Early-life differences in the gut microbiota composition and functionality of infants at elevated likelihood of developing autism spectrum disorder

Abstract

Evidence from cross-sectional human studies, and preliminary microbial-based intervention studies, have implicated the microbiota-gut-brain axis in the neurobiology of autism spectrum disorder (ASD). Using a prospective longitudinal study design, we investigated the developmental profile of the fecal microbiota and metabolome in infants with (n = 16) and without (n = 19) a family history of ASD across the first 36 months of life. In addition, the general developmental levels of infants were evaluated using the Mullen Scales of Early Learning (MSEL) test at 5 and 36 months of age, and with ADOS-2 at 36 months of age. At 5 months of age, infants at elevated-likelihood of ASD (EL) harbored less Bifidobacterium and more Clostridium and Klebsiella species compared to the low-likelihood infants (LL). Untargeted metabolic profiling highlighted that LL infants excreted a greater amount of fecal γ-aminobutyric acid (GABA) at 5 months, which progressively declined with age. Similar age-dependent patterns were not observed in the EL group, with GABA being consistently low across all timepoints. Integrated microbiome-metabolome analysis showed a positive correlation between GABA and Bifidobacterium species and negative associations with Clostridium species. In vitro experiments supported these observations demonstrating that bifidobacteria can produce GABA while clostridia can consume it. At the behavioral level, there were no significant differences between the EL and LL groups at 5 months. However, at 36 months of age, the EL group had significantly lower MSEL and ADOS-2 scores compared to the LL group. Taken together, the present results reveal early life alterations in gut microbiota composition and functionality in infants at elevated-likelihood of ASD. These changes occur before any behavioral impairments can be detected, supporting a possible role for the gut microbiota in emerging behavioral variability later in life.

Fig. 1. Infants at elevated-likelihood of ASD show lower ELCS compared to low-likelihood infants at 36 months of age.

Infants were assessed using the MSEL test at 5- and 36-months of age. a At 5 months of age, no significant differences were observed between groups. b At 36 months of age, the elevated-likelihood group had a significantly lower ELC scores compared to the low-likelihood (p = 0.0002). c Infants in the low-likelihood group had a significant increase in their ELC scores between 5 and 36 months, while this developmental change was not observed in the elevated-likelihood group (d). Boxplots represent first (lower), median and third (upper) quartile. Wilcoxon test: ***p < 0.001. ELCS Early Learning Composite Scores, EL elevated-likelihood, LL low-likelihood.

Fig. 2. Developmental differences in gut microbial diversity and composition between infants at elevated- and low-likelihood of ASD.

a Infants at elevated-likelihood of ASD showed higher alpha diversity during the first year of life, but lower diversity thereafter compared to the low-likelihood group. Dots represent mean estimate per group and whiskers show standard errors. b At 5 months of age, the low-likelihood group presented more Bifidobacterium species (B. breve, B. Bifidum, B. Longum, and B. kashiwanohense), while the elevated-likelihood group harbored more Clostridium related species (C. clostridioforme, C. neonatale, C. difficile, and C. bolteae), B. producta, R. gnavus, and K. variicola. Positive effect sizes indicate higher abundance in the low-likelihood group, while negative effect sizes indicate a greater presence of the taxon in the elevated-likelihood group. Displayed OTUs have p < 0.05 and a circle stroke indicates if the adjusted p value was <0.25. c At 36 months of age, none of the OTUs had an adjusted p value < 0.25, but several species that were more abundant in the low-likelihood group (A. senegalensis, A, timonensis, and I. butyriciproducens), as well as different Bacteroides species that were more abundant in the elevated-likelihood group had absolute effect sizes >0.5. d At 5 months of age, infants at elevated-likelihood of ASD presented significantly less Bifidobacterium, and more Clostridioides and Clostridium compared to infants at low-likelihood of autism. e At 5 months of age, infants at elevated-likelihood of autism had a significantly lower Bifidobacterium/Clostridium ratio compared to the low likelihood one. Boxplots (d, e) represent first (lower), median and third (upper) quartile. Wilcoxon test: *p < 0.05, **p < 0.01. ELCS Early Learning Composite Scores, EL elevated-likelihood, LL low-likelihood.

Fig. 3. Developmental differences in fecal metabolomes between infants at elevated- and low-likelihood of ASD.

Sparse PLS-DA models at 5- and 36-months of age were constructed with integrated metabolites from the full ¹H NMR spectra. Metabolites with VIP > 1 were retained for the visualization of the loading plots of the first principal component. a GABA was identified as the main discriminatory metabolite driving the separation between infants at elevated- and low-likelihood of ASD at 5 months of age. b At 36 months, infants at elevated-likelihood of ASD presented more succinate, tryptophan, maltose, and glucose and less sarcosine compared to infants in the low-likelihood group. c Developmental changes in GABA and d tryptophan concentrations in infants at elevated- and low-likelihood of ASD. Group centroids are represented as * and ellipses represent 95% confidence interval. EL elevated-likelihood, LL low-likelihood.

Fig. 4. Integration of fecal microbial and metabolic profiles of infants at elevated- and low-likelihood of ASD and in vitro studies.

a Network constructed on CLR transformed OTUs and log transformed metabolites at 5 months of age using Spearman correlations. The full network was initially generated using all available OTUs and metabolites and correlations with adjusted p values < 0.25 were retained (full network available in Supplementary Fig. 6). To improve visualization, only Bifidobacterium and Clostridium species of interest and their correlated metabolites are displayed. GABA and acetate were positively correlated to the Bifidobacterium species of interest, while the Clostridium related species correlated with butyrate and glutamate. Red edges indicate positive correlations and blue edges indicate negative correlations. Edge width scaled on the absolute correlation values. b GABA can be produced through the GABA shunt, a closed-loop process that converts the α-ketoglutarate from the TCA cycle into glutamate, then GABA and finally into succinate, which re-enter the TCA cycle. GABA can also be produced from polyamines (e.g., spermidine and putrescine). Three different substrates (glutamate, putrescine, and spermidine) that can be converted into GABA were used in the Bifidobacterium and Clostridium cultures. c Bacterial isolates were cultured for 24 h in monocultures with the added substrate. B. breve and B. scardovii produced glutamate. GABA was produced by B. breve, B. adolescentis and B. scardovii, only when already present in the medium, and consumed by C. difficile and C. bolteae. Putrescine was consumed by Clostridium related species and produced by Bifidobacterium species, only when already present in the medium. No interactions with spermidine were observed. d A positive correlation was found between Bifidobacterium/Clostridium ratio and GABA concentration in fecal samples. e To determine whether the ratio of Bifidobacterium to Clostridium species. influenced the abundance of GABA, different growth conditions were investigated in vitro (Bifidobacterium spp.: Clostridium spp.; 2:1, 1:1, 1:2). GABA abundance was reduced to values comparable to that of the negative control (NC) when a greater proportion (2:1 Clostridium spp.: Bifidobacterium spp.) of Clostridium spp. were present in the cultures. Group means are plotted as line.