 expression was negatively correlated with KCC2 (r = −0.356; P = .02). NKCC1 expression was positively correlated with WNK1 (r = 0.583; P < 10⁻⁷), WNK4 (r = 0.425; P = .001), and STK39 (r = 0.371; P = .01).

In human DLPFC from subjects with PMIs less than 6 hours, immunoreactivity for OXSR1 was highest in layers 3 and 5, intermediate in layers 2 and 6, and low in layers 1 and 4 (Figure 5A and B). OXSR1 immunoreactivity was mainly located in the perinuclear cytoplasmic compartment of neurons, most of which had the morphology of pyramidal cells. In contrast, OXSR1 labeling was severely reduced with PMIs longer than 12 hours (Figure 5C). This finding was confirmed by Western blotting in samples of monkey DLPFC with artificially induced PMIs. At PMIs of 12 and 24 hours, levels of OXSR1 protein were only 60% and 40%, respectively, of that detected at 0 hours (eFigure 2). Thus, a quantitative study of OXSR1 protein levels in subjects with schizophrenia and control subjects could not be conducted.

To determine if the greater expression of OXSR1 and WNK3-1&2 transcripts in the subjects with schizophrenia arose from the same cortical layer and neurons, we used laser microdissection to obtain samples of tissue restricted to layer 3 and samples of 50 pyramidal cells individually cut from layer 3. In the layer 3 tissue samples, the expression levels of OXSR1 and WNK3-1&2 mRNA were significantly correlated (r = 0.88; P < .001) across all 20 subjects, and the within-pair expression differences for OXSR1 and WNK3-1&2 mRNA levels were also significantly correlated (r = 0.92; P < .001). Consistent with these findings, in the samples of layer 3 pyramidal cells, the expression levels of OXSR1 and WNK3-1&2 mRNA were significantly correlated (r = 0.56; P = .001) across all 20 subjects, confirming that both transcripts are expressed in layer 3 pyramidal cells and suggesting that their expression is coregulated. In addition, both transcripts showed greater expression in the subjects with schizophrenia (mean percentage increases: OXSR1, +12.0%; WNK3-1&2, +12.4%), although because of the smaller sample size (n = 10 pairs) these findings did not achieve statistical significance (t = −1.007; P = .15 and t = −0.682; P = .25, respectively). Together, these findings suggest an upregulation of both transcripts in layer 3 pyramidal cells in the illness.

In the DLPFC of subjects with schizophrenia, mRNA expression levels for the chloride transporters NKCC1 and KCC2 were not significantly altered. Although it is possible that our measures of mRNA in total gray matter obscured schizophrenia-associated expression differences in neuronal subpopulation(s) in a different layer, we previously found that neither NKCC1 nor KCC2 mRNA showed a preferential laminar pattern of expression in control human subjects.⁴³

In contrast, the mRNA expression levels of 2 regulatory kinases, OXSR1 and WNK3, were markedly and consistently higher in the DLPFC of subjects with schizophrenia. Although an increase in protein kinase activity may not always follow an increase in the corresponding mRNA transcript, previous reports have shown that increasing WNK3 mRNA levels or silencing OXSR1 expression is associated with alterations in NKCC1 transport activity.^28,44 Thus, our results suggest that the increased expression of OXSR1 and WNK3 are likely to lead to shifts in the activity of the chloride transporters that could substantially alter the nature of cortical GABA neurotransmission in schizophrenia. Indeed, the finding that both transcripts are upregulated in the same layer 3 pyramidal neurons suggests that altered GABA signaling may be particularly prominent in these cells. The findings for OXSR1 and WNK3 are striking for several other reasons. First, the higher levels of these transcripts in subjects with schizophrenia contrast with the more commonly observed lower levels of GABA-related transcripts in the illness.³² Second, the elevated transcript levels in schizophrenia indicate that the findings are not attributable to poorer RNA quality in these subjects, consistent with the excellent brain pH and RIN measures in all subjects used in this study and the absence of group differences in these variables. Third, relative to the matched healthy comparison subjects, almost all of the 42 subjects with schizophrenia had higher levels of both OXSR1 and WNK3 mRNA. The apparent consistency of these findings suggests that they are more likely to be related to a conserved downstream aspect of the disease process (eg, impaired GABA neurotransmission) than to reflect the etiological complexity of schizophrenia.⁸ Fourth, consistent with the findings that these transcripts are not altered in monkeys exposed to antipsychotic medications long-term, the conservation of these alterations across almost all subjects examined strongly suggests that they are not attributable to other factors (eg, substance use, suicide, mood symptoms) that were present in some of the subjects with schizophrenia, an interpretation confirmed by the direct assessment of these factors (Figure 3).

