Direct additive genetics and maternal effect contribute to the risk of Tourette disorder

Abstract

Background

Risk for Tourette disorder, and chronic motor or vocal tic disorders (referenced here inclusively as CTD), arise from a combination of genetic and environmental factors. While multiple studies have demonstrated the importance of direct additive genetic variation for CTD risk, little is known about the role of cross-generational transmission of genetic risk, such as maternal effect, which is not transmitted via the inherited parental genomes. Here, we partition sources of variation on CTD risk into direct additive genetic effect (narrow-sense heritability) and maternal effect.

Methods

The study population consists of 2 522 677 individuals from the Swedish Medical Birth Register, who were born in Sweden between 1 January 1973 and 31 December 2000, and followed for a diagnosis of CTD through 31 December, 2013. We used generalised linear mixed models to partition the liability of CTD into: direct additive genetic effect, genetic maternal effect and environmental maternal effect.

Results

We identified 6227 (0.2%) individuals in the birth cohort with a CTD diagnosis. A study of half-siblings showed that maternal half-siblings had twice higher risk of developing a CTD compared with paternal ones. We estimated 60.7% direct additive genetic effect (95% credible interval, 58.5% to 62.4%), 4.8% genetic maternal effect (95% credible interval, 4.4% to 5.1%) and 0.5% environmental maternal effect (95% credible interval, 0.2% to 7%).

Conclusions

Our results demonstrate genetic maternal effect contributes to the risk of CTD. Failure to account for maternal effect results in an incomplete understanding of the genetic risk architecture of CTD, as the risk for CTD is impacted by maternal effect which is above and beyond the risk from transmitted genetic effect.

INTRODUCTION

Chronic tic disorders (CTDs) are neurological and neurodevelopmental disorders that are diagnosed when motor and/or phonic tics onset in childhood and persist for at least 1 year. Fundamentally, tics are involuntary in nature, often wax and wane, and tic severity and frequency can be influenced by both internal and external factors. A diagnosis of Tourette disorder, perhaps the most well known of the tic disorders, requires the presence of both motor tics and at least one phonic tic. If just motor (or just phonic) tics are present, the diagnosis of chronic motor (or phonic) tic disorder is given. Tourette disorder, chronic motor tic disorder and chronic phonic tic disorder can be considered phenotypic expressions with the same core genetic and biological underpinnings,^1–3 and here, they are referred to as a group as CTDs. The prevalence of CTD is about 0.3%–1%,^4–6 however, it is estimated that for children who meet criteria for CTD, only half have actually received a tic disorder diagnosis, suggesting that prevalence rates may be underestimates.⁷ CTD clusters in families at a higher rate than expected by chance,⁸ consistent with the suggestion of shared genetic architecture and shared risk factors. Indeed, multiple studies demonstrated that between 25% and 77% of the liability of CTD is due to direct additive genetics (narrow-sense heritability).^8–10 Narrow-sense heritability of a phenotype is the proportion of phenotypic variance attributable to the additive genetic effect, while broad-sense heritability is the proportion of phenotypic variance attributable to the total genetic effects (additive, dominant/recessive and epistatic).

Studies have linked environmental and genetic maternal factors before and during pregnancy, for example, maternal alcohol and cannabis use, inadequate maternal weight gain,¹¹ maternal smoking,¹² chronic maternal anxiety and prenatal maternal depression¹³ to the risk of CTD. Such factors could represent maternal effects. Maternal effect is defined as the effect of the mother’s phenotype on the phenotype of the offspring,¹⁴ and it can be partitioned into genetic maternal effect and environmental maternal effect. Genetic maternal effect occurs when maternal genetics impacts the phenotype of the mother which subsequently influences the phenotype of the offspring, independent of the genetics of the child. Environmental maternal effect occurs when the environment impacting the phenotype of the mother (independent of genotype) influences the phenotype of the offspring. These risk categories can be modelled and estimated by contrasting the phenotype risk, here CTD, among maternally linked relatives to the phenotype risk among paternally linked relatives. Failure to account for maternal effect leads to an erroneous inflation of heritability estimates^{15 16} and an imprecise model of risk architecture.

This study sought to achieve a deeper understanding of the risk factors for CTD, with a focus on maternal effect (genetic maternal effect and environmental maternal effect). Using a large population-based prospectively ascertained cohort of Swedish-born individuals and the relevant family data, we partitioned the liability of CTD into three classes: direct genetics, genetic maternal effect and environmental maternal effect.

