We have previously published our protocol for this systematic review, consistent with the Preferred Reporting Items for Systematic review and Meta-analysis (PRISMA) protocols checklist (Liberati et al., 2009; Shamseer et al., 2015). The systematic review protocol (Hill et al., 2021) is registered on PROSPERO (ID: CRD42020158734).

We appraised available evidence surrounding the gut microbiota in ASD, ADHD, and RETT (Hill et al., 2021). Relevant studies were identified using PubMed, Cochrane Library, and OVID electronic databases. Articles published between January 1980 and December 2021 were considered for inclusion and analysis.

We implemented a rigorous search strategy that involved two rounds of literature review, using a three-step method in each round (Hill et al., 2021). The first round involved a review of articles from the database search results; the second round involved a review of works cited by those articles identified in the first round. Titles, abstracts, and full texts were independently screened by at least two authors (LH, JP, MF, VC, EH, MM, or NP). Articles were excluded if they were unrelated or met exclusion criteria (Supplementary Table S1). Inclusion/exclusion criteria were slightly modified for ADHD and RETT studies. The age of the patients was >18 for three of the seven included ADHD studies. In all included studies on RETT, the age range was between 5 and 36 years. We decided to include these studies as they met other inclusion criteria, and they were the only available studies that included human participants (Supplementary Table S1). Two reviewers (MF and LH) completed data extraction using a pre-piloted Microsoft Excel data extraction form. Data extraction was subsequently checked for accuracy and completeness by members of the study team (NP, JP, EH, MM, and VC). One further study was included upon consultation with external content experts (KM and EA) (Luna et al., 2017).

We assessed quality of the included studies using the appropriate Critical Appraisal Skills Program (CASP) checklist based on study methodology. CASP was selected for its quality of appraisal of health-related studies, assessment of transparency of research practice and reporting standards (Long et al., 2020). Level of evidence was assessed using established methods and a “level of evidence score” was established (Obremskey et al., 2005). Two independent reviewers assessed each study and, if discrepancies occurred, a third independent reviewer was included. We applied a semi-quantitative appraisal of CASP checklists based on an approach by Nichols et al. (2020) where study quality was scored out of 20.

The initial search strategy outlined in our protocol (Hill et al., 2021) identified 166 articles. After initial screening and removal of duplicates, 131 articles were screened for inclusion. Finally, after review with external content experts 33 original studies that described the association between intestinal microbiota and ASD, ADHD or RETT in children and adults were included (Figure 1). Twenty-six studies were assessed using the CASP Checklist for Case–Control Studies (Supplementary Table S2), five using the CASP Checklist for Cohort Studies (Supplementary Table S3), and two for RCT studies (Supplementary Table S4).

Scores for each study ranged from 7 (low methodological quality) to 20 (high methodological quality). Two studies (6.9%) had a total score below 10, suggesting weak methodology (Zhang et al., 2018; Ahmed et al., 2020). The majority of studies (75.9%) had a total score between 15 and 20, and eight studies achieved the highest score for methodological quality (De Angelis et al., 2013; Pärtty et al., 2015; Tomova et al., 2015; Strati et al., 2016, 2017; Kang et al., 2018; Zhai et al., 2019; Wan et al., 2020).

Eighteen case–control studies assessed the intestinal microbiota in ASD, seven in ADHD (four cohort studies, three case–control), and three in RETT (one cohort, two case–control; Tables 1–4; Supplementary Tables S5–S8). Geographically, the studies included patients from different countries including the United States, China, India, Australia, Egypt, Germany, Italy, and The Netherlands. The studies included in our review encompassed a total of 2,251 subjects with ASD (among which 10 subjects were diagnosed with PDD-NOS; Supplementary Tables S5, S6), 272 subjects with ADHD (Supplementary Table S7) and 101 subjects with RETT (Supplementary Table S8). Control groups included 1,059 sex- and age- matched neurotypical (NT) subjects in studies of ASD and PDD-NOS, 321 NT subjects in studies of ADHD, and 60 NT subjects in studies of RETT (Supplementary Tables S5–S8). Ages of cases and controls spanned 2 to 18 years old for studies of ASD (Tables 1, 2; Supplementary Tables S5, S6), 6 to 44 years for studies of ADHD (Table 3; Supplementary Table S7), and 5 to 36 years for studies of RETT (Table 4; Supplementary Table S8). Subjects with ASD, PDD-NOS, ADHD and respective NT controls were males and females, while subjects with RETT and their NT controls were all females.

