Available via license: CC BY 4.0
Content may be subject to copyright.
Psychological Medicine
cambridge.org/psm
Original Article
*Full list of PGC-ED working group authors is
available in the online Supplementary
materials.
Cite this article: Johnson JS et al (2022).
Mapping anorexia nervosa genes to clinical
phenotypes. Psychological Medicine 1–15.
https://doi.org/10.1017/S0033291721004554
Received: 21 April 2021
Revised: 23 September 2021
Accepted: 20 October 2021
Key words:
Anorexia nervosa; EHR; pheWAS; PrediXcan;
transcriptomic imputation
Author for correspondence:
Laura M. Huckins,
E-mail: laura.huckins@mssm.edu
© The Author(s), 2022. Published by
Cambridge University Press. This is an Open
Access article, distributed under the terms of
the Creative Commons Attribution licence
(http://creativecommons.org/licenses/by/4.0/),
which permits unrestricted re-use, distribution
and reproduction, provided the original article
is properly cited.
Mapping anorexia nervosa genes to
clinical phenotypes
Jessica S. Johnson1,2 , Alanna C. Cote1,2 , Amanda Dobbyn1,2,3 ,
Laura G. Sloofman4, Jiayi Xu1,2 , Liam Cotter1,2 ,
Alexander W. Charney1,2,3, 5,6 , Eating Disorders Working Group of the
Psychiatric Genomics Consortium1,*, Andreas Birgegård7, Jennifer Jordan8,
Martin Kennedy8, Mikaél Landén7,9 , Sarah L. Maguire10 , Nicholas
G. Martin11 , Preben Bo Mortensen12,13 , Laura M. Thornton14 ,
Cynthia M. Bulik7,14,15 and Laura M. Huckins1,2,3,4, 5,6
1
Pamela Sklar Division of Psychiatric Genomics, Icahn School of Medicine at Mount Sinai, New York, NY 10029,
USA;
2
Department of Genetics and Genomics, Icahn School of Medicine at Mount Sinai, New York, NY 10029, USA;
3
Icahn Institute for Genomics and Multiscale Biology, Icahn School of Medicine at Mount Sinai, New York, NY
10029, USA;
4
Seaver Autism Center for Research and Treatment, Icahn School of Medicine at Mount Sinai,
New York, NY 10029, USA;
5
Department of Psychiatry, Icahn School of Medicine at Mount Sinai, New York, NY
10029, USA;
6
James J. Peters Department of Veterans Affairs Medical Center, Mental Illness Research, Education
and Clinical Centers, Bronx, NY 14068, USA;
7
Department of Medical Epidemiology and Biostatistics, Karolinska
Institutet, Stockholm, Sweden;
8
Department of Psychological Medicine, Christchurch School of Medicine & Health
Sciences, University of Otago, 2 Riccarton Avenue, PO Box 4345, 8140 Christchurch, New Zealand;
9
Institute of
Neuroscience and Physiology, Sahlgrenska Academy at Gothenburg University, SE-413 45 Gothenburg, Sweden;
10
InsideOut Institute, University of Sydney, New South Wales 2006, Australia;
11
QIMR Berghofer Medical Research
Institute, Locked Bag 2000, Royal Brisbane Hospital, Herston, QLD 4029, Australia;
12
The Lundbeck Foundation
Initiative for Integrative Psychiatric Research, Aarhus, Denmark;
13
National Centre for Register-Based Research,
Aarhus University, Aarhus, Denmark;
14
Department of Psychiatry, University of North Carolina at Chapel Hill,
Chapel Hill, NC 27517, USA and
15
Department of Nutrition, University of North Carolina at Chapel Hill, Chapel Hill,
NC 27517, USA
Abstract
Background. Anorexia nervosa (AN) is a psychiatric disorder with complex etiology, with a
significant portion of disease risk imparted by genetics. Traditional genome-wide association
studies (GWAS) produce principal evidence for the association of genetic variants with dis-
ease. Transcriptomic imputation (TI) allows for the translation of those variants into regula-
tory mechanisms, which can then be used to assess the functional outcome of genetically
regulated gene expression (GReX) in a broader setting through the use of phenome-wide asso-
ciation studies (pheWASs) in large and diverse clinical biobank populations with electronic
health record phenotypes.
Methods. Here, we applied TI using S-PrediXcan to translate the most recent PGC-ED AN
GWAS findings into AN-GReX. For significant genes, we imputed AN-GReX in the Mount
Sinai BioMe™Biobank and performed pheWASs on over 2000 outcomes to test the clinical
consequences of aberrant expression of these genes. We performed a secondary analysis to
assess the impact of body mass index (BMI) and sex on AN-GReX clinical associations.
Results. Our S-PrediXcan analysis identified 53 genes associated with AN, including what is,
to our knowledge, the first-genetic association of AN with the major histocompatibility com-
plex. AN-GReX was associated with autoimmune, metabolic, and gastrointestinal diagnoses in
our biobank cohort, as well as measures of cholesterol, medications, substance use, and pain.
Additionally, our analyses showed moderation of AN-GReX associations with measures of
cholesterol and substance use by BMI, and moderation of AN-GReX associations with celiac
disease by sex.
Conclusions. Our BMI-stratified results provide potential avenues of functional mechanism
for AN-genes to investigate further.
Introduction
Anorexia nervosa (AN) is a severe eating disorder with a lifetime prevalence of 0.9–4%, char-
acterized by extreme low body weight, fear of gaining weight, and compensatory weight-loss
behaviors such as dietary restrictions, purging, and excessive exercise (Zipfel, Giel, Bulik,
Hay, & Schmidt, 2015). Despite having one of the highest mortality rate of any psychiatric dis-
order and an increased risk of suicide (Bulik, Flatt, Abbaspour, & Carroll, 2019; Chesney,
Goodwin, & Fazel, 2014; Mitchell & Peterson, 2020), relatively little is known about the
https://doi.org/10.1017/S0033291721004554 Published online by Cambridge University Press
biological mechanisms underlying AN, and effective, evidence-
based treatments are scant, especially for adults (Watson &
Bulik, 2013). Twin studies have established AN-heritability
between 50% and 60%, indicating a considerable contribution of
genetic factors to disease liability (Watson et al., 2019; Yilmaz,
Hardaway, & Bulik, 2015).
Genetic studies of AN provide evidence of both psychiatric
and metabolic etiology. The largest AN genome-wide association
study (GWAS) to date (N
Cases
= 16 992) uncovered eight
AN-associated loci (Watson et al., 2019), and determined single-
nucleotide polymorphism (SNP)-based heritability (h
2
SNP
)of11–
17%, similar to other psychiatric disorders. In addition, genetic
correlations demonstrate significant shared genetic variation
between AN and other psychiatric disorders, including major
depressive disorder (MDD), obsessive-compulsive disorder, anx-
iety disorders (Watson et al., 2019), schizophrenia (Duncan
et al., 2017; Watson et al., 2019), alcohol use disorder
(Munn-Chernoff et al., 2021), as well as with physical activity
(Hübel et al., 2019a; Watson et al., 2019). Significant negative gen-
etic correlation between AN and anthropometric and metabolic
traits [including body mass index (BMI), fat mass, obesity,
type-2 diabetes, leptin, and insulin-related traits; Duncan et al.,
2017; Hübel et al., 2019a; Watson et al., 2019] have also been
observed, further indicating metabolic components to AN disease
risk. Studies of AN polygenic risk scores indicate additional gen-
etic associations of AN risk variants with anthropometric, behav-
ioral, and psychiatric traits (Hübel et al., 2021), and with weight
trajectories even among healthy adults (Xu et al., 2021).
Although GWASs may provide powerful insights into genetic
associations with AN, they cannot pinpoint gene- or tissue-level
associations. Therefore, in this study we apply transcriptomic
imputation (TI) to identify tissue-specific gene associations with
AN. TI approaches leverage gene expression predictor models
derived from large, well-curated gene expression datasets [e.g.
the Genotype-Tissue Expression project (GTEx) (Lonsdale
et al., 2013), CommonMind Consortium (CMC) (Fromer et al.,
2016; Huckins et al., 2019), and Depression Genes and
Networks (DGN) (Battle et al., 2014)]. These models may be
applied to predict genetically regulated gene expression (GReX)
in large genotyped cohorts, without the need to collect tissue sam-
ples (Huckins et al., 2019).
Observational genetic studies are powerful approaches to detect
disease-associated variants, but cannot address subthreshold or pro-
dromal disease states, and do not speak to clinical relevance.
Therefore, we leverage a phenome-wide association study
(pheWAS) design to probe clinical outcomes with AN-associations.
pheWAS effectively query the full electronic health record (EHR) to
identify diagnoses and traits associated with a gene or variant
(Denny et al., 2010; Pendergrass et al., 2011;Smoller,2018), and
have been used to replicate associations and identify new pleiotropic
consequences of GWAS variants, including psychiatric disorders
(Denny et al., 2013;Leppertetal.,2020; Zheutlin et al., 2019).
EHRscontainvastamountoflongitudinal health data, including
diagnoses, medications, laboratory tests, vital signs, and family med-
ical history, and, coupled with genetic data through hospital-based
biobanks, can provide an understanding of the clinical spectrum of
disease and disease progression across the lifetime of the patient.
Exploring the associations of AN-genetics with clinical phenotypes
has the potential to clarify how some of these GWAS variants func-
tionally contribute to AN disease risk, symptomatology, and clinical
presentation.
Here, we explored the associations between AN-GReX and the
clinical phenome. We performed TI using S-PrediXcan on the
most recent PGC-ED AN GWAS (Watson et al., 2019) to first
find GReX associated with AN, and then tested for clinical asso-
ciations of these genes with structured EHR-encoded phenotypes
using pheWAS. We further investigated the effects of BMI and sex
on the GReX-phenotype associations through stratification of bio-
bank individuals. Understanding the clinical associations of aber-
rant gene expression across the phenome may clarify the
biological mechanisms of relevant AN GWAS risk variants.