METHOD

Study population

The study population consists of all live-born singleton children born in Sweden between 1 January 1973 and 31 December 2000, with a known father and a known mother, as defined by the national Medical Birth Register (MBR).¹⁷ Prospective follow-up continued until December 2013, and emigrated individuals identified during the follow-up were excluded from the study. Family relationships were defined by included information from all relatives of each child using the Swedish Multi-Generation Registry.¹⁸

Outcomes and exposure covariate

Diagnostic information about CTD was obtained using the Swedish National Patient Register (NPR), which includes diagnoses from inpatient and outpatient specialists.¹⁹ Sweden has a publicly financed healthcare system, and all visits to a specialist clinician are recorded with a diagnosis code using the International Classification of Diseases (ICD). All psychiatric hospital admissions in Sweden occurring after 1973 are recorded in the NPR. Beginning in 2001, outpatient specialist care is also recorded.²⁰ We used a validated diagnostic strategy developed by Rück et al²¹ to identify CTD cases based on the earliest registered CTD diagnosis code in the NPR (ICD-8: 306,2, ICD-9: 307C and ICD-10: F95). CTD includes Tourette disorder and related CTDs but does not include transient tic disorder. We used sex at birth as a fixed effect (covariate).

Statistical analysis

We used an approach similar to that applied in our previous study of the role of maternal effect on the risk of obsessive–compulsive disorder (OCD).¹⁶ We defined family relationships based on first-degree, second-degree and third-degree relatives: full siblings (FS), paternal and maternal half-siblings (pHS, mHS) and three different cousin types, depending on whether the two parents responsible for the cousin relationship are (1) sisters (maternal parallel cousins, mPC), (2) brothers (pPC: paternal parallel cousins) or (3) brother–sister (CC: cross cousins). We used all possible pairs of siblings and cousins with available family information, born between 1 January 1973 and 31 December 2000. A comparison of risk between mHS and pHS and cousins (mPC vs pPC and CC) delivers the required information to model and estimate maternal effect (online supplemental table S2).

We estimated relative recurrence risks (RRR) of CTD for different relationship types using Cox proportional hazards regression, with attained age as the primary timescale and adjusted for sex. Each individual was followed from 1997 until death, emigration from Sweden, a diagnosis with CTD or end of follow-up on 31 December 2013.

Generalised linear mixed models (GLMMs) were used to partition the liability of CTD into direct genetics, genetic maternal effect, environmental maternal effect and individual variation (residual, R). GLMM has been widely used in quantitative genetics for partitioning the total phenotypic variance of a trait into genetic and environmental components.^{16 22–26} GLMM, as a model with random effects, can handle the analysis of complex pedigrees of varying size and structure, while also accounting for measurement errors and missing values.²⁴ To obtain estimates for GLMM, we used a binary threshold-linear mixed model in a Bayesian framework with a non-informative prior to estimate the proportions of phenotypic variance explained by direct additive genetics and maternal effect (genetic maternal effect and environmental maternal effect).²⁷ We applied a Gibbs sampler implemented in thrgibbs1f90b²⁷ and generated a sample size of 150 000, with 50 000 burn-in, from the posterior distribution of the variance components. Then, we calculated the posterior mean as an estimate of the variance components. The residual variance was fixed during the calculation. We reported the results with 95% credible intervals (CrIs), using the Bayesian highest posterior density interval, which is analogous to two-sided 95% CIs in frequentist statistics.²⁸ For the fixed effect (biologically assigned sex of the child), we used the mean and SD of the posterior to calculate the CI.

In online supplemental material section S1, we used Falconer’s Liability Threshold Model (LTM),^{29 30} a significantly simpler model compared with GLMM, for sensitivity analysis and compared the results to the estimates obtained from GLMM.

RESULTS

According to the ICD instruction manual and Diagnostic and Statistical Manual of Mental Disorders (DSM) guidance, a diagnosis of transient (in DSM 5, ‘provisional’) tic disorder should be updated to a diagnosis of CTD if symptoms persist beyond 1 year. However, some individuals in our cohort had a diagnosis of transient tic disorder that lasted longer than a year. We removed all individuals (n=347) with a diagnosis of transient tic disorder at the end of the follow-up and those with a diagnosis of transient tic lasting for over a year (n=20). The cohort contained 2 522 677 individuals, of which 6227 (0.2%) were diagnosed with CTD (48% female, table 1 and online supplemental figure S1). A total of 967 of the parents in the cohort had a diagnosis of CTD.

Table 1.

