Influence of NOS1 on Verbal Intelligence and Working Memory in Both Patients With Schizophrenia and Healthy Control SubjectsNOS1, Verbal Intelligence, and Working Memory FREE

Schizophrenia (SZ) is a disorder of substantial heritability (approximately 80%) but uncertain underlying pathophysiology.¹ Despite some progress, most of the variance in SZ susceptibility attributable to genetic factors remains to be identified.² A number of recent studies using high-throughput technologies investigating structural and genomic sequence variation in large SZ samples have identified more consistent support for several susceptibility loci.^{3 - 5} As genomic data emerge, important questions are how these variants increase disease liability and whether they are relevant to only the neurobiology of SZ or to neurodevelopment more generally.^{6 - 7}

One putative SZ risk gene is the nitric oxide synthase 1 gene (NOS1 [OMIM 163731]) (also known as neuronal NOS or nNOS), which maps to chromosome 12q24 and synthesizes nitric oxide (NO) in both the central and peripheral nervous system. Nitric oxide is a highly reactive messenger molecule that diffuses freely across membranes, stimulating guanylyl cyclase and modifying protein structure with multiple roles in immune, cardiac, and neurological function. It is produced by different NO synthetase (NOS) enzymes, including neuronal NOS, and is transported to different cellular compartments by adaptor proteins to minimize nonspecific interactions.

Transcription of NOS1 involves 12 alternate untranslated first exons and coding regions consisting of 28 exons covering 240 kilobases. Following modest linkage support to the NOS1 region,^{8 - 11} 4 of 5 published NOS1 candidate gene association studies suggest evidence of association with SZ,^{12 - 15} with the exception being the study by Liou et al.¹⁶ Moreover, in their SZ genome-wide association study of 479 cases and 2937 controls, O’Donovan et al⁴ identified a single-nucleotide polymorphism (SNP) at the NOS1 locus (rs6490121) as being 1 of 12 SNPs with strong initial statistical evidence for association (P = 9.82 × 10⁻⁶). The same allele at this SNP was significantly associated in a replication sample of 1664 cases and 3541 controls of European ancestry but not in a sample of mixed European and Asian ancestry.⁴

Functional evidence for involvement of NO in SZ comes from studies showing abnormal distribution of nitrinergic neurons in frontal and temporal lobes in SZ,^{17 - 18} increased NO metabolites in the serum of patients with SZ,^{19 - 21} and postmortem increased NOS1 messenger RNA in prefrontal cortex of patients.²² Neurally generated NO displays many of the properties of a neurotransmitter and is involved in processes of potential relevance to SZ, including synaptogenesis, long-term potentiation, neurotransmitter release, synaptic plasticity, blood flow, and cell death.^{23 - 24} These functions are dependent on the delivery and localization of NOS1 to multiple intracellular locations by specific adaptor proteins. Presynaptically, NOS1 is linked to the NOS1 (neuronal) adaptor protein gene (NOS1AP [OMIM 605551]) and forms a complex with synapsin, which is involved in the assembly of synaptic vesicle clusters and synaptogenesis. Both NOS1AP and the synapsin II gene (SYN2 [OMIM 600755]) have been implicated in SZ by genetic association studies.^{25 - 29} Postsynaptically, NOS1 is coupled to N-methyl-D-aspartate receptors through postsynaptic density 95, representing another potential mechanism for involvement in SZ susceptibility. Molecular pathway analysis of structural variants implicated in SZ by Walsh et al³⁰ identified a significant excess of disrupted genes involving the NO signaling pathway. Taken together, these studies provide support for the involvement of NO signaling in SZ.

In mouse models, NOS1 knockouts have repeatedly been associated with variance in cognition.^{31 - 32} Notably, phencyclidine hydrochloride–induced cognitive and behavioral deficits that model SZ symptoms (including prepulse inhibition, habituation of acoustic startle, latent inhibition, spatial learning, spatial reference memory, and working memory) can all be prevented by interfering with the production of NO.^{33 - 42} In patients with SZ, Reif et al¹³ reported that 2 of 4 genetic markers tested at the NOS1 loci were associated with variance in performance on a measure of prefrontal function (the continuous performance task) in terms of behavioral performance, P300 peak amplitude, and response latency. Whether this represents a specific or more general effect on cognition or affects cognition in the general population is unknown.

