α7 nicotinic acetylcholine receptor mRNA expression and binding in postmortem human brain are associated with genetic variation in neuregulin 1

Abstract

Studies in cell culture and in animals suggest that neuregulin 1 (NRG1), a probable schizophrenia susceptibility gene, regulates the expression of the α7 nicotinic acetylcholine receptors (nAChRs). We hypothesized that schizophrenia-associated allelic variations within the NRG1 gene, via their effects on NRG1 isoform expression, would be associated with alterations in nAChR α7 receptor levels. We examined the effects of four disease-associated single-nucleotide polymorphisms (SNPs) in the 5′ region of the NRG1 gene on nAChR α7 mRNA transcript expression in both the dorsolateral prefrontal cortex (DLPFC) and hippocampus of normal controls and patients with schizophrenia using quantitative real-time PCR. NRG1 risk alleles at SNPs SNP8NRG221132 and rs6994992 predicted significantly lower nAChR α7 mRNA expression in the DLPFC. Haplotypes containing the risk alleles at the above SNPs were also associated with lower expression of nAChR α7 in the DLPFC. The genotype effect for rs6994992 and the haplotype effect were more pronounced within the schizophrenic patient group. To determine whether receptor levels follow that of mRNA expression, we performed receptor binding and autoradiography using [¹²⁵I] α-bungarotoxin in the DLPFC. Consistent with the mRNA findings, we found a decrease in binding in risk allele carriers of SNP8NRG221132 as compared with heterozygous individuals. Together, these results suggest that the molecular mechanism of the association between NRG1 risk alleles and schizophrenia may include down-regulation of nAChR α7 expression.

INTRODUCTION

Nicotinic acetylcholine receptors (nAChRs) are located on the soma, dendrites and axon terminals of cholinergic, glutamatergic, dopaminergic and GABA neurons (1–3). Twelve genes encoding nAChRs (α2–α10, β2–β4) have been identified thus far (4). The homomeric α7 receptor has low affinity for nicotine and high affinity for α-bungarotoxin (α-BTX) (5). The chromosomal location for the α7 nAChR gene and several polymorphic markers within the gene are weakly linked to schizophrenia and to potential sensory gating deficits associated with the illness (6–9).

Several molecular cascades regulate nAChR expression, including neuregulin 1 (NRG1), a gene implicated in schizophrenia. A haplotype consisting of five SNPs and two microsatellites in the 5′ region of the NRG1 gene is enriched in at least some samples of patients with schizophrenia (10), and several of the markers in this risk region have shown positive association in other clinical samples (11). In postmortem studies, the expression of the NRG1 type I isoform is increased in the dorsolateral prefrontal cortex (DLPFC) and hippocampus of schizophrenic patients (12,13). Genetic variation at one NRG1 SNP (SNP8NRG221132) is associated with NRG1 type I isoform expression in both the DLPFC and hippocampus (13). Another SNP (rs6994992) and a 5′ haplotype consisting of four SNPs from the original deCODE at-risk haplotype is associated with NRG1 type IV isoform expression levels in the hippocampus (13).

Neuregulins up-regulate nAChRs at the neuromuscular junction (14), the developing interneuron synapse (15) and hippocampus (16). The type 1 NRG1 isoform is important in nAChRs’ post-synaptic expression (17). We tested the hypothesis that via their association with altered NRG1 expression, disease-associated SNPs within the NRG1 gene may effect α7 nAChR expression in the brain. Specifically, we examined the association of four SNPs from the original deCODE risk haplotype (10) on α7 nAChR mRNA expression in the hippocampus and DLPFC. Because of prior evidence that two of these SNPs, SNP8NRG221132 and rs6994992, individually affected expression of specific NRG1 isoforms in human brain, we predicted that effects of these SNPs on nAChR α7 would also be found. We also performed receptor-binding assays in the DLPFC to determine whether nAChR α7 receptor density is associated with the same risk SNPs.

RESULTS

The relationship between demographic and tissue-related variables on nAChR α7 expression and receptor binding

Multiple regression analysis revealed that normalized nAChR α7 mRNA expression in DLPFC inversely correlated with age and positively correlated with pH, whereas in the hippocampus in addition to age, it also correlated with RNA integrity number (RIN) and race (adjusted R² > 0.21, F >9.0, P < 0.05). None of the variables had a significant effect on specific [¹²⁵I] α-BTX binding in the DLPFC (adjusted R² = 0.05, F(4,71) = 1.9, P > 0.05).

Multiple regression analyses were performed to test whether any clinical variables independently predicted expression or receptor binding. Normalized nAChR α7 mRNA expression negatively correlated with duration of illness in the hippocampus and DLPFC (adjusted R² > 0.33, F > 4.4, P < 0.02) of schizophrenic patients. None of the clinical variables examined significantly correlated with specific [¹²⁵I] α-BTX binding (adjusted R² = 0.2, F(3,19) = 2.8, P > 0.05).

