Participants

A sample of unrelated individuals with schizophrenia was recruited from out-patient and in-patient clinical settings and from volunteer support organisations within the UK. These individuals were assessed using the Schedule for Assessment of Neuropsychiatric Disorders semi-structured interview (SCAN; Reference Wing, Babor and BrughaWing et al, 1990) together with case-note review wherever possible. The Operational Criteria Checklist (OPCRIT; Reference McGuffin, Farmer and HarveyMcGuffin et al, 1991) and Global Assessment Scale (GAS; Reference Endicott, Spitzer and FleissEndicott et al, 1976) were also completed. High levels of reliability (κ>0.8) were achieved between raters for diagnoses and rating scale items. Controls were unrelated blood donors ascertained from the same regions as the majority of the patients. Given the prevalence of schizophrenia and the fact that people taking regular medication cannot be blood donors in the UK, it was not deemed necessary to screen the control group for schizophrenia to retain statistical power (Reference Owen, Holmans and McGuffinOwen et al, 1997). Ethical approval was granted for this study and informed consent was obtained from all participants.

All study participants were White, with both parents born in the UK or Ireland. All cases of schizophrenia satisfied DSM–IV criteria (American Psychiatric Association, 1994) for consensus lifetime diagnosis of the disorder, made by two independent raters. The following phenotypes, determined a priori, were examined in relation to –86C/T and rs1049353 genotype:

1. (a) age at onset, defined as the age at which psychiatric help for psychotic symptoms was first sought;

2. (b) worst-ever GAS score, ranging from 0 (most severe) to 100 (least severe);

3. (c) poor outcome, defined as either non-continuous or continuous course of illness;

4. (d) standardised, continuous symptom dimensions obtained by factor analysis of OPCRIT psychosis items, corresponding to paranoid delusions and hallucinations; disorganised symptoms; negative symptoms; and first-rank delusions.

Data on tobacco and cannabis use were obtained from interview and case-note records for 657 of the schizophrenia group. Questions regarding the age at which the person first started using cannabis or tobacco were only introduced during the latter part of the sample recruitment, and these data were therefore available for only 22% of cases in which the person reported ever using these substances. Substance use data were not collected initially for the control group, and unfortunately most members of the control group were not asked at the initial interview for permission to contact them again for further information. As a result of this, cannabis use data were available for only 116 controls and tobacco use data for 49 controls.

Genotyping

The CHRNA7 promoter polymorphism –86C/T was genotyped as a restriction-fragment length polymorphism using the restriction enzyme Hph1 (New England Biolabs, Ipswich, Massachusetts, USA). The primers were 5′-agtacctcccgctcacacctcg-3′ and 5′-atgttgagtcccggagctg-3′ as used by Leonard et al (Reference Leonard, Gault and Hopkins2002). The product was amplified using the GC-RICH PCR System (Roche Diagnostics, Basel, Switzerland), and the 272 base pairs fragment was digested with Hph1 resulting in two fragments of 79 bp and 193 bp with the T allele. The products were run out on a 1.5% agarose gel and visualised using ethidium bromide.

The CNR1 polymorphism rs1049353 was genotyped by fluorescence polarisation using an AcycloPrime kit (PerkinElmer, Waltham, Massachusetts, USA) and the output was read on an LJL Biosystems (Sunnyvale, California, USA) plate reader. A 297 bp amplimere was amplified using primers 5′-ttccctcttgtgaaggcact-3′ and 5′tcattgagcatggtaaagtt-3′. The SNP was at position 125. The extension primer used in the fluorescence polarisation assay was 5′-catggttaccttggcaatcttgac-3′. The COMT markers were genotyped using SNaPshot (Applied Biosystems, Foster City, California, USA) using an ABI3100 sequencer. Details of primers and reaction conditions are provided in Appendix 1 at http://www.cardiff.ac.uk/medicine/psychological_medicine/pub_data/comt.htm.