Studies examining the effects of microbiota- or diet-targeted interventions on behavioral outcomes and microbiota composition in subjects with neurodevelopmental disorders are summarized (Table 5; Supplementary Table S9). Two studies assessed the effect of probiotic administration (Tomova et al., 2015; Shaaban et al., 2018), and one study explored the effects of FMT (Kang et al., 2017) on gut microbiota, GI symptoms, and neurodevelopmental outcomes in individuals with ASD. One study assessed the effect of dietary micronutrient supplementation on gut microbiota composition and behavioral symptomatology in subjects with ADHD (Stevens et al., 2019). One study assessed the effect of perinatal administration of probiotic Lactobacillus rhamnosus GG in preventing the onset of ASD or ADHD in childhood (Pärtty et al., 2015). The characteristics of subjects included in the intervention trials are summarized (Supplementary Table S9).

Samples used in the selected studies were derived from stool (31 studies), rectal tissue (1 study), and duodenal tissue (1 study). Samples from seven studies were stored at 4°C for ≤24 h before being frozen for storage (Pärtty et al., 2015; Kantarcioglu et al., 2016; Aarts et al., 2017; Prehn-Kristensen et al., 2018; Stevens et al., 2019; Wang et al., 2020; Szopinska-Tokov et al., 2021). All other studies’ samples were frozen immediately upon collection. Samples from five studies were stored with preservation agents at the time of collection (De Angelis et al., 2013; Son et al., 2015; Rose et al., 2018; Richarte et al., 2021; Thapa et al., 2021). Further details about DNA extraction, sequencing technique, reference database and sequence corrections are described (Supplementary Table S10).

Measurements of neurodevelopmental outcomes in ASD, ADHD, and RETT are summarized in Supplementary Tables S5–S9.

Ten studies compared stool-associated bacterial communities of ASD subjects to NT controls (with no first-degree family relationships to ASD subjects, Table 1), while four studies used NT first-degree siblings of ASD subjects, and two studies included both NT first-degree siblings of ASD subjects and unrelated NT controls (Table 2). Bacterial α-diversity indices were evaluated in five of the 10 studies in which NT controls were unrelated to ASD individuals. Four studies reported reduced, and one study reported increased bacterial α-diversity of ASD-associated microbiota (Table 1). Six of the 10 studies, in which NT controls were unrelated to ASD individuals, evaluated β-diversity indices and showed differences in gut bacterial profiles between ASD subjects and NT controls (Table 1).

Conflicting α- and β-diversity outcomes occurred in three of the four studies in which NT controls were siblings of ASD children as no differences in bacterial α-diversity or composition (β-diversity) were found between groups in these studies (Table 2). No differences in α- or β-diversity were again reported by Yap et al. (2021) and Ahmed et al. (2020), who compared bacterial α- and β-diversities of ASD subjects with both NT siblings and unrelated NT controls. De Angelis et al. (2013) reported significantly increased α-diversity in both ASD and PDD-NOS subjects, and differences in relative abundance of bacterial communities between subjects with ASD, PDD-NOS, and their NT first-degree siblings (Table 2).

The two studies evaluating mucosal biopsy-associated microbiota included only unrelated NT controls (Table 1). Kushak et al. (2017) did not report any differences in bacterial diversity of duodenal mucosa-associated microbiota between ASD subjects and NT controls (Kushak et al., 2017). Luna et al. (2017) reported differences in β-diversity between rectal mucosa-associated microbiota of ASD and NT controls (Luna et al., 2017).

Altered bacterial diversity in ASD children compared to unrelated NT controls might suggest the presence of an ASD-related gut microbiota phenotype (Brito et al., 2019). However further studies with larger sample sizes and with both siblings and family unrelated NT controls are warranted to determine whether children with ASD have an altered microbiota due to the disease.

