Autism Spectrum Disorder (ASD) and Attention-Deficit/Hyperactivity Disorder (ADHD) are complex, frequently co-occurring neurodevelopmental conditions with significant genetic components. While both are highly polygenic, exhibiting substantial genetic overlap, recent studies have revealed discordant relationships with educational attainment (EA). Specifically, increased polygenic risk for ADHD has been associated with lower EA, while increased polygenic risk for ASD has shown a positive correlation with EA. This paradoxical finding necessitates further investigation into the underlying genetic mechanisms. The co-occurrence of ASD and ADHD is well-documented, with estimates suggesting that 15-25% of individuals with ADHD exhibit ASD symptoms, and 40-70% of those with ASD display comorbid ADHD symptomatology. However, the shared etiological mechanisms remain largely unclear. Large-scale genome-wide association studies (GWAS) have identified numerous loci with pleiotropic effects across multiple psychiatric disorders, including ASD and ADHD. Models using exploratory factor analysis and genomic structural equation models support the clustering of ASD and ADHD within the same group of early-onset neurodevelopmental disorders, further highlighting the need to understand the genetic intricacies of their relationship. The genetic correlation between ASD and ADHD diagnoses varies across studies, ranging from 0.36 to 0.87, with even stronger evidence observed for genetic links between the co-occurrence of ASD and ADHD symptoms. However, the discordant relationship with EA presents a crucial puzzle: increased polygenic ASD risk is associated with higher EA, while increased polygenic ADHD risk correlates with lower EA. This discrepancy is especially noticeable in measures of years of schooling and college completion. The existing literature on academic achievement in ASD is varied, with high-functioning individuals potentially achieving higher qualifications despite facing labor market disadvantages. Conversely, observational research consistently links ADHD to poorer school performance and educational outcomes. The mechanisms behind these discordant polygenic associations remain unknown but may involve various biological processes, including different forms of pleiotropy. This study aims to clarify these mechanisms by investigating the genetic overlap between ASD, ADHD, and EA and identifying the specific genetic variants involved. The researchers consider various scenarios, such as independent causal variants for ASD and ADHD’s relationship with EA, ascertainment bias in the recruitment of cases, opposite or identical marker alleles tagging opposite or independent causal variants, and shared risk alleles exhibiting different effects due to biological pleiotropy. The study uses a multivariable regression (MVR) technique to analyze genome-wide summary statistics from large GWAS datasets, allowing for simultaneous estimation of polygenic ASD and ADHD associations with EA while controlling for potential biases.

The literature reviewed extensively documents the genetic complexity of ASD and ADHD, highlighting their high polygenicity and the significant overlap in their genetic architectures. Studies using twin designs and GWAS have consistently shown shared genetic influences between these two disorders, both at the population level and within clinical samples. A major cross-disorder GWAS reported over a hundred loci with pleiotropic effects on both ASD and ADHD, further solidifying the notion of shared genetic etiology. Furthermore, the familial co-aggregation of ASD and ADHD strengthens this evidence, as does the identification of shared copy number variations suggesting involvement of similar biological pathways. However, the existing literature also acknowledges the inconsistent findings regarding the relationship between these disorders and cognitive function and educational attainment. While some studies highlight positive associations between ASD and higher cognitive abilities and educational outcomes, others demonstrate a negative correlation between ADHD and these measures. This study builds on these previous observations, focusing specifically on the discordance between the genetic relationships of ASD and ADHD with EA to elucidate the underlying mechanisms.

This study employed a multivariable regression (MVR) approach to analyze genome-wide summary statistics from large-scale GWAS datasets for educational attainment (EA), autism spectrum disorder (ASD), and attention-deficit/hyperactivity disorder (ADHD). The MVR method is analogous to Mendelian randomization (MR) approaches, but without making explicit causal inferences, acknowledging the possibility of biological pleiotropy. The researchers utilized summary statistics from several consortia, including the Social Science Genetic Association Consortium (SSGAC) for EA, the Psychiatric Genomics Consortium (PGC) and the Lundbeck Foundation Initiative for Integrative Psychiatric Research (iPSYCH) for ASD and ADHD. The study used a bidirectional MVR approach, involving two complementary models: an ASD-MVR model and an ADHD-MVR model. In the ASD-MVR model, a set of independent subthreshold variants from ASD GWAS summary statistics was selected. Each ASD-increasing marker allele was aligned to ASD, ADHD, and EA SNP estimates. The aggregate association effect with EA across all alleles was simultaneously estimated for both ASD and ADHD risk. The ADHD-MVR model followed an analogous procedure using independent subthreshold variants from ADHD GWAS data. These models allowed the researchers to simultaneously estimate the polygenic effects of both ASD and ADHD on EA while controlling for potential biases. Multiple stages of analysis were performed. Initially, the researchers investigated whether discordant association patterns involved independent markers tagging independent ASD and ADHD risk alleles or identical markers tagging shared alleles. This involved a series of ASD-MVR and ADHD-MVR models using different P-value selection thresholds. Subsequently, discovery variant sets were restricted to markers carrying the same risk-increasing allele for both disorders (concordant variants) to test the robustness of the findings. Follow-up analyses replicated the MVR findings using an independent ASD sample from the PGC. To identify specific genomic regions contributing to the discordant association patterns, the researchers employed the gwas-pw method, which estimates the posterior probability of shared or non-shared genetic effects for two traits within independent linkage disequilibrium (LD) blocks. This method helped identify LD blocks carrying genetic markers associated with both ASD and ADHD risk due to biological pleiotropy or high-LD co-localization. The researchers then applied a conditional P-value thresholding approach to identify single variants contributing to the discordant polygenic overlap with EA. This involved systematically assessing the overlap between ASD and ADHD variant sets, creating conditional variant subsets and fitting MVRs with these subsets. Finally, the researchers performed specificity analyses to examine whether the findings extended to general intelligence and other adult-onset psychiatric conditions. They used GWAS summary statistics for general intelligence and disorders like major depressive disorder (MDD), schizophrenia (SCZ), and bipolar disorder (BD) to assess the generalizability of the observed patterns. Genetic correlation analyses were conducted using linkage disequilibrium score regression (LDSC) to further examine the extent of shared genetic influences between different phenotypes.

