Differences in fecal microbial metabolites and microbiota of children with autism spectrum disorders

Highlights

• Isopropanol was higher in feces of children with ASD than neurotypical children.

• Feces of children with ASD have different abundance of possible neurotransmitters.

• Higher fecal p-cresol and possibly lower GABA concentrations for children with ASD.

• Fisher Discriminant Analysis suggests a group of 5 fecal metabolites as a biomarker.

• We confirmed lower microbial diversity and Prevotella in feces of children with ASD.

Abstract

Evidence supporting that gut problems are linked to ASD symptoms has been accumulating both in humans and animal models of ASD. Gut microbes and their metabolites may be linked not only to GI problems but also to ASD behavior symptoms. Despite this high interest, most previous studies have looked mainly at microbial structure, and studies on fecal metabolites are rare in the context of ASD. Thus, we aimed to detect fecal metabolites that may be present at significantly different concentrations between 21 children with ASD and 23 neurotypical children and to investigate its possible link to human gut microbiome. Using ¹H-NMR spectroscopy and 16S rRNA gene amplicon sequencing, we examined metabolite profiles and microbial compositions in fecal samples, respectively. Of the 59 metabolites detected, isopropanol concentrations were significantly higher in feces of children with ASD after multiple testing corrections. We also observed similar trends of fecal metabolites to previous studies; children with ASD have higher fecal p-cresol and possibly lower GABA concentrations. In addition, Fisher Discriminant Analysis (FDA) with leave-out-validation suggested that a group of metabolites-caprate, nicotinate, glutamine, thymine, and aspartate-may potentially function as a modest biomarker to separate ASD participants from the neurotypical group (78% sensitivity and 81% specificity). Consistent with our previous Arizona cohort study, we also confirmed lower gut microbial diversity and reduced relative abundances of phylotypes most closely related to Prevotella copri in children with ASD. After multiple testing corrections, we also learned that relative abundances of Feacalibacterium prausnitzii and Haemophilus parainfluenzae were lower in feces of children with ASD. Despite a relatively short list of fecal metabolites, the data in this study support that children with ASD have altered metabolite profiles in feces when compared with neurotypical children and warrant further investigation of metabolites in larger cohorts.

Introduction

The prevalence of Autism Spectrum Disorders (ASD) continue to increase in the world [1,2]. Genetic, environmental and biological factors play a critical role in neurodevelopment during a mother's pregnancy and immediately after birth [3,4]. The comorbidity of GI symptoms (primarily chronic constipation and/or diarrhea) in ASD is generally estimated to be 30–50%, and is poorly understood [5,6]. The significantly increased presence of GI symptoms in ASD has motivated researchers to explore gut microbial composition in children with ASD and hypothesize about their potential role in contributing to/reflecting ASD symptoms [[7], [8], [9], [10]].

Researchers have focused on whether children with ASD possess a lack of beneficial or an increase of harmful microbes in their gut. Potentially harmful Clostridium species were observed abundant in feces of children with ASD [[11], [12], [13]]. One type of beneficial bacteria, i.e., Bifidobacterium, was reduced in children with ASD [[14], [15], [16]], while another probiotic, i.e., Lactobacillus, was reported to be present in higher concentrations in children with ASD [9,16]. Using next-generation sequencing technology, microbes that might otherwise be ignored have been detected and proposed as either potentially beneficial bacteria (e.g., Prevotella) [7] or harmful bacteria (e.g., Sutterella) [17] in the context of autism. Observations on individual bacterial taxa, however, have not always been congruent between studies. For example, Desulfovibrio and Akkermansia levels in children with ASD were found to be either higher [10,14] or lower [7,15], although multiple testing corrections was not always performed [10]. In some cohorts, no difference in gut microbiota was reported between children with ASD and their neurotypical siblings [18,19], and some studies have found differences between individuals with ASD and their siblings [14], but neurotypical siblings may be different from the general neurotypical population. As previously reviewed [20], these discrepancies may be attributed to the broad spectrum of ASD manifestations, varying molecular technologies used to investigate subject samples, geographical differences between participants (which may result in genetic and/or dietary differences), potential sub-types of gut microbiota within ASD groups, small sample size and inadequate statistical control for testing multiple-hypotheses. Considering the redundant functions of microbes [21], research must expand beyond simply cataloging microbial composition and instead also investigate microbial functions and interrelated pathways, since important microbial functions may be masked when gut microbes are considered individually.

Studies conducted in humans [5,6,16] and animal models of ASD [[22], [23], [24], [25], [26]] suggest that gut microbes and their metabolites may be linked not only to GI problems but also to ASD behavior symptoms. Hsiao et al. [22] demonstrated that 4-ethylphenylsulfate (4EPS) and indolepyruvate concentrations in serum were strikingly increased in the maternal immune activation (MIA) mouse model that displays ASD symptoms and induced anxiety-like behaviors in offspring mice. Bacteroides fragilis modulated 4EPS concentrations most likely by restoring tight-junction integrity [22]. Likewise, stress-induced corticosterone levels were reduced when stressed mice were treated with Lactobacillus rhamnosus [25]. The treatment with beneficial bacteria, Lactobacillus reuteri, ameliorated deficient social behaviors in maternal high-fat diet offspring mice, and oxytocin levels were restored [24]. Oxytocin is a crucial hormone in social behavior and cognition [27], but its therapeutic effect on ASD needs more human clinical trials to draw definite conclusions [28,29].

Despite these compelling observations, studies that examine human gut microbial metabolites are rare in the context of ASD. Functional level analyses such as metagenomics, metatranscriptomics, and metabolomics should follow studies of microbial composition: metabolomics has the advantage that it can provide information about the final products of microbial functions. Metabolomics with fecal samples can provide clues on gut microbial metabolism. However, most metabolomics studies focus on urine and blood metabolites [[30], [31], [32], [33], [34], [35]]. Only a few studies have investigated fecal metabolites and attempted to correlate them with gut microbial structure in children with ASD. Wang et al. [36] observed elevated short-chain fatty acids (SCFA) and ammonia concentrations in feces of children with ASD, although De Angelis et al. [14] and Adams et al. [16] reported reduced total SCFA in children with ASD. De Angelis et al. [14] also observed altered levels of neurotransmitters-glutamate and GABA in fecal samples. Glutamate was highest in children with autism and GABA was lowest in children with Pervasive developmental disorder not otherwise specified (PDD-NOS). In addition, phenol substances including p-cresol were higher in feces of children with autism and PDD-NOS [14], but phenol and p-cresol levels were comparable between children with ASD and controls in the other cohort [36].

In this study, we received fecal samples from children with ASD and neurotypical controls, and obtained quantitative levels of 59 fecal metabolites and gut microbial profiles using ¹H-NMR spectroscopy and pyrosequencing, respectively. We detected some fecal metabolites that may be present at significantly different concentrations between children with ASD and neurotypical children. We also performed correlation tests between fecal metabolite and individual bacterial phylotypes, and postulated their potential implications in the detection, etiology, and treatment of GI and ASD symptoms in children with ASD.