All for One and One for All: Heterogeneity of Genetic Etiologies in Neurodevelopmental Psychiatric Disorders

Abstract

Alexandre Dumas’ famous phrase All for One and One for All recapitulates our current understanding of the genomic architecture of neurodevelopmental psychiatric disorders (NPD), like autism spectrum disorder, bipolar disorder, and schizophrenia. Many rare genomic variants of large effect size have been identified; all of them together can explain a significant proportion of NPD. In parallel, one rare genomic variant can cause all of the above NPD. Finally, common genomic variants of individually small effect size can be combined to further explain risk for NPD. How do we reconcile different genomic variants accounting for one clinical diagnosis, and different clinical diagnoses arising from a single genomic variant? Here, we discuss a framework to understand genetic and clinical heterogeneity in NPD.

Introduction

Although marked progress has been made in identifying the genetic basis of many neurodevelopmental psychiatric disorders (NPD), such as autism spectrum disorder (ASD), schizophrenia (SCZ), bipolar disorder (BPD), and attention deficit hyperactivity disorder (ADHD), there is still a clear disconnect between conventional nosological classifications and their biological underpinnings. Arising from the need among clinicians to find a common language to guide and standardize treatment, a clinical diagnosis of a NPD is made based on a consensus understanding of clusters of co-occurring behavioral symptoms, as outlined in the Diagnostic and Statistical Manual of Mental Disorders (DSM-5) [1]. While often useful for the aforementioned purpose, descriptive clinical diagnoses do not provide information about the underlying biology and cause of disease, nor do they acknowledge the considerable overlap among these diagnoses. Moreover, there is a vast diversity of clinical presentations within each of these clinical diagnoses that stems from the different nature and degree of compromise within affected neurobehavioral domains, as well as the presence of additional psychiatric and other medical comorbidities.

For example, no two individuals with ASD are alike despite sharing the same clinical diagnosis [2], and even in light of the clinical utility of this symptom-based diagnostic classification, the richness of individual narratives, strengths, and challenges can be overlooked by such an approach. Furthermore, our ability to find precision health treatments rooted in the biology of these conditions is hampered in the absence of a clear link between a clinical diagnosis and the underlying biology. In a way, we are often treating the behavioral equivalent of fever, and could learn from how this observable diagnosis was reclassified as a symptom, rather than a disease, with multiple different causes that would ultimately dictate different and specific treatment options and prognoses. To overcome the limitations of using categorical clinical diagnoses as descriptors, rather than focusing on etiologic underpinnings and quantitative, dimensional traits, the National Institute of Mental Health (NIMH) has proposed the Research Domain Criteria, (RDoC) [3]. Studies based on RDoC principles focus on neurobehavioral, domain-based classifications that cut across different clinical diagnoses, informed by neuroscience, and are shifting our understanding of the architecture of mental health conditions [4–7]. However, heterogeneity continues to be the norm in NPD, extending from clinical presentation to genetic roots.

All for One: Uncovering Genetic Heterogeneity in NPD

NPD with a high heritability, such as ASD, SCZ, BPD, and ADHD, each affect over 1% of the population worldwide, and have a major impact on affected individuals, their families, and society [8,9]. Genetic factors have long been implicated in the pathogenesis of NPD [10]; high degrees of concordance in monozygotic twins, along with observed family recurrence rates that significantly exceed population frequencies, have long pointed to primary genetic underpinnings in NPD [11–16]. The search for common variants of individually small effect size, as well as rare variants with individually large effect size, has led to the identification of a rapidly increasing number of genetic factors that increase the risk of developing these disorders [17]. Progress has been particularly notable for the discovery of rare, highly penetrant copy number variants (CNVs) and single nucleotide variants (SNVs) that result in loss of gene function, with hundreds of reports over several decades linking specific genomic variants to NPD clinical diagnoses [10,18,19].

