Uncovering Eating Disorder Genetics Through Large Scale Sequencing | WCPG 2024 Poster

Franjo Ivankovic^1-3, Kristin N. Javaras^1,4, Rocky Stroud^1,5, Faith Jennings^1,5, Kaitlin Pennels¹, Chirstine Stevens¹, Pamela Morales Cedillo⁶, Beatriz Camarena Medellin⁶, Mark J. Daly^1-3, Benjamin M. Neale^1-3

¹Stanley Center for Psychiatric Research, Broad Institute of MIT and Harvard, Cambridge, MA;
²Program in Medical and Population Genetics, Broad Institute of MIT and Harvard,
Cambridge, MA;
³Analytic and Translational Genetics Unit, Massachusetts General Hospital, Boston, MA;
⁴Division of Women’s Mental Health, McLean Hospita, Belmont, MA;
⁵Department of Epidemiology, Harvard T.H. Chan School of Public Health, Cambride, MA;
⁶Instituto Nacional De Psiquiatría Ramón de la Fuente Muñiz, Ciudad de México, México

Both anorexia (AN) and bulimia (BN) nervosa are marked by a substantial genetic component, with family study heritability estimates of 0.58-0.76 for AN and 0.30-0.83 for BN. Despite such heritability estimates, discovery of genetic risk factors for AN and BN remain limited. Genome-wide association studies are limited in power and resolution, and discoverability of genetic risk factors is hindered by study-design limitations such as potential confounding by BMI due to current AN nosology. The Genetics of Eating Disorders study at the Broad Institute is currently recruiting participants for deep-phenotyping assessments and DNA sequencing analysis to further elucidate the biopathological mechanisms underpinning eating disorders and associated comorbidities.

Our analysis is focusing on existing samples from large biobanks (FinnGen, N=4,616; All of Us, N=2,500; MGB biobank, N=1,000), representing one of the largest collective cohorts in eating disorder genome sequencing so far. We are additionally planning to sequence additional 10,000 AN/BN cases and 10,000 controls over the next 2 years, recruited in the clinics across the United States, Mexico, and Europe, as well as direct-to-participant (DTP) recruitment in the United States. To facilitate the assessment of DTP participants, we have developed the McLean-Stanford-Washington Eating Disorder (MSWED) questionnaire, a screening tool to identify lifetime diagnosis of eating disorders, which was initially validated in the Mexican clinically ascertained cohorts (N=350) and is currently undergoing validation against SCID-5 in the United States. The phenotypic analysis of prospectively recruited individuals in INP Mexico and US DTP is shown in this poster. Analyzed phenotypes include compulsive exercise (CET), disordered eating (EDE-Q), anxiety (GAD-7), depression (PHQ-7), obsessive-compulsive symptoms (OCI-R), and PTSD (PCL-5).

We have identified about 22,000 samples for sequencing, shown in TABLE 1, 977 of which have been received for sequencing and 807 of which have been sequenced so far. Extensive phenotypic data collected for Mexico (N=960) and US DTP (N=1,632). EDE-Q scores (FIGURE 1) and CET scores (FIGURE 2) show higher eating disorder related and compulsive exercise symptoms among individuals with reported eating or feeding disorder diagnoses compared to those without. Overall EDE-Q and CET scores were equivalent among individuals with a reported eating or feeding disorder diagnosis, regardless of the diagnosis. OCI-R score (FIGURE 3) did not substantially differ with respect to the eating or feeding disorder diagnosis status. Preliminary validation of MSWED resulted in high specificity (AN=0.93, BN=0.97) and lower sensitivity (AN=0.72, BN=0.51), as intended. Additional results for the entirety of phenotyping battery are available online via the QR code included at the bottom of this poster.

The Genetics of Eating Disorders study at the Broad Institute is currently undergoing active recruitment, phenotyping, and DNA sequencing of participants with eating disorders and controls.
Initial analyses of common and rare variants in cohorts where sequencing data are available are underway. We are expanding these cohorts with recruitment and genome sequencing of additional samples from ancestrally diverse clinically recruited and community-based participants.
MSWED validation suggests that it is highly specific, making it a valuable tool for identification of eating disorder cases for genetic research in community-based participants.