The relationship between diagnosis and smoking history on nAChR α7 subunit expression

On the basis of the preceding regression analysis, age and pH were included as covariates in the mRNA expression analysis within DLPFC, while age, RIN and race were used as covariates in the hippocampal analysis. With these covariates, normalized nAChR α7 expression did not significantly differ based on diagnosis, smoking history or an interaction of two variables in either the DLPFC (Fig. 1A) or hippocampus (Fig. 1B) (all F values <1.8; all P-values >0.1). There was no difference in normalized nAChR α7 mRNA expression comparing normal controls to patients with schizophrenia or based on smoking history in the either DLPFC or in the hippocampus, when the analysis was performed without covariates. Of the variables examined, pH in the DLPFC (t = 2.29, P =0.02) and RIN in the hippocampus (t = 2.0, P = 0.05) significantly differed between controls and patients. However, including these factors as covariates had no impact on the finding.

Figure 1.

The effect of disease and smoking variables on normalized nAChR α7 subunit mRNA expression and [¹²⁵I] α-bungarotoxin ([¹²⁵I] α-BTX)-specific binding. mRNA expression in the DLPFC (A) and hippocampus (B) did not significantly differ between normal controls (Control) and patients with Schizophrenia (Schiz) or due to nicotine exposure (P > 0.2). Additionally, [¹²⁵I] α-BTX-specific binding in the DLPFC (C) did not significantly differ in diagnosis or due to smoking history. Graphs represent mean ± SEM.

The relationship between 5′ NRG1 SNPs and nAChR α7 subunit mRNA expression in the DLPFC

A two-way ANCOVA was performed including age and pH as covariates to investigate the association of the four individual NRG1 SNPs (see Table 1) with normalized nAChR α7 expression. Where a genotype effect was observed, we examined the effect of tissue-related and demographic variables between the different genotype groups within each diagnostic group. Within the normal control group, postmortem interval (PMI) significantly differed between the genotype carriers of NRG221132 (t = −2.65, P = 0.01), while there was a trend for a significant difference in the patient group (t = 1.87, P = 0.07). There were no significant differences in any of the tissue-related or demographic variables between the genotypic groups for rs6994992 in either normal controls or in schizophrenic patients. We found a main effect of genotype for SNP8NRG221132 (F(1,82) = 6.3, P = 0.01) in the combined (patients and normal controls) cohort. Individuals who were homozygous for the risk-associated G allele had a 30% decrease in nAChR α7 transcript expression compared with carriers of the A allele (Fig. 2A). There was no significant diagnosis by genotype interaction (F(1,82) = 0.3, P = 0.6) (Fig. 2B).

Figure 2.

Association between SNP8NRG221132 and nAChR α7 subunit mRNA expression in the DLPFC of normal controls and patients with schizophrenia. (A) A significant effect of genotype on the normalized nAChR α7 subunit mRNA expression was observed in the whole cohort, with homozygous G allele carriers having significantly lower amount compared with heterozygous G/A individuals (P = 0.01). (B) When the genotype groups are split by diagnosis, a similar trend for expression is observed in both normal controls and patients with schizophrenia. There was no significant diagnosis by genotype interaction. Graphs represent mean ± SEM.

Table 1.

Allelic determination of deCODE SNP markers

deCODE ID	Major(1)/minor (2) allele	dbSNP ID allele
SNP8NRG221132	aG/A	N/A
SNP8NRG221533	T/Ca	N/A
SNP8NRG241930	aC/T	N/A
SNP8NRG243177	C/Ta	rs6994992

^aDisease-associated allele.

There was a trend for a main effect of genotype at rs6994992 on normalized nAChR α7 expression in the DLPFC of the combined cohort (F(2,81) = 2.3, P = 0.1) and a significant interaction of genotype and diagnosis (F(2,81) = 6.04, P = 0.004). A post-hoc comparison found a significant difference in expression between schizophrenic patients who are C/C carriers compared with those who are heterozygous carriers (P = 0.007). A dose allele effect for this SNP as reported in the hippocampus by Law et al. (13) was not observed for nAChR α7 expression and since the T/T genotype within the schizophrenic population was too small (n = 4) to perform statistical analysis, we grouped individuals who were carriers of the risk (T) allele to increase our statistical power. With this secondary analysis, we observed a main effect of genotype at rs6994992 (F(1,83) = 4.3, P = 0.04), whereby risk allele carriers exhibited 18% lower nAChR α7 mRNA expression than those homozygous for the C allele (Fig. 3A). In addition, an interaction of genotype with diagnosis was observed (F(1,83) = 10.4, P = 0.002) with schizophrenic patients carrying the risk allele (n = 22) having significantly lower expression of mRNA than controls (n = 41) with the same genotype (P = 0.02) (Fig. 3B). Since there is virtually no linkage disequilibrium (LD) between rs6994992 and SNP8NRG221132 in African American individuals who make up majority of our sample, the effects of each of these SNPs on nAChR α7 mRNA expression are likely to be independent at the level of the gene.

Figure 3.

