Gut microbiota changes in children with autism spectrum disorder: a systematic review

Autism spectrum disorder (ASD) refers to a developmental and neurobehavioral condition characterized by deficits in social communication and social interaction across multiple contexts with restricted, repetitive patterns of behaviour, interests, or activities [1]. Recent data suggest that as many as 1 in 59 children are diagnosed with ASD, although other reports not using parental report and school age children generally show a prevalence of 1% globally, with little regional variations in developed countries within North America, Western Europe, Central Latin America, and Asia Pacific [2‐7].

There is no single known cause for all ASD-related behaviors. Current research alludes to multifactorial etiologies including genetic risk factors, de novo mutations, gene-environment interactions, and environmental factors such as in utero exposures and perinatal events [2, 8]. Due to reports suggesting that children with ASD have increased prevalence of gastrointestinal symptoms including constipation, diarrhea, and abdominal discomfort, researchers have started to examine the differences in gut microbiome composition in these children [9‐12].

Longitudinal studies on adults with ASD indicate that 37 to 59% have poor outcomes [13]. The average lifetime cost of supporting an individual with ASD is estimated to be at least USD$1.4 million in the United States and £0.92 million in the United Kingdom [14]. When a child has concurrent intellectual disability, this cost rises to USD$2.4 million and £1.5 million, respectively [14]. While autism-specific behavioral therapies have strong data supporting outcome improvement, there has not been reliable evidence on the effectiveness of environmental modifications including diet, antifungals, fecal microbiota transplants, heavy metal chelation, and vaccine avoidance. The intention of this review is not to discuss potential ways for intervention through gut microbiome modulation. Rather, it is to take a closer look at whether the plethora of literature published provides consistent evidence on features of gut microbiome alterations associated with ASD and to establish the strength of evidence.

In total, 26 papers [30‐55] were selected for this systematic review (PRISMA Fig. 1; search criteria Additional file 1: Table S1). Two studies [52, 53] belonged to the same cohort but were both included in the synthesis of this systematic review because the gut microbes examined were different.

A summary of the papers used is provided in Table 1, including methodological techniques and the rigor of how ASD had been diagnosed. The detailed results from each study are presented in different tables (Tables 2, 3, 4 and 5), grouped according to bacterial taxonomic classification for easy comparison.

In the works reviewed, children diagnosed with ASD have various forms of dysregulation of the microbiome when compared to siblings or unrelated children without the ASD profile. Since each individual study describes a restricted and different bacterial composition, direct comparison between strains with similar classification is limited. However, the data follow a more consistent pattern for a few strains. Relative and absolute Clostridia clusters I, II, and XI are not found to be decreased in the gut microbiome of children with ASD when compared to those without. Similarly, the relative and absolute abundances of Firmicutes at the phylum level, Streptococcus at the genus level, Prevotella species, and Bifidobacterium species are not increased in children with ASD versus non-sibling controls. Of note, in all studies reviewed including intervention ones, the absolute abundance of Bifidobacterium species is significantly decreased in children with ASD compared to non-sibling controls, and the species is also significantly increased after intervention. Despite some recognizable patterns, the majority of microorganisms reviewed from phyla to species have disparate results across different studies. Hence, to date, gut microbial composition by itself does not provide a predictive biomarker for ASD and the single technology of high-throughput sequencing will need to be integrated with multiple sources of omics data (e.g. proteomics, transcriptomics, metabolomics, microRNAs and exosomes) to produce potential signatures for the spectrum of symptoms in individuals with ASD.

Although a direct causal mechanism of microbiome in the etiology of ASD in humans cannot be validated at this time, the gut microbiome likely alters brain functions through various other mechanisms, including environmental factors (e.g. in utero exposure to infection, maternal conditions, and medications), host genetics, host immune response regulation [12, 57, 58], excretion of metabolites such as tyrosine analogues, p-cresol, 4-ethylphenylsulfate, indoles, lipopolysaccharides and free amino acids [59-62], regulation of neurotransmitters and their receptors [21, 63], or neuroactive compounds [61, 62, 64].