Multiple studies reported distinctive alterations in bacterial taxa among subjects with ASD; however, these results were not consistent. A reduction in the relative abundance of the genus Prevotella in ASD was reported across five studies assessing stool-associated microbiota (two studies included NT siblings), and in the study by Kushak et al. (2017), which investigated duodenal mucosa-associated microbiota (Tables 1, 2). The results from bacterial cultures in the study by De Angelis et al. (2013), reported increased relative abundance of Prevotella in ASD compared to NT subjects (Supplementary Table S6). Faecalibacterium and Streptococcus levels were decreased in ASD subjects in three (De Angelis et al., 2013; Kang et al., 2018; Pulikkan et al., 2018) and four studies (De Angelis et al., 2013; Kushak et al., 2017; Kang et al., 2018; Zhang et al., 2018), respectively (Tables 1, 2). Bacterial culture, 16S rDNA sequencing, and targeted PCR revealed higher levels of genus Clostridium in stool samples of ASD patients in the studies by De Angelis et al. (2013) (Table 2; Supplementary Table S6) and Tomova et al. (2015) (Table 5; Supplementary Table S9). In ASD-associated microbiota, Lactobacillus was increased in three studies (Tomova et al., 2015; Strati et al., 2017; Pulikkan et al., 2018) and decreased in De Angelis et al. (2013) (Tables 1, 2, 5). Three studies reported increased abundance of genus Bacteroides in ASD stool samples (De Angelis et al., 2013; Zhai et al., 2019; Ahmed et al., 2020), while Kushak et al. (2017) showed decreased abundance of Bacteroides in duodenal mucosa-associated microbiota from ASD subjects (Tables 1, 2).

At the species level, increased Bacteroides fragilis and B. vulgatus (De Angelis et al., 2013), Clostridium sp. (De Angelis et al., 2013; Finegold et al., 2017)., Lactobacillus sp. (Tomova et al., 2015). and reduced Faecalibacterium prausnitzii (De Angelis et al., 2013), Romboutsia timonensis (Yap et al., 2021), Prevotella sp. (De Angelis et al., 2013) were detected in ASD-associated microbiota (Tables 1, 2). Conflicting results were reported for Bifidobacterium abundance in ASD subjects compared to NT first-degree siblings. Decreased and increased abundance of the genus Bifidobacterium in ASD subjects compared to NT siblings were reported by De Angelis et al. (2013) and Pulikkan et al. (2018), respectively (Table 2). At the species level, reduced B. adolescentis, and Bifidobacterium sp. were reported by De Angelis et al. (2013) and Ahmed et al. (2020) (Table 2), whereas Tomova et al. (2015) observed an increased abundance of Bifidobacterium sp. in their cohort of ASD children (Table 5; Supplementary Table S9). In all studies there was no difference in the abundance of Bifidobacterium between children with ASD and unrelated NT subjects.

In mucosa-associated microbiota from duodenal biopsies of subjects with ASD, significant alterations occurred at the genus level, with reduced abundance of Prevotella, Bacteroides, Streptococcus, Neisseria, and increased abundance of Devosia, Oscillospira, Actinomyces, and Peptostreptococcus. These findings recapitulate some of the changes observed in the stool-associated microbiota from ASD subjects (Kushak et al., 2017). Positive correlations between Bacteroides, Faecalibacterium, and Clostridium and the activity of duodenal enzymes such as maltase, lactase and sucrase were also observed in ASD (Kushak et al., 2017; Table 1).

The study of Luna et al. (2017) reported increased abundance of bacterial species from Clostridiales in rectal mucosa-associated microbiota from ASD subjects. In particular, bacterial species Clostridium lituseburense, Lachnoclostridium bolteae, Lachnoclostridium hathewayi, Clostridium aldenense, and Flavonifractor plautii (Table 1), were significantly more abundant in ASD microbiota compared to NT controls. This study also observed a decrease in Dorea formicigenerans and Blautia luti (Table 1) as well as reduced Sutterella species (Table 1) in ASD-associated microbiota. Together these data showed major perturbations in the abundance of bacterial genera and species belonging to the phyla Firmicutes and Bacteroidetes in ASD-associated microbiota.