Association between rs6994992 and nAChR α7 subunit mRNA expression in the DLPFC of normal controls and patients with schizophrenia. (A) A significant effect of genotype on the normalized nAChR α7 subunit mRNA expression was observed in the whole cohort, with T allele carriers having significantly lower amount compared with C/C individuals (P = 0.04). (B) Splitting the genotypes by diagnosis revealed a significant interaction of diagnosis and genotype (P = 0.002) with schizophrenic patients carrying the risk allele having significantly lower expression of mRNA than controls who are risk allele carriers (P = 0.02). Graphs represent mean ± SEM.

In contrast, SNPs SNP8NRG221533 and SNP8NRG241930, which have previously been shown to have no effect on NRG1 mRNA expression (13), had no effect on nAChR α7 subunit mRNA abundance in the DLPFC, and there were no significant diagnosis by genotype interactions (all F values <3.0; all P >0.09). These SNPs are only in marginal LD with SNP8NRG221132 (13) and would be expected to be independent in their effects. They are in only mild–moderate LD with RS6994992. No main effect of genotype and no diagnosis by genotype interactions were observed for negative controls SNPs within the DLPFC (all F values <1.8; all P-values >0.2).

The relationship between 5′ NRG1 SNPs and nAChR α7 subunit expression in the hippocampus

No main effects of genotype were observed for any of the four 5′ NRG1 SNPs from the original risk haplotype on nAChR α7 expression in the hippocampus when examining the schizophrenics alone, normal controls alone and in the combined sample (all F values <2.6; all P-values >0.1). The two negative control SNPs also were not associated with nAChR α7 subunit mRNA expression in the hippocampus (all F values <1.3; all P-values >0.3). In our hippocampus analysis, samples of 61 normal controls and 28 schizophrenic patients also were in the sample used for the analysis of NRG1 isoforms in Law et al. (13). We did not find a correlation between the mRNA levels of any of the normalized NRG1 isoforms and nAChR α7 mRNA expression in either the hippocampus or the DLPFC (data not shown).

The relationship between NRG1 deCODE haplotype and nAChR α7 subunit mRNA abundance

The four markers examined in this study comprise a haplotype showing a positive association with expression of type IV NRG1, and were found to be in significant LD in Caucasians (13). To test whether this portion of the deCODE risk haplotype (designated here as HAP4) was associated with nAChR α7 subunit mRNA expression, we compared Hap4 carriers (diplotypes: hap1/hap4, hap2/hap4, hap3/hap4 and hap4/hap4) to non-Hap4 carriers (diplotypes: hap1/hap1, hap1/hap2, hap1/hap3, hap2/hap2, hap3/hap3 and hap2/hap3) (see Table 2). No main effect of diplotype or diagnosis by diplotype interaction was observed (F(1,82) < 2.8, P > 0.1) in either the DLPFC or the hippocampus (F(1,81) < 0.6, P > 0.4) in the combined cohort.

Table 2.

Haplotype determination. Based on deCODE SNP allele carrier status, every person (Caucasians and African Americans) in the cohort was assigned a diplotype (haplotype pair) using the SNPHAP program. Hap4 carriers were any individual with the diplotypes (hap1/hap4, hap2/hap4, hap3/hap4 and hap4/hap4), while non-hap4 carriers were diplotypes (hap1/hap1, hap1/hap2, hap1/hap3, hap2/hap2, hap3/hap3 and hap2/hap3)

Haplotype	SNP8NRG221132	SNP8NRG221533	SNP8NRG241930	rs6994992
hap1	G	T	G	C
hap2	G	T	G	T
hap3	G	T	T	C
hap4	G	C	G	T

Association analysis of this haplotype may be impacted by ethnic diversity in our sample. Hap4 is more frequent in Caucasians than African Americans, who have most likely undergone more genetic recombination at this locus (13). Law et al. (13) reported a separate haplotype (hap2) that also contains the risk alleles at SNP8NRG221132 and rs6994992, both of which are associated with altered NRG1 and nAChR α7 expression (see above). While hap4 is infrequent, hap2 is more common and overrepresented in African Americans with schizophrenia (13). Thus, hap2 may represent a risk haplotype in the African American sample. To increase our statistical power by including more African Americans and addressing the possibility that racial groups may have slightly different risk haplotypes, we combined hap2 and hap4 carriers and compared their mRNA expression with that of hap1 and hap3 carriers. With this analysis, a main effect of diplotype was observed in the combined cohort in the DLPFC, with hap2/hap4 carriers having 21% lower expression of nAChR α7 than individuals carrying other haplotypes (F(1,82) = 5.3, P = 0.02) (Fig. 4A). An interaction of diagnosis and haplotype carrier status was observed (F(1,82) = 10.0, P = 0.002) with schizophrenic patients carrying the risk haplotype (n = 21) having significantly lower expression of nAChR α7 mRNA than controls (n = 40) with the risk haplotype (P = 0.01) (Fig. 4B).

Figure 4.