Additional information about this study and enrollment instructions are available at: https://www.broadinstitute.org/EDgenetics.

Summaries of DNA sample collections and sequencing, excluding already sequenced samples from large biobanks (e.g., All of Us).

Cohort	NEstimated	NCollected	NSequenced
Finish Biobank, Finland	4,616	–	–
Instituto Nacional de Psiquiatría, Mexico	6,000	422	422
Careggi University Hospital, Italy	210	60	60
Kings College London, UK	5,000	–	–
Nationwide Children’s Hospital, USA	1,000	37	37
MGB Biobank, USA	500	–	–
MGB ED Study, USA	500	–	–
Columbia University Medical Center, USA	64	64	–
University of Florida Health, USA	95	–	–
Direct-to-Participant	4,000	394	288

In linear models associating specific diagnoses with EDE-Q total scores, controlling for study arms and sex assigned at birth, we notice equivalent and significant effects across specific eating or feeding diagnoses on the EDE-Q total score. The models are broken down in the table below.

Diagnosis	Diagnosis	Intercept	Sex_Male	Source_DTP
Anorexia Nervosa	21.28 (<0.001)	17.49 (<0.001)	-2.81 (<0.1)	-11.23 (<0.001)
OSFED Anorexia Nervosa	17.97 (<0.001)	15.97 (<0.001)	-3.06 (<0.05)	-7.87 (<0.001)
Bulimia Nervosa	24.54 (<0.001)	18.80 (<0.001)	-2.73 (<0.01)	-13.96 (<0.001)
OSFED Bulimia Nervosa	18.48 (<0.001)	15.85 (<0.001)	-2.65 (<0.05)	-7.75 (<0.001)
Binge-Eating Disorder	18.30 (<0.001)	16.46 (<0.001)	-2.79 (<0.05)	-8.97 (<0.001)
OSFED Binge-Eating Disorder	15.97 (<0.001)	15.61 (<0.001)	-3.32 (<0.05)	-7.05 (<0.001)
OSFED Purging Disorder	19.37 (<0.001)	16.00 (<0.001)	-3.22 (<0.05)	-7.90 (<0.001)
OSFED Night Eating Syndrome	20.20 (<0.001)	15.60 (<0.001)	-3.55 (<0.05)	-6.99 (<0.001)
USFED	18.76 (<0.001)	16.47 (<0.001)	-3.58 (<0.05)	-8.77 (<0.001)
ARFID	17.27 (<0.001)	15.75 (<0.001)	-3.57 (<0.05)	-7.27 (<0.001)
Pica	14.90 (<0.001)	15.61 (<0.001)	-3.32 (<0.05)	-7.07 (<0.001)
Rumination Disorder	17.10 (<0.001)	15.56 (<0.001)	-3.26 (<0.05)	-6.97 (<0.001)

Unlike the EDE-Q total score, EDE-Q restraint score was only significantly associated with the intercept and specific eating or feeding disorder diagnosis for all diagnoses considered. Sex assigned at birth and study arms were not significantly associated with the EDE-Q restraint score in our multiple regression model.

Similarly to the EDE-Q total score, EDE-Q eating concern score was significantly associated with the intercept, sex assigned at birth, study arm, and specific eating or feeding disorder diagnosis for all diagnoses considered.

EDE-Q weight concern score was strongly associated with the intercept, diagnosis, and sex assigned at birth consistently across all specific eating or feeding disorder diagnoses considered.

EDE-Q shape concern score was strongly associated with the intercept, diagnosis, and sex assigned at birth consistently across all specific eating or feeding disorder diagnoses considered. In case of bulimia nervosa model, study arm was also significantly associated with EDE-Q shape concern score.

In our multiple linear regression models, the CET total score was significantly associated with the intercept term and specific diagnoses. Additionally, the CET total score was also strongly associated with the study arm term, indicating DTP US individuals reporting an average of 12-point higher CET total score than the Mexican counterparts.

OCI-R current total score was only significantly associated with the intercept term. No significant associations between OCI-R current total score and specific eating or feeding disorder diagnosis, sex assigned at birth, or study arm were observed.