The purpose of this study was to investigate the effect of NOS1 on cognition in SZ and control populations. We selected a single SNP at NOS1 (rs6490121) from the initial genome-wide association study by O’Donovan et al⁴ to which our cases had contributed as replication samples. We chose this SNP alone as it was the only SNP in the study by O’Donovan et al⁴ that surpassed a statistical significance threshold of P < 1.0 × 10⁻⁵ and was replicated with the same allele G in the European sample. We hypothesized that the genotypes carrying the risk allele would be associated with poorer performance on a selected set of cognitive domains typically impaired in SZ, namely IQ, episodic and working memory, and attentional control.⁴³

The Irish sample consisted of 349 cases and 230 controls. Ninety-one of the patients were genotyped as part of the prior genome-wide association study⁴ ; the remaining patients and all of the control subjects were independent of that study. Cases consisted of clinically stable patients with a DSM-IV diagnosis of SZ (n = 294) or schizoaffective disorder (n = 55) recruited from 5 sites across Ireland. Inclusion criteria required that participants were aged 18 to 65 years, had no history of comorbid psychiatric disorder, had no substance abuse in the preceding 6 months, had no prior head injury with loss of consciousness, and had no history of seizures. Diagnosis was confirmed by trained psychiatrists using the Structured Clinical Interview for DSM-IV Axis I Disorders–Patient Edition.⁴⁴ Additional diagnostic details and clinical sample characteristics ascertained at the time of the interview, including symptom severity (scores on the Scale for the Assessment of Positive Symptoms and the Scale for the Assessment of Negative Symptoms^{45 - 46} ) and medication dosage, are detailed elsewhere.⁴⁷

The healthy control sample was recruited on the basis of responses to local media advertisements. Control participants were included only if they were aged between 18 and 65 years and, based on clinical interview, satisfied the criteria of having no history of major mental health problems, intellectual disability, acquired brain injury, or substance misuse in the preceding 6 months based on self-report. Control participants were also excluded from the study if they reported having a first-degree relative with a history of psychosis. All patients and control assessments were conducted in accordance with the relevant ethics committees' approval from each participating site. All patients and control subjects were of Irish ancestry (ie, had 4 grandparents born in Ireland) and all provided written informed consent.

The German sample consisted of 240 clinically stable patients with a DSM-IV diagnosis of SZ and 1344 healthy control subjects, all of whom were genotyped as part of the previous study.⁴ Patients were ascertained from mental health services in the Munich area and all participants provided written informed consent. Inclusion criteria were a diagnosis of SZ (symptom duration >6 months) and being aged 18 to 65 years. Exclusion criteria included a history of head injury or neurological diseases. Detailed medical and psychiatric histories were collected, including a clinical interview using the Structured Clinical Interview for DSM-IV Axis I Disorders⁴⁴ and the Structured Clinical Interview for DSM-IV Axis II Personality Disorders⁴⁸ to evaluate lifetime Axis I and II diagnoses. Four physicians and 1 psychologist rated the Structured Clinical Interview for DSM-IV Axis I Disorders and Structured Clinical Interview for DSM-IV Axis II Personality Disorders responses, and all measurements were double-rated by a senior researcher (A.C.). Among the patients, 68% were of strict German descent (ie, all 4 grandparents were born in Germany) and the other 32% were German Caucasian. Of the 240 patients, 216 completed the full Wechsler Adult Intelligence Scale–Revised⁴⁹ assessment and 232 completed a further comprehensive neuropsychological battery.

