Understanding anhedonia from a genomic perspective

Abstract

Anhedonia, or the decreased ability to experience pleasure, is a cardinal symptom of major depression that commonly occurs within other forms of psychopathology. Supportive of long-held theory that anhedonia represents a genetically-influenced vulnerability marker for depression, evidence from twin studies suggests that it is moderately-largely heritable. However, the genomic sources of this heritability are just beginning to be understood. In this review, we survey what is known about the genomic architecture underlying anhedonia and related constructs. We briefly review twin and initial candidate gene studies before focusing on genome-wide association study (GWAS) and polygenic efforts. As large samples are needed to reliably detect the small effects that typically characterize common genetic variants, the study of anhedonia and related phenotypes conflicts with current genomic research requirements and frameworks that prioritize sample size over precise phenotyping. This has resulted in few and underpowered studies of anhedonia-related constructs that have largely failed to reliably identify individual variants. Nonetheless, the polygenic architecture of anhedonia-related constructs identified in these studies has genetic overlap with depression and schizophrenia as well as related brain structure (e.g., striatal volume), providing important clues to etiology that may usefully guide refinement in nosology. As we await the accumulation of larger samples for more well powered GWAS of reward-related constructs, novel analytic techniques that leverage GWAS summary statistics (e.g., genomic structural equation modeling) may currently be used to help characterize how the genomic architecture of anhedonia is shared and distinct from that underlying other constructs (e.g., depression, neuroticism, anxiety).

INTRODUCTION

It has long been speculated that psychiatric heterogeneity hinders the identification of etiologic mechanisms and stalls treatment advances (Gottesman & Gould, 2003; Kotov et al., 2017). Plausibly, amalgamations of symptoms that define psychiatric disorders may reflect shared and/or unique complex pathophysiologies that are obfuscated in traditional case/control studies. This is further complicated by common psychiatric comorbidity as well as symptoms, characteristics, and correlates that are often shared across diagnoses (Cross-Disorder Group of the Psychiatric Genomics Consortium, 2019; Radonjić et al., 2021). Concerns that within-disorder heterogeneity and across-disorder similarity have hindered research progress in psychiatry has led to research conceptualizations (e.g., endophenotype, intermediate phenotypes) advocating for the study of more homogenous quantifiable phenotypes whose pathophysiologies and genomic architecture will be presumably less heterogenous and hence more scientifically tractable (Gottesman & Gould, 2003; Hasler et al., 2004; Meyer-Lindenberg & Weinberger, 2006). These conceptual frameworks have inspired a variety of efforts to refine nosology (e.g., hierarchical taxonomies, machine learning-informed approaches; Feczko & Fair, 2020; Kotov et al., 2017) and promoted funding for studies of fundamental dimensions of behaviors such as the U.S. National Institute of Mental Health (NIMH)’s Research Domain Criteria (RDoC) initiative, which is a framework for investigating mental health within a comprehensive consideration of typical and atypical functioning. The RDoC research strategy focuses on major domains of human functioning and dysfunction within these domains associated with psychopathology (for more information on the U.S. NIMH RDoC initiative see: (Insel et al., 2010; Morris & Cuthbert, 2012).

Anhedonia, which reflects reduced pursuit of reward and diminished reactivity to pleasurable stimuli, has long been heralded as a critical neuropsychiatric phenotype for in-depth study that is rooted in deficits across various subcomponents of the Positive Valence Systems (PVS) domain of the RDoC including reward anticipation, response to reward, reward learning, habitual reward, reward delay, and reward effort (Hasler et al., 2004; Kendler, 2017; Morris & Cuthbert, 2012; addition information regarding RDoC PVS domains can be found https://www.nimh.nih.gov/research/research-funded-by-nimh/rdoc/constructs/positive-valence-systems). Anhedonia is a core clinical feature of depression that commonly presents across various other forms of psychopathology including schizophrenia, posttraumatic stress disorder, and substance use disorder. Highlighting its potential import, across psychopathologies, anhedonia has been associated with increased severity and comorbidity as well as reduced response to treatment (Destoop et al., 2019; Kasch et al., 2002; Kendler, 1997). Despite evidence that anhedonia and related constructs are moderately-largely heritable (Hess et al., 2016), the genomic sources of this heritability remain poorly understood. In this chapter, we review contemporary genetic studies of anhedonia and related RDoC PVS components. We begin by broadly contextualizing current research frameworks in psychiatric genetics and how they are and are not consistent with RDoC. Next, we delve into the genetics of anhedonia by first providing a broad overview of heritability estimates derived from twin studies and briefly discussing initial candidate gene studies that have generally not been replicated. We then focus predominantly on emerging evidence from nascent genome-wide association studies (GWASs) of anhedonia and related constructs as well as polygenic scoring approaches. Finally, we discuss challenges associated with genetic work on anhedonia and how addressing them may broadly inform our understanding of anhedonia to ultimately influence psychiatric nosology, treatment, and prevention.

