Gut microbiome of children with Autism Spectrum Disorder associates with clinical severity

Autism spectrum disorder (ASD) is a heterogeneous neurodevelopmental condition. Although the gut microbiome has been implicated in ASD, prior studies have insufficiently addressed clinical heterogeneity. Here, we perform shotgun metagenomic sequencing on stool samples from 720 children with ASD (median age 8 years, 87% male) and 151 matched controls.

Children with ASD are stratified into three severity groups using 14 symptom subscales covering core features and co-occurring psychopathologies. Microbiome composition is significantly associated with clinical severity and individual symptoms; however, the proportion of microbiome variance explained by these clinical factors is modest (R2 = 0.0017–0.0065). Children with higher severity show reduced abundance of Lachnospiraceae bacterium AM48_27BH, a short-chain fatty acid producer, together with reduced predicted capacity of GABA-related modules (GABA synthesis II, putrescine degradation and GHB degradation), while NAD+ salvage pathways show trend-level associations with severity.

Sensory atypicalities have the largest number of associated microbial markers and show trend-level links to reduced predicted capacity for degradation of excitatory neurotransmitters. These findings demonstrate that gut microbiome composition varies across the ASD severity spectrum and that different symptom domains differ in their relevance to microbiome alterations. Whether alterations in GABAergic metabolism contribute to clinical severity, and whether microbial features are preferentially linked to sensory atypicalities, remain to be tested in mechanistic and longitudinal studies.

The authors would like to thank the Child and Adolescent Psychiatric Team of Alice Ho Miu Ling Nethersole Hospital for their effort in recruiting participants for the present study. H.M.T. is an Adjunct Professor at the Nanjing Medical University, PR China and a Visiting Associate Professor at the National University of Singapore, Singapore. This study was funded in part by the InnoHK initiative of the Innovation and Technology Commission of the Hong Kong Special Administrative Region Government, and the D.H.

Chen Foundation (F.K.L.C. and S.C.N.). S.C.N. has received research funding from the New Cornerstone Science Foundation (grant no. L. was supported by Innovation and Technology Commission (ITC) Research Talent Hub (RTH) program.

Authors affiliated with MagIC are partially supported by the InnoHK initiative of the Innovation and Technology Commission of the Hong Kong Special Administrative Region Government. These authors contributed equally: Oscar WH Wong, Suisha Liang. These authors jointly supervised this work: Siew C.