Despite the large number of known genetic etiologies of NPD, including the relatively common 22q11.2 deletion observed in ~1/4,000 individuals in the general population and ten times more frequently in people with schizophrenia [20,21], we have only recently been able to uncover and appreciate their collective contribution to NPD. As shown in Figure 1A, we now know that genetic heterogeneity, where many different genomic variants can cause the same clinical diagnosis, is a core feature of NPD, diminishing the evidence that specific genetic changes are exclusively associated with any single clinical diagnosis. This recent progress is largely due to a rapidly changing landscape, as powerful microarray and next generation sequencing technologies have allowed pathogenic CNVs and SNVs to be detected throughout the genome, with particular relevance for NPD such as ASD, SCZ, BPD, and ADHD [18,19,22–25]. These technological advances have also fostered global genomic consortia, supported by the NIH and other sources, that are revolutionizing our understanding of rare and common genomic variation in NPD. For instance, the ClinGen project has allowed pooling and expert curation of rare genomic variants and genes that might otherwise remain unclassified, if considered in isolation, and the Psychiatric Genomics Consortium (PGC) has accelerated our ability to understand common and rare genomic contributors to NPD[26,27]. In addition, a recent initiative, the Genes to Mental Health (G2MH) network, has a goal of collating and harmonizing genetic and quantitative phenotypic data across many different rare NPD genomic variants to better characterize and predict the risk for particular clinical diagnoses[28]. Success stories of large collaborations abound and are shedding light on the importance of collaboration as we enter the era of precision health in psychiatry [29–35].

Figure 1.

Examples of genetic and clinical heterogeneity in neurodevelopmental psychiatric disorders (NPD). A) Multiple CNVs can lead to a clinical diagnosis of ASD. The size of the CNV circles in the periphery represents the frequency of individuals with ASD that have a particular CNV; B) A single CNV, like a 22q11.2 deletion, can result in multiple NPD. The circles in the periphery represent some of the NPD phenotypes that can be observed in individuals with a 22q11.2 deletion; the size of the circle represents the proportion of 22q11.2 deletion individuals with a particular NPD diagnosis.

One of the major outcomes of these research collaborations is the clinical impact they have on our growing understanding of the role rare genomic variants play in NPD. Although none account for more than ~1% of the cases, in aggregate, rare CNVs and SNVs may explain up to 35% of ASD [10,19,36–38] and up to 10% of schizophrenia [10] and are often marked by pleiotropy, impacting multiple organ systems [39,40]. Therefore multiple medical professional societies now recommend genetic testing as a key element in the assessment of individuals with NPD [41–44], including adults [45], as the diagnostic yield is significant and can result in changes to medical care and provide more accurate recurrence risk estimates for families. In addition, rare variants offer distinct opportunities to illuminate the molecular and cellular underpinnings of these disorders due to their large effect size, as well as the potential for the development of novel therapeutic strategies for these conditions [28,40].

One for All: Clinical Heterogeneity of Genomic Variants

Although individuals meeting criteria for clinical NPD diagnoses share core behavioral commonalities, they can present with a wide range of functional impairments, behavioral symptoms, and medical comorbidities. This high degree of clinical variability relates in part to differential phenotypic impacts of the many known and still unknown genetic etiologies that cause NPD. For example, individuals with 22q11.2 deletions are more likely to develop psychosis than those with deletions of 16p11.2; conversely, people with ASD stemming from a pathogenic SNV in PTEN are less likely to have been born with structural cardiac anomalies than those with a 22q11.2 deletion; however, those with PTEN SNV have a significant cancer risk that requires active medical surveillance. Moreover, even individuals with the same genomic variant may present with different behavioral diagnoses; for example, some individuals with a 17q12 deletion may have an ASD diagnosis, while others, even within the same family, may have other NPD diagnoses, such as BPD [46]. In general, this heterogeneity is likely influenced by factors such as background common genetic variation, additional rare genomic changes of large effect, mosaicism, epigenetics, and stochastic developmental variation. An example of the diversity of clinical presentations associated with a given rare genetic change, in this case 22q11.2 deletions, is highlighted in Figure 1B.

In addition to the clinical heterogeneity observed for NPD in individuals with the same genetic diagnosis, most rare genomic variants of large effect size also have effects on other body systems. [40]. New genetic etiologies for NPD are being discovered at a rapid rate, from both research studies and clinical testing on samples referred for genetic testing. Pleiotropy and variable expressivity are proving to be the norm, rather than the exception, but paradoxically, individuals with conditions such as ASD and SCZ who have significant medical comorbidities are often excluded from large research studies in order to decrease the phenotypic heterogeneity of the studied population [47]. This strategy has the disadvantage of reducing the generalizability of results, as the stringently-selected population from which conclusions are drawn often does not generally represent the patients in typical clinical settings. Therefore, the contribution of neuropsychiatric risk for specific rare genomic variants that have marked medical consequences may be underappreciated as these can be screened out of study populations.