Contemporary Psychiatric Genetics Frameworks: Relevance to RDoC

In contrast to the study of fundamental dimensions of behavior (e.g., RDoC) that have largely eschewed disorder-guided investigation, the field of psychiatric genetics has predominantly focused on case/control studies and easily quantifiable, yet heterogeneous, psychiatrically-relevant phenotypes (e.g., dichotomous responses to “Would you describe yourself as someone who takes risks?”; Strawbridge et al., 2018) in an effort to bolster sample size (Sullivan & Geschwind, 2019). Initial excitement that the mapping of the human genome would quickly identify clinically actionable mechanistic targets and usher in an era of personalized medicine in psychiatry (Craddock & Owen, 1996) was met with largely unreplicable candidate gene research (with notable exceptions in the field of substance use disorders: e.g., ADH1B rs1229984 and alcohol use disorder, OPRM1 rs1799971 and opioid use disorder; Gelernter & Polimanti, 2021) and initial underpowered GWASs that largely produced null associations (e.g., Sullivan et al., 2009). Constrained by high initial costs of molecular genetic research, this early underpowered work attempted to confront these limitations by focusing on intermediate phenotypes (e.g., brain function, structure) consistent with an RDoC perspective (Bogdan et al., 2017); while many intermediate phenotypes (e.g., brain structure, circulating biomarkers) have proven to characterized by larger genetic association estimates than psychiatric diagnoses, effects overall are small and require large samples to reliably detect (Grasby et al., 2019; Ligthart et al., 2018). Following technological advancements and related cost reductions in genome-wide genotyping, as well as sober realizations that associations between single common genetic variation and complex behavioral and biological traits are small in magnitude, the pendulum in psychiatric genetics shifted away from fundamental dimensions of behavior (e.g., RDoC) to focus predominantly on easily assayed and widely available phenotypes (e.g., psychiatric diagnoses, brief widely used questionnaires) that enable the formation of large collaborative datasets and the generation of single large studies (Sullivan & Geschwind, 2019). This transition has prioritized a top-down scientific approach wherein initial GWASs are conducted on higher-order heterogenous constructs with subsequent follow-up studies applying these results to more finely grained behavioral constructs by attempting to define functional consequences of these variants and their associations with intermediate phenotypes, as well as phenotypic associations with their polygenic architecture. While the focus of psychiatric genetics remains on heterogenous clinical constructs, more recently large scale GWASs have begun to address heterogeneity by conducting GWASs of individual disorder symptoms and construct items (e.g., each item contributing to neuroticism) that have revealed both shared and distinct architectures (Mallard et al., 2021; Nagel, Watanabe, et al., 2018). Further, the establishment of large-scale studies that include deep phenotyping (e.g., UK Biobank, ABCD Study, All of Us), has begun to establish large datasets that have assessed RDoC-related constructs across levels of analysis (i.e., behavior, biomarkers, brain; Elliott et al., 2018).

Genetic Studies of Anhedonia and Related Constructs

Twin and Family Studies.