Methods
Transcriptomic imputation
We performed TI using S-PrediXcan (Barbeira et al., 2018) on the
largest available summary statistics of AN (16 992 cases/55 525
controls) (Watson et al., 2019). We tested for the association of
GReX using available GTEx (Lonsdale et al., 2013), DGN
(Battle et al., 2014), and CMC (Fromer et al., 2016; Hoffman
et al., 2019) predictor models (Barbeira et al., 2018; Gamazon
et al., 2015; Huckins et al., 2019) across 50 tissues with AN
case–control status. We established two thresholds for signifi-
cance; first, Bonferroni correction for all genes tested within
each tissue ( p
tissue
, online Supplementary Table S1), and second,
correcting for all tissues and genes tested ( p
Experiment
=3.75×
10
−8
). We performed a gene-set analysis for 53 significant
AN-genes using FUMA (v1.3.6) GENE2FUNC (Watanabe,
Taskesen, van Bochoven, & Posthuma, 2017), including all
S-PrediXcan genes (N= 28 454) as the background genes. We
defined significant gene sets using p
FDR
< 0.05.
BioMe™
The Mount Sinai BioMe™Biobank includes genotype and EHR
data from 31 704 individuals (online Supplementary Table S2).
Individuals were genotyped on the Illumina Global Screening
Array; quality control (QC), and imputation of the genotyping
data for BioMe™is described elsewhere (Zheutlin et al., 2019).
Ancestry QC of BioMe™individuals using principal component
analysis in PLINK (Chang et al., 2015; Purcell et al., 2007)is
described in the online Supplemental Methods. After QC, a
total of 31 585 individuals were available for analysis.
pheWAS
We calculated GReX for all S-PrediXcan p
tissue
significant genes
(N= 53) across 45 tissues from GTEx, DGN, and CMC in
BioMe™using PrediXcan-2 (Barbeira et al., 2018; Gamazon
et al., 2015). We excluded sex-specific tissues (ovary, uterus,
vagina, prostate, and testis) from this analysis, and genes with
GReX variance (gVAR) less than 0.002 (online Supplementary
Table S3C).
Logistic regressions between AN-GReX and categorical pheno-
types were performed using the pheWAS R package (Carroll,
Bastarache, & Denny, 2014), adjusting for age, sex, and the first
five genotype-derived principal components. pheWAS were run
individually for each BioMe™ancestry cohort, and results were
meta-analyzed using an inverse-variance based approach in
METAL (Willer, Li, & Abecasis, 2010). We required at least 10
occurrences of each phenotype in each population group for
2 Jessica S. Johnson et al.
https://doi.org/10.1017/S0033291721004554 Published online by Cambridge University Press
inclusion in the analysis, and an overall effective sample size N
eff
> 100 (Equation (1)). We excluded associations with significant
Cochran’sQ-test for heterogeneity scores across cohorts (pHet
> 0.001).
Effective sample size (N
eff
):
Neff =4
(1/NCases)+(1/NControls )(1)
‘Encounter Diagnoses’were recorded at each patient visit using
the International Classification of Disease (ICD) coding system.
Phecodes were assigned from Encounter Diagnoses by grouping
ICD-9 and ICD-10 diagnostic codes (Wu et al., 2019). We defined
cases as individuals with at least two counts of a code. Those with
zero counts were considered ‘controls’, and those with only one
count were set to missing. After QC, our dataset included 2178
unique Encounter Diagnosis codes and 1093 unique phecodes. Due
to the high correlation between Encounter Diagnosis and phecode
files, we combined all results from both pheWAS and performed
anFDRcorrectioninRtodeterminesignificance(p
FDR
<0.05).
In addition to diagnostic codes, BioMe™EHR data include
allergies, vital signs (weight, height, blood pressure, pulse, pulse
oximetry, respirations, and temperature), lab results, family history,
personal history, medication use, obstetrics and gynecology out-
comes (OB/GYN), and other social and behavioral traits.
Summaries, descriptive statistics, and QC descriptions for these phe-
notypes can be found in online Supplementary Methods. We tested
for AN-GReX-pheWAS associations within each category, establish-
ing significance using a Bonferroni correction (Equation (2)).
Bonferroni-correction for EHR data at tissue- and experiment-
significance thresholds:
PTissue =0.05
NPhenotypes
PExperiment =0.05
NPhenotypes ×NTissues
(2)
Stratification by BMI
Given the nature of AN symptomology, we tested whether our
significant AN-GReX associations varied based on BMI. We
stratified BioMe™individuals into three BMI categories: low
(<first BMI quartile, N= 6976), mid (first—third BMI quartiles,
N= 13 972) and high (>third BMI quartile, N= 9350; online
Supplementary Fig. S1A). Quartiles were defined based on
ancestry- and sex-specific BMI distributions within our data
(online Supplementary Fig. S1B; Table S2; Supplemental
Methods). We repeated our pheWAS within these categories,
for all AN-GReX associations reaching significance in our overall
analysis. BMI-stratified pheWAS were adjusted for age, sex, and
the first five genotype-derived PCs.
Stratification by sex
To determine whether our associations varied based on sex, we
stratified the BioMe™cohort by females (N= 17 968) and
males (N= 12 417), with sex determined by genotype, and
repeated our AN-GReX pheWAS analyses within each group.
Here, we included sex-specific tissues (ovary, uterus, vagina, pros-
tate, and testis) for the respective sex. Sex-stratified pheWAS were
adjusted for age and the first five genotype-derived PCs.
Testing for hidden case contamination
It is possible that some of the clinical associations within our
study stem from undiagnosed ED and AN cases, rather than a dir-
ect effect of gene expression on phenomic expression. We term
this ‘diagnostic contamination’. To test whether this effect may
drive the associations we observe, we simulated the effects of diag-
nostic contamination occurring at rate pwithin our biobank sam-
ple, with effect βon GReX.
Full derivations and simulations are shown in our online
Supplementary Material; briefly, we derive the expression
among controls, and cases contaminated at rate p; the difference
in expression between the two groups; the expected variance
among cases, and pooled across all samples, and the expected stat-
istical significance (T-score, and pvalue). Next, we simulated gene
expression distributions for (i) 1000 cases and 1000 controls; (ii)
1000 cases and 10 000 controls; (iii) 1000 cases and 30 000 con-
trols; each for a range of p(0.1, 0.5, 1, 2, 5, 10, 20, 30, 40, 50%)
and β(1/10, 1/5, 1/4, 1/3, 1/2, 1, 2, 3, 4, 5, 10) values. We repeated
our simulations 10 000 times at each case–control proportion- p-
βcombination, and demonstrated that our formulae accurately
estimate the desired values (online Supplementary Fig. S2).
In order to test whether diagnostic contamination may
account for the associations observed within our pheWAS, we cal-
culated the expected impact of diagnostic contamination for two
genes (NCKIPSD-Aorta; SEMA3F-Spinal Cord) with the largest
effect sizes βobserved in our S-PrediXcan analysis, under two
scenarios:
(1) Diagnostic contamination occurs among our cases only,
assuming the normal population rate for AN: 0.9% among
women; 0.3% among men.
(2) Contamination occurs at significantly higher levels than
might be expected in the population (0.6, 1.2, 3, 6%), and
all of these samples fall into our case category.
For each scenario, we repeated our calculation using 500 and
1000 cases, matched with 1000, 5000, and 30 000 controls.
Results
S-PrediXcan identifies gene–tissue associations with AN
We applied S-PrediXcan to PGC-AN GWAS summary statistics,
and identified 218 gene–tissue associations, including 53 unique
genes across 12 loci (Fig. 1). Two loci on chromosomes 2 and 3
reached experiment-wide significance ( p< 3.75 × 10
−8
), while 10
loci reached tissue-specific significance (Fig. 2,Table 1; online
Supplementary Table S3A). Our S-PrediXcan AN-genes are sig-
nificantly enriched in gene sets for traits including inflammatory
bowel disease (IBD: Crohn’s and ulcerative colitis) ( p
adj
= 1.3 ×
10
−25
), sleep duration ( p
adj
= 6.9 × 10
−17
), blood protein levels
(p
adj
= 3.2 × 10
−16
), intelligence ( p
adj
<1.5×10
−6
), regular
attendance at a religious group, gym, or sports club ( p
adj
<1.46 × 10
−10
), mood instability ( p
adj
= 3.3 × 10
−9
), AN ( p
adj
=
7.4 × 10
−8
), and body fat distribution (P
adj
<5.0×10
−7
) (online
Supplementary Table S4A and B).
AN-GReX is associated with autoimmune and
autoinflammatory diseases
We performed a pheWAS to identify clinical associations with
dysregulation of our AN-genes. We identified 16
Psychological Medicine 3
https://doi.org/10.1017/S0033291721004554 Published online by Cambridge University Press
FDR-significant gene–tissue associations with four Phecode- and
Encounter Diagnosis phenotypes (FDR < 0.05): type 1 diabetes
(N
Cases
= 408), celiac disease (N
Cases
= 63), peptic ulcer (N
Cases
=
254), and unspecified immunodeficiency (N
Cases
= 55) (Fig. 3;
online Supplementary Table S5). These associations are driven
in part by the MHC-gene CLIC1; predicted upregulation of
CLIC1 in spleen was associated with celiac disease and type 1
diabetes (p= 1.30 × 10
−11
,p=5.08×10
−20
, respectively) in the
overall cohort. We did not observe any significant effect of patient
BMI group on these associations (online Supplementary Fig. S3A
and C). Upregulation of CLIC1 in females was significantly asso-
ciated with type 1 diabetes (N
Cases
= 230) and celiac disease
(N
Cases
= 53) (Females-Spleen-CLIC1,p< 1.49 × 10
−9
) (online
Supplementary Fig. S3B and D) .
Fig. 1. Graphical depiction of S-PrediXcan and
PrediXcan TI and pheWAS analyses. (a)We
used S-PrediXcan predictor models for 50 differ-
ent tissue types to impute GReX in Watson et al.
(2019) AN GWAS and found 53 genes whose
GReX was associated with AN. We then (b)
imputed GReX for those 53 AN-genes in our
BioMe™cohort and (c) performed a pheWAS
across available EHR phenotypes. The pheWAS
analyses were run within each ancestral popula-
tion and then meta-analyzed using an inverse-
variance approach in METAL. Secondary ana-
lyses included stratifying individuals in BioMe™
by BMI and sex, and running the pheWAS ana-
lyses within each stratification group.