Family Matters: Understanding Proband Phenotypes Based on Parental Traits

Family background has long been known to play a significant role in influencing phenotypic variability in genetic disorders. For example, parental height is an important determinant of adult stature in girls with Turner syndrome, while IQ correlations between probands and their first-degree relatives have now been reported for several genetic syndromes, including 16p11.2 and 22q11.2 deletions [48,49]. In individuals with a pathogenic genomic variant, the phenotypic outcomes are influenced by background genomic factors resulting in variable expressivity [50].

For example, in two unrelated individuals, the same pathogenic genomic variant will confer the same magnitude of deleterious impact (“shift”) on a quantitative trait (Figure 2), such as reciprocal social behavior. However, depending on the expected family “starting point” for this quantitative trait, the variant may shift one of these individuals with the pathogenic variant into the range of significant social deficits, thus fulfilling diagnostic criteria for ASD, while the familial starting point in the other individual provides a relative protective effect, decreasing the chance for ASD. Of note, even in the second individual, the shift still has an effect, decreasing the reciprocal social behavior from where it would have been without the pathogenic variant. A family-based approach to phenotypic characterization is therefore essential to contextualize the impacts of rare pathogenic variants on various functional domains, including protective effects that mitigate NPD risk in some families.

Figure 2.

Relationship of rare and common variants to the distribution of an example quantitative trait. The distributions between the general population and individuals with a rare genomic variant, such as a 22q11.2 deletion, are compared. Individuals with a rare variant demonstrate a deleterious “shift” from the expected values in the general population for a given quantitative trait, which can be modified by the additive effects of common variants, expressed as polygenic risk scores (PRS), resulting in an overall high or low risk for a particular trait (for this example, the PRS for a given trait in the general population and in individuals with a rare variant are assumed to follow the same distribution.

Polygenic Risk Score as a Measure of Genetic Background

Family history, as a proxy for background genetic variation, is one of the strongest and most easily implementable genomics tools in clinical practice. Now, using newly developed genomic-based tools, we can measure the aggregate impact of this background genetic variation on specific phenotypes. Recent genome-wide association studies (GWAS) have confirmed that a substantial component of NPD heritability is distributed over thousands of common variants of small effect size [51]. The aggregate impact that these individual low-risk genetic variants confer on a given trait can be quantified in the form of a polygenic risks score (PRS) [52]. In the PRS framework, individuals are scored on their genotypes for multiple, common, low-impact risk alleles that have been weighted by their reported effect sizes in an external GWAS. PRS for different NPD have been demonstrated to strongly associate with disease status, such as the PRSSCZ for SCZ which explains ~7% of variance in SCZ liability in European populations⁵³.

Recently, PRS derived from common forms of complex disease have been applied to understand apparently incomplete penetrance of rare Mendelian forms, such as familial breast cancer and obesity, and confirm that polygenic background is an important component of the genetic architecture, even in the presence of a dominant monogenic mutation [53,54]. Consistent with this principle, the clinical expression of SCZ in individuals with a 22q11.2 deletion is influenced by common risk variants identified in the general population [55], highlighting the joint contribution of rare and common variation on a quantitative trait, as shown in Figure 2. Approaches such as this one provide a unifying understanding of how common and rare variation work together to explain the heterogeneity and convergence of genetic factors on psychiatric risk; a clinical vignette applying this principle is shown in Table 1.

Table 1.