Healthy control participants of German descent (ie, all 4 grandparents being German) were randomly selected from the general population of Munich, Germany, and contacted by mail. Control subjects for this study were included only if they were aged between 18 and 65 years. To exclude subjects with central neurological diseases and psychotic disorders or subjects who had first-degree relatives with psychotic disorders, several screenings were conducted before the volunteers were enrolled in the study. First, subjects who responded were initially screened by telephone for the absence of neuropsychiatric disorders. Second, detailed medical and psychiatric histories were assessed for both themselves and their first-degree relatives by using a semistructured interview. Third, if no exclusion criteria were fulfilled, the subjects were invited to a comprehensive interview including the Structured Clinical Interview for DSM-IV Axis I Disorders–Patient Edition⁴⁴ and the Structured Clinical Interview for DSM-IV Axis II Personality Disorders⁴⁸ to validate the absence of any lifetime psychotic disorder. Additionally, the Family History Assessment Module⁵⁰ was conducted to exclude psychotic disorders among first-degree relatives. A neurological examination was also conducted to exclude subjects with current central nervous system impairment. In volunteers older than 60 years, the Mini-Mental Status Test⁵¹ was performed to exclude subjects with possible cognitive impairment. All of the 1344 control participants completed the full Wechsler Adult Intelligence Scale–Revised assessment, and 364 completed tests from a further extensive neuropsychological battery.

This study was designed so that identical or near identical tests of the phenotypes of general cognitive function (IQ), episodic memory, working memory, and attention were used for both the Irish discovery samples and the German replication samples. Only 1 verbal and 1 visuospatial measure for each cognitive function were tested in the Irish samples. Replication in the German samples was then sought for the cognitive functions showing significant association in the discovery sample.

General cognitive functioning (IQ) was measured in the Irish sample using selected subtests (vocabulary, similarities, block design, and matrix reasoning) from the Wechsler Adult Intelligence Scale, third edition,⁵² yielding full-scale, verbal, and performance IQs. For the German samples, IQ was indexed by the German version of the Wechsler Adult Intelligence Scale–Revised⁴⁹ using all 11 verbal or performance subtests (vocabulary, comprehension, information, digit span, arithmetic, similarities, block design, picture completion, picture arrangement, object assembly, and digit symbol coding).

Verbal episodic memory and visual episodic memory were assessed in the Irish samples using the logical memory and faces subtests, respectively, from the Wechsler Memory Scale, third edition.⁵³ In the German samples, these were assessed using the logical memory and visual memory scores from the German version of the Wechsler Memory Scale–Revised.^{54 - 55}

Verbal working memory and spatial working memory were assessed in the Irish samples using the letter-number sequencing task from the Wechsler Memory Scale, third edition, and the spatial working memory task from the Cambridge Neuropsychological Test Automated Battery, Expedio Version.⁵⁶ In the German samples, working memory was measured using the composite digit span and spatial span score from the Wechsler Memory Scale–Revised and the N-back task (National Institute of Mental Health version).⁵⁷

Attentional control was assessed in the Irish samples using the continuous performance test, identical pairs version.⁵⁸ In the German sample, it was assessed using the 3-7 Continuous Performance Test.⁵⁹ Full descriptions of all memory and attention tasks are provided in the eText.

The SNP rs6490121 was genotyped in the German samples and a proportion of the Irish samples using the iPLEX Gold system (Sequenom, Inc, San Diego, California) (further details are provided in the article by O’Donovan et al⁴ ). The call rate for the iPLEX genotyping was greater than 99%, and all Irish and German case and control samples were in Hardy-Weinberg equilibrium (P > .05). The remainder of the Irish samples (n = 529) was genotyped using a Taqman SNP Genotyping Assay on a 7900HT Sequence Detection System (Applied Biosystems Inc, Foster City, California). The call rate for the Taqman genotyping was greater than 95%, and both case and control samples were in Hardy-Weinberg equilibrium (P > .05). Along with the Irish samples, a number of HapMap CEU DNA samples (http://www.hapmap.org) were genotyped for rs6490121 for quality control purposes and were all found to be concordant with available online HapMap data for this SNP.