Alterations of the host immune responses by gut microbiota are closely linked to ASD-related symptoms. The implicated cytokine pathways include, and not limited to, IL-5, IL-15, IL-17, IL-17a, IL-10, IL-1b, TNF-α, TGF-β1 and IFNγ [12, 18, 65, 66]. Interestingly, the gut microbiota has recently been shown to influence the immune system directly via activation of the vagus nerve [67, 68]. Furthermore, gut microbiota-derived short-chain fatty acids (SCFAs), such as propionic acid [69, 70] and butyric acid [71, 72], produced by bacterial fermentation of carbohydrates have immunomodulatory properties, e.g. upregulating genes associated with immune activation [69], regulating T cells and cytokine production [70], microglia homeostasis during developmentally sensitive periods [73], and neuronal excitability [74], and have recently been used in vivo in the treatment of inflammatory conditions such as inflammatory bowel diseases [75]. In addition to understanding microbiome composition differences in children with ASD, there is a need to investigate the patterns of dysregulation in their immune responses as well as to look more upstream at the maternal immune response during pregnancy. Prior literature has substantiated that infections during pregnancy have been correlated with increased frequency of neurodevelopmental disorders in offspring [16, 17, 76‐78]. Specifically, there is an association between ASD and maternal infection requiring hospitalization during pregnancy, elevated C-reactive protein, and a family history of autoimmune diseases. Thus, future studies will need to explain the bidirectional and possibly transgenerational roles of microbiome alterations and immune pathways on behaviours.

A promising development in this field points to the need to consider interactions between host genetics and microbial composition. Differences in microbiome diversity have been shown to be partially attributed by genotype and sex [79‐83]. In a rodent model, Tabouy et al. [84] used the Shank3 KO mice and demonstrated that specific bacterial species (i.e. L. reuteri) were sensitive to an autism-related mutation, were decreased in abundance, and positively correlated with the expression of gamma-aminobutyric acid (GABA) receptor in the brain. Treatment with L. reuteri resulted in an increase of both GABA receptor gene expression and protein levels in brain regions of mice, which also corresponded to improvements in social engagement. It is noteworthy to mention that there is a paucity of research examining the interactions of host genetics and microbial dysregulation in humans with ASD. Perhaps it is worthwhile to isolate individuals with the same autism-related genotype and investigate for potential dysbiosis in their microbiome, along with changes in gene expression and/or in brain structure. Likewise, studies suggesting therapeutic potential for probiotic treatment has currently looked at individuals with the ASD profile as a whole. Future studies may consider subgroup analysis (e.g. responders vs non-responders) to understand the potential differences between subgroups.

Lastly, the gut microbiome’s contribution to neurological development and regulation has been implicated and demonstrated in animal models [85]. For example, gnotobiotic animals demonstrate heightened hypothalamic-pituitary response, elevated plasma adrenocorticotropic hormone and corticosterone, and reduced brain-derived neurotrophic factor (BDNF) expression levels in the cortex and hippocampus [86]. Absence of colonization results in differential expression of proteins involved in synaptogenesis [87] and atypical development [88]. Subsequent microbial colonization reverses these processes. Furthermore, gut microbiota manufactures neuroactive chemicals and influences levels of circulating 5-hydroxytryptamine (5-HT) and serotonin, thereby altering fetal neuronal cell synaptogenesis [89] and neuronal morphogenesis [90], respectively. Although mounting evidence is accumulating for microbiome’s role in neural development, the precise nature of how multiple systems interact or overlap remain poorly defined.