Study cohort including individuals born between 1973 and 2000

	Total	Siblings			Cousins
		Full	mHS	pHS	mPC	pPC	CC
Participants	2 522 677	1 822 046	304 241	307 597	942 594	921 702	922 483
Male, count (%)	1 304 898 (52)	943 141 (52)	156 264 (51)	158 176 (51)	487 815 (52)	476 928 (52)	476 980 (52)
Female, count (%)	1 127 779 (48)	878 905 (48)	147 977 (49)	149 421 (49)	454 779 (48)	444 774 (48)	445 503 (48)
Diagnosed with CTD	6227	3957	1128	1035	2075	2083	2219
Male, count (%)	4843 (77)	3092 (78)	916 (81)	797 (77)	1612 (78)	1621 (78)	1723 (78)
Female, count (%)	1384 (23)	865 (22)	212 (19)	238 (23)	463 (22)	462 (22)	496 (22)
Population frequency	0.002	0.002	0.004	0.003	0.002	0.002	0.002
Male	0.004	0.002	0.003	0.003	0.002	0.002	0.002
Female	0.001	0.001	0.001	0.001	0.001	0.001	0.001
Incidence rate*	0.9	0.8	1.3	1.2	0.8	0.8	0.8

Individuals can be in included in more than one relationship type.

^*Per 10 000 person-years.

CC, cross-cousins; mHS, maternal half-siblings; mPC, maternal parallel cousins; pHS, paternal half-siblings; pPC, paternal parallel cousins.

We used Cox regression to compare the risk among different relationship types (table 2). The RRR between FS was 17.6 (95% CI 14.0 to 20.0). MHS had higher RRR than pHS (7.1 vs 3.6; table 2). Analysis of familial risk—the probability that an individual has an affected relative of a specific type—revealed a similar pattern (online supplemental table S1).

Table 2.

Relative recurrence risk (RRR) of CTD for different relationship types

Family types	RRR (95% CI)
Full siblings	17.6 (14.0 to 20.0)
Maternal half-siblings	7.1 (4.8 to 9.6)
Paternal half-siblings	3.6 (1.8 to 5.1)
Maternal parallel cousins	2.1 (1.3 to 2.8)
Paternal parallel cousins	2.1 (1.2 to 2.8)
Cross cousins	1.9 (1.2 to 3.0)

CTD, chronic tic disorder.

We used GLMM to partition and estimate the contribution of direct genetics and of maternal effect (genetic maternal effect and environmental maternal effect) to risk for CTD. The GLMM that best explained the data was direct genetics+genetic maternal effect+environmental maternal effect, as determined by Bayes factor analyses (table 3, and the online supplemental material section S2 and online supplemental table S5). Based on this model, 60.7% (95% CrI 58.5% to 62.4%) of the liability for CTD was due to direct genetics (narrow-sense heritability), 4.8% (95% CrI 4.4% to 5.1%) due to genetic maternal effect and 0.5% (95% CrI 0.2% to 7%) due to environmental maternal effect. We used LTM to evaluate the sensitivity of the results (online supplemental file section S1) and estimated 48.6% direct genetics and 6.9% genetic maternal effect (online supplemental table S3). In a model with only direct genetics, 62.2%–64.3% of the phenotypic variation was estimated due to direct genetics; with genetic maternal effect added, the estimation was 43.6%–60.7% (using GLMM and LTM).

Table 3.

The proportion of phenotypic variance explained by different models, and the estimated coefficient for the fixed effect

Model	DG (95% CrI)	GME (95% CrI)	EME (95% CrI)	R (95% CrI)	SEX (95% CrI)
DG	64.3% (61.0% to 67.3%)	–	–	35.7% fixed	0.3 (0.2 to 0.4)
DG+GME + EME	60.7% (58.5% to 62.4%)	4.8% (4.4% to 5.1%)	0.5% (0.2% to 0.7%)	34% fixed	0.3 (0.2 to 0.4)

CrI, credible interval; DG, direct additive genetic; EME, environmental maternal effect; GME, genetic maternal effect; R, individual variation (residual); SEX, sex of the child.

Of 6227 individuals with CTD, 1012 had comorbid OCD (16%). We removed the 1012 individuals with OCD from the cohort and estimated 46.3% direct genetics and 5.4% genetic maternal effect using LTM (6.9% genetic maternal effect before removing OCD cases vs 5.4% genetic maternal effect after removing OCD cases; online supplemental table S4).

DISCUSSION

This is the first study to identify genetic maternal effect as a risk factor for CTD. Information for estimating maternal effect comes largely from the contrast of recurrence risk through maternal versus paternal lineages, an approach made feasible through use of large multigenerational data from the Swedish data registers.

The analysis of RRR of CTD was consistent with maternal effect as a contributor to CTD risk architecture. The CTD RRR for mHS was almost double the RRR for pHS. Using GLMM, and under the liability threshold assumption, we estimated that 4.8% of the variance in risk for CTD was explained by genetic maternal effect, 0.5% by environmental maternal effect and 60.7% by direct genetics, while adjusting for the sex of the individual. Estimate of fixed effects showed that females had 0.3 times the risk relative to males (95% CrI 0.2 to 0.4) similar to accepted CTD demographic data where boys are 3–4 more likely to have CTD compared with girls.