Few studies assessed changes in inferred bacterial function using 16S profiles and Phylogenetic Investigation of Communities by Reconstruction of Unobserved States (PICRUSt) to predict the presence of Kyoto Encyclopedia of Genes and Genomes (KEGG) orthologs across functional and metabolic pathways (Box A1; Kanehisa, 2000). Zhang et al. (2018) observed significant enrichments of several bacterial genes associated with D − Arginine and D − ornithine metabolism, ether lipid metabolism, bacterial chemotaxis, phosphotransferase system (PTS) and flagellar assembly genes in ASD subjects. Steroid hormone biosynthesis, glycosaminoglycan degradation and lipoic acid metabolism were more abundant in NT controls (Zhang et al., 2018; Table 1). Conversely, Kang et al. (2018) assessed genes in the metabolic pathways associated with p-cresol, vitamin K, and GABA but did not find any significant differences between ASD subjects and NT controls (Table 1). The study by Rose et al. (2018) differentiated NT or ASD subjects with GI symptoms (NT-GI and ASD-GI, respectively), compared with NT or ASD without GI symptoms (NT-no-GI and ASD-no-GI, respectively; Supplementary Table S5). Amino sugars and nucleotide sugars pathways related to the gut microbiota were significantly upregulated in ASD-GI compared to ASD-no-GI subjects. Starch and sucrose metabolism pathways were downregulated in ASD-no-GI subjects compared to ASD-GI subjects (Rose et al., 2018; Table 1). Zhai et al. (2019) described 40 metabolic pathways found to be statistically significantly different between subjects with ASD and NT controls. Bacteroides, Oscillospira and Sutterella were positively correlated with carbon fixation pathways in prokaryotes, the citrate cycle, and gas degradation (pathways downregulated in ASD), and negatively correlated with porphyrin and chlorophyll metabolism (pathways upregulated in ASD). Ether lipid metabolism and sporulation were upregulated in ASD and negatively associated with Parabacteroides. A negative correlation was found between Bacteroides and Sutterella and carotenoid biosynthesis metabolism (Table 1). Yap et al. (2021) assessed the potential functional profiling of ASD-associated microbiota by using the Microba Gene and Pathway Profiler (MGPP) tool against the Microba Genes (MGENES) database. The study assessed metabolic pathways associated with Romboutsia timonensis, as this species was found to be less abundant in ASD subjects. The predictive analysis showed reduced abundance of microbial genes responsible for the metabolism of amino acids (L-glutamine, L-lysine, L-methionine, and L-threonine), purines and pyrimidines, carbohydrates (galactose), bacterial spore germination and dsDNA digestion in ASD subjects compared to NT siblings and unrelated NT controls (Table 2). These results suggest potential alterations in the transcription of microbial genes associated with ASD. Further analyses (e.g., metatranscriptomics) will be required to evaluate protein expression associated with these microbial differences.

Two studies investigated possible correlations between intestinal microbiota changes and ASD severity. Ahmed et al. (2020) assessed ASD subjects with mild–moderate (n = 35) and severe ASD (n = 6). No statistically significant differences were found across most taxa, except for Firmicutes taxa which were significantly higher in patients with severe ASD (Figure 2; Table 2). This was also described by Tomova et al. (2015), who found a lower Bacteroidetes to Firmicutes ratio in children with severe ASD, compared to children with milder ASD and NT controls, respectively (Supplementary Table S9). This study also found children with severe ASD had higher Desulfovibrio and Clostridia compared to children with a milder ASD phenotype (Tomova et al., 2015) (Supplementary Table S9). A strong correlation between Desulfovibrio and the ADI restricted/repetitive behavior subscale score (a measure of ASD severity) was reported by Tomova et al. (2015) (Figure 2; Supplementary Table S9). These studies suggest a possible association between the phylum Firmicutes and ASD severity.

Three studies investigated alterations in fungi composition of stool samples from ASD subjects (Table 1). Strati et al. (2016) observed a significant difference in the composition of the gut mycobiome between subjects with ASD and unrelated NT controls with enrichment in relative abundance of the genus, Candida in ASD. The study by Kantarcioglu et al. (2016) also reported increased levels of Candida species in stool specimens from ASD subjects, and Hughes and Ashwood (2018) observed an increased concentration of anti-C. albicans immunoglobulin (IgG) in serum samples (Hughes and Ashwood, 2018; Figure 2). Further studies are required to assess the role of fungi in ASD.