4 Jessica S. Johnson et al.
https://doi.org/10.1017/S0033291721004554 Published online by Cambridge University Press
BMI moderates the effect of MGMT-GReX on cholesterol levels
Next, we looked at the association of AN-GReX and measures of
highest recorded, lowest recorded and mean total cholesterol
(mg/dl), high-density lipoprotein (HDL) cholesterol (mg/dl),
and low-density lipoprotein (LDL) cholesterol (mg/dl) (Table 2).
In the full cohort, RPS26 and SUOX GReX were significantly
associated with the highest HDL measures, and MGMT GReX
Fig. 2. S-PrediXcan results for PGC-ED AN GWAS (2019)(N
Cases
= 16 992, N
Controls
= 55 525). S-PrediXcan TI of the PGC-ED AN GWAS summary statistics to determine
GReX, with tests for association of GReX with AN disease status. Manhattan plot of S-PrediXcan gene–tissue associations with AN for 50 tissues. Each point repre-
sents a different gene–tissue association result; i.e. the same gene may have multiple points within a peak. Experiment-wide significance threshold of p< 3.75 ×
10
−8
(solid line); tissue-specific significance threshold of p< 2.45 × 10
−5
(dotted line).
Table 1. S-PrediXcan loci results
Chromosome Genes Zscore pvalue
Top finding
(gene) Top finding (tissue)
1FBLIM1 −4.851060 1.23 × 10
−6
FBLIM1 Liver
2ASB3, CHAC2, GPR75 −5.780228 7.46 × 10
−9
GPR75 Whole Blood
3SPINK8, PFKFB4, TMEM89, SLC26A6, CELSR3, NCKIPSD,
ARIH2OS, ARIH2, P4HTM, WDR6, DALRD3, NDUFAF3,
USP19, LAMB2, CCDC71, CCDC36, C3orf62, GPX1, NICN1,
DAG1, APEH, MST1, MST1R, RNF123, RBM6, RBM5,
SEMA3F, TUSC2
7.321127 2.46 × 10
−13
WDR6 Skin, Sun Exposed
3PROS1, DHFRL1 4.687382 2.77 × 10
−6
DHFRL1 DGN Whole Blood
6CLIC1 4.814629 1.47 × 10
−6
CLIC1 Tibial Nerve
10 MGMT, EBF3 −5.086391 3.65 × 10
−7
MGMT Thyroid
11 ARNTL 4.534957 5.76 × 10
−6
ARNTL Whole Blood
12 SUOX, RPS26, SLC26A10 4.628307 3.69 × 10
−6
RPS26 Uterus
17 TNFSF12, LINC00324 4.982641 6.27 × 10
−7
TNFSF12 Breast Mammary Tissue
20 SLC2A10 4.921176 8.60 × 10
−7
SLC2A10 Tibial Nerve
22 KREMEN1 4.872670 1.10 × 10
−6
KREMEN1 Tibial Nerve
22 SCO2 −4.560015 5.11 × 10
−6
SCO2 Brain, Cerebelium
Summary of S-PrediXcan of PGC-ED AN GWAS results. Genes associated with AN from each locus are shown, as well as the top finding Zscore, pvalue, gene and tissue for each locus. pvalues
shaded in bold indicate experiment-wide significant loci ( p< 3.75 × 10
−8
).
Psychological Medicine 5
https://doi.org/10.1017/S0033291721004554 Published online by Cambridge University Press
was associated with the highest cholesterol at within-tissue sig-
nificance (N= 23 357) ( p< 5.1 × 10
−4
)(Fig. 4C). Associations
of AN-GReX with lipid measures varied across BMI-stratified
groups; predicted downregulation of MGMT in the stomach,
liver, esophagus, and cells was associated with the highest total
cholesterol and LDL levels, and in dorsolateral prefrontal cortex
and hippocampus with mean cholesterol and LDL among
high-BMI individuals (N
High BMI
= 6910) (High-BMI-Sto
mach-MGMT,p= 1.55 × 10
−7
)(Fig. 4C; online Supplementary
Table S6).
AN-GReX is associated with tobacco use
Multiple genes were associated with categorical and continuous
measures of tobacco use in the overall cohort at within-tissue
significance (online Supplementary Fig. S4, Table S7,
Supplementary Results). We found moderating effects of patient
BMI on tobacco use: among High-BMI groups, different genes
were associated with the number of years of tobacco use at
experiment-wide significance (N
High BMI
= 8129) (High-BMI-
Pituitary-CCDC36,p= 4.2 × 10
−6
; High-BMI-Colon, Transverse-
P4HTM,p= 3.4 × 10
−5
; High-BMI-Stomach-P4HTM,p= 2.2 ×
10
−5
) (online Supplementary Fig. S4, Table S7).
AN-GReX is associated with measures of pain score and pain
location
We tested for association between AN-GReX and (1) pain location,
(2) pain score overall, and (3) pain score by body location.
AN-GReX was associated with multiple measures of pain and
Fig. 3. AN-GReX associations with BioMe™Diagnosis codes. GReX-Tissue-Phenotype associations for (a) encounter diagnosis ICD-10 codes (N= 2178) and (b) phe-
codes (N= 1093) for the Bio Me™cohort (N= 30 585). Diagnosis codes are plotted along the x-axis and grouped by category, with the −log(10) pvalue associations
along the y-axis. FDR-significant diagnosis codes are labeled (FDR-adjusted p< 0.05 ). FDR-significant pvalue threshold p= 3.4 ×10
−7
(blue dashed line).
6 Jessica S. Johnson et al.
https://doi.org/10.1017/S0033291721004554 Published online by Cambridge University Press
Table 2. pheWAS AN-GReX tissue associations with cholesterol phenotypes
Stratification category Gene Tissue Effect S.E.p
(A) AN-GReX pheWAS associations with highest recorded total cholesterol (mg/dl)
High BMI MGMT Stomach −7.2 1.4 1.6 × 10
−7
High BMI MGMT Brain, nucleus accumbens basal ganglia −11.5 2.3 4.0 × 10
−7
High BMI MGMT Colon, transverse −6.1 1.3 1.2 × 10
−6
High BMI MGMT Brain, caudate basal ganglia −7.2 1.5 2.1 × 10
−6
High BMI MGMT Cells, EBV-transformed lymphocytes −8.0 1.7 2.8 × 10
−6
High BMI MGMT Thyroid −5.6 1.2 2.9 × 10
−6
High BMI MGMT Visceral omentum adipose −5.7 1.2 2.9 × 10
−6
High BMI MGMT CommonMind DLPFC −13.6 2.9 3.7 × 10
−6
High BMI MGMT Lung −5.9 1.3 6.3 × 10
−6
High BMI MGMT Esophagus, mucosa −5.6 1.3 9.3 × 10
−6
High BMI MGMT Brain, hippocampus −7.6 1.7 1.0 × 10
−5
High BMI MGMT Brain, cerebellum −7.8 1.8 1.9× 10
−5
High BMI MGMT Spleen −3.7 0.9 2.7 × 10
−5
High BMI MGMT Tibial nerve −5.2 1.3 3.7 × 10
−5
High BMI MGMT Liver −5.5 1.4 1.3 × 10
−4
High BMI MGMT Brain, cerebellar hemisphere −4.2 1.2 2.9 × 10
−4
Overall MGMT Stomach −2.7 0.7 2.9 × 10
−4
(B) AN-GReX pheWAS associations with highest recorded LDL cholesterol (mg/dl)
High BMI MGMT Stomach −5.2 1.1 2.6 × 10
−6
High BMI MGMT Brain, nucleus accumbens basal ganglia −7.9 1.8 1.5 × 10
−5
High BMI MGMT Brain, hippocampus −5.8 1.4 3.1 × 10
−5
High BMI MGMT Colon, transverse −4.1 1.0 4.0 × 10
−5
High BMI MGMT Cells, EBV-transformed lymphocytes −5.7 1.4 4.2 × 10
−5
High BMI MGMT Brain, caudate basal ganglia −4.9 1.2 7.4 × 10
−5
High BMI MGMT Thyroid −3.8 1.0 7.4 × 10
−5
High BMI MGMT Visceral omentum adipose −3.8 1.0 9.4 × 10
−5
High BMI MGMT CommonMind DLPFC −9.1 2.4 1.2 × 10
−4
High BMI MGMT Esophagus, mucosa −3.9 1.0 1.2 × 10
−4
High BMI MGMT Lung −3.8 1.1 3.5 × 10
−4
High BMI MGMT Brain, cerebellum −5.2 1.5 4.1 × 10
−4
High BMI MGMT Tibial nerve −3.5 1.0 6.3 × 10
−4
High BMI MGMT Liver −3.8 1.1 8.2 × 10
−4
Mid BMI SLC26A10 Aorta artery −4.7 1.4 8.8 × 10
−4
(C) AN-GReX pheWAS associations with highest recorded HDL cholesterol (mg/dl)
Overall SUOX Thyroid 0.9 0.3 1.7 × 10
−4
Overall SUOX Brain, caudate basal ganglia 1.4 0.4 1.9 × 10
−4
Overall RPS26 Breast mammary tissue −0.7 0.2 1.9 × 10
−4
Overall SUOX Tibial artery 1.4 0.4 2.0 × 10
−4
Overall RPS26 Brain, nucleus accumbens basal ganglia −0.7 0.2 3.1 × 10
−4
Overall SUOX Tibial nerve 0.8 0.2 3.5 × 10
−4
Overall SUOX Pituitary 1.0 0.3 4.6 × 10
−4
Female SUOX Brain, caudate basal ganglia 1.8 0.5 4.7 × 10
−4
(Continued)
Psychological Medicine 7
https://doi.org/10.1017/S0033291721004554 Published online by Cambridge University Press
Table 2. (Continued.)
