Association of Essential Tremor With Novel Risk Loci

Key Points Question Can common genetic variants associated with essential tremor (ET) be identified? Findings In this genome-wide association study and meta-analysis including genetic data on 483 054 individuals, 5 genome-wide significant loci were associated with risk of ET and common variants were associated with approximately 18% of ET heritability. Meaning Findings of this study may help identify new genes and inform ET biology.

under a genotype-phenotype protocol. This protocol was approved by the NIH Combined Neurosciences IRB. Written consent was obtained for all participant prior to recruitment.

Mayo Clinic Florida
Patients and control subjects were recruited under the Mayo Clinic IRB Committee approved University. The patients were diagnosed with essential tremor (ET) by movement disorders specialists according to the Movement Disorders Society ET diagnostic criteria. Both sporadic and familial cases were collected. The patients with other than ET neuroglial symptoms and signs were excluded from this study. Also the patients in which ET was part of a broader phenotype and the patients who first presented with ET and later develop additional clinical/neurological features such as rest tremor, rigidity, bradykinesia, dystonia, tremor, chorea, and others were excluded from this study. The controls were mainly opportunistic (not specifically sought for this study), and mainly included spouses, other family members, and caregivers. A subset of controls was recruited through electrodiagnostic laboratory and included the patients who were electro-diagnostically studied for the median neuropathy at the wrist, ulnar neuropathy at the elbow, peroneal neuropathy at the fibular head, or minor complains consistent with peripheral neuropathy.

University of British Columbia (UBC) / University of Saskatchewan
Patients were recruited from the Movement Disorder Clinic Saskatchewan (MDCS). Diagnosis was made based on postural and/or kinetic tremor of upper limbs or head tremor with no other known neurological cause. In certain cases, brain pathological studies were performed following autopsy to identify known tremor causes.

University of Navarra (Spain)
The study was approved by the University of Navarra's Ethical Committee and all informed consent was obtained to patients prior to their inclusion in the study. Patients were diagnosed by a movement disorder neurologist with either 'definite' or 'probable' ET based on current criteria used by the Department of Neurology from the Clinica Universidad de Navarra, Pamplona, Spain.
Only one affected subject from each family with the familial form of the disease were included in the study. Controls were unrelated healthy subjects or spouses from PD cases, recruited within the same university center or from the community. Exclusion criteria were patients with other neurological disorders or a positive family of neurodegenerative disease.

Hopsital Universitario del Sureste (Arganda del Rey) and University of Extremadura
Participants were recruited from the University Hospitals of Sureste (Arganda del Rey, Madrid) and Principe de Asturias (Alcalá de Henares, Madrid). Diagnosis of ET patients were based on 'definite' ET criteria from MDS. Only patients without other neurological disorders and without thyroid dysfunction were included. All ET patients had at least 1 first-degree relative with ET and only 1 family member per family was included. Control particpiants were recruited from the Infanta Cristina University Hospital (Badajoz) and from the Clinica Universitaria de Navarra (Pamplona). These were healthy, gender-matched, unrelated participants without tremor. Informed consent was obtained before inclusion in the study. Ethical approval was given by the Ethics Committees of the Infanta Cristina University Hospital and the Principe de Asturias University Hospital.

University of Genoa (Genova, Italy)
ET patients with 'definite' or 'probable' ET based on criteria from the Tremor Investigation Group 2,3 were recruited at the collaborating Neurological Centres (Naples and Bologna) by specialized neurologists. Assessment of patients involved neurological examination, identification of tremor type and its topography as well as recording of associated signs or symptoms, drugs in use, videotaping and other paraclinical investigations. Self-reported age of onset was recorded.
Samples from patients and relatives (when possible) were obtained after written informed consent was given. Age-and gender-matched controls were recruited among the spouses of Italian ET patients or other subjects without neurological dysfunctions (as assessed by the same neurologists and protocols previously mentioned).

CARTaGENE
The CARTaGENE cohort was used as additional controls for the study. CARTaGENE is population-based biobank of individuals from Quebec, Canada 4 . It is a long-term cohort of over 20,000 participants that consent to visiting assessment sites to provide health and sociodemographic information. Individuals without neurological or psychiatric disorders were used as controls. Further details about CARTaGENE are described in Awadalla et al (2013) 4 .

Quality control and association analyses for clinical cohorts
The samples from both clinical cohorts 1 and 2 that were genotyped on the Axiom CEU Array and Illumina GSA array respectively. and followed standardized quality control (QC), imputation, and post-imputation QC. Briefly, samples were removed if >2% missingness, autosomal heterozygous deviation (Fhet <0.2), and failed sex check 5 . Low quality SNPs were removed based on Hardy-Weinberg Equilibrium (P>1E-6), SNP missingness < 0.02 after sample removal. Samples were mapped against the 1000 Genomes Project phase 3 reference panel after pruning and removing SNPs from high-LD regions, and only individuals of inferred-European ancestry were retained 6 .
No relatedness filter was done because a linear-mixed model was used subsequently to account for relatedness. Imputation was done using the Sanger Imputation Server with Eagle v2.3.5 and the Haplotype Reference Consortium Reference Panel v1.1 7,8 . A Bayesian linear-mixed model was done using BOLT-LMM 2.3.4 including 20 principal components (PCs) and sex as covariates to accelerate convergence 9 . The non-infinitesimal model was used if there was an increase in power.

Description of population-level cohorts
Two separate biobanks, 23andMe and the UK BioBank were used for the study.

23andMe
The 23andMe samples consisted of individuals who sent saliva samples to 23andMe Inc. and agreed to partake in research and answered questions related to ET. Participants involved provided informed consent and answered questions in accordance to their human subjects' protocol, which was approved and reviewed by Ethical and Independent Review services, an AAHRPP-accredited institutional review board. 23andMe provided summary statistics of their ET GWAS that included cases that responded "yes" to the question "Have you ever been diagnosed with a neurological condition?" and later indicated that they had been diagnosed with ET or indicated that they had been diagnosed with Parkinson's disease but were later re-diagnosed with ET. Cases were excluding individuals genotyped on the 23andMe v1 genotyping chip and individuals younger than 30 years old.
Controls were recruited from all eligible participants. The age, sex, genotyping platform and survey response time were matched against the cases that met the same criteria but excluded cases that were genotyped on the 23andMe v1 and subjects under 30 years old. Controls were iteratively Participants were restricted to a set of individuals with European ancestry through an analysis of local ancestry. Briefly, the algorithm partitions phased genomic data into intervals of 300 SNPs.
Iteratively for each window, a support vector machine (SVM) was used to characterize each haplotype into reference populations (1000 Genomes, HapMap, Human Genome Diversity Project, 23andMe customers reporting four grandparents from same country). To account for relatedness, a segmental identity-by-descent (IBD) estimation algorithm was used. Individuals that shared roughly 20% of the genome were considered related. When selecting for case/control, cases were preferentially retained. Imputation was done with an imputation panel combining the 1000 Genomes Phase 3 haplotypes with the UK10K reference panel 10 . Participants from the v5 platform and prior, the tool Finch was used to phase participant data. For samples from the v5 array, Eagle2 was used.
Association tests for genotyped data was done using a logistic regression assuming additive effects. For imputed data, the dosage was used over hard-call genotypes. Quality control was done independently for dosage and imputed data. The directly genotyped SNPs were flagged for failed QC if they were only genotyped on v1 or v2 platforms due to smaller sample size. SNPs with a Hardy-Weinberg (P<10E-20) or a call rate of <90% were flagged. Any SNPs that were flagged by ANOVA of genotypes against a factor dividing genotyping date into 20 roughly equal bins were