Association between hap4/hap2 carrier status and normalized nAChR α7 subunit mRNA in the DLPFC. (A) A significant effect of haplotype on the normalized nAChR α7 subunit mRNA expression was observed in the whole cohort, with hap2/hap4 carriers having significantly lower amount compared with those individuals carrying hap1 and hap3 (P = 0.02). (B) Splitting the genotypes by diagnosis revealed a significant interaction of diagnosis and genotype (P =0.002) with schizophrenic patients carrying the risk haplotype having significantly lower expression of mRNA than controls who are risk haplotype carriers (P = 0.01). Graphs represent mean ± SEM.

[¹²⁵i] α-BTX binding in DLPFC

Non-specific binding was subtracted from total binding to determine specific binding for [¹²⁵I] α-BTX (Fig. 5A). [¹²⁵I] α-BTX binding was distributed homogenously across all cortical layers of the BA 46 (Fig. 5B) and did not differ rostrocaudally in the combined cohort (F(2,94) = 0.3, P = 0.7) (Fig. 5C). Therefore, specific binding from all three levels was averaged.

Figure 5.

Binding with [¹²⁵I] α-BTX throughout the rostrocaudal extent of the DLPFC. (A) Specific binding was determined by subtracting non-specific binding from total binding as shown by representative autoradiographs of the human DLPFC. (B) [¹²⁵I] α-BTX total binding throughout the rostrocaudal extent of the DLPFC is shown for a normal control subject. (C) Statistical analysis using ANOVA revealed that there is no significant difference in specific binding throughout the rostrocaudal extent. Graph shows means ± SEM.

The relationship between diagnosis and smoking history on [¹²⁵i] α-BTX binding in the DLPFC

Since none of the demographic or tissue-related factors affected average specific binding, ANOVA was performed on the whole cohort without covariates. Average [¹²⁵I] α-BTX-specific binding did not significantly differ between normal controls and patients with schizophrenia in BA 46 (P = 0.9) or with smoking history (P = 0.4) and there were no interactions of the two (F(1,72) = 0.01, P = 0.9) (Fig. 1C). Age and PMI significantly differed between patients with schizophrenia and normal controls; however, including these as covariates revealed no differences in specific binding.

The relationship between 5′ NRG1 SNPs and [¹²⁵i] α-BTX binding in the DLPFC

A two-way ANOVA was performed to examine the association of the four individual NRG1 SNP markers with specific [¹²⁵I] α-BTX binding. We observed a main effect of genotype for SNP8NRG221132 (F(1,71) = 5.94, P = 0.017) in the combined cohort. Individuals homozygous for the risk-associated G allele had a 20% decrease in specific binding compared with A allele carriers (Fig. 6A). There was no significant diagnosis by genotype interaction (F(1,71) = 1.44, P = 0.23) (Fig. 6B). Of the demographic and tissue-related variables examined, pH significantly differed between the two allelic groups; however, including this as a covariate had no impact on the significance level.

Figure 6.

Association between SNP8NRG221132 and [¹²⁵I] α-BTX-specific binding in the DLPFC. (A) A significant effect of genotype on specific binding (fmol/mg tissue) was observed in the whole cohort, with homozygous G allele carriers having significantly lower levels compared with heterozygous G/A individuals (P = 0.02). (B) When the genotype groups were split by diagnosis, a similar trend for binding is observed in both normal controls and patients with schizophrenia. There was no significant diagnosis by genotype interaction. Graphs represent mean ± SEM.

None of the other SNPs from the deCODE haplotype were associated with [¹²⁵I] α-BTX binding in the DLPFC, and there were no significant interactions of genotype with either diagnosis or smoking (all F values <0.5; all P-values >0.4). Additionally, no main effect of diplotype or interaction with diagnosis was observed on receptor-binding levels (F values <0.02; P-values >0.9).

DISCUSSION

One of the many functions of NRG1 is the regulation of nAChR α7 expression (14–16). We tested the hypothesis that schizophrenia-associated SNPs within the NRG1 gene may have downstream effects on nAChR α7 mRNA expression and receptor density via their effects on NRG1 isoform expression. Diagnostic comparisons between normal controls and patients with schizophrenia did not reveal any alterations in nAChR α7 mRNA expression in the hippocampus or alterations in mRNA or receptor-binding levels within the DLPFC. However, we found that nAChR α7 mRNA expression is significantly lower in the DLPFC of individuals carrying risk alleles at two schizophrenia-associated NRG1 SNPs, SNP8NRG221132 and rs6994992. Individuals with haplotypes containing risk alleles for the above two SNPs (13) were also found to have significantly decreased nAChR α7 mRNA expression. In addition, in the DLPFC, the risk allele at SNP8NRG221132 was associated with significantly lower [¹²⁵I] α-BTX binding. These findings suggest that NRG1 polymorphisms that increase risk for schizophrenia are also associated with decreased α7 subunit mRNA and receptor levels in the DLPFC.