Stratification category Gene Tissue Effect S.E.p
Overall SUOX Brain, anterior cingulate cortex (BA24) 2.0 0.6 5.1 × 10
−4
Overall RPS26 CommonMind DLPFC −1.2 0.3 5.2 × 10
−4
Overall RPS26 Skeletal muscle −0.6 0.2 5.4 × 10
−4
Overall SUOX Esophagus, mucosa 4.2 1.2 5.5 × 10
−4
Overall SUOX Skin, not sun exposed 1.2 0.3 6.1 × 10
−4
Overall RPS26 Brain, putamen basal ganglia −0.7 0.2 6.3 × 10
−4
Overall RPS26 Minor salivary gland −1.0 0.3 6.5 × 10
−4
Overall SUOX Small intestine, terminal ileum 1.3 0.4 6.8 × 10
−4
Overall SUOX Esophagus, muscularis 1.4 0.4 7.6 × 10
−4
Overall RPS26 Tibial nerve −0.6 0.2 7.6 × 10
−4
Overall SUOX Lung 1.0 0.3 7.7 × 10
−4
Overall SUOX Pancreas 1.0 0.3 7.9 × 10
−4
Overall RPS26 Brain, substantia nigra −0.9 0.3 8.0 × 10
−4
Overall RPS26 Brain, anterior cingulate cortex (BA24) −0.6 0.2 8.1 × 10
−4
Overall RPS26 Brain, hypothalamus −0.7 0.2 8.1 × 10
−4
Overall SUOX Heart, left ventricle 2.2 0.7 8.1 × 10
−4
Overall RPS26 Aorta artery −0.7 0.2 8.2 × 10
−4
Overall RPS26 Brain, spinal cord (cervical c-1) −0.6 0.2 8.5 × 10
−4
Overall SUOX Spleen 0.9 0.3 8.8 × 10
−4
Overall SUOX Breast mammary tissue 3.7 1.1 9.2 × 10
−4
Overall RPS26 Whole blood −0.7 0.2 9.5 × 10
−4
Overall RPS26 Lung −0.6 0.2 9.9 × 10
−4
(D) AN-GReX pheWAS associations with mean total cholesterol (mg/dl)
High BMI MGMT CommonMind DLPFC −9.5 2.1 4.8 × 10
−6
High BMI MGMT Brain, spinal cord (cervical c-1) −4.9 1.1 6.6 × 10
−6
High BMI MGMT Brain, caudate basal ganglia −4.8 1.1 7.9 × 10
−6
High BMI MGMT Brain, hippocampus −5.2 1.2 1.8 × 10
−5
High BMI MGMT Lung −3.9 0.9 2.2 × 10
−5
High BMI MGMT Esophagus, mucosa −3.7 0.9 2.6 × 10
−5
High BMI MGMT Brain, cerebellum −5.1 1.3 6.8× 10
−5
High BMI MGMT Liver −3.9 1.0 9.1 × 10
−5
High BMI MGMT Spleen −2.4 0.6 1.4 × 10
−4
High BMI APEH Skin, sun exposed 8.9 2.6 5.8 × 10
−4
(E) AN-GReX pheWAS associations with mean LDL cholesterol (mg/dl)
High BMI MGMT Brain, hippocampus −4.1 1.0 2.6 × 10
−5
High BMI MGMT CommonMind DLPFC −7.1 1.7 4.1 × 10
−5
High BMI MGMT Brain, spinal cord (cervical c-1) −3.7 0.9 4.2 × 10
−5
High BMI MGMT Esophagus, mucosa −2.9 0.7 1.0 × 10
−4
High BMI MGMT Brain, caudate basal ganglia −3.4 0.9 1.2 × 10
−4
High BMI MGMT Brain, cerebellum −3.9 1.1 2.8 × 10
−4
High BMI MGMT Lung −2.8 0.8 2.8 × 10
−4
High BMI MGMT Liver −3.0 0.8 3.1 × 10
−4
Mid BMI SLC26A10 Aorta artery −3.4 1.0 7.8 × 10
−4
(Continued)
8 Jessica S. Johnson et al.
https://doi.org/10.1017/S0033291721004554 Published online by Cambridge University Press
location across all stratification groups (Fig. 4D–F;online
Supplementary Table S8). Our results revealed a large number of
AN-GReX-pain associations specific to the Low-BMI category
(Fig. 4D–F; online Supplementary Table S8); in particular, we
note 60 gene–tissue associations with presence of foot pain
(N
Cases
= 433), one of which passed experiment-wide significance
(Low-BMI-Small intestine-GPX1,p=7.6×10
−6
), which may indi-
cate a propensity to excess exercise, or exercise-related injuries
among these individuals.
Upregulation of CLIC1 is associated with glucagon medication
Upregulation of CLIC1 was associated with glucagon, a hormone
used to treat severe hypoglycemia, in the overall cohort (N
Cases
=
129) (Overall-Spleen-CLIC1,p= 4.20 × 10
−9
) and in females
(N
Cases
= 78) (Females-Spleen-CLIC1,p= 1.4 × 10
−8
) (online
Supplementary Fig. S5, Table S9). We additionally find significant
associations between the upregulation of CLIC1 and insulin
medications in females (Females-Subcutaneous adipose-CLIC1,
p< 9.6 × 10
−7
).
Identifying potential case contamination effects
Our BioMe™patients have not all been explicitly assessed for eat-
ing disorders, and information regarding eating disorder diagno-
ses and assessments earlier in life may be omitted from the
records due to incomplete clinical history assessments.
Therefore, it is possible that diagnostic contamination within
some of our sample is responsible for the associations observed
within our data. We tested whether such contamination may
drive the associations observed in our study, assuming the follow-
ing different contamination scenarios (online Supplementary
Table S10), for two genes with large S-PrediXcan effect sizes: (i)
that contamination occurs among cases, assuming the highest
common estimate for AN prevalence (0.9% among women,
0.3% among men); (ii) that contamination occurs within our bio-
bank, assuming the highest common estimate for AN prevalence
(0.6%, 190 cases), and that all of these samples fall into our ‘case’
category.
Our model indicates that diagnostic contamination among
cases is unlikely to result in significant pheWAS associations in
our data. Contamination occurring at population ED prevalence
(0.6%) did not result in significant associations for any of the
Table 2. (Continued.)
Stratification category Gene Tissue Effect S.E.p
(F) AN-GReX pheWAS associations with mean HDL cholesterol (mg/dl)
Mid BMI RP11-3B7.1 Lung −2.1 0.6 4.2 × 10
−4
Male RP11-3B7.1 Lung −2.0 0.6 4.9 × 10
−4
Overall RP11-3B7.1 Lung −1.5 0.4 6.2 × 10
−4
Male GPX1 Brain, caudate basal ganglia 1.2 0.3 8.3 × 10
−4
Overall RPS26 Brain, nucleus accumbens basal ganglia −0.5 0.1 9.4 × 10
−4
Overall RPS26 Breast mammary tissue −0.5 0.1 9.4 × 10
−4
(G) AN-GReX pheWAS associations with lowest total cholesterol (mg/dl)
Female PFKFB4 Skin, sun exposed 5.7 1.7 7.4 × 10
−4
High BMI MGMT CommonMind DLPFC −7.3 2.2 7.5 × 10
−4
High BMI MST1 Stomach 5.0 1.5 8.7 × 10
−4
(H) AN-GReX pheWAS associations with lowest LDL cholesterol (mg/dl)
Mid BMI RBM5 Pituitary 8.3 2.5 7.2 × 10
−4
Overall MST1 Liver 1.4 0.4 8.6 × 10
−4
(I) AN-GReX pheWAS associations with lowest HDL cholesterol (mg/dl)
High BMI RP11-477N3.1 Lung 4.2 1.0 3.4 × 10
−5
Female SEMA3F Spleen 3.3 0.9 3.9 × 10
−4
High BMI RP11-477N3.1 Skeletal muscle 2.4 0.7 3.9 × 10
−4
Mid BMI RP11-3B7.1 Lung −2.0 0.6 4.5 × 10
−4
Overall SEMA3F Spleen 2.3 0.7 7.0 × 10
−4
Overall RP11-3B7.1 Lung −1.4 0.4 8.0 × 10
−4
High BMI RP11-477N3.1 Tibial artery 2.0 0.6 9.1 × 10
−4
Mid BMI CCDC36 Lung −1.7 0.5 9.2 × 10
−4
pheWAS associations with highest, lowest, and mean total cholesterol, LDL cholesterol and HDL cholesterol measures for all stratification groups: overall (N= 23 661), high-BMI (N= 6910),
mid-BMI (N= 11 243), low-BMI (N= 5324), females (N= 13 907), and males (N= 9669). AN-GReX association with phenotypes of (A) highest recorded total cholesterol (mg/dl), (B) highest
recorded LDL cholesterol (mg/dl), (C) highest recorded HDL cholesterol (mg/dl), (D) mean total cholesterol (mg/dl), (E) mean LDL cholesterol (mg/dl), (F) mean HDL cholesterol (mg/dl), (G)
lowest recorded total cholesterol (mg/dl), (H) lowest recorded LDL cholesterol (mg/dl), and (I) lowest recorded HDL cholesterol (mg/dl). Bonferroni experiment-wide significant associations
are marked in bold ( p
Experiment
= 0.05/(9 phenotypes × 45 tissues) = 1.23 × 10
−4
). Bonferroni tissue-specific value threshold p< 0.0056 (0.05/9 phenotypes).
Psychological Medicine 9
https://doi.org/10.1017/S0033291721004554 Published online by Cambridge University Press
Fig. 4. Context-specific pheWAS associations. (a) Concordance of context-specific experiment-wide significant AN-GReX clinical associations with AN direction of
effect. We compared the direction of effect (DoE) for each experiment-wide gene–tissue-pheWAS association with the DoE of that gene–tissue pair for AN from
our S-PrediXcan results (online Supplemental Methods). For those phenotypes concordant with AN, this may indicate that genetic regulation of those
AN-genes is more similar to individuals with AN in individuals with those clinical outcomes. (b) Schematic of the proportion of concordance of AN-GReX
pheWAS associations with AN S-PrediXcan associations. Associations with similar direction of effect to AN (green) identified as ‘concordant’, associations with
opposite direction of effect ( purple) identified as ‘discordant’.(c) Context-specific associations of AN-GReX with lipid phenotypes of highest recorded, lowest
recorded, and mean measures of total cholesterol (mg/dl), HDL cholesterol (mg/dl), and LDL cholesterol (mg/dl). Experiment-wide significant threshold set at
p= 0.05(9 phenotypes × 45 tissues) = 1.2 × 10
−4
. Tissue-specific threshold set at 0.05/(9 phenotypes) = 0.0056. Context-specific associations of AN-GReX with pain
location for (d) experiment-wide significant associations ( p= 0.05/(99 phenotypes × 45 tissues) = 1.1 × 10
−5
), (e) all context-specific associations (experiment-wide
and tissue-specific) with generalized pain, and ( f) with foot pain. Tissue-specific threshold for pain location set at 0.05/(99 phenotypes) = 5.0 × 10
−4
.
