Anorexia nervosa (AN) is a debilitating and often refractory eating disorder that is unique among psychiatric disorders insofar as nutrition is key to recovery. Treatment options and efficacy are limited with no approved medications for AN. Genetic studies are clarifying the etiology of AN, with the goal of eventually informing the development of innovative personalized pharmacologic, nutritional, microbial, and behavioral interventions. We present the current state of genome-wide and epigenome-wide association studies, gut microbiome research, and functional genomics investigations and discuss translating this knowledge into clinical practice.
Key points
- •
Both genetic factors and environmental factors influence the risk of developing anorexia nervosa (AN).
- •
A polygenic risk score cannot predict whether a person will develop AN.
- •
Functional genomic studies will help us understand biological mechanisms underlying AN-associated genes to inform therapeutic development.
- •
Epigenetics and gut microbiome studies can provide insights into AN development and maintenance and contribute to biomarker identification for AN prognosis.
- •
The registered dietitian is an integral member of the AN treatment team and will benefit from genetic studies to eventually develop and implement personalized interventions.
| AN | anorexia nervosa |
| BMI | body mass index |
| ESRRA | estrogen-related receptor <SPAN role=presentation tabIndex=0 id=MathJax-Element-1-Frame class=MathJax style="POSITION: relative" data-mathml='α’>𝛼α α |
| EWAS | epigenome-wide association study |
| GWAS | genome-wide association study |
| HDAC4 | histone deacetylase 4 |
| IBD | inflammatory bowel disease |
| iPSC | induced pluripotent stem cells |
| MTHFR | methylenetetrahydrofolate reductase |
| NNAT | neuronatin |
| PRS | polygenic risk scores |
| RDs | registered dietitians |
| T2D | type 2 diabetes |
| WES | whole-exome sequencing |
| WGS | whole-genome sequencing |
Background
Anorexia nervosa (AN) is an eating disorder (ED) marked by extreme weight loss (or failure to gain weight), dangerously low weight ( Box 1 ), a seeming inability to recognize the seriousness of the condition, and a fear of weight gain. AN is subclassified into (1) restricting subtype (anorexia nervosa [AN-R]) in which weight is lost or low weight maintained solely through energy restriction and/or excessive exercise, and (2) binge eating/purging subtype (anorexia nervosa [AN-BP]) in which energy restriction is coupled with regular binge eating and/or purging behavior (eg, self-induced vomiting and laxatives). AN occurs in up to 4% of female individuals and 0.3% of male individuals, with peak onset during adolescence (although onset can occur at any age).
Low body weight is pathognomonic of AN. A controversial diagnostic label (atypical AN) purportedly captures individuals who meet all of the criteria of AN except low body weight. The problems with both the label and the conceptualization of atypical AN have been pointed out. Genetic research may play a central role in determining a fundamental difference between individuals with the behavioral and cognitive features of AN who do and do not attain dangerously low body weights. The term “atypical” can lead to the unfortunate perception of an illness being not sufficiently severe to warrant treatment. This is not the case as individuals with this presentation have many of the same somatic consequences as those with AN have in terms of physical symptoms (eg, malnutrition, amenorrhea, hypotension, low body temperature, and slow heart rate). One hypothesis is that individuals with AN and the “atypical” presentation may share the psychiatric genetic underpinnings of AN but differ on the anthropometric/metabolic genetic risk factors.
AN has adverse effects on almost every bodily system and carries significantly elevated mortality risk and risk of death by suicide. Treatment is only moderately effective, especially in adults, and there is no Food and Drug Administration-approved medication. , For young people, family-based treatment is the evidence-based treatment of choice, although access to specialist therapists is limited and it is not effective for all individuals or appropriate for all families. Relapse is common, with weight often returning to, or exceeding, previous nadirs.
Genetic research on anorexia nervosa
Like other psychiatric disorders, AN is a “complex trait” with polygenic underpinnings, meaning that risk is accumulated across hundreds or even thousands of common disease-associated alleles, all with small effect sizes. These risk alleles may act through influencing gene expression or other functional effects on higher order biology, and the effects may be either risk increasing or protective. Importantly, genetic risk factors do not act alone, but rather act in concert with environmental factors. An individual with fairly high genetic risk may never develop AN if they are not exposed to environmental risk factors such as dieting or teasing; on the other hand, a person with fairly low genetic risk might be barraged by environmental factors that push them over the threshold for developing the illness. To fully understand how genetics and environment interact to influence risk for AN, one has to embrace complexity and probability.