The precise mechanism by which specific NRG1 isoforms influence α7 nAChR mRNA and protein levels is unknown. The EGF-like domain peptide used for many in vitro studies is common to all known NRG1 isoforms (18). The NRG1 type I β1 isoform has an Ig-like domain at its amino terminus and may up-regulate nAChRs concentrated in the post-junctional membrane of the neuromuscular junction (14). Consistent with the observation that NRG1 Ig heterozygous mice have a 50% reduction in the density of post-synaptic nAChRs at the neuromuscular junction (17), chronic treatment with NRG1- β1 isoform has been associated with increased [¹²⁵I] α-BTX binding in hippocampal slices (16). This up-regulation of nAChRs most likely occurs through a forward-signaling pathway involving either the ErbB2 or ErbB4 receptors. However, acute treatment with NRG1-β1 down-regulated α7 receptors in mice hippocampal interneurons (19). Based on these previous observations combined with the fact that type I and type IV NRG1 belong to the Ig-containing class of NRG1s, we would have predicted that allelic variations in NRG1 that are associated with increased levels of these isoforms would also be associated with increased levels of nAChR α7 mRNA and receptor binding. However, we did not find this relationship between NRG1 genotype and α7 nAChR mRNA and protein levels in the DLPFC and hippocampus. Indeed, our results suggest the opposite relationship.

In human brain, normal controls homozygous for the risk allele at SNP8NRG221132 had lower hippocampal expression of the NRG1 type I isoform compared with other genotypes (13). However, schizophrenic subjects homozygous for the risk allele at SNP8NRG221132 had higher expression of the NRG1 type I isoform in the hippocampus (13). Similar effects were reported for the DLPFC (13). Our mRNA and receptor-binding data in the DLPFC are consistent with the findings of Law et al. (13) in controls and are consistent with the effects of NRG1 on nAChR α7 expression in experimental models (i.e. homozygosity at the risk allele predicts lower NRG1 type I and lower nAChR α7 expression). However, our findings are counter to earlier results in schizophrenic subjects in the DLPFC and in both groups in the hippocampus (13). We have no obvious explanation for these apparent discrepancies. The degree to which other factors may alter these genotype–mRNA level relationships in postmortem human brain tissues are unknown. A combination of factors, including high levels of chronic nicotine exposure, antipsychotic treatment and/or environmental deprivation in schizophrenic subjects might differentially impact on these relationships between subject groups. These uncertainties notwithstanding, we are struck nonetheless by the fact that genotype at this risk associated SNP in NRG1 predicts variation in nAChR α7 mRNA expression.

Carriers of the risk allele for rs6994992 and hap4 have been shown previously to have significantly higher NRG1 type IV isoform mRNA expression in the hippocampus (13). We did not find an effect of this SNP or of hap4 on nAChR α7 mRNA expression in the hippocampus. In the DLPFC, however, we found that in the entire sample, carriers of the risk allele at rs6994992 had significantly lower nAChR α7 mRNA expression. We were unable to observe an effect of hap4 alone on nAChR α7 mRNA expression in the DLPFC, perhaps because of the smaller sample size compared with the Law et al. study (13) and the impact of racial differences in haplotype frequencies. In an effort to accommodate potential racial differences in risk haplotypes, and since two of the NRG1 SNPs that independently affect nAChR α7 mRNA expression are found in both hap2 and hap4, we combined these two haplotype groups. In the combined haplotype analysis, we observed a main effect of haplotype on nAChR α7 mRNA expression in the DLPFC. Unfortunately, since the effects of genetic variation at rs6994992 on type IV isoform expression in the DLPFC are unknown, any predictions on the directionality of either rs6994992 or hap4 on nAChR α7 expression in this region would be premature.

Consistent with previous studies of the prefrontal cortex (20,21), we found no change in nAChR α7 transcript levels between normal controls and schizophrenic subjects in the DLPFC. There are no published reports on [¹²⁵I] α-BTX binding in the DLPFC comparing normal controls to schizophrenia subjects. However, while we observed no differences in transcript levels in the hippocampus, there are reports that receptor levels in the hippocampus are significantly reduced in patients with schizophrenia (22). This discrepancy lends credence to the supposition that there might be post-translational factors regulating nAChR α7 protein levels independent of the regulation of mRNA transcription, at least in the hippocampus. Because we observed an association between schizophrenia risk alleles for NRG1 with lower nAChR α7 mRNA and binding levels in the DLPFC, we might have expected a difference between normal controls and schizophrenic subjects. However, in this study, there were more risk allele carriers in the normal control group than in patients so any expected diagnostic effect might be masked by the distribution of allelic frequencies. Moreover, as mentioned before, environmental factors specific to schizophrenia subjects might obscure the effect of NRG1 susceptibility alleles on nAChR α7 mRNA and/or receptor levels. For example, both atypical and typical anti-psychotics increased acetylcholine release in rat hippocampus (23), while another study found that prolonged risperidone treatment significantly decreased [¹²⁵I] α-BTX binding levels in the rat hippocampus and neocortex (24). Finally, as mentioned previously, the link between NRG1 and nAChR α7 mRNA expression may be different in schizophrenic patients.