10 Jessica S. Johnson et al.
https://doi.org/10.1017/S0033291721004554 Published online by Cambridge University Press
case–control scenarios in our model (online Supplementary
Table S10); nor did contamination at 1.2% or 3%. Assuming
6% contamination resulted in potentially significant associations
(estimated p< 2.8 × 10
−8
; see Methods and online
Supplementary Methods for more details).
Discussion
Our analysis identified novel AN-genes associated with metabolic,
anthropometric, autoimmune, and psychiatric phenotypes (online
Supplementary Table S4B). In particular, the experiment-wide
significant locus on chromosome 3 overlaps with a known
GWAS peak for IBD (de Lange et al., 2017), and includes genes
associated with Crohn’s and ulcerative colitis (online
Supplementary Table S4B) (Hedman et al., 2019; Mårild et al.,
2017; Raevuori et al., 2014; Zerwas et al., 2017), in line with earlier
AN GWAS showing genetic overlap with autoimmune disease
(Gibson & Mehler, 2019). Variants in one of our significant
S-PrediXcan genes, GPR75, have been recently shown to have a
protective effect against obesity, and have been associated with
lower body weight overall (Akbari et al., 2021).
To our knowledge, our results include the first AN-association
within the MHC locus (CLIC1, chloride intracellular 1), a region
that has been associated with many other psychiatric (Ripke et al.,
2014; Stahl et al., 2019) and autoimmune disorders. In particular,
CLIC1 has previously been identified as associated with schizo-
phrenia (The Autism Spectrum Disorders Working Group of
The Psychiatric Genomics Consortium, 2017), autism (The
Autism Spectrum Disorders Working Group of The Psychiatric
Genomics Consortium, 2017), MDD (Zhu et al., 2019), post-
traumatic stress disorder (Marchese et al., 2021), neuroticism
(Baselmans et al., 2019) and depressive phenotypes (Baselmans
et al., 2019). Importantly, CLIC1 variants have also been asso-
ciated with complement component C4 and C3 protein levels
in the blood (Yang et al., 2012), which through the complement
cascade (Sekar et al., 2016) are involved in immunological func-
tions of pathogen clearance and in synaptic pruning and neuronal
connectivity (Stephan, Barres, & Stevens, 2012). CLIC1 encodes a
chloride ion channel protein involved in many necessary cellular
functions including the regulation of cell membrane potential,
and the proliferation and differentiation of cells (Li et al., 2018),
including a role in axonal outgrowth of neurons (Carlini et al.,
2020).
We also identified multiple genes that have been previously asso-
ciated with decreased sleep and insomnia phenotypes (including a
core circadian clock gene ARNTL), and with a range of psychiatric
disorders (Alloy, Ng, Titone, & Boland, 2017;Boivin,2000;Bunney
&Bunney,2013;Bunneyetal.,2015;Etainetal.,2014;Huckins
et al., 2017; Karatsoreos, 2014;Kripke,Mullaney,Atkinson,&
Wolf, 1978; Mansour et al., 2009;Melo,Abreu,LinharesNeto,de
Bruin, & de Bruin, 2017; Meyrer, Demling, Kornhuber, & Nowak,
2009; Murray, Allen, Trinder, & Burgess, 2002), including eating
disorders (Allison, Spaeth, & Hopkins, 2016), as well as with satiety
andhunger(Arble,Bass,Laposky,Vitaterna,&Turek,2009;
Herpertz et al., 2000).
In a patient population, expression of AN genes is associated
with metabolic and autoimmune phenotypes
Unlike GWAS, which include carefully constructed case–control
cohorts, pheWAS encompass all individuals within a healthcare
system, including patients with subthreshold or partial presenta-
tions of a disorder, or individuals with commonly comorbid or
co-diagnosed conditions. Importantly, because individuals are
not ascertained for any specific disorder, they represent a more
comprehensive clinical picture of the comorbidities and symp-
tomatology associated with AN gene expression.
Examining the consequences of aberrant predicted gene
expression among these patients (i.e. testing for GReX associa-
tions) may reveal clinical and biological consequences of these
genes; for example, studying whether AN-associated genes have
anthropometric and metabolic consequences among adults with
no evidence for previous AN- or ED-diagnoses may disentangle
whether certain endophenotypes present as a cause or conse-
quence of AN. Food avoidance and restriction may arise due to
gastrointestinal (GI) complaints and distress that provoke these
behaviors and precede development of AN (Zucker & Bulik,
2020). Similarly, autoimmune disorders of the GI tract, such as
celiac disease and Crohn’s disease, show a bidirectional relation-
ship with AN, with previous diagnosis of a GI-associated auto-
immune disorders increasing the risk of AN and vice versa
(Hedman et al., 2019). Our pheWAS results of AN-GReX associa-
tions with GI symptoms such as abdominal pain, ascites, and pep-
tic ulcer, as well as GI-related autoimmune disorders such as
celiac disease, suggest AN-GReX may contribute directly to
these diseases and phenotypes, and that food aversive behaviors
and gastric distress may be genetically regulated in these indivi-
duals, rather than occurring as a consequence of AN. We found
a sex-specific association of CLIC1-GReX with celiac disease in
females (online Supplementary Fig. S3B), indicating sex-specific
regulation of AN-genes may be contributing to the disparity in
autoimmune diagnoses and symptomatology between the sexes
(Fuchs et al., 2014; Ludvigsson & Murray, 2019). Our main
pheWAS diagnosis results of celiac disease and type 1 diabetes
further show concordance with AN in the direction of effect of
the tissue GReX, indicating a similarity in the genetic regulation
of AN-genes between patients with these disorders and indivi-
duals with AN (online Supplementary Fig. 4A; Supplemental
Results). Although these associations could be the consequence
of undiagnosed AN-individuals in our biobank, they more likely
reflect real biological associations of expression of those particular
genes with the phenotype.
Our results further confirm the contribution of metabolic fac-
tors to AN etiology; we see very robust association of AN-GReX
with type 1 diabetes, the hyperglycemic hormone glucagon and
various forms of insulin. The association of AN with metabolic
traits and abnormalities has been fairly well established, including
with insulin resistance (r
g
=−0.29), fasting insulin (r
g
=−0.24),
leptin (r
g
=−0.26) and type 2 diabetes (r
g
=−0.22), and a positive
genetic correlation with HDL cholesterol (r
g
= 0.21) (Adams,
Reay, Geaghan, & Cairns, 2021; Ilyas et al., 2019; Watson et al.,
2019). Notably, our results point to a similar role of aberrant gly-
cemic regulation in the etiology of AN. Future analyses including
EHR-derived lab results (LabWAS) studies may further elucidate
AN-genes associated with abnormal metabolic regulation and
clinical features.
Context-specific associations reveal a role for BMI in the
regulation of AN-gene expression
We stratified our pheWAS analyses by BMI in order to observe
whether the effects of predicted AN-gene expression on clinical
outcomes were moderated by BMI status. Understanding how
Psychological Medicine 11
https://doi.org/10.1017/S0033291721004554 Published online by Cambridge University Press
BMI influences associations between AN-GReX and the clinical
phenome may give us further insight into the biological mechan-
isms leading to those outcomes and the conferral of AN disease
risk. The context-specific associations we see between BMI status
and the association of AN-GReX with multiple phenotypes indi-
cate the potential differences in AN-gene regulation in an envir-
onment of higher or lower BMI.
Among our BMI-stratified pheWAS analyses, we identified a
number of associations between AN-GReX and foot pain, exclu-
sive to the Low-BMI group. We hypothesize that these associa-
tions may stem from exercise-related injuries among this group.
Multiple genes in our study, many of which were associated
with foot pain in Low-BMI individuals, have been previously
associated with regular gym attendance (online Supplementary
Table S4B, p
adj
=3.4×10
−10
), or measures of physical
activity. Excessive and compulsive exercise is a behavior often
seen in individuals with AN (Bulik, 2004; Shroff et al., 2006),
and evidence is suggestive that hyperactivity increases risk of
chronic AN (Achamrah, Coëffier, & Déchelotte, 2016; Strober,
Freeman, & Morrell, 1997). Similarly, studies have shown a gen-
etic correlation between AN and physical activity (r
g
= 0.17)
(Hübel et al., 2019b; Watson et al., 2019). These genes may reflect
a genetic liability to compulsive physical activity. Given the
known association of AN disease with low bone mineral density
and propensity for bone fracture (Solmi et al., 2016), our results
may reflect the result of genes associated with compulsive exercise
and bone mineral density contributing to increased osteoarticular
pain.
We saw distinct modulation by BMI of MGMT-GReX-
cholesterol associations in our cohort. Variants in the MGMT
gene have been shown to be associated with AN (Watson et al.,
2019), as well as anthropometric phenotypes such as waist-to-hip
ratio and body height (Kichaev et al., 2019). CTNNB1, which was
associated with both highest and lowest weight measures in our
cohort, distinctly with measures of highest weight in High-BMI
individuals, has been previously shown to be associated with
lean body mass (Hübel et al., 2019a). We saw distinct associations
of PROS1-GReX with lowest weight in Low-BMI individuals,
which is a gene involved in many biological processes, including
inflammation (Suleiman, Négrier, & Boukerche, 2013). Our
results appear to be due to the effect of BMI on AN-GReX, rather
than the direct effect of BMI/weight on the phenotypes (online
Supplementary Fig. S4B). Future studies are needed to assess the
specific role of BMI on AN-gene regulation.