Early research on AN genetics began with family and twin studies. The relative risk of developing AN in female relatives of individuals with AN was reported to be 11.3, compared to female relatives of controls. Twin studies estimated AN heritability to be between 22% and 74%, with unique environmental factors accounting for the remaining variance in liability and little or no contribution from shared environmental factors (0%–14%). Consistent replication of this pattern of results transformed beliefs about causes of AN from a focus on sociocultural and family factors, to ensuring that genetics was an integral component of causal models.
Genome-Wide Era
Genome-wide association study
The current status of genome-wide association study (GWAS) for AN stands on the shoulders of several earlier small studies. Despite low power, these samples yielded early evidence for a polygenic etiology for AN, and laid the foundation for later global collaborative studies. The latest published AN GWAS included 16,992 individuals with AN and 55,525 controls. Although still fairly small, results were nonetheless intriguing, identifying 8 significant genomic regions (or loci) that were associated with AN, several of which had been implicated in previous GWAS of metabolic/anthropometric, neuropsychiatric, autoimmune, and sex hormone phenotypes.
Genetic correlations
Another readout of GWAS is “genetic correlations.” These quantify the extent to which the genetic factors that influence AN also influence other GWAS traits. A positive correlation means that the same genetic risk factors are operating across 2 traits in the same direction (eg, increasing risk of both traits), while a negative correlation means that genetic risk factors are protective for one trait, while increasing risk for the other. Some observed genetic correlations were expected; for example, AN was positively genetically correlated with other psychiatric disorders like obsessive compulsive disorder, depression, anxiety, and schizophrenia. These results tell us that what we see clinically (frequent co-occurrence of AN and other psychiatric disorders) may be because some of the same genetic variants that increase the risk for AN also increase the risk for these other psychiatric disorders.
Another positive genetic correlation was between AN and physical activity. One common and difficult to treat characteristic of AN is driven exercise and generally high activity, which is often attributed to willful movement to burn calories or to stay warm. This result suggests that people with AN may be genetically predisposed to high activity, which could partially explain why it is such a difficult symptom to control.
Finally, we identified an unexpected array of negative genetic correlations between AN and anthropometric (body measurement) and metabolic traits. AN was negatively genetically correlated with body mass index, obesity, fasting insulin, body fat percentage, and type 2 diabetes (T2D). This means that many of the same genetic factors that increase the risk for AN protect against developing obesity, high body fat, and T2D, among other metabolic risk factors. The transformative aspect of these findings was that from a genetic perspective, not only is AN psychiatric disorder but also appears to have fundamental metabolic underpinnings that could contribute to the inexplicable ability to attain extremely low body weights and the challenges with regaining and maintaining restored weight. This opens new avenues for treatment development that focus on metabolism. Valuable confirmatory evidence emerged from a study of individuals in large biobanks who did not have AN (eg, Bio Me Biobank at Mount Sinai in New York, UK Biobank) but had a higher genetic risk of AN. These individuals, who did not have a recorded AN diagnosis, but who had higher genetic risk of AN, were more likely to have a weight loss trajectory over time, providing further evidence that genetic components of AN reflect not only psychiatric traits but also weight regulation.
Polygenic risk scores
GWAS enables calculation of a metric called polygenic risk scores (PRS) that captures the load of disease-associated genetic risk alleles that an individual carries for a particular disease. Notably, a PRS is not a genetic test for a trait or illness. In fact, as illustrated in Fig. 1 , there are many people with high AN PRS who never develop the disease, and there are many people with fairly low AN PRS who do develop the disease. An important research question is what factors prevent an individual with a high AN PRS from developing AN. These may be other genetic factors, protective environmental factors, or even epigenetic factors. This complexity is precisely why PRS cannot be considered to be a “genetic test” for AN.
Variants: from Common to Rare
GWAS typically include variants that are commonly observed in the population and that have only small effects on disease risk. To identify variants with large effects on disease risk, researchers must query rare variants. Costly approaches such as whole-exome sequencing (WES) or whole-genome sequencing (WGS) read individual genotype at every single base pair location for all protein-coding regions (1% of the genome) or the whole genome (approximately 3 billion base pairs!). No large-scale whole genome sequencing study for AN exists yet. However, a few smaller studies have been reported.
Lombardi and colleagues performed a WES study in 2 families of male patients with AN ( Box 2 ) and identified the neuronatin ( NNAT ) gene as a potential gene of interest. Low-frequency risk variants identified in this gene were replicated by another group via sequencing. The NNAT gene has been previously implicated in body mass index (BMI) and adipose accumulation in mouse models, making it a potential metabolism-related gene underlying AN.