Twin studies, which assume uniform additive effects of segregating loci, have been used for a century to estimate latent sources of genetic influence on psychiatrically-relevant traits. Overall, twin work suggests that the vast majority of complex psychiatric and related traits (e.g., brain structure, personality) have moderate to high heritability (Polderman et al., 2015; ~0.30–0.80). Despite long-held theory that anhedonia may be a genetically influenced vulnerability factor for depression (Loas, 1996; Snaith, 1993), there has been limited research on the heritability of anhedonia and related behavioral constructs, with the exception of self-reported impulsivity and broad spectrum externalizing behavior (Linnér et al., 2020), which will not be considered here. Much like the broader literature of complex psychiatrically-relevant traits, the heritability of self-reported anhedonia (Berenbaum et al., 1990; Berenbaum & McGrew, 1993; Clementz et al., 1991; Dworkin & Saczynski, 1984; Franke et al., 1993; Hay et al., 2001; Heath et al., 1994; Katsanis et al., 1990; Kendler et al., 1991; Thaker et al., 1993) as well as related RDoC-related PVS constructs (e.g., reward learning, delay discounting, corticostriatal brain structure, reward-related ventral striatum activation; Anokhin et al., 2011, 2015; Bogdan & Pizzagalli, 2009; Galimberti et al., 2013; Grimm et al., 2014; Guffanti et al., 2019; Li et al., 2019; Liu et al., 2016; Melhorn et al., 2016) ranges from moderate to large. As the heritability work on these constructs has predominantly been conducted in small samples that produce wide 95% confidence intervals overlapping with one another, it is unclear from this literature to what extent the heritability of distinct anhedonia-related constructs differs from one another, making it difficult to prioritize particular aspects of reward processing for further investigation based on heritability.

Due to aforementioned concerns about psychiatric heterogeneity, some studies have examined the heritability of specific subtypes of depression, based on symptom constellations, comorbid conditions, age of onset, and general outcomes (Harald & Gordon, 2012). Melancholic depression, one of the most widely studied subtypes, is characterized by anhedonia, early morning wakening, weight loss, psychomotor disturbances (agitation or retardation), and excessive guilt (Angst et al., 2007; Parker et al., 2017). Findings regarding the heritability of melancholic depression largely suggest that melancholic depression is less uniquely heritable than other subtypes, such as atypical depression (Kendler, 1997; Klein et al., 2002; Lamers et al., 2016; Maier et al., 1991), and is instead suggestive of higher familial liability for any form of depression.

Candidate Gene Studies.

Similar to RDoC-like approaches, initial hypothesis-driven candidate gene efforts adopted a bottom-up scientific approach by studying single variants with putative functional consequences (e.g., 5-HTTLPR with SLC6A4 (Caspi et al., 2003; Lesch et al., 1996), COMT rs4680 (Egan et al., 2001), MAOA promoter variant (Caspi et al., 2002), ADH1B rs1229984 (Thomasson et al., 1991)) within pathways (e.g., serotonin, dopamine, etc.) known to be associated with psychiatric phenotypes. While some of these variants have been validated using GWASs (e.g., ADH1B rs1229984; Gelernter et al., 2014), the vast majority have not been replicated in the extensive psychiatric genetics literature (Border et al., 2019; Duncan & Keller, 2011; Johnson et al., 2017). Prior candidate gene investigations of single common variants in this context initially reported associations with anhedonia and related constructs (e.g., reward learning, delay discounting, reward-related brain function) in small samples (Corral-Frías et al., 2016; Dillon et al., 2010; Docherty & Sponheim, 2008; Dreher et al., 2009; Grimm et al., 2014; Troisi et al., 2011); however, while some evidence of convergence emerged, there have been few replication attempts and given what has been learned in the broader field of psychiatric genetics, a high degree of skepticism exists.

Today, the vast majority of candidate gene research is done following the identification of variants from a GWAS, consistent with a top-down approach in which variants are first associated with heterogenous psychiatric constructs and then characterized according to their associations with intermediate phenotypes (Bogdan et al., 2018). This work has linked variants associated with depression to anhedonia and related neural constructs (e.g., responsiveness to reward; (e.g.., Wetherill et al., 2019; Lancaster et al., 2019). However, the small effect sizes associated with common single genetic variants and extensive evidence that complex traits are undergirded by an extensive polygenic architecture, this work has largely transitioned to the of polygenic scores that allow for the non-specific aggregation of genomic risk (see Polygenic Studies section below). Further, much like studies of regional differences in neural phenotypes that account for global metrics, it will be important for single variant work to test whether single variant associations exist above and beyond the polygenic architecture.

GWAS.