Despite our findings on the association of lower nAChR α7 mRNA expression with rs6994992 and hap4 in the DLPFC, we did not find a relationship of these genetic variations with [¹²⁵I] α-BTX binding. This may be due to the inherent variability in radioligand binding assays. In agreement with results by Sugaya et al. (25), our Scatchard analysis of receptor binding in the human DLPFC showed two binding sites for [¹²⁵I] α-BTX (See Supplementary Fig. S1). With 5 nM of [¹²⁵I] α-BTX, our specific binding in the DLPFC in normal controls was similar to that reported for other human cortical regions (26,27). Our Scatchard analysis also suggested that while 5 nM saturates the high affinity (presumably the homomeric α7) receptors, it also binds to a small percentage of lower affinity sites. Although the nature of this lower affinity site is unknown, it may represent a functional heteromeric α7* receptor complex composed of either α7 mRNA splice variants (28–31) or other nAChR subunits (32–34), increasing the variability of our data set which may have complicated the rs6994992 and hap4 analyses. There might also be post-translational factors regulating nAChR α7 protein levels independent of mRNA transcription. Finally, another explanation for this inconsistency is that our binding assay measures receptor density not only on the cell surface, but also of internalized receptors; therefore, mRNA levels may more accurately assess the receptor synthesis rate, which in turn may be closer to the functional state of gene expression.

NRG1 genotype appears to have an effect on nAChR α7 mRNA in the DLPFC, but not in hippocampus. There are well-demonstrated regional differences in the regulation of nAChRs in human brain, in response to nicotine exposure (35). Moreover, while haloperidol alone did not affect binding of [³H] methyllycaconitine (MLA), a nAChR α7 receptor antagonist, in combination with nicotine, it increased cortical binding and decreased hippocampal binding (36). Studies in rodents and primates reported especially high densities of α7 nAChRs in the hippocampus (36–39) that decrease with age (40) suggesting that these receptors may play a regionally specific role in brain development and maturation. Additionally, the anterior hippocampus may be selectively involved in the pathophysiology of schizophrenia (41–43); therefore, an abnormality in the anterior hippocampus would be missed in an assay based on tissue homogenates taken from the entire hippocampus. Slide-based studies of the hippocampus might resolve some of these issues.

While nicotine exposure is associated with an up-regulation of the α4β2 receptors, its effect on α7 receptors is not as clear. In mice, chronic nicotine exposure led to a modest elevation in [¹²⁵I] α-BTX binding and occurred only at high doses of nicotine (44). On the other hand, in rats, cortical [¹²⁵I] α-BTX binding was unchanged after 7 and 14 days of continuous nicotine infusion (45). In humans, there is even less consensus regarding effects of nicotine on α7 receptors. While α7 receptor protein levels were significantly increased in the hippocampus (35), the same group reported no changes in frontal cortex associated with smoking. Another study (46) reported that [¹²⁵I] α-BTX binding was unaltered in the hippocampus of smokers. Our lack of an association between α7 nAChR mRNA and binding levels and smoking history agrees with the latter study.

Functional relationships between NRG1 and nAChR neurotransmission may explain some of the intermediate phenotypes associated with schizophrenia. Dysfunction of DLPFC and hippocampus contributes to schizophrenia symptoms (47–49). Within the DLPFC, nAChR α7 receptors are located on pyramidal neurons and GABAergic interneurons (50–52), while in the hippocampus they are located on GABAergic interneurons (53,54). Abnormalities in nAChR function in the DLPFC and hippocampus may lead to cognitive and memory impairments (55,56) and sensory gating deficits (57). Infusions of the nAChR α7 antagonist MLA into the rat ventral hippocampus causes memory deficits (56,58,59). In humans, nAChR agonists improve attention, learning and memory (55,60–63) and sensory gating deficits (64,65). Ventricular infusion of a cholinergic toxin leads to impairment in long-term potentiation (LTP). This effect was reversed and the induction of LTP was restored by a nicotine infusion (66). In addition, NRG1β may de-potentiate LTP via internalization of α-amino-3-hydroxy-5-methylisoxazole-4-propionic acid (AMPA) receptors (67), an effect reversed by inhibition of ErbB receptors (67). Therefore, the effect of neuregulins on LTP might occur by affecting nAChR α7 transcription through ErbB receptor-mediated forward-signaling. The cognitive and electrophysiological deficits associated with schizophrenia might be due to impaired nicotinic neurotransmission mediated in part by disease-associated SNPs for NRG1.

In summary, our results provide evidence that disease-associated variants in the 5′ regulatory region of the NRG1 gene that are associated with NRG1 type I and type IV levels also are associated with decreased nAChR α7 mRNA expression and receptor density in the DLPFC, but not the hippocampus. The mechanism by which this occurs requires additional investigation. The results of the current study demonstrate how susceptibility genes may interact to contribute to the pathophysiology of schizophrenia.