Five studies did not find any significant differences in bacterial α- or β-diversities indices between ADHD subjects and NT controls, including those studies where the age of the subjects was >18 years (Aarts et al., 2017; Zhai et al., 2019; Szopinska-Tokov et al., 2020; Wan et al., 2020; Richarte et al., 2021; Szopinska-Tokov et al., 2021; Table 3). Prehn-Kristensen et al. (2018) observed significantly decreased richness and evenness in the gut microbiota of ADHD subjects and their mothers compared to unrelated NT controls. The same study observed significant differences in bacterial profiles between ADHD subjects and NT controls, and between ADHD subjects and their mothers (Prehn-Kristensen et al., 2018; Table 3). Wang et al. (2020) found changes in bacterial α-diversity in ADHD subjects and NT controls, with no significant differences in bacterial composition overall (Table 3). Only two studies evaluated first-degree NT siblings of ADHD subjects, however NT siblings were mixed with unrelated NT controls and data were analyzed as a single group (Aarts et al., 2017; Szopinska-Tokov et al., 2021). Diversity measures showed no differences in bacterial communities between ADHD subjects and NT controls. Changes in α-diversity observed in the studies by Prehn-Kristensen et al. (2018) and Wang et al. (2020) need further validation using larger cohorts.

The principal phyla characterizing the bacterial kingdom of the human microbiota, Firmicutes, Bacteroidetes, Actinobacteria, and Proteobacteria, were identified across all ADHD studies. One study found a significant increase of Actinobacteria, accompanied by a decrease of Firmicutes in ADHD gut microbiota (Aarts et al., 2017). Further taxonomic analyses showed that taxa from families and genera belonging to the phyla Firmicutes and Bacteroidetes were primarily associated with ADHD diagnosis (Figure 2; Table 3). At least two studies reported reduced abundance of bacteria from the family Oscillospiraceae (Aarts et al., 2017; Wan et al., 2020) (renamed from Ruminococcaceae (Tindall, 2019)), and the genera Faecalibacterium (Jiang et al., 2018; Wan et al., 2020) and Coprococcus (e.g., Coprococcus_2) (Aarts et al., 2017; Szopinska-Tokov et al., 2021) within the ADHD gut microbiota (Table 3). Within the phylum Bacteroidetes, the abundance of bacteria from the genera Bacteroides, Alistipes, Odoribacter, and Parabacteroides, as well as from the family Bacteroidaceae, were reported to be increased in ADHD subjects (Table 3). Increased abundance of Egghertella, Neisseria (phylum Proteobacteria), Bifidobacterium (phylum Actinobacteria), and Fusobacterium (phylum Fusobacteria) were also observed in the gut microbiota of ADHD subjects. At the species level, two studies reported increased amounts of Bacteroides uniformis and Bacteroides ovatus in ADHD gut microbiota (Aarts et al., 2017; Wang et al., 2020). Increased relative abundance of Bacteroides vulgatus, Bacteroides caccae, Paraprevotella xylaniphila, Odoribacter splanchnicus (phylum Bacteroidetes) as well as Bifidobacterium longum, Bifidobacterium adolescentis (phylum Actinobacteria), and Sutterella stercoricanis (phylum Proteobacteria) were observed in ADHD-associated microbiota (Table 3). Each set of findings were reported in only one study; further validation across larger samples is needed.

Changes in inferred metabolic function of the intestinal microbiota in ADHD subjects were assessed by two studies using PICRUSt-KEGG analysis. Wan et al. (2020) found an enrichment in the enzymatic pathways involved in dopamine turnover at the postsynaptic neuron as well as in genes encoding the catalytic subunit of protein phosphatase-1 (PP1), threonine synthase, and 6-pyruvoyl-5,6,7,8-tetrahydropterin, and depletion of the gene encoding 4-hydroxy threonine-4-phosphate dehydrogenase in ADHD-associated microbiota (Wan et al., 2020; Table 3). PP1 catalytic subunit has been described to increase synaptic sodium ion flux and may impact transmembrane sodium/chloride-dependent transporter dopamine receptors (Navratna and Gouaux, 2019). Abnormal Enterococcus and Odoribacter were found to be associated with alterations in the dopamine metabolic pathway in ADHD patients compared to NT controls (Wan et al., 2020). Aarts et al. (2017) reported increased abundance of the enzyme cyclohexadienyl dehydratase (CDT) in the gut microbiota of ADHD subjects. CDT is responsible for the production of phenylalanine, a precursor for the synthesis of dopamine and noradrenaline. Predicted CDT was found to be positively correlated with Bifidobacterium in ADHD. Reward anticipation, a key symptom in ADHD mediated by dopamine, was negatively correlated with predicted CDT in ADHD subjects (Aarts et al., 2017; Table 3). This study suggests that Bifidobacterium may enhance production of phenylalanine via the enzyme CDT in patients with ADHD. Several studies have assessed the role of phenylalanine in ADHD with conflicting results, showing decreased concentrations of phenylalanine in ADHD and no association (Bornstein et al., 1990; Baker et al., 1991; Bergwerff et al., 2016). In contrast, among patients with the metabolic condition, phenylketonuria, toxic phenylalanine build-up may lead to ADHD symptoms (Stevenson and McNaughton, 2013; Ashe et al., 2019; Beckhauser et al., 2020).