Although AN afflicts female individuals far more often than male individuals (ratio 9:1), we do not know why. We presume that multiple factors are operative, including cultural norms, hormones, gut microbiota, and genetics. For example, Hübel and colleagues reported a negative genetic correlation between body fat and AN that was markedly stronger in female individuals than male individuals (−0.44 vs −0.26), indicating that some of the same genetic factors affecting body fat regulation may have a stronger influence on AN etiology in female individuals than in male individuals. Although the effect of AN genetic risk modeled by the polygenic risk score does not appear to differ by sex, , specific genetic domains (eg, anthropometric/metabolic domain, rather than the psychiatric domain) could contribute to the sex difference observed in AN. A major hurdle to uncovering sex differences is small samples of male individuals with AN in current genetic studies (eg, only 3% of studied samples in the latest AN GWAS). Inclusion of more male participants in future genetic studies is essential to understand sex differences.
Only one WGS study in patients with AN has been reported in a 20 person family. One rare missense mutation on the estrogen-related receptor <SPAN role=presentation tabIndex=0 id=MathJax-Element-2-Frame class=MathJax style="POSITION: relative" data-mathml='α’>𝛼α
α
( ESRRA ) gene was identified. WES of another included family identified a rare missense variant in the histone deacetylase 4 ( HDAC4 ) gene. This gene encodes the HDAC4 protein, which is a repressor of transcription factors including ESRRA. Overall, this study suggested the involvement of estrogen in AN etiology as well as its potential mechanism through transcription factors and gene expression regulation.
Beyond Genome-Wide Association Study
Although GWAS is a powerful tool to identify disease-associated genetic loci, what do we do with these loci once identified? To translate genetic variants into clinically actionable targets, we must understand which genes these variants affect, and when and where the effects occur (eg, in which cells and in which tissues). Mapping these variants to specific genes with precise location and timing, in theory, will facilitate the development of effective pharmacologic targets while minimizing off-target effects.
Genetic variants may impact expression of a phenotype (here, AN risk) through their impact on gene expression. Genetic (dys) regulation of gene expression is well-studied. Researchers seek to identify risk variants that are associated with dysregulated gene expression in particular tissues or cell types of interest. We term these variants expression quantitative trait loci. , These may be used either to annotate GWAS variants with functional data or to provide predicted gene expression estimates in disease-relevant tissues.
Using this approach on the latest AN GWAS, Johnson and colleagues identified 53 AN genes expressed in brain and other tissues with known roles in metabolic, gastrointestinal, and autoimmune traits, including blood protein levels, body fat distribution, and inflammatory bowel disease (IBD). Taking IBD as an example, these shared genes between AN and IBD may contribute to the development of IBD in patients with AN, which has been observed in epidemiologic studies. , Moreover, by calculating genetically predicted expressions for these same genes in an independent patient cohort, Johnson and colleagues showed that patients with higher levels of these genes also had higher risks for conditions comorbid with AN, including abdominal pain and peptic ulcer.
To expand our understanding of AN development beyond statistical inference and toward causal evidence, it is vital that researchers test the effects of putatively involved genes in animal and human cell models. To date, several genes have been tested in mouse models, including brain derived neurotrophic factor ( BDNF ), sirtuin 1 ( SIRT1 ), solute carrier family 32 member 1 ( SLC32A1 ), solute carrier family 18 member A3 ( SLC18A3 ) , tet methylcytosine dioxygenase 3 ( TET3 ), and HDAC4 . Notably, HDAC4 was identified in a previously described family study of AN ; a transgenic mouse study that modeled a mutation on HDAC4 demonstrated an associated loss of interest in high-fat food.
To study cell-type-specific mechanisms in AN, human induced pluripotent stem cells (iPSC) can be used to model various brain cell types, such as neurons, microglia, and astrocytes. Further, genome-editing technologies, such as clustered regularly interspaced short palindromic repeats (CRISPR) and CRISPR-associated protein 9 (CRISPR-Cas9), can be coupled with iPSC-derived cells to explore molecular and cellular consequences of target genes. iPSC-derived cells with patient genetic background also present a unique opportunity for future drug screening. Negraes and colleagues demonstrated that AN can be modeled in a dish. This field is currently growing and holds great potential in understanding AN etiology to inform therapeutic development and future personalized interventions. A recent review by Huckins and colleagues has put forth a battery of experimental genomic techniques that can be employed in human iPSC models to study AN-associated genes.