Association between NOS1 rs6490121 and the cognitive intermediate phenotypes of general cognitive function, episodic memory, working memory, and attentional control was tested using a general factorial design in SPSS statistical software version 14.0 (SPSS Inc, Chicago, Illinois). In the original article identifying this variant,⁴ there was no difference in the genotypic vs allelic model; as there was no evidence on which to test a specific dominant or recessive model and as sample sizes allowed, our analysis was based on a comparison of all 3 genotype groups. The NOS1 genotype (GG vs AG vs AA) and diagnosis (cases vs controls) were entered as fixed effects. Scores for each neuropsychological subtest were entered as the dependent variables, with age and sex included as covariates as appropriate.

Clinical and demographic characteristics by genotype for each sample appear in Table 1. For the discovery control sample, NOS1 genotypes were not associated with age, sex, or years of education. For the patient group, AA genotype carriers were observed to be older than the AG carriers by an average of 4 years (t = −2.94; uncorrected P = .03); the GG group did not differ significantly in age from the other genotype groups. To investigate whether possible differences in symptom severity rating between genotype groups might influence cognitive performance in cases, we conducted a principal components analysis with varimax rotation of all scores on the Scale for the Assessment of Positive Symptoms and the Scale for the Assessment of Negative Symptoms, yielding 3 factors of positive, negative, and disorganized symptoms (Table 1, eTable 1, and eTable 2). We did not observe any differences on any of the symptom factors associated with genotype. We also did not observe differences in medication dosage associated with NOS1 genotype (all P > .05). A similar analysis of Positive and Negative Syndrome Scale⁶⁰ scores in the German sample also failed to identify any differences in symptom severity associated with NOS1; medication dosage was similarly nonsignificant (all P > .05).

Analysis of verbal IQ in the Irish sample revealed a significant overall effect of genotype (Table 2). Tukey post hoc analysis revealed that this effect was driven by differences between the GG and AA groups (t = −6.36; P = .04). An effect size estimate (partial η²) suggests that genotype explains approximately 2% in cases and approximately 3.5% in controls. In clinical terms, carriers of the risk GG allele performed on average 5 or 6 standard score verbal IQ points below carriers of the AG and AA genotypes (Figure 1). No genotype × diagnostic group interaction was observed (even though, as expected, patients performed significantly lower than control subjects on this and all other cognitive tests [Table 2]). Although the same general pattern of results was observed for performance and full-scale IQ scores, these differences were not statistically significant (P > .05).

Analysis of verbal IQ in the German sample also revealed a significant overall effect of genotype (Table 2), with carriers of the risk GG allele performing significantly below AA and AG carriers. Based on post hoc analysis, the largest difference was between the GG and AG genotype groups (t = 2.87; P = .01), with the difference between the GG and AA groups just failing to achieve statistical significance (t = 2.34; P = .051). In this sample, an interaction effect with affected status was also observed (F = 11.51; P < .001); as illustrated in Figure 1, the association with genotype appears to be strongest in cases. An effect size estimate (partial η²) suggests that genotype explains approximately 5% of variance in verbal IQ in cases and 0.2% in controls.

Following this replication of association between NOS1 genotype and verbal IQ, we investigated post hoc whether NOS1 was also associated with the 2 other indices of IQ in the German sample—performance and full-scale IQ. A similar pattern of differences associated with genotype was observed (Table 2). Genotype × diagnosis interactions were observed for both scores (performance IQ: F = 11.52; P < .001; full-scale IQ: F = 8.95; P < .001) such that the effect of genotype was apparent in cases only (Figure 1). For cases, the estimate of effect size was 6.9% for performance IQ and 6.4% for full-scale IQ.

As an additional secondary analysis, we further sought to determine whether the overall effect of genotype was comparable when cases and controls were considered separately. For the Irish samples, verbal IQ was not significantly associated with NOS1 when cases and controls were considered individually (P > .05), suggesting insufficient power to analyze these samples separately. In the German samples, the GG genotype was associated with significantly poorer verbal IQ in patients (F = 5.54; P = .005) but not control subjects; this pattern of association in patients but not control subjects was also observed for performance IQ in patients (F = 7.84; P = .001) and full-scale IQ in patients (F = 7.26; P = .001).