Although the hypothesis-free, phenome-wide design of our
study allows for powerful detection of clinical and biological asso-
ciations of AN-risk genes, the same design also bears some not-
able limitations and caveats. One key caveat is the lack of
diagnostic detail available for the patients in our study. EHR ana-
lyses leverage large, existing data sets to rapidly amass cohorts for
analysis, to yield insights into whole phenome consequences of
genotype and GReX associations; however, the scale and scope
of these studies precludes deep phenotyping or performing
clinical interviews. This lack of diagnostic precision may arise
from a number of factors.
First, we make use of ICD codes within the medical record in
order to infer diagnoses; since these are primarily used by clini-
cians for billing purposes, they likely provide an imperfect
proxy for true disease state. In order to mitigate spurious results
stemming from imperfectly assigned codes, we require ‘cases’to
have at least two instances of an ICD code; individuals with
only one code were set to missing. Furthermore, for each code
we restricted our meta-analysis to include only phenotypes with
at least 10 cases and 10 controls, and required a total effective
sample size >100 for inclusion in our final analysis.
Second, it is possible that our patients regularly receive treat-
ment at multiple different hospital systems; as such, we may be
capturing only partial data for each of our patients. In order to
mitigate this, we restricted our analysis to patients with multiple
data points within our EHR. Future analyses that seek to harmon-
ize or meta-analyze patient data across EHR (e.g. NYC-CDRN,
PsycheMERGE) are ongoing, and will be vital to disentangling
this effect further.
Importantly, our patients have not all been explicitly assessed
for EDs, and information regarding ED diagnoses and assess-
ments earlier in life may be omitted from the records due to
incomplete clinical history assessments. Therefore, it is possible
that case-contamination within some of our sample is responsible
for the associations observed within our data. In order to address
this, we performed a simple experiment to simulate the effect sizes
expected if undiagnosed AN contamination drives our result. We
estimated the expected association statistics observed due to pos-
sible levels of contamination among pheWAS cases. Our model
indicates that diagnostic contamination at 0.06, 1.2, or 3% is
insufficient to account for the effect sizes observed within our
pheWAS analysis. Moreover, since these genes are selected due
to their large S-PrediXcan effect sizes, we expect that the contam-
ination effects observed for these two genes will be greater than
any others in our study; as such, we do not believe that our find-
ings are attributable to hidden case contamination.
Our results illustrate the clinical outcomes associated with dif-
ferences in AN-gene expression. Characterization of the pheno-
typic associations with AN-gene expression in a clinical setting
can give us more insight into the biological mechanisms under-
lying AN and, consequently, how to diagnose and treat the dis-
order. By understanding the associations of AN-gene expression
with symptomatology, prodromal or subthreshold disease states,
we may gain insights into the biology of the disease, and perhaps
identify therapeutic targets and opportunities for clinical inter-
ventions. For example, if GI complaints are truly the consequence
of aberrant AN-gene expression, and contribute to disordered eat-
ing due to GI distress, treatment of those symptoms may help
alleviate other AN symptoms or prevent development of later
AN (Riedlinger et al., 2020; Wiklund et al., 2019). An understand-
ing of the clinical associations of AN-gene expression can further
augment the definition of AN, and could allow clinicians to more
broadly identify individuals at greater risk of AN, or those who
present with symptom constellations that do not yet meet the
established diagnostic threshold.
Supplementary material. The supplementary material for this article can
be found at https://doi.org/10.1017/S0033291721004554.
Acknowledgements. We are deeply grateful for the mentorship of Pamela
Sklar, whose guidance and wisdom we miss daily. We strive to continue her
legacy of thoughtful, innovative, and collaborative science.
Financial support. JJ and LMH were supported by funding from the
Klarman Family Foundation (2019 Eating Disorders Research Grants
Program) and the NIMH (R01MH118278). CMB is supported by NIMH
(R01MH120170; R01MH124871; R01MH119084; R01MH118278; and
R01MH124871); Brain and Behavior Research Foundation Distinguished
Investigator Grant; Swedish Research Council (Vetenskapsrådet, award:
538-2013-8864); Lundbeck Foundation (Grant no. R276-2018-4581). This
study was supported in part through the resources and staff expertise provided
12 Jessica S. Johnson et al.
https://doi.org/10.1017/S0033291721004554 Published online by Cambridge University Press
by the Charles Bronfman Institute for Personalized Medicine and The
BioMe
TM
Biobank Program at the Icahn School of Medicine at Mount
Sinai. Research reported in this paper was supported by the Office of
Research Infrastructure of the National Institutes of Health under award num-
bers S10OD018522 and S10OD026880. The content is solely the responsibility
of the authors and does not necessarily represent the official views of the
National Institutes of Health.
Conflict of interest. CMB reports: Shire (grant recipient, Scientific Advisory
Board member); Idorsia (consultant); Pearson (author, royalty recipient);
Equip Health Inc. (clinical advisory board). ML declares that over the past
36 months, he has received lecture honoraria from Lundbeck
Pharmaceutical (no other equity ownership, profit-sharing agreements, royal-
ties, or patent). The remaining authors declare no competing interests.
References
Achamrah, N., Coëffier, M., & Déchelotte, P. (2016). Physical activity in
patients with anorexia nervosa. Nutrition Reviews,74(5), 301–311. https://
doi.org/10.1093/nutrit/nuw001.
Adams, D. M., Reay, W. R., Geaghan, M. P., & Cairns, M. J. (2021).
Investigation of glycaemic traits in psychiatric disorders using Mendelian
randomisation revealed a causal relationship with anorexia nervosa.
Neuropsychopharmacology,46(6), 1093–1102. https://doi.org/10.1038/
s41386-020-00847-w.
Akbari, P., Gilani, A., Sosina, O., Kosmicki, J. A., Khrimian, L., Fang, Y.-Y., …
Lotta, L. A. (2021). Sequencing of 640 000 exomes identifies GPR75 variants
associated with protection from obesity. Science,373(6550), eabf8683.
https://doi.org/10.1126/science.abf8683.
Allison, K. C., Spaeth, A., & Hopkins, C. M. (2016). Sleep and eating disorders.
Current Psychiatry Reports,18(10), 92. https://doi.org/10.1007/s11920-016-
0728-8.
Alloy, L. B., Ng, T. H., Titone, M. K., & Boland, E. M. (2017). Circadian
rhythm dysregulation in bipolar Spectrum disorders. Current Psychiatry
Reports,19(4), 21. https://doi.org/10.1007/s11920-017-0772-z.
Arble, D. M., Bass, J., Laposky, A. D., Vitaterna, M. H., & Turek, F. W. (2009).
Circadian timing of food intake contributes to weight gain. Obesity,17(11),
2100–2102. https://doi.org/10.1038/oby.2009.264.
Barbeira, A. N., Dickinson, S. P., Bonazzola, R., Zheng, J., Wheeler, H. E.,
Torres, J. M., …Im, H. K. (2018). Exploring the phenotypic consequences
of tissue specific gene expression variation inferred from GWAS summary
statistics. Nature Communications,9(1), 1825. https://doi.org/10.1038/
s41467-018-03621-1.
Baselmans, B. M. L., Jansen, R., Ip, H. F., van Dongen, J., Abdellaoui, A., van
de Weijer, M. P., …Bartels, M. (2019). Multivariate genome-wide analyses
of the well-being spectrum. Nature Genetics,51(3), 445–451. https://doi.org/
10.1038/s41588-018-0320-8.
Battle, A., Mostafavi, S., Zhu, X., Potash, J. B., Weissman, M. M., McCormick,
C., …Koller, D. (2014). Characterizing the genetic basis of transcriptome
diversity through RNA-sequencing of 922 individuals. Genome Research,
24(1), 14–24. https://doi.org/10.1101/gr.155192.113.
Boivin, D. B. (2000). Influence of sleep-wake and circadian rhythm distur-
bances in psychiatric disorders. Journal of Psychiatry & Neuroscience:
JPN,25(5), 446–458.
Bulik, C. M. (2004). Genetic and biological risk factors. In J. K. Thompson
(Ed.), Handbook of eating disorders and obesity ( p. 3). New Jersey: Wiley.
Bulik, C. M., Flatt, R., Abbaspour, A., & Carroll, I. (2019). Reconceptualizing
anorexia nervosa. Psychiatry and Clinical Neurosciences,73(9), 518–525.
https://doi.org/10.1111/pcn.12857.
Bunney, B. G., & Bunney, W. E. (2013). Mechanisms of rapid antidepressant
effects of sleep deprivation therapy: Clock genes and circadian rhythms.
Biological Psychiatry,73(12), 1164–1171. https://doi.org/10.1016/j.biop-
sych.2012.07.020.
Bunney, B. G., Li, J. Z., Walsh, D. M., Stein, R., Vawter, M. P., Cartagena, P., …
Bunney, W. E. (2015). Circadian dysregulation of clock genes: Clues to rapid
treatments in major depressive disorder. Molecular Psychiatry,20(1), 48–55.
https://doi.org/10.1038/mp.2014.138.
Carlini, V., Verduci, I., Cianci, F., Cannavale, G., Fenoglio, C., Galimberti, D.,
& Mazzanti, M. (2020). CLIC1 protein accumulates in circulating monocyte
membrane during neurodegeneration. International Journal of Molecular
Sciences,21(4), 1484. https://doi.org/10.3390/ijms21041484.
Carroll, R. J., Bastarache, L., & Denny, J. C. (2014). R pheWAS: Data analysis
and plotting tools for phenome-wide association studies in the R environ-
ment. Bioinformatics,30(16), 2375–2376. https://doi.org/10.1093/bioinfor-
matics/btu197.
Chang, C. C., Chow, C. C., Tellier, L. C., Vattikuti, S., Purcell, S. M., & Lee, J. J.
(2015). Second-generation PLINK: Rising to the challenge of larger and
richer datasets. GigaScience,4,7.https://doi.org/10.1186/s13742-015-
0047-8.
Chesney, E., Goodwin, G. M., & Fazel, S. (2014). Risks of all-cause and suicide
mortality in mental disorders: A meta-review. World Psychiatry,13(2), 153–
160. https://doi.org/10.1002/wps.20128.
de Lange, K. M., Moutsianas, L., Lee, J. C., Lamb, C. A., Luo, Y., Kennedy, N.