None of the studies in our review stratified the microbiota data according to the ADHD severity. Some correlations between bacterial taxa and ADHD symptoms, however, were reported. The relative abundance of B. uniformis, B. ovatus, and S. stercoricanis correlated with ADHD symptoms (Wang et al., 2020), and Coproccocus_2 with inattention symptoms, even though this association was not statistically significant (p = 0.055) (Szopinska-Tokov et al., 2020) (Figure 2; Table 3). A negative correlation occurred between hyperactivity scores and α-diversity in ADHD patients (Prehn-Kristensen et al., 2018), and between hyperactivity and impulsivity and Faecalibacterium levels (Figure 2; Table 3). No correlations were identified between microbiota changes and attention deficit and impulsivity (parental rating assessed by the German ADHD Rating Scale, FBB-HKS), or clinical symptoms assessed by the CBCL. A positive correlation between Bacteroides and levels of hyperactivity and impulsivity was observed (Figure 2; Table 3). The authors further assessed intestinal microbiota of mothers of ADHD patients and found patterns of similarly reduced α-diversity, which correlated with maternal ADHD symptoms as compared to NT maternal controls. There were no differences between α-diversity and ADHD symptoms between fathers of ADHD patients and fathers of NT controls (Prehn-Kristensen et al., 2018). The study of Richarte et al. (2021) did not find any correlation between bacterial taxa (family, genus) and ADHD rating scale scores; however, some correlations between bacterial taxa among ADHD subjects were identified (Table 3). Together these findings suggest relationships between ADHD symptoms and Bacteroides, yet the biological mechanisms behind these remain unknown.

Two studies from Italy showed a reduction in α- and β-diversity between RETT subjects and NT controls (Strati et al., 2016; Borghi et al., 2017). The study by Borghi et al. (2017) did not reach statistical significance. The authors suggested that this result was affected by the small sample size (Table 4). A third study, conducted in United States, did not find any differences in α- and β-diversities between RETT and NT controls; however, changes in bacterial richness, diversity, and composition were observed within RETT subjects based on pubertal stage, severity of disease and dietary habits (Thapa et al., 2021) (Table 4). These studies indicate a possible change in bacterial diversity associated with RETT. However further analyses in larger cohorts are required to confirm these results.

Most changes in the abundance of bacterial taxa associated with RETT occurred within the Firmicutes and Bacteroidetes phyla, followed by Actinobacteria and Proteobacteria phyla (Figure 2; Table 4). In general, RETT microbiota was characterized by reduced abundance of bacterial genera from the family Oscillospiraceae (Ruminococcus, Faecalibacterium, Oscillibacter, Sporobacter) and several genera from the phylum Bacteroidetes (Prevotella, Barnesiella, Alistipes, Odoribacter, Butyricimonas).

Conflicting findings were reported for the genus Bacteroides (Strati et al., 2016; Borghi et al., 2017). An increased relative abundance of the genera Lactobacillus, Enterococcus (phylum Firmicutes), Egghertella, Bifidobacterium, Actinomyces (phylum Actinobacteria) were observed in the gut microbiota of RETT subjects. Strati et al. reported decreased abundance of bacteria from the genus Bilophila and increased levels of bacteria in genera Escherichia and Shigella within phylum Proteobacteria (Strati et al., 2016). At the species level, subjects with RETT in the study by Borghi et al. (2017) showed decreased abundance of F. prausnitzii, Roseburia sp., Prevotella sp. and an increased abundance of Clostridium sp., and Sutterella sp., compared to NT controls. The study by Strati et al. (2016) also reported increase abundance of Bifidobacterium longum in RETT-associated microbiota (Table 4). Overall, these results suggest the presence of bacterial changes across multiple bacterial phyla that are associated with RETT.