As the field of psychiatric genetics has prioritized large datasets of heterogeneous clinical constructs (see Contemporary Psychiatric Genetics Frameworks above), it is unsurprising that there have been few GWASs of anhedonia and related constructs and that they have typically been conducted in small samples. Indeed, contrasting recent well-powered investigations of heterogeneous psychiatric constructs including depression (Howard et al., 2019), neuroticism (Nagel, Jansen, et al., 2018), externalizing behavior (Linnér et al., 2020), and risk taking (Karlsson Linnér et al., 2019) that have begun to reliably identify single loci and characterize the broad polygenic architecture of these phenotypes, we are only aware of 11 GWASs assessing anhedonia and/or more precise PVS constructs (Jia et al., 2016; May-Wilson et al., 2021; Ortega-Alonso et al., 2017; Pain et al., 2018; Ren et al., 2018; Sanchez-Roige et al., 2018; Service et al., 2012; Tomppo et al., 2012; Verweij et al., 2010; Ward et al., 2019; Wingo et al., 2017), of which only 3 (Ward et al., 2019; Sanchez-Roige et al, 2018, May-Wilson et al., 2021) have more than 12,000 participants (Table 1), which we highlight here. These anhedonia-related GWASs may be divided into those assessing anhedonia directly in the context of depression or psychosis assessments as well as RDoC PVS domain components, and they have generally been conducted among those of European ancestry (primarily from the United States and United Kingdom), with one study conducted among those with African American ancestry.

Table 1.

GWAS Studies of Anhedonia and related Positive Valence System Constructs

Reference	N	Anhedonia/PVS Phenotype	Significant Loci (#)	Genes identified
Verweij et al., 2010	5,117	Reward dependence	0	N/A
Service et al., 2012	11,000	Reward dependence	0	N/A
(Tomppo et al., 2012)	4,561	Revised Social Anhedonia Scale, Revised Physical Anhedonia Scale	0	N/A
(Jia et al., 2016)	1,544	Neural activity during reward anticipation	0	N/A
(Ortega-Alonso et al., 2017)	4,269	Revised Physical Anhedonia Scale, Revised Social Anhedonia Scale	0	N/A
Wingo et al., 2017	2,522	Positive affect (PANAS)	1	LINC01221
(Pain et al., 2018)	6,579	Psychotic-like experience domains	1	IDO2
(Ren et al., 2018)	759	Composite measure of “interest-activity” generated from items on Montgomery-Asberg Depression Scale, Hamilton Depression Rating Scale, and Beck Depression Inventory	18	PRPF4B, TAOK3, NPAS3, EYS,LARP4, LOC153910, FAM19A2, COX16, CDH18, STAB2, NOTCH4, ADGRG6, FAM46A, NOX3, LOC105374974
(Sanchez-Roige et al., 2018)	23,217	Delay discounting	1	GPM6B
(Ward et al., 2019)	375,275	Question from PHQ-9	11	RGS1, RGS2 EPHB1, CTNNA3, GRM5, DISC1FP1 NCAM1, DRD2 PRKD1 SLC8A3 ISLR2, NRG4 DCC
(May-Wilson et al., 2021)	161,625	Food “liking” ratings	173	ADH1B, TSNARE1, SLC39A8, PARP1, LINC01833, GDF15, OR4K17, TAS2R38, FGF21, among other

In the largest GWAS of anhedonia (n=375,275) to our knowledge, Ward and colleagues (2019) identified 11 independent loci associated with state anhedonia assessed using the Patient Health Questionnaire-9 (i.e., “Over the past two weeks, how often have you had little interest or pleasure in doing things?”) in the UK Biobank (middle – late adulthood). While several of these loci reside near genes that have previously been implicated in reward-related clinical constructs (e.g., EPHB1 associated with antidepressant response, schizophrenia, and Parkinson’s Disease; DRD2 well characterized role in reward), none showed replicable associations with state anhedonia in an independent UK Biobank subsample (n=17,120). Single loci did not replicate in this study, although genomic liability to state anhedonia was significantly shared with major depressive disorder (r_g=0.77 [95% CI 0.71, 0.83]), schizophrenia (r_g=0.28 [0.21, 0.35]), and bipolar disorder (r_g=0.12 [0.05, 0.19]), but not Parkinson’s Disease or OCD. Simply put, these findings suggest that the overall genomic architecture that is associated with state anhedonia during mid-later life largely overlaps with depression, has moderate overlap with schizophrenia, and a small, but significantly overlap with bipolar disorder. These genetic correlations with discrete psychiatric disorders highlight the potential utility of GWASs of specific symptoms shared across disorders and RDoC constructs to identify shared and unique genomic architectures that may ultimately aid in the characterization of pathophysiologies across various forms of psychopathology to refine nosology and treatments (see also Future Directions below).