For working memory, we observed a significant difference associated with genotype in verbal working memory as measured in the Irish samples by scores on the letter-number sequencing task of the Wechsler Memory Scale, third edition (Table 2). While the association was statistically significant as a main effect (F = 4.02; P = .02), as Figure 2 illustrates and a statistical interaction of genotype × diagnosis confirms, this association is clearly driven by differences in the control group. For the control group, the risk GG allele carriers performed 3 scaled score points (1 SD based on norms for the test) lower than the AA carriers on this measure of verbal working memory, with the between-group difference significant between GG carriers and both other genotype groups (GG < AG: t = −2.21; P = .01; GG < AA: t = −3.05; P < .001). A significant main effect of genotype was also observed with the spatial working memory task used (Table 2), in the absence of an interaction with diagnosis. Post hoc analysis reveals the difference as largest between the AG and AA groups (t = 0.31; P = .02).

Using 2 different measures of working memory in the German samples, the risk GG allele was again associated with poorer performance on both measures (Table 2). While association with genotype was observed as a main effect, in the absence of a statistical interaction with diagnosis, the effect size appeared slightly larger in cases than in controls (N-back partial η² = 2.2% vs 1.1%, respectively).

We again sought to determine whether the overall effect of genotype was comparable when cases and controls were considered separately. For the Irish samples, differences on both working memory tasks were associated with NOS1 genotype in controls (letter-number sequencing task: F_2,163 = 9.02; P < .001; spatial working memory task: F_2,163 = 4.99; P = .008) but not patients (P > .05). When the German case and control samples were considered separately, the association was seen in the patient group but not the control group on the working memory score of the Wechsler Memory Scale, third edition (F = 5.49; P = .005); for the N-back task, the effects of genotype were no longer apparent when cases and controls were considered separately.

In the Irish samples, no differences associated with NOS1 genotype were observed on either the verbal or nonverbal memory tests used (the logical memory and Cambridge Neuropsychological Test Automated Battery paired associates learning tests, respectively). Similarly, when the effects of NOS1 on attention were investigated using the intradimensional-extradimensional task (scores for block 8) and the continuous performance test, identical pairs version task (using d′ values for each of the 3 conditions), no significant associations were observed. As no association with NOS1 genotype was observed in the Irish samples (all P > .05), we did not investigate these aspects of cognition in the German samples.

Variants at NOS1 have previously been associated with both increased SZ risk and variation in cognitive performance. The involvement of NOS1 in cognition is well established from animal models, with the N-methyl-D-aspartate–NO–cyclic guanine monophosphate signaling cascade being linked to neurotransmitter release and synaptic plasticity in multiple genetic and pharmacological studies.⁶¹ From the known biology of SZ, NOS-driven negative feedback on N-methyl-D-aspartate receptor function and inhibition of synaptic reuptake of dopamine positions it at the crossroads between 2 messenger systems whose mutual regulation is a leading hypothesis for SZ.²³

This study shows that a NOS1 SNP, identified as a putative risk variant by a recent SZ genome-wide association study, is associated with variation in verbal intelligence and working memory in 2 large independent samples. Significant issues for studies addressing the cognitive effects of previously identified risk variants in SZ have included small sample size, lack of a suitable replication sample within the same study design, and a heavy multiple testing burden. One approach to the issue of multiple testing is statistical (eg, Bonferroni) correction; however, the expected effect size based on previous molecular studies of cognition and the nonindependence of measures of cognition serve to limit the value of these statistical approaches. Instead, our approach was to seek independent replication of significant results using 2 large independent case and control samples, both of which contributed to the original identification of the NOS1 rs6490121 variant studied here.⁴ Second, only a tightly selected group of cognitive functions (IQ, memory, and attention), each measuring a core cognitive deficit in SZ, were selected for analysis on the basis of their availability in both samples. Finally, this study was based on a single associated allele and results obtained were in the same expected direction in both samples, with homozygous G carriers—the associated allele in the original genome-wide association study—performing more poorly.