MATERIALS AND METHODS

Human postmortem tissue collection

Postmortem brains were obtained with informed consent from the legal next of kin (protocol #90-M-0142 approved by the NIMH/NIH Institutional Review Board) (68). Clinical characterization, diagnoses and macro- and microscopic neuropathological examinations were performed on all cases using a standardized paradigm (69). Toxicological analysis was conducted on every case. All control subjects were free of a history of psychiatric illness or significant alcohol or drug abuse. Positive toxicology was not an exclusion criterion for schizophrenic cases. Control and patient cohorts differed by sex, smoking history and manner of death. Smoking history (regular use of cigarettes) was rated as ‘yes’ or ‘no’ at the time of death based on medical records, family interviews and/or nicotine levels. Manner of death was categorized as natural, homicide, suicide, accident or unknown based upon reports from the Medical Examiners’ offices. These factors did not significantly affect normalized nAChR α7 mRNA expression or binding.

Tissue retrieval and processing

Brains were hemisected, cut into 1.5 cm coronal slabs, flash frozen and stored at −80°C. For quantitative RT–PCR (qRT–PCR), hippocampal and DLPFC dissections were performed as previously described (69). For receptor binding and autoradiography, DLPFC blocks from the contralateral hemisphere were cut into 14 µm thick sections, mounted on double-subbed slides (Fisher Scientific, Pittsburgh, PA) and stored at −80°C. Three pairs of slides were used from the rostrocaudal extent of the DLPFC for each subject (at the midpoint and ∼320 µm rostral and caudal to the midpoint). Nissl stained sections from each brain were examined to verify that the sections were within the demarcations of BA46 using criteria from Rajkowska et al. (70). Pulverized cerebellum was used for pH measurement.

RNA extraction and qRT–PCR

Reverse transcription and qRT–PCR were performed as described previously (13,69). Transcript expression was measured using an ABI Prism 7900 sequence detection system (ABI, Foster City CA, USA), following a previously published protocol (71). All measurements were performed in triplicate with gene expression calculated as the mean of triplicates.

Probes/primers

For quantification of nAChR α7 mRNA expression, a specific combination of primer and probes (Applied Biosystems, Foster City, CA, USA) was used. A Taqman probe/primer set, (Cat No. Hs01063372_m1) spanning exons 4 and 5, which specifically recognizes the full-length nAChR α7 transcript (and not CHRNFAM7A) (72) was used. β-glucuronidase (GUSB), β-actin and β-2-microglobulin (B2M) (Applied Biosystems, Assays-on Demand, Cat No. HS99999908, Hs99999903, Hs99999907, respectively) were used as control genes for normalization, by applying the geometric mean of the three control genes as the normalization factor.

Receptor-binding autoradiography and saturation binding

Due to high non-specific binding of the [¹²⁵I]α-BTX in tissue homogenates, a full Scatchard binding analysis (20 points) was performed on slide-mounted human DLPFC sections (n = 5 normal controls) over a concentration range of 0.01–30 nM following published protocols by Clarke et al. (73). Based on these results (see Supplementary Fig. S1) and studies performed in rats and mice (73,74), subsequent receptor-binding autoradiography was performed using 5 nM concentration of [¹²⁵I] α-BTX following a published protocol (75) with slight modifications. This concentration saturates the homomeric α7 receptors (73,74). Slide-mounted sections were warmed to room temperature until dry and preincubated (30 min) at room temperature in buffer containing 50 mm Tris–HCl (pH 7.4), 1 mg/ml BSA, with or without 2.5 mm nicotine bitartrate (Sigma-Aldrich, St. Louis, MO) for either non-specific binding or total binding, respectively. After the preincubation procedure, sections were air dried (30 min) followed by incubation (2 h) in buffer containing 5 nM [¹²⁵I] α-BTX (Specific activity: 91.57 Ci/mmol, Amersham Biosciences) ± 2.5 mm nicotine. Slides were subsequently washed 4 × 15 min in ice-cold buffer, dipped twice in ice-cold deionized water, air dried and apposed to Kodak Biomax MR film with ¹²⁵I microscales (Amersham Biosciences, Arlington Heights, IL) for three days.

Film quantification

For receptor-binding autoradiography, films were developed; images were scanned and quantitative densitometric analysis was performed using Image J (version 1.36b http://rsb.info.nih.gov/ij/index.html). The density of binding, calculated in fmol/mg tissue was based on co-exposed [¹²⁵I] microscale standards and adjustments made according to the ligand’s specific activity. Specific binding was determined by subtracting non-specific binding from total binding (see Fig. 5A for representative autoradiograms). Two independent measures of the middle frontal gyrus (BA46) were made and the analysis was performed on the average of the two measurements.

