Genetic overlap and causal associations between smoking behaviours and mental health

Cigarette smoking is a modifiable behaviour associated with mental health. We investigated the degree of genetic overlap between smoking behaviours and psychiatric traits and disorders, and whether genetic associations exist beyond genetic influences shared with confounding variables (cannabis and alcohol use, risk-taking and insomnia). Second, we investigated the presence of causal associations between smoking initiation and psychiatric traits and disorders. We found significant genetic correlations between smoking and psychiatric disorders and adult psychotic experiences. When genetic influences on known covariates were controlled for, genetic associations between most smoking behaviours and schizophrenia and depression endured (but not with bipolar disorder or most psychotic experiences). Mendelian randomization results supported a causal role of smoking initiation on psychiatric disorders and adolescent cognitive and negative psychotic experiences, although not consistently across all sensitivity analyses. In conclusion, smoking and psychiatric disorders share genetic influences that cannot be attributed to covariates such as risk-taking, insomnia or other substance use. As such, there may be some common genetic pathways underlying smoking and psychiatric disorders. In addition, smoking may play a causal role in vulnerability for mental illness.


Supplementary Figure S7. Generalised Summary-Based Mendelian Randomization analyses between psychiatric disorders and smoking initiation
Scatterplots with the x-axis displaying instrumental variable effects on the exposure (bzx) and the y-axis displaying the instrument-outcome association (bzy). Regression lines included for reference.
Supplementary Figure S8. Generalised Summary-Based Mendelian Randomization between positive psychotic experiences in adulthood and smoking initiation Scatterplots with the x-axis displaying instrumental variable effects on the exposure (bzx) and the y-axis displaying the instrument-outcome association (bzy). Regression lines included for reference. The ALSPAC sample: Pregnant women resident in Avon, UK and with an expected delivery date between 1 st April 1991 and 31 st December 1992 was invited to participate in the ALSPAC study. The initial sample consisted of 14,775 children. Informed consent for the use of data collected via questionnaires and clinics was obtained from participants following the recommendations of the ALSPAC Ethics and Law Committee at the time. Consent for biological samples was collected in accordance with the Human Tissue Act (2004). The number of genotyped individuals who completed items on psychotic experiences (after exclusions) was 3,951 -4,019. The ALSPAC study website contains details of all the data that is available through a fully searchable data dictionary and variable search tool (http://www.bristol.ac.uk/alspac/researchers/our-data/).

Schizotypy in adulthood
GWAS on four continuous schizotypy scales assessed during middle adulthood in the Northern Finland Birth Cohort 1996 (NFBC) 11 when participants were aged 31 years were obtained from the authors (N 3,967 -4.057). 3 Perceptual aberrations were assessed using the Perceptual Aberration Scale, 12 hypomania using the Hypomanic Personality Scale, 13 social anhedonia with the Revised Social Anhedonia Scale and physical anhedonia using the Revised Physical Anhedonia Scale. 14

Psychotic experiences in adults
The presence of lifetime positive PE were assessed in the UK Biobank using four dichotomous items as part of a mental health questionnaire completed by 157,397 participants aged 40-69 years. Participants reported an average age of PE onset of 31.6 (s.d. = 17.6) years. Summary statistics for individuals of European ancestry were obtained from the Neale Lab (http://www.nealelab.is/uk-biobank) on experiences of auditory hallucinations, visual hallucinations, delusions of persecution and delusions of reference.

Psychiatric disorders
Summary statistics were obtained from the Psychiatric Genomics Consortium (https://www.med.unc.edu/pgc/results-and-downloads) meta-GWAS for schizophrenia 4 (N= 105,318), major depressive disorder 6 (N = 173,005 excluding 23andMe participants) and bipolar disorder 5 (N = 41,653). Diagnosis of schizophrenia was based on DSM-IV criteria for schizophrenia or schizoaffective disorder. Major depression diagnoses were based on clinical interviews, obtained from electronic healthcare records or based on self-report in some UK Biobank participants. Different clinical interview formats were used to diagnose Bipolar disorder, described in full elsewhere. 5

Covariates in genomic multiple regression
Publicly available summary statistics were obtained for lifetime cannabis use (N = 162,082), 15 alcohol consumption (N = 537,349 excluding 23andMe participants), 1 risk taking (Neale Lab; N = 348,549 UK Biobank participants) and insomnia (N= 113,006). 16 Cannabis use was a binary phenotype assessed using self-report items on whether participants had ever used cannabis. Alcohol consumption came from participant reports on the average number of weekly drinks they drank. Risk taking was assessed with the item "Would you describe yourself as someone who takes risks?" (UK Biobank data-field 2040). Insomnia was from an item from the UK Biobank (data-field 1200) with participants who indicated that they usually have trouble falling asleep at night or wake up in the middle of the night classed as cases.

Mendelian randomization
Mendelian randomization (MR) is a method used to test for a causal relationship between an exposure and outcome trait by using instrumental variables as proxies for the exposure trait. In MR, genetic variants that are robustly associated with the exposure (based on GWAS results) are used as instrumental variables. The random nature of Mendelian segregation of genetic variants during meiosis means that the extent to which unmeasured confounders influence the outcome is not expected to differ between those who inherited a specific copy of a genetic variant and those who did not (analogous to randomization during randomized controlled trials).
In MR, a causal effect of an exposure (X) on an outcome (Y) is calculated as the ratio of the effect size of an instrumental variable (Z) on the outcome over its effect on the exposure: = / , where is the causal effect of the exposure on the outcome, is the effect of the instrumental variable on the outcome and is its effect on the exposure. To overcome the small effect sizes of individual genetic variants, an aggregate effect can be obtained using multiple variants as instrumental variables.