Archaeal, fungal, viral, as well as bacterial, functional makers for autism spectrum disorder in children

By Vijay Kumar MalesuReviewed by Danielle Ellis, B.Sc.Jul 11 2024

In a recent study published in Nature Microbiology, a group of researchers investigated the associations between multikingdom gut microbiome components and functional markers with autism spectrum disorder (ASD) (a complex neurodevelopmental condition characterized by social, cognitive, and behavioral impairments) through metagenomic sequencing of children's fecal samples.

Background 

ASD causes are believed to involve a combination of genetic and environmental factors. Recent studies suggest that the gut microbiome plays a significant role in ASD by modulating the gut-brain axis and neuroimmune networks. Altered gut microbiota compositions have been observed in children with ASD, and interventions like fecal microbiota transplants from healthy donors have shown symptom improvements.

Most research has focused on bacterial components, but new metagenomic technologies reveal the importance of studying archaea, fungi, and viruses. Further research is needed to fully understand the multikingdom interactions and their contributions to ASD pathogenesis.

About the study 

In the present study, children under 12 years old, both neurotypical and with ASD, were recruited from the Child and Adolescent Psychiatric Clinic between December 2021 and December 2023. ASD diagnosis was based on Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5) criteria. Neurotypical children were matched by age and sex and screened using the Chinese Autism Spectrum Quotient Child Version. Exclusions included individuals with mental retardation, neurological disorders, psychosis, depressive disorders, major medical illnesses, recent probiotic or antibiotic use, and certain medications.

Comprehensive participant profiles covered demographics, physical and psychiatric conditions, gastrointestinal (GI) disorders, medication history, parental parameters, and dietary patterns. To test marker specificity, an independent hospital ASD cohort and a community ASD cohort were established for validation, alongside cohorts for ADHD and atopic dermatitis.

Stool samples were collected using preservative media, ensuring the integrity of microbial deoxyribonucleic acid (DNA) and ribonucleic acid (RNA). DNA extraction and sequencing were performed on an Illumina NovaSeq system, followed by quality filtering and mapping to various genomes.

Microbial profiles were analyzed using Kraken 2, Bracken, and HUMAnN, with data transformed for microbiome-phenotype association assessments. Machine learning models, trained using random forest classifiers, were tested in independent validation cohorts and public datasets to ensure robustness. 

Study results 

A total of 1,627 children aged 1-13 years (24.4% female) from five independent cohorts were recruited for this study. Extensive phenotypic data, including 236 factors such as age, sex, body mass index (BMI), diet, medication, comorbidities, psychiatric disorders, GI symptoms, family characteristics, and technical factors, were collected. Metagenomic sequencing was performed on fecal samples from these children, including 709 children with ASD and 374 neurotypical controls in the discovery cohort.

An independent hospital cohort of 172 fecal samples (82 ASD, 90 neurotypical) and a community cohort of younger children (116 ASD, 60 neurotypical) were used for validation. Additionally, 237 fecal metagenomes from published datasets and non-ASD cohorts of children with attention deficit hyperactivity disorder (ADHD) (n=118) and atopic dermatitis (n=78) were analyzed for further validation and specificity testing.

At the functional level, host phenotype factors explained 17.1% and 15.7% of the variation in microbiome pathways and microbial genes, respectively. A diagnosis of ASD ranked as the top factor accounting for variation in both microbiome pathways and microbial genes. After adjusting for confounders, 27 differential Kyoto Encyclopedia of Genes and Genomes Orthology (KO) genes (23 decreased, 4 increased) and 12 differential pathways (9 negative, 3 positive associations with ASD) were identified. Ubiquinol-7 and thiamine diphosphate biosynthesis pathways were notably reduced in children with ASD compared to neurotypical children, supporting their potential role in ASD pathogenesis.

Single-kingdom microbial markers for ASD diagnosis were evaluated, with the microbial pathway model showing the strongest predictive ability (area under curve (AUC) 0.87), followed by microbial genes (AUC 0.86), bacteria (AUC 0.85), archaea (AUC 0.76), fungi (AUC 0.74), and viruses (AUC 0.68). A multikingdom model combining these features showed superior performance with an AUC of 0.91, indicating higher diagnostic accuracy for detecting ASD. The 31 microbial markers identified included several bacteria and pathways contributing to the diagnostic accuracy, such as the ubiquinol-7 biosynthesis pathway and thiamine diphosphate biosynthesis pathways.

External validation of the 31-marker panel in an independent hospital cohort maintained an AUC ranging from 0.55 to 0.87, with the ensembled model ranking highest. Further testing in a younger cohort showed consistent performance, with the model achieving an AUC of 0.89. The panel also demonstrated reproducibility across different populations, with an AUC of 0.78 in public datasets, confirming its applicability across sexes and geographical locations.

The specificity of the multikingdom marker panel was validated in non-ASD cohorts, showing lower AUC values in children with ADHD and atopic dermatitis, supporting the panel's specificity for ASD. The depletion of ubiquinol-7 and thiamine diphosphate biosynthesis genes in the gut microbiota was consistently observed across cohorts, highlighting their strong association with ASD.

Conclusions 

To summarize, this study analyzed over 1,600 metagenomes from five independent cohorts, showing that archaeal, fungal, viral species and functional microbiome pathways can differentiate children with ASD from neurotypical children.

A model based on 31 multikingdom markers achieved high predictive values for ASD diagnosis. The reproducibility across ages, sexes, and cohorts underscores their potential as diagnostic tools.