A conserved OXSR1 binding motif present on NKCC1¹⁷ has also been detected on WNK1 and WNK4⁴⁵ and the activation of NKCC1 by OXSR1 is regulated through upstream interactions with WNK1 and WNK4.^{23,25- 27,46} Consistent with these observations, we found strong correlations across all subjects between OXSR1 and WNK1 (r = 0.478; P < .01) and between WNK1 and WNK4 (r = 0.516; P < .001) mRNA levels. The expression differences between subjects with schizophrenia and control subjects for WNK1 and WNK4 kinases clustered together, consistent with the idea that WNK1 and WNK4 regulate each other's activity.⁴⁷ Nevertheless, neither WNK1 nor WNK4 transcript expression level was significantly altered in subjects with schizophrenia, despite suggestions of overexpression (+7%; P = .36 and +12.6%; P = .17, respectively), indicating that the overexpression of the OXSR1 transcript could be primarily responsible for an increase in NKCC1 transporter activity.

WNK3 is a potent activator of NKCC1^28- 29 and they are colocalized in neurons.²⁸ Furthermore, WNK3 possesses an OXSR1/STK39 consensus binding motif,⁴⁵ suggesting that an increase in WNK3 expression in schizophrenia could lead to increased NKCC1 transporter activity through OXSR1. Consistent with this hypothesis, we found very strong correlations (r = 0.744; P < 10⁻⁷) between OXSR1 and WNK3 mRNA levels across all subjects as well as within-pair expression differences in layer 3 samples (r = 0.92; P < .001), suggesting that their effects on chloride transporter activity are synergistic.

WNK3 can inhibit KCC2 activity regardless of cellular tonicity,^28- 29 suggesting that the elevated WNK3 expression in subjects with schizophrenia could be associated with lower KCC2 activity. Consistent with this interpretation, KCC2 and WNK3 expression levels were inversely correlated (r = −0.356; P = .02) across all 84 subjects. Additionally, in our data set, both KCC2 and WNK3 clustered together, suggesting that indeed both KCC2 and WNK3 belong to the same regulatory pathway. Consistent with this hypothesis, KCC2 and WNK3 share the same developmental trajectory.^28,48

Because kinase-dead WNK3 activates KCC2,^28- 29 KCC2 inhibition triggered by the increase in WNK3 could be indirect, through the potential inhibition of a yet to be identified phosphatase or the activation of another kinase. Coimmunoprecipitation studies have shown that WNK3 can interact with WNK4 via their carboxy termini⁴⁹ and it has been suggested that WNK4 could act as a regulator of WNK3 activity and therefore KCC2/NKCC1 activity.⁵⁰ In our data set, WNK4 gene expression did display an upregulated trend (+12.6%; P = .17) but its expression level in all 84 subjects was correlated with NKCC1 and not with WNK3. This suggests that NKCC1 and KCC2 regulation by WNK3 may happen through different regulatory pathways. Finally, the strong Pearson correlation between OXSR1 and WNK3 (0.744; P < .01) suggests that the regulation of both NKCC1 and KCC2 activities could be linked through the interaction of these 2 kinases.

The combined and coordinated greater expression of OXSR1 and WNK3 in schizophrenia would be expected to increase both NKCC1 and KCC2 phosphorylation levels. Because NKCC1 and KCC2 are activated through phosphorylation and dephosphorylation, respectively,^24,51- 52 our findings suggest that schizophrenia is associated with both increased NKCC1 activity and decreased KCC2 activity. Interestingly, the latter finding may be consistent with preliminary reports of decreased levels of KCC2 mRNA in the hippocampus of patients with schizophrenia.^53- 54 This combined shift in transporter activities would be expected to increase intracellular chloride levels, reducing or reversing the net inward flow of chloride ions when GABA_A receptors are activated, and thus altering the nature of synaptic GABA neurotransmission.

Early in development, the ratio of NKCC1 to KCC2 is high, resulting in high intracellular chloride content and a depolarizing excitatory response on GABA_A receptor activation. Later, as the expression of NKCC1 declines and KCC2 increases, the level of intracellular chloride decreases⁵⁵ and GABA_A receptor activation triggers chloride influx, hyperpolarization, and inhibition. Interestingly, an increase in OXSR1 and WNK3 kinase activity could promote a change of transporter activities (ie, increased NKCC1 to KCC2 ratio) and a shift in the nature of GABA neurotransmission to be more similar to that observed earlier in development. Specifically, our findings suggest that schizophrenia is associated with a tendency for GABA neurotransmission to be less strongly inhibitory in the DLPFC, at least in the neurons that have elevated levels of OXSR1 and WNK3 expression.