Analysis

The reference participants for the analyses were those with genotypes that were CC homozygous for –86C/T, GG homozygous for rs1049353, AA homozygous for rs737865, AA homozygous for rs165599, and homozygous for the Met allele at Val¹⁵⁸Met within COMT. Only 0.4% of our participants were homozygous for the T allele at the –86C/T locus, and they were therefore grouped with the C/T heterozygotes.

Logistic regression was used to examine associations between dichotomous outcomes and genotypes. A dominance genetic model, as described above, was examined for –86C/T, whereas additive models were used for the CNR1 and COMT variants (Reference LewisLewis, 2002). For the study of continuous phenotypic outcomes, linear regression was used. However, for age at onset, where assumptions of normality were not met, data were ln-transformed prior to regression modelling. Statistical interactions on a multiplicative scale between substance use and genotype on risk of schizophrenia were investigated using a likelihood ratio test within the logistic regression models. For Val¹⁵⁸Met, however, as no association was observed between this SNP and cannabis use in the Dunedin cohort (Reference Caspi, Moffitt and CannonCaspi et al, 2005), we used a case-only approach to investigate possible gene–environment interactions because this is statistically more powerful (Reference Khoury and FlandersKhoury & Flanders, 1996). The case-only analysis was also used for rs737865 and rs165599 within COMT. Haplotypes for COMT were examined using UNPHASED, version 3.0 (Reference DudbridgeDudbridge, 2003).

This study had greater than 95% power to detect an additive genetic effect with an odds ratio of 1.4 or above at α=0.05 for the CNR1 and COMT variants examined. This study also had greater than 95% power to find an association between –86C/T variation and schizophrenia based on frequencies of CC genotype of 0.91 in the control group and 0.84 in the schizophrenia group, as observed by Leonard et al (Reference Leonard, Gault and Hopkins2002). The interaction odds ratio previously reported for cannabis and Val¹⁵⁸Met was 3.5 (Reference Caspi, Moffitt and CannonCaspi et al, 2005), and our case-only approach had more than 90% power to detect an interaction odds ratio of as low as 1.5, at α=0.05.

Sensitivity analysis

Some participants were likely to have started using tobacco or cannabis after the onset of schizophrenia and it is possible that this could obscure and complicate interpretation of results from this study. Examination of the association between schizophrenia and genotypes was therefore repeated with analyses restricted to cases where the onset of substance use was reported to be at least 1 year prior to age at schizophrenia onset.

CHRNA7

The –86C/T genotypes were in Hardy–Weinberg equilibrium in both the schizophrenia group (χ²=0.01, P=0.76) and the control group (χ²=0.01, P=0.92). As shown in Table 1, there was no evidence for any association between –86C/T geno-type and schizophrenia (CT/TT genotypes OR= 1.07, 95% CI 0.77–1.49; P=0.70). There was little evidence of any difference in the effect of genotype on schizophrenia between those who smoked (schizophrenia group n=473, controls n=24; OR=3.0, 95% CI 0.4–22.9) and those who did not (schizophrenia group n=186, controls n=25; OR=1.7, 95% CI 0.4–7.7; interaction likelihood ratio test χ²=0.21, d.f.=1, P=0.65). As tobacco use data were available only for a small proportion of the control group, a more powerful case-only analysis was also used, and this also failed to provide any evidence for interaction (n=659; odds ratio for tobacco use by CHRNA7 genotype 0.89, 95% CI 0.53–1.48).