The utility of the cognitive intermediate phenotypes approach in SZ has been well articulated as potentially representing less complex genetic architecture than the broader disease phenotype, often shared by unaffected relatives, and thus capable of increasing power to detect genetic effects.⁶² While verbal IQ and working memory are themselves complex processes likely to be influenced by many genetic variants (eg, DTNBP1),^{63 - 65} the observed effect of the NOS1 genotype on cognition is not simply statistically significant. In both samples, the size of the association for verbal IQ in clinical terms is between 0.3 and 0.5 SD. For working memory, the differences between genotype groups represent a difference of approximately 1 SD. By comparison with findings of association between cognitive performance and other gene variants, the observed effect size is considerably higher than that reported in the recent meta-analysis of catechol O-methyltransferase and IQ⁶⁶ but comparable to reports of the association between dysbindin and general cognitive ability in patients and control subjects⁶³ and spatial working memory in patients.⁶⁴ While the overall pattern was the same across the Irish and German samples, the effect size varied between patients and healthy control subjects. Such differences may reflect differences related to sample size or ascertainment strategy (eg, the somewhat better-than-expected overall performance of German patients and Irish control subjects, respectively).

That NOS1 genotype was associated with both verbal IQ and working memory but not attentional control is noteworthy. We have previously argued that deficits in working memory, unlike attentional control, correlate with the general cognitive decline apparent in SZ.⁶⁷ Working memory has been characterized as the common factor in Spearman g, a measure of general cognitive function derived from the statistical correlation between cognitive tasks. Alternatively, working memory may simply be another name for g given the overlap between working memory and psychometric tests of g.⁶⁸ This is supported by the evidence that various measures of working memory correlate with g at a level corresponding to the reliability of the measure used.^{69 - 71} Our association with indices of both (verbal) intelligence and working memory is consistent with this evidence and suggests that for NOS1, verbal IQ and working memory may represent overlapping or at least psychometrically similar constructs. Providing empirical support for this point, covarying for the influence of either cognitive variable resulted in the association of the other with NOS1 becoming nonsignificant (eTable 3 and eTable 4). These findings are consistent with cognitive studies in both animal and human models of NOS1 where a general rather than specific effect on cognition is suggested.^{13 ,31 - 32 ,38 - 41}

The implicated SNP (rs6490121), located in intron 10 of NOS1, has no obvious functional effect and may reflect a proxy association with 1 or more other causal variants. Based on HapMap CEU data, rs6490121 is not in high linkage disequilibrium (r² > 0.80) with any other common SNP at this locus. NOS1 is characterized by complex transcriptional regulation. Twelve alternative first exons have been identified (exon 1a-1l).⁷² Exons 1d and 1f, along with the exon 1g transcript, are the most commonly expressed NOS1 transcripts in the brain.⁷³ A dinucleotide variable-number tandem repeat is located in the core promoter region of exon 1f.⁷⁴ The short alleles of this variant, termed NOS1 Ex1f variable-number tandem repeat, are associated with decreased transcriptional activity of the NOS1 exon 1f promoter, alterations in the neuronal transcriptome affecting other putative SZ susceptibility genes (RGS4 and GRIN1), and effects in electrocortical measures of prefrontal functions involved in cognitive control.⁷⁵ Interestingly, data from our laboratory indicate that the risk G allele of rs6490121 is in linkage disequilibrium (D′ = 0.70, r² = 0.26, HapMap CEU) with the short alleles of the NOS1 Ex1f variable-number tandem repeat, highlighting a putative functional mechanism of cognitive performance dysregulation at NOS1 for future study.