The key findings of this study demonstrate a complex interplay between genetic variation associated with educational attainment (EA), autism spectrum disorder (ASD), and attention-deficit/hyperactivity disorder (ADHD): 1. **Shared Genetic Variation:** EA-related genetic variation is significantly shared across ASD and ADHD architectures, involving identical marker alleles at the same loci. This suggests a substantial degree of pleiotropy, where genetic variants influence multiple traits simultaneously. 2. **Discordant Polygenic Associations:** Despite the shared genetic variation, the polygenic association profile with EA is discordant for ASD and ADHD. Specifically, ASD shows a positive correlation with EA, while ADHD exhibits a negative correlation. This indicates independent, potentially opposing effects on EA from ASD and ADHD risk alleles. 3. **Single-Variant Level Findings:** Analysis at the single-variant level suggests that the discordance arises from either biological pleiotropy or co-localization of different risk variants within the same genomic region. This is particularly highlighted by the involvement of MIR19A/19B microRNA mechanisms. 4. **Polygenic Pleiotropy:** At the polygenic level, the findings strongly indicate a polygenic form of pleiotropy. The combination of the same risk alleles across the genome results in different aggregate effects on EA, reflecting independent polygenic influences for ASD and ADHD. 5. **Effect Cancellation:** The discordant polygenic associations suggest that when ASD and ADHD summary statistics are meta-analyzed, the opposing effects on EA across shared regions can lead to a cancellation of signals and attenuation of the overall genetic correlation with EA. 6. **High-Confidence Genomic Regions:** Analysis identified three independent genomic regions with strong evidence for shared genetic effects for ASD and ADHD, implying biological pleiotropy or high-LD co-localization. These regions contain genes previously linked to ASD or ADHD risk or potentially single-variant co-localization. 7. **Single-Variant Identification:** A conditional P-value thresholding approach pinpointed 83 loci significantly associated with both ASD and ADHD, capturing larger polygenic effects than observed in discovery analyses. These loci were enriched for microRNA targets, particularly MIR19A/19B, implying a regulatory role for microRNAs in the observed pleiotropic effects. 8. **Robustness and Replication:** The findings are robust, replicating across independent samples, showing little evidence for variance inflation or attenuation of signal in the multivariate analyses, and remaining consistent even when excluding variants with opposing effects on ASD and ADHD. 9. **Specificity Analyses:** Analyses confirmed the observed discordant patterns extend to general intelligence, and further exploratory analysis suggested the potential for EA-related polygenic pleiotropy to contribute to the genetic architecture of several adult-onset psychiatric disorders.

This study's findings provide compelling evidence for a polygenic form of pleiotropy in the relationship between EA, ASD, and ADHD. The observation of shared genetic variation underlying both ASD and ADHD, yet exhibiting discordant effects on EA, challenges simplistic models of genetic overlap between these disorders. The significant enrichment for MIR19A/19B microRNA targets suggests a regulatory role for these microRNAs in the complex interplay between these traits. These microRNAs are known key regulators of neurodevelopmental processes, and their influence on multiple traits supports the notion of biological pleiotropy. The observed effect cancellation, where the opposing effects on EA from ASD and ADHD risk alleles lead to attenuated overall genetic correlation, highlights the importance of considering multivariate relationships when analyzing complex traits. Traditional univariate approaches may mask important insights into the underlying genetic architecture. The study's findings underscore the limitations of solely relying on genetic correlation measures and emphasize the need for more sophisticated multivariate methods that can disentangle complex genetic effects. The replication of the findings using independent samples and the robustness observed across various analytical approaches significantly strengthen the validity of the study's conclusions. The extension of the findings to general intelligence further supports the broad impact of these genetic mechanisms on cognitive functions and educational outcomes. The exploration of adult-onset disorders suggests the wider relevance of polygenic pleiotropy in psychiatric illness, a realm requiring further investigation.

This research demonstrates that shared genetic variation underlies the relationship between EA, ASD, and ADHD, but the effects are discordant and independent, indicating a polygenic form of pleiotropy. The involvement of microRNA mechanisms, particularly MIR19A/19B, suggests a crucial regulatory role in these complex relationships. The study emphasizes the importance of multivariate analytical approaches to unveil the intricacies of genetic architectures in psychiatric disorders and highlights the need for further research to fully elucidate the mechanisms and implications of polygenic pleiotropy in the context of neurodevelopment and complex traits. Future research might focus on larger datasets to further validate the role of identified miRNAs, investigate gene-environment interactions, and explore the implications of these findings for clinical diagnosis and intervention strategies.

While the study employed robust methodology and large datasets, certain limitations should be noted. The reliance on GWAS summary statistics means that individual-level data were not directly analyzed, potentially limiting the depth of investigation into certain aspects of genetic architecture. The reliance on case-control designs for ASD and ADHD also presents limitations in terms of generalizability, and the availability of independent samples for replication could have been broader. The exploratory analyses for adult-onset disorders should be interpreted cautiously due to the increased sample overlap.