The variety of protocols for sampling and characterization of microbial ecology among included studies also warrant discussion. Since the human microbiome exhibits considerable spatial and temporal variability, single samples obtained from a specific anatomical site may not be representative of its true diversity at any given time and may especially fail to capture rarer or less abundant taxa. Heterogeneity also exists with regards to workflows for specimen storage and processing, and factors such as shipping time and ambient temperature are established to influence the microbial composition in poorly-handled specimens. In terms of experimental procedures, high-throughput nucleic acid-based interrogation represent the most common technique used in included studies. However, interpretation of the collective results across studies may be constrained by the lack of standardization of experimental protocols and is further hampered by suboptimal inter-platform agreement and measurement reliability. Finally, with regards to the comparison of microbial constituents between ASD cases and controls, the issue of multiple testing looms large. For these and other reasons, it is essential that the salient findings summarized in the present review are externally validated by independent laboratories.

Autism spectrum disorder is a neurobiological disorder which is potentially a result of disruptions in normal brain growth very early in development. The studies reviewed have not reported on the birth or pre-diagnosis microbiome of children with ASD. Instead, studies generally report bacterial diversity after children are diagnosed with ASD. It is hard to determine the directionality of the association between microbiome differences and dietary habits. It is possible that children with ASD have greater likelihood of having more unique preferences in certain diets and this limited diet variety may account for microbiome differences. One study suggests that children with ASD may have an increased intake of chia seeds in smoothies, which is associated with specific microbiome findings [48]. Children with ASD are also sometimes placed on non-specific gluten-free, casein-free diets, which easily change one’s gut microbiome composition.

The literature currently lacks prospective studies that follow a child from prior to ASD diagnosis, preferably as an infant, with repeated objective assessment of ASD symptomatology and its trajectory at the same time as stool collection for microbiome. Given the long duration of such prospective studies, it is unlikely that the same environmental conditions such as diet, exposure to antibiotics or other medications, pets in the home, exposure to livestock, and limits on travel may be imposed on the participants, which will further complicate interpretation of microbial samples. Nonetheless, ongoing investigations, such as the National Institutes of Health (NIH) Environmental influences on Child Health Outcomes (ECHO) study, have already started the collection of infant microbiotas with planned serial samples. When these studies are complemented with mechanistic experiments in animal models, they can be powerful in giving insight into human biology.

Research studies of this kind requires the involvement of professionals with clinical expertise in children with ASD. In this review, only a few studies have involved developmental specialists and psychologists who are apt in monitoring changes in ASD symptoms [32, 48]. Parent-reported questionnaires, while important to provide a summary of behaviors within the home setting, are not as objective compared to experienced observations in standardized assessments by psychologists or developmental-behavioral pediatricians. The heterogeneous nature of ASD is also a challenge in review studies. Further, the diagnostic criteria for ASD and classification of ASD into subtypes have been updated in 2013. Older studies classifying children into Asperger Disorder, PDD-NOS, and Autistic Disorder are based on the older edition of DSM-IV and not the DSM-5. There are studies to support that these diagnoses do not translate directly to an ASD diagnosis on DSM-5 [1, 91]. Future studies should consider a rigorous diagnosis of ASD and a description of the variety of ASD symptomatology in the participants, along with documentation of diet, intake of probiotics, antibiotics, travels, and episodes of gastrointestinal symptoms.

In summary, we provide data to show that the current literature on dysbiosis in children with ASD does not provide a predictive signature for the condition or symptoms. However, researchers may take note of the general consistencies found in composition changes of Prevotella, Firmicutes as a whole, three Clostridia clusters, C. perfringens, and Bifidobacterium in children with ASD to design future studies and to look deeper into the influence of these microorganisms on multi-system pathways.

The relationship of the microbiome and social behaviors is multifaceted and complex involving not only environmental factors and immune responses, but also the genetic background of the host. Further suggestions for future research include confirming the potential therapeutic qualities of specific microbial reconstitution in humans, dissecting the overlapping pathways between the microbiome and various organ systems, as well as the use of microbial metabolome and other omics platforms to study this topic.