A provisional model summarizing these potential interactions between NKCC1 and KCC2 chloride transporters and OXSR1 and WNK3 kinases is shown in Figure 6. In normal adult neurons (Figure 6A), NKCC1 levels are low and KCC2 levels are high resulting in a low level of intracellular chloride. GABA binding to GABA_A receptors triggers an inward flow of chloride leading to hyperpolarization. In schizophrenia (Figure 6B), greater levels of OXSR1 and WNK3 result in greater phosphorylation of both chloride transporters, and a consequent increase in NKCC1 and decrease in KCC2 activities, leading to a higher level of intracellular chloride. This change in intracellular chloride concentration has the potential to reduce the hyperpolarizing effect triggered by GABA binding and therefore to substantially change the nature of GABA neurotransmission.

The affected cortical neurons are likely to include pyramidal cells, given the immunocytochemical evidence that OXSR1 is preferentially localized in pyramidal neurons in layers 3 and 5 and the correlated expression of OXSR1 and WNK3-1&2 mRNA in layer 3 pyramidal neurons. Effective synchronization of networks of pyramidal neurons at the gamma band frequencies associated with working memory requires strongly hyperpolarizing inputs to a network of pyramidal cells so that the postsynaptic pyramidal neurons then escape from inhibition simultaneously and fire subsequent action potentials in unison, producing synchronized activity.^56- 58 Thus, a shift to increased NKCC1 and decreased KCC2 activity in pyramidal neurons in schizophrenia could alter the reversal potential of the GABA_A receptor current to be closer to the resting membrane potential, reducing the strong inhibition required for gamma oscillations,⁶ and thus contributing to the alterations in the prefrontal gamma oscillations associated with impaired working memory in schizophrenia.⁴

The increased expression of OXSR1 and WNK3 could also contribute to other circuitry disturbances in the DLPFC. For example, vesicular glutamate transport appears to be highly dependent on the chloride content outside of synaptic vesicles.⁵⁹ Thus, the change in intracellular chloride concentration predicted by our findings could also alter vesicular glutamate transporter activity and contribute to alterations in glutamate transmission and synaptic plasticity reported in schizophrenia.⁶⁰

Recent work also indicates that KCC2, through its interaction with the dendritic cytoskeleton, regulates spine maturation.⁶¹ Thus, a decrease in KCC2 activity, secondary to increased OXSR1 and WNK3 expression, could contribute to impaired spine maturation and thus to lower dendritic spine density in the DLPFC.³⁰ Consistent with this interpretation, we observed higher OXSR1 and WNK3 expression in layer 3 pyramidal cells where the spine deficit in schizophrenia is most pronounced.⁶²

In summary, we report substantial increases in the expression of the regulatory kinases OXSR1 and WNK3 in the DLPFC that appear to be highly conserved across subjects with schizophrenia. This combination is likely to alter the relative activities of the chloride transporters NKCC1 and KCC2, producing a higher than normal intracellular chloride concentration, such that activation of GABA_A receptors results in less chloride influx and a smaller increase in hyperpolarization, or depending on the chloride reversal potential, could even lead to an outflow of chloride ions, and some degree of membrane depolarization of the postsynaptic cell.

Correspondence: David A. Lewis, MD, Department of Psychiatry, University of Pittsburgh, 3811 O’Hara St, W1651 BST, Pittsburgh, PA 15213 (lewisda@upmc.edu).

Submitted for Publication: January 22, 2010; final revision received June 29, 2010; accepted July 4, 2010.

Published Online: September 6, 2010. doi:10.1001/archgenpsychiatry.2010.114

Author Contributions: Dr Lewis had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Financial Disclosure: Dr Lewis currently receives investigator-initiated research support from the Bristol-Myers Squibb Foundation, Bristol-Myers Squibb, Curidium Ltd, and Pfizer and in 2007 to 2009 served as a consultant in the areas of target identification and validation and new compound development to AstraZeneca, BioLine Rx, Bristol-Myers Squibb, Hoffmann-La Roche, Eli Lilly, Merck, Neurogen, and SK Life Science.

Funding/Support: This work was supported by grants MH-084053 and MH-043784 from the National Institutes of Health.

Additional Contributions: Mary Brady, BA, assisted with the graphics and Kelly Rogers, MS, and Anthony Cipriano, BS, assisted with the immunocytochemistry runs. We express our deep gratitude to the families of the brain donors who made this research possible.