From Genetics to Epigenetics
Epigenetics studies how the environment can cause gene expression changes without direct changes in our DNA sequence through mechanisms including DNA methylation and histone modifications. Epigenetic changes are heritable but can be altered by the environment. Therefore, whereas our genetic information remains largely the same throughout life (without considering a very low rate of mutations), epigenetic studies provide insight into the dynamic relationship between genetic risk factors and AN in the context of changes in food intake or exercise intensity. Therefore, the study of epigenetics could identify longitudinal changes over the disease course that studies of static DNA cannot provide.
The largest epigenome-wide association study (EWAS) of AN included only 194 patients with AN and 64 controls, substantially smaller than the current GWAS (16,992 cases and 55,525 controls). Among these patients with AN, 49 were remitted for 1 year, enabling investigation of DNA methylation changes before and after recovery. Notably, identified genes in the EWAS study converge with those identified in the published GWAS for gene function in the nervous, metabolic, and immune systems, suggesting an underlying psycho-metabo-immuno etiology of AN. Future studies combining signatures from both GWAS and EWAS may help identify AN subtypes with distinct etiologic roots.
Another striking finding in this EWAS was, for recovered patients, none of their AN-related methylation changes differed from controls, indicating reversibility of methylation alterations with AN remission. However, given the small sample size, replication with larger sample size is essential. Nevertheless, this suggests that epigenetic markers, such as DNA methylation, may serve as biomarkers for AN prognosis or treatment response.
Nutrigenetics and Nutrigenomics
EDs are unique among psychiatric disorders insofar as nutrition is central to both pathology and treatment. Nutrigenetics considers genetic variations that alter the individual physiologic response to nutrient intake, whereas nutrigenomics focuses on how nutrients modulate human gene expression. One example of nutrigenetics is the increased folate requirement conferred by polymorphisms in the methylenetetrahydrofolate reductase ( MTHFR ) gene. Given folate’s role in the synthesis of neurotransmitters such as dopamine that are involved in psychiatric disorders, MTHFR polymorphisms have been studied for their role in depression, schizophrenia, and bipolar disorder, among others.
An example of nutrigenomics is the binding of micronutrients (eg, vitamin D) to promotor regions on DNA to initiate the transcription of genes essential for bodily functions, such as calcium metabolism and bone health. Nutrigenomic mechanisms can also act through epigenetic modifications mentioned earlier. Further research in this area of research for AN may have direct translational potential into clinical practice given the accessible and modifiable nature of nutritional interventions.
From Host Genetics to Microbial Genetics
Another area that can be important to EDs is the gut microbiota—a collective term for all microorganisms living in the human intestine—that can be heavily influenced by dietary changes. The gut microbiome refers to genetic material of these microbes. The documented involvement of gut microbiota in host weight regulation has inspired the study of gut microbiota in EDs. Similar to human GWAS and EWAS, most gut microbiome studies have focused on AN, and most measure relative abundances of microbial taxa per se, with only 2 exploring the function of microbial genetics. , Altered gut microbial composition has been frequently observed in patients with AN. In a study by Fouladi and colleagues, inpatient treatment partially normalized the gut microbiota in patients with AN, and the elevated microbial genetic capacity of fermentation observed in patients with AN was restored to control levels after therapeutic renourishment. This study showed that diversity in microbial genetics, in particular the metabolic genetic pathways, differentiated AN inpatients at admission, at discharge, and non-AN controls, whereas diversity in microbial abundance alone could not, highlighting the importance of studying microbial genetics and its potential as a prognostic biomarker. Further, as much of the variation in gut microbiota composition is individual-specific, future research effort in integrating individual microbial and genetic information to develop personalized treatment plans is warranted to improve AN treatment efficacy, given that an average of 1 out of 3 patients with AN may experience relapse and typically within the first year after treatment.
Clinical implications
How to Apply Genetic Knowledge in Clinics to Benefit Patients and Their Families?
Although genetic research on AN is very much a work in progress, results can have immediate benefits for clinicians, families, and individuals with AN ( Fig. 2 ). Table 1 includes a list of questions that genetic knowledge can be applied to address patient and family concerns and ultimately improve treatment outcomes. AN (and other EDs) are unique in that nutrition plays a central role in treatment and recovery and registered dietitians (RDs) are essential members of the treatment team. Accordingly, we focus on the critical role that RD plays in the person-centered treatment of AN.
