Person-focused approach explains distinct autism genetic subtypes

New study shows that autism is not one condition, but four, each driven by different genes and brain development patterns, helping to reshape diagnosis and care.

Study: Decomposition of phenotypic heterogeneity in autism reveals underlying genetic programs. Image credit: chrupka/Shutterstock.com

A paper published in Nature Genetics provides a new glimpse into the role of genetic variation in the range of clinical symptoms found in autism spectrum disorder (ASD). The team of researchers found that autism classes can be clinically distinguished. Each is associated with unique patterns of gene dysregulation, reflecting distinct molecular-level perturbations caused by class-specific sets of mutations.

Introduction

ASD is a neurodevelopmental condition characterized by difficulties with social communication and interaction, often demonstrating restricted and repetitive behavior patterns, interests, or activities. With an increasing ASD burden each year, the phenotypic and genetic differences in the ASD pool are becoming more obvious.

A systematic evidence-based association study of genetic and phenotypic data has not matched the genetic and phenotypic complexity of ASD. The current study used a large sample of autism phenotypes to reveal phenotypic classes and the underlying genetic heterogeneity.

A person-centric method is essential in this analysis, rather than focusing on single traits, since each co-occurring trait inevitably affects the other. Only then can they be properly mapped to their genotypic origin. Such an approach provides a clearer picture of how these developmental disruptions interact and evolve, allowing for an informed prognosis.

This approach avoids the limitation of traditional trait-centric analyses and better captures the complex, interacting nature of ASD symptoms in real individuals.

About the study

The current study used genotypic and phenotypic data from the SPARK cohort of 5,392 people. They collected phenotypic data from standard diagnostic questionnaires (the Social Communication Questionnaire-Lifetime (SCQ)^, Repetitive Behavior Scale-Revised (RBS-R), and the Child Behavior Checklist 6–18 (CBCL), combined with the developmental milestone history.

Researchers applied a generative finite mixture modelling (GFMM) framework to analyze 239 features, allowing individuals to be grouped into clinically meaningful phenotypic classes based on their overall trait profiles.

Study results

The model distinguished autism classes using the seven core traits: limited social communication, restricted and/or repetitive behavior, attention deficit, disruptive behavior, anxiety and/or mood symptoms, developmental delay (DD), self-injury, and the severity of symptoms. They assigned each of 239 phenotype features associated or co-occurring with each trait.

Four ASD phenotypic classes emerged. One class (Social/behavioral) had both severe social communication deficits and restricted or repetitive behavior, compared to other ASD children. They also showed disruptive behavior and attention deficit, with anxiety, but normal developmental rates.

The Mixed ASD class distinctively shows developmental delay, despite some features of social communication deficits, restricted/repetitive behavior, and self-injury. The other two classes were Moderate challenges, with a lower ASD score for all seven core traits than other ASD children but above non-autistic siblings, and Broadly affected, with a higher score than other ASD children.

These classes were validated by their agreement with the reported co-occurring conditions, parental narratives, and medical history. For instance, the Broadly affected class was much more likely to have all co-occurring conditions: attention-deficit hyperactivity disorder (ADHD), obsessive-compulsive disorder (OCD), delayed language development, anxiety, and depression.

Conversely, the Mixed ASD class had the lowest odds of anxiety, depression, or ADHD. Yet they were at high risk for language and intellectual delay, and motor abnormalities, concordant with their delayed development and high occurrence of restricted/repetitive behavior. In contrast, the Social/behavioral class had a higher risk of ASD-associated ADHD, anxiety, and depression diagnoses.

Mixed ASD and Broadly affected classes were usually diagnosed the earliest. They were most likely to be receiving a variety of therapies, due to the highest cognitive impairment and poor language skills. The Broadly affected class also showed the most significant co-occurring conditions and the highest use of clinical interventions.

An independent cohort of autistic subjects also confirmed the validity of these classes. The model showed a high correlation in feature enrichment patterns across the SPARK and SSC cohorts, highlighting its robustness and generalizability.

The scientists then investigated genetic influences using polygenic scores (PGS) for autism and five well-accepted genome-wide association studies (GWAS) for autism-related conditions or traits. They explored both inherited and newly arising genetic variations. The findings revealed genetic differences according to the four classes.

The Broadly affected and Social/behavioral classes had higher ADHD PGS signals than for other classes or non-autistic siblings. Depression-related PGS and diagnosis rate were both highest in the Social/behavioral class. This class also had the highest number of high-impact variants in neuronal genes, mainly expressed after birth. However, ASD PGS did not differ significantly between classes due to high within-group variance, highlighting the limited explanatory power of current common-variant-based ASD scores.

The Broadly affected class, which showed the highest rate of cognitive impairment, delayed development, and the lowest educational status, had the lowest IQ PGS. Thus, “co-occurring conditions were associated with common genetic variation that significantly differed among the four identified classes.”

All four classes had more mutations than non-autistic siblings, and mutations were unevenly distributed between the classes. The Broadly affected class had excessive loss-of-function (LoF) or missense mutations, and the social/behavioral class was the least enriched.

Whilst all four classes showed more mutations than non-ASD siblings, the Broadly affected class had the highest high-impact de novo mutations. In contrast, the Mixed ASD class showed increased rare inherited variation alongside de novo mutations, indicating a stronger inherited genetic component.

ASD-specific gene set analysis showed a higher burden of new LoF mutations associated with developmental delay. High-impact mutations in a relatively small group of genes contribute to cognitive impairment. The greater the development delay in the class, the higher the odds for new LoF mutations.

Fragile X mental retardation protein (FMRP) target gene mutations were especially enriched in the Broadly affected class and the Mixed ASD class. FMRP in both Broadly affected and Mixed ASD classes was linked to developmental delay and cognitive deficit. The Broadly affected class showed additional enrichment for mood and behavioral traits such as anxiety, hyperactivity, and aggression.

Tracing the molecular pathways affected by these mutations showed that each class reflected specific pathway disruptions. For instance, disruption of microtubule activity, chromatin organization, and DNA repair was enriched in the Social/behavioral class, compared to neuronal action potential and membrane depolarization in Mixed ASD.

In the Mixed ASD class, LoF mutations affected prefrontal cortical neuronal genes expressed primarily during fetal and early newborn life. This class, therefore, had the most developmental delay and the earliest diagnosis, compared to the Social/behavioral class, where postnatal gene expression was disrupted.

The Broadly affected class showed gene dysregulation spanning all developmental stages and cell types, especially of FMRP target genes and highly constrained genes. In contrast, the Moderate challenges class had enrichment for variants in genes with lower evolutionary constraint, which may explain the milder developmental impact.

Conclusions

The study demonstrated the value of a person-centered rather than a trait-centric approach to ASD genotype-phenotype analysis. The four phenotype-based classes described here agreed with reported clinical features and can be applied to any clinical cohort. Importantly, it suggests that ASD phenotypes do not reflect a spectrum of intellectual disability.

The classes also had separate genetic signals and differed in the timing of gene dysregulation during the developmental trajectory. These differences correlate with the degree of delay in development and the outcome.

The study provides a framework to investigate the neurobiological mechanisms underlying distinct ASD presentations, supported by genetic and molecular data across developmental stages.

These findings suggest new research directions to understand the neurobiological mechanisms underlying different ASD presentations, allowing for more precise diagnosis and management of these conditions.

Download your PDF copy now!