The other 2 large GWASs of anhedonia-relevant reward-related behavioral traits also show intriguing genetic correlations. For example, Sanchez-Roige and colleagues conducted a GWAS of delay discounting using data from 23,217 23andMe (www.23andme.com) customers (Sanchez-Roige et al., 2018). This GWAS identified a single variant (rs6528024) in an intron of the gene GPM6B (which encodes membrane glycoprotein) on the X chromosome associated with delay discounting performance. Much like the GWAS of anhedonia by Ward et al., (2001), this study also showed genetic correlations with delay discounting across psychiatric disorders (ADHD r_g=0.37 [0.15, 0.59], depression r_g=0.47 [0.14, 0.80]; schizophrenia; r_g=−0.22 [−0.35, −0.08]), although the directionality of this relationship with schizophrenia (negative) differed from results found by Ward and colleagues. Sanchez-Roige et al. (2018) also found that delay discounting shared genomic liability with related constructs (e.g., lifetime smoking r_g=−0.32 [0.08, 0.56], neuroticism r_g=−0.18 [0.02, 0.34], cognition measures: college attainment r_g=−0.93 [−1.22, −0.64], years of education r_g=−0.67 [−0.85, −0.49], and childhood IQ r_g=−0.63 [−0.96, −0.45]). Finally, in a recent preprint GWAS of 161,625 UK Biobank participants, May-Wilson and colleagues (2021) identified 173 loci (61 showed nominal evidence of replication) associated with food liking and reported that the genomic architecture of liking highly palatable foods is associated with reduced striatal volumes, which have been previously associated with anhedonia and other brain metrics; correlations with psychiatric diagnoses were, however, not examined.

The smaller and less well-powered studies of anhedonia-related phenotypes have largely identified individual loci that have as of yet not been replicated. In the smallest GWAS among them (n=759), Ren and colleagues identified 18 partially-independent (i.e., LD R2<.50) genome-wide significant variants associated with an interest-activity composite score derived from depression assessments in a treatment seeking sample of depressed individuals; however, none of these replicated in an independent treatment sample of 1,351, with some even showing nominally significant associations in the opposite direction. Of the three GWASs conducted on anhedonia in the context of psychosis assessments (Ns 4,269–6,297; Ortega-Alonso et al., 2017; Pain et al., 2018; Tomppo et al., 2012), only the study by Pain et al., identified a single genome-wide significant locus, which however did not replicate in an independent sample and was characterized by relatively low minor allele frequency (0.015). Along similar lines, two GWASs of reward dependence (n = 5,117; Verweij et al., 2010; n=11,000; Service et al., 2012), which were conducted prior to the development of now typically follow-up techniques (e.g., genetic correlation as estimated using LD score regression), failed to identify any significant loci. A GWAS of positive affect as assessed using the Positive and Negative Affect Schedule among 2,522 African American participants identified a single genome-wide significant locus (rs322931) associated with wellbeing; although there was convergent evidence in a small sample (n=55) also linking this variant to ventral striatum reactivity to emotional stimuli, this variant has yet to be replicated in other GWASs of related constructs (Wingo et al., 2017). Finally, a GWAS of a coordinated network of brain activity to reward anticipation evoked by the monetary incentive delay task among 1,544 adolescents identified no genome-wide significant loci, though highlighted the potential role of VPS4A, which regulates catecholaminergic systems (Jia et al., 2016).

Initial GWASs of anhedonia and related RDoC PVS constructs have yet to reliably identify individual loci as a whole, with the exception of the recent GWAS preprint of food liking (May-Wilson et al., 2021). This may be attributable to the lack of power of these GWASs and/or heterogeneity of the anhedonia construct more generally. Despite the overall lack of association with single common genetic variation, these GWASs are beginning to characterize an extensive polygenic architecture that is shared with disorders characterized by anhedonia.

Polygenic Studies.