Table 1 Association between CHRNA7 (–86C/T) and CNR1 (rs1049353) genotypes and schizophrenia

	Number (%) by genotype			Total	OR (95% CI)
-86C/T	CC	CT	TT
Control group	548 (88.7)	68 (11.0)	2 (0.3)	618	1.0
Schizophrenia group	660 (88.0)	87 (11.6)	3 (0.4)	750	1.07 (0.77-1.49)1
rs1049353	GG	GA	AA
Control group	335 (48.7)	298 (43.3)	55 (8.0)	688	1.0
Schizophrenia group	407 (51.1)	319 (40.0)	71 (8.9)	797	0.97 (0.82-1.13)2

There were 123 in the schizophrenia group with data relating to age of first using tobacco, and 104 (85%) of these claimed to have started using tobacco prior to the onset of schizophrenia. In the sensitivity analysis there was similarly little evidence of any difference in the effect of genotype on schizophrenia between non-smokers and those smoking prior to illness onset (n=110; OR=2.7, 95% CI 0.3–22.3; interaction χ²=0.1, d.f.=1, P=0.73).

Another way of presenting these data is to examine the relationship between tobacco use and schizophrenia stratified by –86C/T genotype. Tobacco use was strongly associated with schizophrenia in the whole sample (OR=4.4. 95% CI 3.3–6.0; P<0.001), with no evidence of any interaction when stratified by genotype (CC genotype OR=2.6, 95% CI 1.4–4.7; CT/TT genotypes OR=4.6, 95% CI 0.4–53.0; interaction likelihood ratio test as above, P=0.65). Tobacco use was not associated with –86C/T genotype (OR=0.9, 95% CI 0.5–1.5).

Results for associations between –86C/T genotype and various phenotypes within schizophrenia are presented in Table 2. There was weak evidence (P=0.07) that participants with the CT/TT genotypes had a younger age of onset, by approximately 2 years on average, than those homozygous for the C allele.

Table 2 Effect estimates for phenotype characteristics according to CHRNA7 (–86C/T) and CNR1 (rs1049353) genotypes in participants with schizophrenia

	-86C/T1		rs10493532
	β (95% CI)	P	β (95% CI)	P
Age at onset3	-0.06 (-0.13 to 0.01)	0.07	0.01 (-0.02 to 0.05)	0.43
GAS score	-0.9 (-2.7 to 0.9)	0.34	0.02 (-0.86 to 0.90)	0.97
Paranoid delusions	-0.05 (-0.3 to 0.2)	0.70	-0.09 (-0.21 to 0.03)	0.14
Disorganised symptoms	0.11 (-0.1 to 0.4)	0.40	0.07 (-0.05 to 0.19)	0.26
Negative symptoms	0.01 (-0.2 to 0.3)	0.96	0.03 (-0.09 to 0.15)	0.61
First-rank delusions	-0.01 (-0.3 to 0.2)	0.91	-0.03 (-0.15 to 0.09)	0.62
Poor outcome	OR=1.30 (0.8 to 2.1)	0.29	OR=1.03 (0.81 to 1.29)	0.83

CNR1

Genotypes at rs1049353 were in Hardy–Weinberg equilibrium in both the schizophrenia group (χ²= 0.56, P=0.44) and controls (χ²=1.0, P=0.36). As shown in Table 1, there was no evidence for any association between rs1049353 genotype and schizophrenia (odds ratio for linear trend of genotypes 0.97, 95% CI 0.82–1.13; P=0.66).

There was little evidence of any difference in the effect of rs1049353 genotype on schizophrenia between those who did not use cannabis (schizophrenia group n=445, controls n=93; OR=1.04, 95% CI 0.73–1.47) and those who did (schizophrenia group n=261, controls n=23; OR=0.92, 95% CI 0.48–1.75; interaction χ²=0.11, d.f.=1, P=0.74). As cannabis use data were again available for only a small proportion of the control group, a case-only analysis was used, and this also failed to provide any evidence for interaction (n=706; odds ratio for cannabis use by CNR1 genotype 0.83, 95% CI 0.65–1.05).

As part of the sensitivity analysis, there were 71 individuals in the schizophrenia group with data relating to age of first using cannabis, and 64 (90%) of these reported first use prior to onset of schizophrenia. As in the main analysis, there was little evidence of any difference in the effect of rs1049353 genotype on schizophrenia between those who did not use cannabis and those who claimed to have used cannabis at least 1 year prior to illness onset (n=614; OR=0.84, 95% CI 0.40–1.78; interaction χ²=0.26, d.f.=1, P=0.61).