Previously, we showed that direct additive genetics accounted for 35% (95% CrI 32.3% to 36.9%) of the total variance in risk for OCD and genetic maternal effect for 7.6% (95% CrI 6.9% to 8.3%). Since OCD and CTD share aetiological overlap,⁸ we sought to determine whether the observed maternal effect for CTD could be explained by maternal effect for CTD comorbid with OCD. After removing individuals with OCD from the cohort, we still observed a large maternal effect for CTD. Future studies using multivariate analysis are warranted to draw a more comprehensive picture of the common genetic correlation between these two disorders while accounting for shared maternal effect.

This study had several strengths and some limitations: (1) In this study, we used data from the Swedish national MBR, which created a genetically homogeneous sample and is expected to minimise the risk of confounding due to population stratification; (2) Our population study was based on clinical diagnoses of CTD by a specialist, which would also be expected to reduce case misclassification and biases; (3) environmental maternal effect can be partitioned into shared (among offspring of the same mother) and unshared environmental maternal effect. However, the nature of the data and the statistical model used in this study captures only shared environmental effects. To estimate unshared environmental maternal effect, such as maternal illness during one pregnancy, individual information to that effect would need to be available; (4) The data are likely right-censored (some individuals may have received a diagnosis after study follow-up, and therefore, these diagnoses would be missing from our study population), in particular for individuals born in the later years of the study. This would contribute to an underdiagnosis of CTD and decrease the observed prevalence.

Risk for CTD derives from a complex combination of both nature (genome) and nurture (environment). Here, we demonstrated that genetic maternal effect (also discussed as maternal genetic nurture³¹) plays an important role in the risk of CTD. Maternal effects act as environmental components affecting the risk of CTD in children and can arise from a variety of scenarios. Studies have suggested that prenatal maternal factors, for example, maternal smoking during pregnancy, maternal anxiety and maternal prenatal depression, can increase the risk of CTD in children.^{12 13} Factors such as these, and others yet to be identified, could comprise the genetic maternal effect on CTD risk identified here. However, it remains possible that the association between these prenatal factors and the risk of CTD could be due to genetic confounding (shared genetic risk) behind these factors.

The exact biological mechanism of maternal effect for CTD is currently unknown. A variety of mechanisms for maternal effect have been observed in other work. A recent review article by Edwards et al discussed five major maternal mechanisms and their influence on offspring phenotype in general: maternal androgen levels, photoperiod (melatonin), microbiome, immune regulation and milk composition.³² Another mechanism of maternal effect could be due to additional mRNAs or proteins passed from the mother to the fetus during pregnancy that remain in the offspring after birth.^{33 34}

Studies have shown that both common and rare variation, including de novo mutations, contribute to the risk of CTD.^{10 35} The number of de novo mutations found in an offspring’s genome could increase with maternal age³⁶ and potentially other maternal effects. Another source of mutation is postzygotic mosaicism. Mosaic mutations in genes can occur early in the development of an embryo. Indeed, from the very first postzygotic division, asymmetrical cellular lineage trees can be identified in humans.³⁷ This poses an additional challenge to the field as we continue to seek to gain a deeper understanding of the maternal effects underlying CTD.

In summary, while a large portion of the variation in risk for CTD can be explained by genetics, environmental factors, and in particular maternal effect, play an important role in the aetiology of this disorder. Future studies to understand the mechanism of maternal effect, the role of sex-specific offspring sensitivity to maternal effect and early predictors of CTD (eg, maternal autoimmune disorders, infections during pregnancy, maternal smoking during pregnancy and other prenatal factors) is critical because early detection allows for early intervention, which can decrease impairment and result in more positive long-term outcomes for the child.³⁸ We and others will be examining these directions in future studies.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• The aetiology of Tourette disorder encompasses a combined contribution of genetic and environmental factors.

WHAT THIS STUDY ADDS

• We demonstrated that while a large portion of the variation in risk for Tourette disorder can be explained by genetics, maternal effect (maternal genetic nurture) plays an important role in the aetiology of this disorder. This updates our conceptualisation of the risk architecture for Tourette disorder.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• These results will provide a rationale for assessing the role of specific maternal genes and loci on Tourette disorder risk.

Data availability statement

Data may be obtained from a third party and are not publicly available. Data cannot be shared publicly owing to restrictions by law. Data are available from the National Medical Registries in Sweden after approval by the Swedish Ethical Review Authority.