The identification of single variants associated with psychiatrically-relevant phenotypes holds tremendous potential to identify novel pathways and molecular mechanisms of disease that may be leveraged for treatment; however, with few exceptions (e.g., APOE rs429358 & rs7412 haplotypes & Alzheimer’s Disease; Corder et al., 1993), they do little to inform individual risk or represent individual differences in molecular function in follow-up studies. Indeed, the vast majority of single loci associated with psychiatrically relevant traits are characterized by small effects (e.g., odds ratios < 1.1; Bogdan et al., 2018). Nonetheless, the additive effects of common variants, when weighted by their association with a phenotype of interest in an adequately powered GWAS, are reliably predictive of variance in the same and related traits (Bogdan et al., 2018; van Rheenen et al., 2019; Wray et al., 2021). These additive scores, which we refer to as polygenic scores (PGSs) have been applied to independent samples to explore how the polygenic architecture for a given trait is correlated with other phenotypes. Notably, the effect size of PGSs are small (typically predicting 0.01–3% of variance), although new techniques (e.g., PRS-CS; Ge et al., 2019) and more strongly powered discovery GWASs have increased effect size estimates (Bogdan et al., 2018). Optimistically, current PGSs are typically most predictive of phenotypes most proximal to the phenotype in the original GWAS as opposed to intermediate phenotypes; for example, educational attainment PGSs are most predictive of educational attainment and less predictive of more homogenous constructs such as cognitive performance (Lee et al., 2018), highlighting the need for additional GWAS of intermediate phenotypes (e.g., Grasby et al., 2019). Further, because PGSs rely on linkage disequilibrium (LD) patterns, which differ ancestrally, a PGS derived from a discovery GWAS in one ancestry is a less informative predictor in other ancestral groups, though novel approaches integrating GWAS data and external panels from multiple ancestries are showing promise (e.g., PRS-CX; (L. Duncan et al., 2019; Peterson et al., 2019; Ruan et al., 2021)). Importantly, the phenotypic presentation of depressive symptoms, including anhedonia, can vary cross-culturally (Chentsova-Dutton et al., 2015); as such, cultural variations in the experience of anhedonia will be important to consider in future PGS studies that include data from multiple ancestries. PGS studies require samples of 300 for adequate power and in all likelihood require samples much larger to reliably detect effects (Ge et al., 2019). Here, we summarize studies that have evaluated how PGSs of anhedonia-related constructs generated from the GWASs reviewed above are correlated with anhedonia and other related phenotypes, as well as how genomic liability to other phenotypes (i.e., depression, schizophrenia, and C-reactive protein [CRP]) are associated with anhedonia.

We are aware of only two studies that have evaluated associations between polygenic risk for anhedonia and psychiatrically-relevant phenotypes in an independent sample form the original discovery GWAS (but see also Ren et al., for the application of a polygenic risk score within the sample from which it was derived). These two studies have shown that polygenic risk for state-anhedonia (derived using the Ward et al., 2019 discovery GWAS in UK Biobank) is associated with brain structure in an independent UK Biobank imaging sample (n=17,120 in Ward et al., 2019; n=19.952 in Zhu et al., 2021). More specifically, these studies found that polygenic risk for state anhedonia was associated with state anhedonia (R²=0.004) as well as smaller total gray matter volume and larger total white mater volume, consistent with many studies linking polygenic risk for psychiatric phenotypes to global brain metrics. Interestingly, these global brain metrics were also associated with reported anhedonia. There was less evidence that anhedonia polygenic risk was associated with regional variability in gray matter structure; anhedonia polygenic risk was associated with thinner cortex in the insula, parahippocampal cortex, and superior temporal gyrus. Although significant associations were observed with white matter integrity, none of these were robust to the inclusion of potential confounds (e.g., stress, socioeconomic status, alcohol use). Collectively, these studies suggest that associations between global brain metrics and anhedonia may be partially genomic in origin.

In contrast to limited studies evaluating polygenic risk for anhedonia, several studies have investigated whether polygenic risk for related phenotypes (i.e., depression, schizophrenia, bipolar, inflammation) are associated with reported anhedonia and related behavioral and neural indices. For example, building upon the genetic correlations showing the genomic architecture of anhedonia is shared with clinical diagnoses characterized by anhedonia, Pain and colleagues (2018) found that polygenic risk for schizophrenia and depression, but not bipolar disorder, is associated with anhedonia at nominal levels of significance in a sample of 6,579.