NRG1 genotype determination

Cerebellar DNA was extracted using a standard protocol supplied by PUREGENE (Gentra Systems, Minneapolis, MN). Four markers from the original deCODE haplotype (SNP8NRG221132; SNP8NRG221533; SNP8NRG241930; rs6994992) (Table 1) were used for single SNP analysis and for haplotype analyses. Details on the specific SNP sequences and Taqman probe/primer sets used for genotyping are described elsewhere (10,13). Two 3′ NRG SNPs (rs10954867, rs7005288), not associated with schizophrenia and with no effect on the expression of NRG1 isoforms types I and IV (13), were chosen as negative controls. LD between SNPs was determined using the program LDMAX/GOLD (76). The program SNPHAP written by David Clayton (version 1.0, http://www-gene.cimr.cam.ac.uk\clayton\software\) was used to calculate haplotype frequencies and to assign individual diplotypes (Table 2).

Statistical analysis

For all mRNA analyses, the subjects were under the age of 90 years with a RIN of 3.8 or higher for the hippocampus and RIN of 4.0 or higher for the DLPFC and included only Caucasians and African Americans. Therefore, all mRNA analyses in the hippocampus were performed on 64 normal controls and 30 schizophrenic patients (Table 3) and in the DLPFC on 61 normal controls and 30 schizophrenic patients (Table 3). For receptor binding and autoradiography, a subset of DLPFC samples was available and included 49 normal controls and 27 patients with schizophrenia (Table 3). Statistical analyses were performed using Statistica [StatSoft, Inc., STATISTICA (data analysis software system) version 7.1 www.statsoft.com]. By the Kolmogorov–Smirnof test, the data followed a normal distribution and therefore parametric analyses were used for all comparisons. Multiple regression analyses were performed to assess the contributions of age, sex, pH, PMI, RIN and smoking history on endogenous control genes, nAChR α7 mRNA expression and receptor binding. Multiple regression analyses were also performed to assess the effects of age at onset of illness, age at first hospitalization, lifetime neuroleptic exposure, total daily dose and final neuroleptic dose (converted to Chlorpromazine equivalents), on mRNA expression and receptor binding in the patients with schizophrenia. Comparisons between diagnostic groups were made using univariate ANCOVA for mRNA expression and receptor binding with diagnosis and smoking history as independent variables. Effects of allelic variation on gene expression and receptor binding were examined using univariate ANCOVA with genotype and diagnosis as independent variables. Inclusion of race as a factor in the ANCOVA for either diagnostic or genotype comparisons showed no affect on nAChR α7 expression or binding. To increase power for statistical analysis of SNPs with minor allele frequencies <10%, individuals heterozygous and homozygous for the rare allele were grouped as minor allele carriers. All experiments were conducted blind to diagnosis. Bonferroni correction for multiple SNP genotyping was not performed as our analysis was based on an a priori hypothesis that there is biological evidence for an association of NRG1 with nicotinic receptor α7 expression. Moreover, three of the SNPs in the deCODE haplotype are in LD and therefore, not considered to be independent observations.

Table 3.

Characteristics of subject cohort used for mRNA expression and receptor binding studies. Data shows mean ± SD

	n	Race	Sex	Age (years)	PMI (h)	pH	RIN	Smoking hx
Hippocampus mRNA
Controls	64	42AA/22C	45M/19F	41 ± 15	30 ± 15	6.6 ± 0.3	5.9 ± 1.0	44N:20Y
Schizophrenia patients	30	18AA/12C	22M/8F	46 ± 15	36 ± 18	6.6 ± 0.2	5.4 ± 1.0	7N:23Y
DLPFC mRNA
Controls	61	39AA/22C	41M/20F	43 ± 14	32 ± 14	6.6 ± 0.3	6.7 ± 1.2	45N:16Y
Schizophrenia patients	30	18AA/12C	16M/14F	48 ± 15	37 ± 18	6.4 ± 0.3	6.8 ± 1.1	6N:24Y
DLPFC receptor binding
Controls	49	34AA/15C	30M/19F	43 ± 13	34 ± 14	6.5 ± 0.3	—	35N:14Y
Schizophrenia patients	27	19AA/8C	19M/8F	52 ± 15	39 ± 21	6.4 ± 0.3	—	9N:18Y

AA, African Americans; C, Caucasians; N, non-smoker; PMI, postmortem interval; RIN, agilent RNA integrity number (on a scale of 1–10); Y, smoker.

SUPPLEMENTARY MATERIAL

Supplementary Material is available at HMG Online.

FUNDING

This research was supported in its entirety by the Intramural Research Program of the NIMH, NIH. Dr Amanda J. Law is a UK Medical Research Council Career Development Fellow and NARSAD Young Investigator.

ACKNOWLEDGEMENTS

The authors would like to thank Amy Deep-Soboslay, M.Ed. and Llewellyn B. Bigelow, M.D. of the Clinical Brain Disorders Branch, GCAP, IRP, NIMH for their efforts in clinical diagnosis and demographic characterization, and Mary M. Herman, M.D. and her staff for the neuropathological screening of all subjects included in this study. Special thanks to Vesna Imamovic and Bianca Iglesias for their excellent technical assistance.

Conflict of Interest statement. None declared.