Presenting these data in another way, there was a strong association between cannabis use and schizophrenia in this sample (OR=2.6, 95% CI 1.8–3.7; P<0.001), with no evidence of any difference when stratified by rs1049353 genotype (GG genotype OR=2.3, 95% CI 1.2–4.4; GA genotype OR=3.1, 95% CI 1.3–7.2; AA genotype OR=1.1, 95% CI 0.2–4.6; interaction likelihood ratio test as above, P=0.33). There was no evidence for any association between rs1049353 genotype and various phenotypes within schizophrenia (Table 2).

COMT

There was no evidence for any association between Val¹⁵⁸Met genotype and cannabis use in our sample of 493 persons with schizophrenia (OR=0.98, 95% CI 0.76–1.27, P=0.89). Results were almost identical when restricting the analysis to participants who first used cannabis at least 1 year prior to their illness onset and who had first used it by age 18 years or earlier (n=338; OR=0.76, 95% CI 0.41–1.40; P=0.38). Similarly, there was no evidence that variation at rs737865 or rs165599 was associated with cannabis use in the case-only analysis, even when restricting the analysis to first use of cannabis at least 1 year prior to illness onset and first use by age 18 years or earlier (rs737865, OR=1.09, 95% CI 0.56–2.00; rs165599, OR=1.09, 95% CI 0.57–2.08). There was no evidence of overall haplotype association with cannabis use (χ²=4.7, d.f.=7, P=0.69) or of specific association with the rs737865–4680–165599 GGG haplotype (χ²=0.001, d.f.=1, P=0.98).

CHRNA7 and tobacco use

There have been few association studies of polymorphisms within CHRNA7 and schizophrenia to date. Leonard et al (Reference Leonard, Gault and Hopkins2002) screened the core promoter region of the full-length gene and reported an association between schizophrenia and variant –86C/T. Although we found no evidence for an association between the promoter SNP –86C/T and schizophrenia, CT/TT genotypes occurred slightly more frequently in participants with schizophrenia than in controls, in a direction consistent with the findings by Leonard et al (Reference Leonard, Gault and Hopkins2002). However, we observed a much smaller difference in CC frequency of less than 1%, as opposed to the 7% reported in the original study (Reference Leonard, Gault and HopkinsLeonard et al, 2002).

People with schizophrenia commonly display evidence of sensory attention impairments (Reference Adler, Pachtman and FranksAdler et al, 1982; Reference Leonard, Adams and BreeseLeonard et al, 1996), including deficits in pre-pulse inhibition and P50 gating response (Reference Braff and SaccuzzoBraff & Saccuzzo, 1985; Reference Braff, Grillon and GeyerBraff et al, 1992; Reference Waldo, Cawthra and AdlerWaldo et al, 1994). Improvements in such neurophysiological deficits in people with schizophrenia following cigarette smoking have been reported (Reference Adler, Hoffer and WiserAdler et al, 1993; Reference Olincy, Ross and YoungOlincy et al, 1998), with similar improvements observed following nicotine administration in animal models (Reference Bickford and WearBickford & Wear, 1995; Stevens et al, Reference Stevens, Freedman and Collins1996, Reference Stevens, Kem and Mahnir1998). Specific agonists of the α₇ receptor (CHRNA7) normalise sensory gating deficits in animal models (Reference Stevens, Kem and MahnirStevens et al, 1998), whereas evidence for genetic linkage to the P50 deficit and, to a lesser extent, to schizophrenia, has been reported for chromosome band 15q14, an area that contains CHRNA7 (Reference Coon, Plaetke and HolikCoon et al, 1993; Reference Freedman, Coon and Myles-WorsleyFreedman et al, 1997; Reference Leonard, Gault and MooreLeonard et al, 1998).