A., …Barrett, J. C. (2017). Genome-wide association study implicates
immune activation of multiple integrin genes in inflammatory bowel dis-
ease. Nature Genetics,49(2), 256–261. https://doi.org/10.1038/ng.3760.
Denny, J. C., Bastarache, L., Ritchie, M. D., Carroll, R. J., Zink, R., Mosley, J. D.,
…Roden, D. M. (2013). Systematic comparison of phenome-wide associ-
ation study of electronic medical record data and genome-wide association
study data. Nature Biotechnology,31(12), 1102–1110. https://doi.org/10.
1038/nbt.2749.
Denny, J. C., Ritchie, M. D., Basford, M. A., Pulley, J. M., Bastarache, L.,
Brown-Gentry, K., …Crawford, D. C. (2010). pheWAS: Demonstrating
the feasibility of a phenome-wide scan to discover gene-disease associations.
Bioinformatics,26(9), 1205–1210. https://doi.org/10.1093/bioinformatics/
btq126.
Duncan, L., Yilmaz, Z., Gaspar, H., Walters, R., Goldstein, J., Anttila, V., …
Bulik, C. M. (2017). Significant locus and metabolic genetic correlations
revealed in genome-wide association study of anorexia Nervosa. The
American Journal of Psychiatry,174(9), 850–858. https://doi.org/10.1176/
appi.ajp.2017.16121402.
Etain, B., Jamain, S., Milhiet, V., Lajnef, M., Boudebesse, C., Dumaine, A., …
Bellivier, F. (2014). Association between circadian genes, bipolar disorders
and chronotypes. Chronobiology International,31(7), 807–814. https://doi.
org/10.3109/07420528.2014.906445.
Fromer, M., Roussos, P., Sieberts, S. K., Johnson, J. S., Kavanagh, D. H.,
Perumal, T. M., …Sklar, P. (2016). Gene expression elucidates functional
impact of polygenic risk for schizophrenia. Nature Neuroscience,19(11),
1442–1453. https://doi.org/10.1038/nn.4399.
Fuchs, V., Kurppa, K., Huhtala, H., Collin, P., Mäki, M., & Kaukinen, K.
(2014). Factors associated with long diagnostic delay in celiac disease.
Scandinavian Journal of Gastroenterology,49(11), 1304–1310. https://doi.
org/10.3109/00365521.2014.923502.
Gamazon, E. R., Wheeler, H. E., Shah, K. P., Mozaffari, S. V.,
Aquino-Michaels, K., Carroll, R. J., …Im, H. K. (2015). A gene-based asso-
ciation method for mapping traits using reference transcriptome data.
Nature Genetics,47(9), 1091–1098. https://doi.org/10.1038/ng.3367.
Gibson, D., & Mehler, P. S. (2019). Anorexia nervosa and the immune system
–A narrative review. Journal of Clinical Medicine,8(11), 1915. https://doi.
org/10.3390/jcm8111915.
Hedman, A., Breithaupt, L., Hübel, C., Thornton, L. M., Tillander, A., Norring,
C., …Bulik, C. M. (2019). Bidirectional relationship between eating disor-
ders and autoimmune diseases. Journal of Child Psychology and Psychiatry,
60(7), 803–812. https://doi.org/10.1111/jcpp.12958.
Herpertz, S., Albers, N., Wagner, R., Pelz, B., Köpp, W., Mann, K., …
Hebebrand, J. (2000). Longitudinal changes of circadian leptin, insulin
and cortisol plasma levels and their correlation during refeeding in patients
with anorexia nervosa. European Journal of Endocrinology,142(4), 373–379.
https://doi.org/10.1530/eje.0.1420373.
Hoffman, G. E., Bendl, J., Voloudakis, G., Montgomery, K. S., Sloofman, L.,
Wang, Y.-C., …Roussos, P. (2019). CommonMind Consortium provides
transcriptomic and epigenomic data for schizophrenia and bipolar dis-
order. Scientific Data,6(1), 180. https://doi.org/10.1038/s41597-019-
0183-6.
Psychological Medicine 13
https://doi.org/10.1017/S0033291721004554 Published online by Cambridge University Press
Hübel, C., Abdulkadir, M., Herle, M., Loos, R. J. F., Breen, G., Bulik, C. M., &
Micali, N. (2021). One size does not fit all. Genomics differentiates among
anorexia nervosa, bulimia nervosa, and binge-eating disorder. The
International Journal of Eating Disorders,54(5), 785–793. https://doi.org/
10.1002/eat.23481.
Hübel, C., Gaspar, H. A., Coleman, J. R. I., Finucane, H., Purves, K. L.,
Hanscombe, K. B., …Breen, G. (2019a). Genomics of body fat percentage
may contribute to sex bias in anorexia nervosa. American Journal of Medical
Genetics. Part B, Neuropsychiatric Genetics,180(6), 428–438. https://doi.org/
10.1002/ajmg.b.32709.
Hübel, C., Gaspar, H. A., Coleman, J. R. I., Hanscombe, K. B., Purves, K.,
Prokopenko, I., …Breen, G. (2019b). Genetic correlations of psychiatric
traits with body composition and glycemic traits are sex- and age-
dependent. Nature Communications,10(1), 5765. https://doi.org/10.1038/
s41467-019-13544-0.
Huckins, L. M., Dobbyn, A., McFadden, W., Wang, W., Ruderfer, D. M.,
Hoffman, G., …Stahl, E. A. (2017). Transcriptomic imputation of bipolar
disorder and bipolar subtypes reveals 29 novel associated genes. BioRxiv.
https://doi.org/10.1101/222786.
Huckins, L. M., Dobbyn, A., Ruderfer, D. M., Hoffman, G., Wang, W.,
Pardiñas, A. F., …Stahl, E. A. (2019). Gene expression imputation across
multiple brain regions provides insights into schizophrenia risk. Nature
Genetics,51(4), 659–674. https://doi.org/10.1038/s41588-019-0364-4.
Ilyas, A., Hübel, C., Stahl, D., Stadler, M., Ismail, K., Breen, G., …Kan, C.
(2019). The metabolic underpinning of eating disorders: A systematic
review and meta-analysis of insulin sensitivity. Molecular and Cellular
Endocrinology,497, 110307. https://doi.org/10.1016/j.mce.2018.10.005.
Karatsoreos, I. N. (2014). Links between circadian rhythms and psychiatric dis-
ease. Frontiers in Behavioral Neuroscience,8, 162. https://doi.org/10.3389/
fnbeh.2014.00162.
Kichaev, G., Bhatia, G., Loh, P.-R., Gazal, S., Burch, K., Freund, M. K., …Price,
A. L. (2019). Leveraging polygenic functional enrichment to improve
GWAS power. American Journal of Human Genetics,104(1), 65–75.
https://doi.org/10.1016/j.ajhg.2018.11.008.
Kripke, D. F., Mullaney, D. J., Atkinson, M., & Wolf, S. (1978). Circadian
rhythm disorders in manic-depressives. Biological Psychiatry,13(3), 335–
351.
Leppert, B., Millard, L. A. C., Riglin, L., Davey Smith, G., Thapar, A., Tilling,
K., …Stergiakouli, E. (2020). A cross-disorder PRS-pheWAS of 5 major
psychiatric disorders in UK biobank. PLoS Genetics,16(5), e1008185.
https://doi.org/10.1371/journal.pgen.1008185.
Li, B., Mao, Y., Wang, Z., Chen, Y., Wang, Y., Zhai, C., …Chen, J. (2018).
CLIC1 promotes the progression of gastric cancer by regulating the
MAPK/AKT pathways. Cellular Physiology and Biochemistry,46(3), 907–
924. https://doi.org/10.1159/000488822.
Lonsdale, J., Thomas, J., Salvatore, M., Phillips, R., Lo, E., Shad, S., …Moore,
H. F. (2013). The genotype-tissue expression (GTEx) project. Nature
Genetics,45(6), 580–585. https://doi.org/10.1038/ng.2653.
Ludvigsson, J. F., & Murray, J. A. (2019). Epidemiology of celiac disease.
Gastroenterology Clinics of North America,48(1), 1–18. https://doi.org/10.
1016/j.gtc.2018.09.004.
Mansour, H. A., Talkowski, M. E., Wood, J., Chowdari, K. V., McClain, L.,
Prasad, K., …Nimgaonkar, V. L. (2009). Association study of 21 circadian
genes with bipolar I disorder, schizoaffective disorder, and schizophrenia.
Bipolar Disorders,11(7), 701–710. https://doi.org/10.1111/j.1399-5618.
2009.00756.x.
Marchese, S., Cancelmo, L., Diab, O., Cahn, L., Aaronson, C., Daskalakis, N. P.,
…Feder, A. (2021). Altered gene expression and PTSD symptom dimen-
sions in world trade center responders. MedRxiv.https://doi.org/10.1101/
2021.03.05.21252989.
Mårild, K., Størdal, K., Bulik, C. M., Rewers, M., Ekbom, A., Liu, E., …
Ludvigsson, J. F. (2017). Celiac disease and anorexia nervosa: A nation-
wide study. Pediatrics,139(5), 1–9. https://doi.org/10.1542/peds.2016-
4367.
Melo, M. C. A., Abreu, R. L. C., Linhares Neto, V. B., de Bruin, P. F. C., & de
Bruin, V. M. S. (2017). Chronotype and circadian rhythm in bipolar dis-
order: A systematic review. Sleep Medicine Reviews,34,46–58. https://doi.
org/10.1016/j.smrv.2016.06.007.
Meyrer, R., Demling, J., Kornhuber, J., & Nowak, M. (2009). Effects of night
shifts in bipolar disorders and extreme morningness. Bipolar Disorders,
11(8), 897–899. https://doi.org/10.1111/j.1399-5618.2009.00767.x.
Mitchell, J. E., & Peterson, C. B. (2020). Anorexia Nervosa. The New England
Journal of Medicine,382(14), 1343–1351. https://doi.org/10.1056/
NEJMcp1803175.
Munn-Chernoff, M. A., Johnson, E. C., Chou, Y.-L., Coleman, J. R. I.,
Thornton, L. M., Walters, R. K., …Agrawal, A. (2021). Shared genetic
risk between eating disorder- and substance-use-related phenotypes:
Evidence from genome-wide association studies. Addiction Biology,26
(e12880), 1–20. https://doi.org/10.1111/adb.12880.