Clinical Vignettes Demonstrating Genetic and Clinical Heterogeneity in NPD

	Genetic Heterogeity - ASD
		Clinical Heterogeneity - 22q11.2 deletion
	Billy	Michael	Delilah
Overview	17 year-old male, socially driven but unconventional social interactions, loves music and trains	17 year-old male, shy and imaginative, loves Harry Potter movies and Outer Space	7 year-old girl, enjoys playing football and reading mistery novels, has an imaginary friend
Early medical comorbidities	Supravalvular aortic stenosis	No concerns	Surgically-corrected ventricular septal defect
Early surveillance	Concerns at age 4 because of irritability, lack of spontaneous play, and attention difficulties	Concerns at age 12 because of irritability, lack of social connections, and paranoia	No concerns for neurodevelopmental issues
Initial NPD Diagnosis	ASD	ASD	None
Genetics	7q11.23 deletion identified at 4 years of age	22q11.2 deletion identified at at 14 years of age after being placed on aripiprazole and having cardiac arrhythmia. High risk PRS for schizophrenia	22q11.2 deletion identified at birth because of congenital heart defect. Low risk PRS for schizophrenia
Genetically informed interventions	Early detection led to psychopharmacologic management of ADHD symptoms in conjunction with Billy’s pediatric cardiologist, in this case resulting in avoidance of alpha-2 agonists and use of stimulant medication	Late detection led to cardiac complications stemming from atypical antipsychotic use in the setting of cardiac problems. High risk PRS led to increased surveillance for prodromal psychotic symptoms and psychoeducation for the family.	Low risk PRS led to stratification of risk and contextualization of symptoms that could have been potentially mis-perceived as delusions (imaginary friend). Developmental surveillance was continued, with no concerns for psychotic symptoms and avoidance of antipsychotic medications, which would have been particularly deleterious given her cardiac history.

Context Matters: Impact of genetic changes on adaptive functioning

Although much attention has been focused on understanding the heterogeneity of phenotypes associated with CNVs and SNVs in research settings, the impact of these rare genetic changes can be quite diverse across different populations and cultures. In addition to the potential role of common genetic variation as a modulatory factor influencing the expression of neurobehavioral phenotypes, there is one factor that could play a key role that often receives less attention in the genomic era: cultural and environmental expectations leading to different levels of adaptive functioning[56–58]. This shift away from isolated cognitive functioning and towards adaptive functioning has been embraced by the last edition of the DSM-5, where, for example, the severity of intellectual disability is classified based on conceptual, social, and practical skills[1]. These frameworks could help identify areas of strength and reflect more closely the interplay between genetic vulnerabilities and environmental expectations for people with rare CNVs, helping bolster precision medicine interventions for this population.

Opportunities to Move Forward

There is a striking dissonance between the technological advances that have propelled the recent genomic discoveries in NPD and the current behavioral assessments used to study these populations, which have remained unchanged for decades and focus on descriptive clustering of symptoms in diagnoses that, although clinically useful, remain distant from the underlying biology. These assessments are often administered in one or a few points in time, and in person, which leads to several constraints: 1) we are only obtaining phenotypic information at one single moment in the life of people with life-long, dynamic mental health conditions, 2) we are bound by the availability of highly-trained professionals to administer these assessments, and 3) we are limited by geographical proximity to the assessors. All of these limitations are particularly troublesome for research in rare genetic conditions where we are still trying to understand the natural history and evolution of symptoms across domains in individuals who are often geographically and linguistically distant, with only a few researchers working on any given rare genetic condition with diverse protocols.

In addition, research on common genetic risk factors for NPD has often been limited to people of European backgrounds, limiting our ability to extrapolate PRS derived in this setting to people from diverse ethnic backgrounds. There is now a critical need to develop standardized and scalable approaches across diverse populations that will allow us to understand the behavioral, neurological, and medical consequences of rare NPD variants, with a focus on how this information can be ultimately used to impact clinical management [28], rooted in the frameworks of implementation and dissemination science [59]. Ideally, this information will be obtained serially across multiple points in time, fueled by recent advances in wearable technologies and ecological momentary assessments, and be acquired digitally and remotely, allowing to breach geographical barriers and diverse populations, and with less reliance on human administration of diverse batteries of tests, eliminating some of the existing bottlenecks.

Overall, synergistic advances in genomic technologies, large international collaborations, and reconceptualization of phenotypic classifications are allowing us to understand the genomic bases of NPD. Evidence for the integration of rare and common genomic variation and their combined role in risk and resilience for NPD is increasing. Challenges remain, including the time-limited nature of phenotyping assessments that only provide a snapshot over the course of the natural history of NPD, the difficulties of deploying assessments across diverse geographical regions, and the lack of ethnic diversity of the populations studied that impacts their generalization. However, the collaborative international environment focused on data sharing and knowledge transfer has encouraging opportunities to address these challenges and move forward to implement the use of genomic information as a key, and routine, component of precision psychiatry.