To explore potential mechanisms by which NO could exert an effect on cognitive processes, we screened experimentally validated protein-protein interactions of NOS1 using the protein-protein interaction databases (STRING [Search Tool for the Retrieval of Interacting Genes/Proteins],⁷⁶ DIP [Database of Interacting Proteins],⁷⁷ IntAct,⁷⁸ and BIND [Biomolecular Interaction Network Database]⁷⁹ ) and identified 19 confirmed human binary interactions. With potential relevance to SZ susceptibility are genes involved in presynaptic synaptogenesis (NOS1AP and syntrophin [SNTA1] [OMIM 601017]) and postsynaptically through the postsynaptic density 95–NOS1 receptor complex including direct NOS1 interaction with GRIN1 and GRIN2b. Elements of PSD95 signaling cascades have been targeted in SZ genetic association studies including erbB4/neuregulin signaling and the N-methyl-D-aspartate receptor complex, which is involved in long-term potentiation, memory, and learning. There is some support from these studies for involvement of GRIN2B, GRM3, erbB4, and NRG1 in SZ susceptibility. Negative studies have been reported for PSD95 (studies by Kawashima et al⁸⁰ and Tsai et al⁸¹ ) and NOS1AP (studies by Zheng et al,⁸² Puri et al,⁸³ and Fang et al⁸⁴ ). de Quervain and Papassotiropoulos⁸⁵ identified variation at GRIN2B and GRM3 as genes contributing to the memory formation signaling cascade involved in episodic memory performance. Additionally, NOS1 activators calmodulin and protein kinase C, alpha, although not directly associated with SZ risk, have also been associated with memory in humans.^{85 - 86} More systematic studies of genes involved in NOS1 regulation and adaptor proteins linked to synaptic function and plasticity are warranted, and we plan to screen functional variants at genes from these pathways for additive and epistatic effects on cognition and SZ susceptibility.

The findings reported here raise interesting questions. Is NOS1 an SZ susceptibility gene? Or, is NOS1 perhaps a modifier gene that influences cognitive deficits without altering disease liability?^{6 ,87} The present data do not provide evidence one way or another for association between NOS1 and SZ. Including the additional Irish samples in the meta-analysis of NOS1 by O’Donovan et al⁴ does not change the significance of the findings of that study (ie, NOS1 remains moderately associated with SZ in the European-only sample [P = .005, 2-tailed], but not in the total European and Asian sample combined). One interpretation of these findings is in terms of cognitive reserve—the idea that individuals with lower IQ are more vulnerable to or more likely to express the features of neurodevelopmental, neurodegenerative, or acquired brain injury.⁸⁸ In SZ, evidence of low intelligence as a risk factor is long established, with both prospective birth cohort studies^{89 - 91} and conscription studies^{92 - 96} consistently reporting lower intelligence in children and adolescents who will later develop SZ. However, differences in education (often used as a proxy measure of premorbid IQ) were observed as a trend in only the Irish samples and were absent in the German samples, thus only weakly supporting lower premorbid IQ as the mechanism of risk associated with NOS1. Instead, given the association between NOS1 and cognition in both patients and control subjects shown here, a more parsimonious explanation is that NOS1 influences cognition generally, an effect that is observed to some extent in both cases and controls. In terms of the intermediate phenotype approach in psychiatric genetics, these findings highlight the benefits of cognitive studies of putative SZ variants for understanding the genetics of fundamental cognitive and neurodevelopmental processes. It remains an open question as to the degree to which these genes are coding for disease-specific processes, either as mediated through or independent of the effects on cognition reported here.

Correspondence: Gary Donohoe, DClinPsych, PhD, Neuropsychiatric Genetics Research Group, Department of Psychiatry, Trinity Health Sciences Bldg, St James's Hospital, Dublin 8, Ireland (gary.donohoe@tcd.ie).

Submitted for Publication: September 30, 2008; final revision received January 14, 2009; accepted February 23, 2009.

Financial Disclosure: None reported.

Funding/Support: Recruitment of the patients from Munich, Germany, was partially supported by GlaxoSmithKline. Genotyping of the German samples was funded by grants from the Medical Research Council and Wellcome Trust. Recruitment and genotyping of the Irish samples were supported by the Wellcome Trust and Science Foundation Ireland. Dr Donohoe was supported by a research frontiers grant from Science Foundation Ireland and a Young Investigator Award from NARSAD. Dr Walters was supported by a clinical training fellowship from the Medical Research Council.

Additional Contributions: We sincerely thank all patients who contributed to this study and all staff who facilitated their involvement. We are grateful to the Genetics Research Centre GmbH, an initiative by GlaxoSmithKline and Ludwig-Maximilians-Universität.