In an RDoC-inspired study among 83 participants, Guffanti and colleagues (2019) find that polygenic risk for depression is not associated with self-reported anhedonia or behavioral reward learning, but it shows nominally significant stress-related reduction in reward prediction errors in the ventral striatum and putamen, as well as reductions in ventral striatum and putamen volume (the association with putamen volume would survive correction for multiple testing). These findings are consistent with theoretical speculation that intermediate phenotypes such as neural measures would be more closely associated with genetic architecture than more homogenous distal phenotypes such as self-reported anhedonia, although prior PGS studies suggest that they are typically most predictive of the most proximal phenotype to the original GWAS (e.g., in this case depression or depression symptoms), even if that phenotype represents a more heterogenous construct. Nonetheless, in another study of 478 college students, Mareckova and colleagues find that a transcriptomically-informed PGS based on 76 genes was not directly related to anhedonia but showed an indirect association through neural responses to neutral faces (2020).

Finally, Kappelmann and colleagues examined associations between anhedonia and PGSs of immune-metabolic markers (Kappelmann et al., 2021), supported by widespread evidence of elevated inflammatory signaling across psychopathology (Baumeister et al., 2014). Inflammatory markers act upon several neural systems, and consequences include a reduction in the availability of dopamine in the basal ganglia (Felger, 2017), plausibly implicating it in anhedonia symptomatology. Supportive of a potential etiologic mechanism, administration of proinflammatory cytokines in humans and non-human animals induces an anhedonic-like phenotype, alters the structure of the striatum, and reduces dopamine signaling capacity and striatal responses to reward (Felger et al., 2013; Harrison et al., 2016; Treadway et al., 2019). Interestingly, across samples (Ns=1,058–110,0101), higher polygenic risk for CRP was associated with reduced anhedonia, while the PGS for TNFα was correlated with higher levels of anhedonia (Kappelmann et al., 2021). Although immune pathways act on reward-related neural circuits implicated in anhedonia, these contradictory findings suggest that it remains unclear to what extent genetic liability to immune signaling is a contributing mechanism.

Conclusions and Future Directions

In contrast to a strong theoretical history highlighting the importance of anhedonia to psychopathology, studies of its genomic architecture remain uncommon and largely underpowered. While specific loci remain elusive, the polygenic architecture of anhedonia is showing expected correlations with related disorders and phenotypes (e.g., depression, striatal brain volume). Initial hopes that leveraging intermediate phenotypes would generate large boosts in power that would permit the use of smaller samples have not been empirically demonstrated (e.g., Grasby et al., 2020). However, it remains plausible that loci identified by intermediate phenotype approaches may be more tractable for understanding psychiatric etiology and treatment. As is now clear in psychiatric genetics, larger samples are needed to be able to reliably estimate expected small effects, even for intermediate phenotypes. Large scale nationwide data collection projects (e.g., UK Biobank, Adolescent Brain Cognitive Development [ABCD] Study, HEALthy Brain and Child Development [HBCD] Study, All of Us) that incorporate deep phenotyping (e.g., neuroimaging, biomarkers, behavioral assessments) are importantly contributing to the genomic investigation of intermediate phenotypes and increasing sample sizes available to investigators. In the meantime, as we wait for larger samples to accumulate, we may leverage currently available GWAS summary statistics from studies of reward processing phenotypes and combine them with larger efforts characterizing related constructs that are more heterogeneous. For instance, we recently used genomic structural equational modeling, which identifies the shared and unique genetic architectures of multiple GWAS phenotypes, to show that genomic risk for specific forms of substance use disorders is largely shared and independent of the genomic architecture associated with substance use (Hatoum et al., 2021). Moreover, we found that this shared genomic risk for addiction was also shared with trait vulnerability to neurobehavioral stages of addiction (i.e., impulsivity, neuroticism, executive function), which mediates associations between substance psychopathology and non-substance psychopathology. A similar approach could be used to characterize the shared and distinct genomic architecture of anhedonia, and depression as well as depression related traits not typically characterized by reward deficits (e.g., neuroticism, and anxiety). This may bolster the genomic signal and allow for anhedonia-related and non-anhedonia-related genomic risk for depression to be estimated as we continue to accumulate larger samples needed for the genomic study of anhedonia. It is the hope that the replicated results emerging from genomic studies of anhedonia will uncover novel treatment targets for this condition that has been linked to poor disease course, treatment failures and increased suicide risk.