Murray, G., Allen, N. B., Trinder, J., & Burgess, H. (2002). Is weakened circa-
dian rhythmicity a characteristic of neuroticism? Journal of Affective
Disorders,72(3), 281–289. https://doi.org/10.1016/s0165-0327(01)00465-7.
Pendergrass, S. A., Brown-Gentry, K., Dudek, S. M., Torstenson, E. S., Ambite,
J. L., Avery, C. L., …Ritchie, M. D. (2011). The use of phenome-wide asso-
ciation studies (pheWAS) for exploration of novel genotype-phenotype rela-
tionships and pleiotropy discovery. Genetic Epidemiology,35(5), 410–422.
https://doi.org/10.1002/gepi.20589.
Purcell, S., Neale, B., Todd-Brown, K., Thomas, L., Ferreira, M. A. R., Bender,
D., …Sham, P. C. (2007). PLINK: A tool set for whole-genome association
and population-based linkage analyses. American Journal of Human
Genetics,81(3), 559–575. https://doi.org/10.1086/519795.
Raevuori, A., Haukka, J., Vaarala, O., Suvisaari, J. M., Gissler, M., Grainger, M.,
…Suokas, J. T. (2014). The increased risk for autoimmune diseases in
patients with eating disorders. PLoS One,9(8), e104845. https://doi.org/
10.1371/journal.pone.0104845.
Riedlinger, C., Schmidt, G., Weiland, A., Stengel, A., Giel, K. E., Zipfel, S., …
Mack, I. (2020). Which symptoms, complaints and complications of the
gastrointestinal tract occur in patients with eating disorders? A systematic
review and quantitative analysis. Frontiers in Psychiatry,11, 195. https://
doi.org/10.3389/fpsyt.2020.00195.
Ripke, S., Neale, B. M., Corvin, A., Walters, J. T. R., Farh, K.-H., Holmans, P.
A., …O’Donovan, M. C. (2014). Biological insights from 108
schizophrenia-associated genetic loci. Nature,511(7510), 421–427. https://
doi.org/10.1038/nature13595.
Sekar, A., Bialas, A. R., de Rivera, H., Davis, A., Hammond, T. R., Kamitaki, N.,
…McCarroll, S. A. (2016). Schizophrenia risk from complex variation of
complement component 4. Nature,530(7589), 177–183. https://doi.org/10.
1038/nature16549.
Shroff, H., Reba, L., Thornton, L. M., Tozzi, F., Klump, K. L., Berrettini, W. H.,
…Bulik, C. M. (2006). Features associated with excessive exercise in women
with eating disorders. The International Journal of Eating Disorders,39(6),
454–461. https://doi.org/10.1002/eat.20247.
Smoller, J. W. (2018). The use of electronic health records for psychiatric phe-
notyping and genomics. American Journal of Medical Genetics. Part B,
Neuropsychiatric Genetics,177(7), 601–612. https://doi.org/10.1002/ajmg.b.
32548.
Solmi, M., Veronese, N., Correll, C. U., Favaro, A., Santonastaso, P., Caregaro,
L., …Stubbs, B. (2016). Bone mineral density, osteoporosis, and fractures
among people with eating disorders: A systematic review and meta-analysis.
Acta Psychiatrica Scandinavica,133(5), 341–351. https://doi.org/10.1111/
acps.12556.
Stahl, E. A., Breen, G., Forstner, A. J., McQuillin, A., Ripke, S., Trubetskoy, V.,
…Sklar, P. (2019). Genome-wide association study identifies 30 loci asso-
ciated with bipolar disorder. Nature Genetics,51(5), 793. https://doi.org/10.
1038/s41588-019-0397-8.
Stephan, A. H., Barres, B. A., & Stevens, B. (2012). The complement system:
An unexpected role in synaptic pruning during development and disease.
Annual Review of Neuroscience,35, 369–389. https://doi.org/10.1146/
annurev-neuro-061010-113810.
Strober, M., Freeman, R., & Morrell, W. (1997). The long-term course of severe
anorexia nervosa in adolescents: Survival analysis of recovery, relapse, and
outcome predictors over 10–15 years in a prospective study. International
Journal of Eating Disorders,22(4), 339–360. https://doi.org/10.1002/(SICI)
1098-108X(199712)22:4<339::AID-EAT1>3.0.CO;2-N.
Suleiman, L., Négrier, C., & Boukerche, H. (2013). Protein S: A multifunc-
tional anticoagulant vitamin K-dependent protein at the crossroads of
14 Jessica S. Johnson et al.
https://doi.org/10.1017/S0033291721004554 Published online by Cambridge University Press
coagulation, inflammation, angiogenesis, and cancer. Critical Reviews in
Oncology/Hematology,88(3), 637–654. https://doi.org/10.1016/j.critrevonc.
2013.07.004.
The Autism Spectrum Disorders Working Group of The Psychiatric Genomics
Consortium. (2017). Meta-analysis of GWAS of over 16000 individuals with
autism spectrum disorder highlights a novel locus at 10q24.32 and a signifi-
cant overlap with schizophrenia. Molecular Autism,8(1), 21. https://doi.org/
10.1186/s13229-017-0137-9.
Watanabe, K., Taskesen, E., van Bochoven, A., & Posthuma, D. (2017).
Functional mapping and annotation of genetic associations with FUMA.
Nature Communications,8(1), 1826. https://doi.org/10.1038/s41467-017-
01261-5.
Watson, H. J., & Bulik, C. M. (2013). Update on the treatment of anorexia ner-
vosa: Review of clinical trials, practice guidelines and emerging interven-
tions. Psychological Medicine,43(12), 2477–2500. https://doi.org/10.1017/
S0033291712002620.
Watson, H. J., Yilmaz, Z., Thornton, L. M., Hübel, C., Coleman, J. R. I., Gaspar,
H. A., …Bulik, C. M. (2019). Genome-wide association study identifies eight
risk loci and implicates metabo-psychiatric origins for anorexia nervosa. Nature
Genetics,51(8), 1207–1214. https://doi.org/10.1038/s41588-019-0439-2.
Watson, Hunna J, Yilmaz, Zeynep, Thornton, Laura M., Hübel, Christopher,
Coleman, Jonathan R. I, Gaspar, Héléna A., …Bulik, Cynthia M. (2019).
Genome-wide association study identifies eight risk loci and implicates
metabo-psychiatric origins for anorexia nervosa. Nature Genetics,51(8),
1207–1214. http://dx.doi.org/10.1038/s41588-019-0439-2.
Wiklund, C. A., Kuja-Halkola, R., Thornton, L. M., Hübel, C., Leppä, V., &
Bulik, C. M. (2019). Prolonged constipation and diarrhea in childhood
and disordered eating in adolescence. Journal of Psychosomatic Research,
126, 109797. https://doi.org/10.1016/j.jpsychores.2019.109797.
Willer, C. J., Li, Y., & Abecasis, G. R. (2010). METAL: Fast and efficient
meta-analysis of genomewide association scans. Bioinformatics,26(17),
2190–2191. https://doi.org/10.1093/bioinformatics/btq340.
Wu, P., Gifford, A., Meng, X., Li, X., Campbell, H., Varley, T., …Wei, W.-Q.
(2019). Mapping ICD-10 and ICD-10-CM codes to phecodes: Workflow
development and initial evaluation. JMIR Medical Informatics,7(4),
e14325. https://doi.org/10.2196/14325.
Xu, J., Johnson, J. S., Birgegård, A., Jordan, J., Kennedy, M. A., Landén, M., …
Huckins, L. M. (2021). Exploring the clinical consequences and genetic aeti-
ology of adult weight trajectories. MedRxiv.https://doi.org/10.1101/2021.10.
04.21264526.
Yang, X., Sun, J., Gao, Y., Tan, A., Zhang, H., Hu, Y., …Mo, Z. (2012).
Genome-wide association study for serum complement C3 and C4 levels
in healthy Chinese subjects. PLoS Genetics,8(9), e1002916. https://doi.
org/10.1371/journal.pgen.1002916.
Yilmaz, Z., Hardaway, J. A., & Bulik, C. M. (2015). Genetics and epigenetics of
eating disorders. Advances in Genomics and Genetics,5, 131–150. https://
doi.org/10.2147/AGG.S55776.
Zerwas, S., Larsen, J. T., Petersen, L., Thornton, L. M., Quaranta, M., Koch, S.
V., …Bulik, C. M. (2017). Eating disorders, autoimmune, and autoinflam-
matory disease. Pediatrics,140(6), e20162089. https://doi.org/10.1542/peds.
2016-2089.
Zheutlin, A. B., Dennis, J., Karlsson Linnér, R., Moscati, A., Restrepo, N.,
Straub, P., …Smoller, J. W. (2019). Penetrance and pleiotropy of polygenic
risk scores for schizophrenia in 106 160 patients across four health care sys-
tems. American Journal of Psychiatry,176(10), 846–855. https://doi.org/10.
1176/appi.ajp.2019.18091085.
Zhu, Z., Zhu, X., Liu, C.-L., Shi, H., Shen, S., Yang, Y., …Liang, L. (2019).
Shared genetics of asthma and mental health disorders: A large-scale
genome-wide cross-trait analysis. European Respiratory Journal,54(6), 1–
12. https://doi.org/10.1183/13993003.01507-2019.
Zipfel, S., Giel, K. E., Bulik, C. M., Hay, P., & Schmidt, U. (2015).
Anorexia nervosa: Aetiology, assessment, and treatment. The Lancet.
Psychiatry,2(12), 1099–1111. https://doi.org/10.1016/S2215-0366(15)
00356-9.
Zucker, N. L., & Bulik, C. M. (2020). On bells, saliva, and abdominal pain or
discomfort: Early aversive visceral conditioning and vulnerability for anor-
exia nervosa. International Journal of Eating Disorders,53(4), 508–512.
https://doi.org/10.1002/eat.23255.
Psychological Medicine 15
https://doi.org/10.1017/S0033291721004554 Published online by Cambridge University Press