Despite this support, from a variety of sources, that CHRNA7 is a good candidate gene for schizophrenia, there is weak evidence at present that variation within this gene is associated with the disorder (Reference Riley, Makoff and Mogudi-CarterRiley et al, 2000; Reference Xu, Pato and TorreXu et al, 2001; Reference Leonard, Gault and HopkinsLeonard et al, 2002; Reference Gault, Hopkins and BergerGault et al, 2003; Reference Li, Liao and ChenLi et al, 2004; Reference Fan, Ma and LiFan et al, 2006). However, given the findings from experimental studies of the effect of nicotine on neurophysiological deficits in both animal models and humans, as described earlier, it may be that any association between CHRNA7 and schizophrenia is mediated by impairments in sensory gating or other related physiological responses. In the study by Leonard et al (Reference Leonard, Gault and Hopkins2002), presence of the T allele at –86C/T was also associated with reduced inhibition of the P50 response in the control group, and although two other studies did not replicate this finding (Reference Gault, Hopkins and BergerGault et al, 2003; Reference Houy, Raux and ThibautHouy et al, 2004), one reported an association between P50 sensory gating response and another promoter variant, –194G/C (Reference Houy, Raux and ThibautHouy et al, 2004). There is a clear need for research into CHRNA7 variation in relation to neurophysiological deficits in well-designed and adequately powered studies to address this further.

CNR1, COMT and cannabis

We found no evidence of association between the CNR1 locus rs1049353 and schizophrenia, consistent with the overall findings previously reported for this variant from two much smaller studies (Reference Leroy, Griffon and BourdelLeroy et al, 2001; Reference Ujike, Takaki and NakataUjike et al, 2002), although one of these reported an association in a subgroup analysis (Reference Leroy, Griffon and BourdelLeroy et al, 2001). Two studies have reported associations between schizophrenia and variation within an (AAT) _n microsatellite approximately 20 kb upstream of the translational start site of CNR1 (Reference Ujike, Takaki and NakataUjike et al, 2002; Reference Martinez-Gras, Hoenicka and PonceMartinez-Gras et al, 2006). However, different alleles were associated with increased risk in these two studies, and the association in one of the studies was again observed only for a subgroup of participants, this time with hebephrenic schizophrenia.

The CNR1 gene is located on 6q14–15, a region of replicated linkage for schizophrenia (Reference Lewis, Levinson and WiseLewis et al, 2003). There are four SNPs within CNR1 on HapMap that have a heterozygosity in European populations greater than 0.1; three of these are in the 3′ untranslated region whereas rs1049353 is a synonymous SNP found within exon 1. The relatively small size of CNR1, the limited variation within the gene and its linkage disequilibrium structure mean it is unlikely that a substantial effect on schizophrenia risk is conferred by variation within this gene, given our findings and the lack of other consistent associations reported to date.

We also failed to find any supporting evidence for a differential effect of cannabis use on psychosis risk according to variation at Val¹⁵⁸Met within COMT. In the Dunedin study evidence for an interaction was observed only for people first using cannabis by age 18 years, but not for those using it after this age (Reference Caspi, Moffitt and CannonCaspi et al, 2005). One explanation proposed for this was that there may exist a sensitive or even critical period of risk when the influence of cannabis exposure is moderated by COMT genotype. In our study we failed to find evidence for an interaction between cannabis use and COMT genotype even when restricting the analysis to participants who claimed to have first used cannabis by the same cutoff period of age 18 years, despite more than adequate statistical power to replicate the original findings. Furthermore, in contrast to the findings by Caspi et al (Reference Caspi, Moffitt and Cannon2005), cannabis use by age 18 years was actually less common in participants with schizophrenia homozygous for the Val allele compared with those heterozygous for this allele or homozygous for Met (5.3%, 6.4% and 8.7% respectively), although this was not significantly different.