Two studies investigated the potential metabolic function of gut bacteria using PICRUSt-KEGG analysis. Strati et al. (2016) observed an enrichment in carbohydrate and propanoate metabolism (pathways of SCFA production) in the gut microbiota of RETT subjects. In the study by Borghi et al. (2017), enzymes for carbohydrate and lipid metabolism were depleted, whereas amino acids pathways, butanoate and propanoate metabolism were enriched in RETT subjects (Table 4). These data suggest that the bacterial metabolic pathways responsible for SCFA production are altered in RETT. As these findings were only derived from two studies, larger cohorts of patients are required for further validation.

Specific correlations between intestinal microbial dysbiosis and disease severity in RETT were only assessed by Thapa et al. (2021), who found a significant reduction in α-diversity and differences in β-diversity between subjects with severe RETT compared to mild and moderate phenotypes (Figure 2; Table 4). Further studies are needed to understand whether there is a link between the composition of bacterial communities and RETT severity.

One study investigated alterations in the gut mycobiota associated with RETT (Strati et al., 2016). Fungal β-diversity measurements showed a significant difference between RETT subjects and NT controls (Table 4). Metataxonomic analysis led to the identification of 77 fungal genera that differed between groups, including Candida, Penicillium, Aspergillus, Malassezia. Among identified genera, only Candida was increased in RETT subjects compared with NT controls (Figure 2; Table 4).

Several studies assessed the concentration of microbial-derived metabolites in stool samples from ASD subjects and evaluated possible correlations with changes in microbial taxa associated with ASD. De Angelis et al. (2013) found that the metabolism of free amino acids (FAA) differed between patients with PDD-NOS and ASD compared to NT first-degree sibling controls. Levels of FAA and volatile organic compounds (VOCs) of subjects with PDD-NOS were more similar to NT controls than those of ASD on principal component analysis (PCA) plots (De Angelis et al., 2013).

Glutamine was highest in fecal samples of subjects with ASD compared to NT controls while indole, a microbial metabolite of tryptophan (which converts to serotonin) and 3-methylindole were increased in ASD and PDD-NOS subjects (De Angelis et al., 2013; Table 2). Kang et al. (2018) found trimethylamine, malonate, methylamine, choline, and 3- phenylpropionate weakly correlated with age in ASD subjects but were not statistically significant after correction for multiple hypothesis testing. Fecal concentrations of GABA, a major inhibitory neurotransmitter, were lower in stool of children with ASD in this cohort, although differences were only marginally significant. Concentrations of glutamate, a major excitatory neurotransmitter, were comparable between groups. GABA to glutamate ratio was observed to be marginally lower in stool of ASD subjects (Kang et al., 2018; Table 1).

Phenol compounds (phenol, 4-(1,1- dimethylethyl)-phenol, and para-Cresol) and methyl esters were increased in the stool of both PDD-NOS and ASD subjects compared to NT controls (De Angelis et al., 2013). This was also reported by Kang et al. (2018), who found significantly higher concentrations of p-cresol in stool of children with ASD compared to neurotypical controls. p-Cresol is introduced to the body either through skin, inhalation, or microbially produced in the intestine from dietary tyrosine or toluene substrates. The authors found no significant correlation between p-cresol and tyrosine in subjects’ stool (Spearman correlation, r = 0.08, p = 0.62) (Kang et al., 2018). p-cresol concentrations were also found to be significantly and negatively correlated with age within the ASD group (Spearman correlation, r = 0.47, p = 0.02), whereas no significant correlation was observed within NT subjects (Spearman correlation, r = 0.22, p = 0.34) (Kang et al., 2018) (Supplementary Table S5).

The study of Kang et al. (2018) involved multiple testing corrections (permutation-based non-parametric significance test, adjusted p = 0.022) among 59 metabolites and found that isopropanol was the only metabolite that was significantly different between ASD and NT controls after multiple testing corrections. In this same study, metabolite combinations from the 25 most differently abundant metabolites were evaluated to determine if they could support differentiation of ASD subjects from NT controls. Only five out of 23 children with ASD (78% sensitivity), and four out of 21 neurotypical controls children would not be classified as belonging to their true group (81% specificity). Further, using a more stringent threshold (>30% taxa membership in ASD subjects), cross-validation showed 70% sensitivity and 95% specificity (Kang et al., 2018). This analysis showed higher concentrations of caprate and aspartate in stool of subjects with ASD, while nicotinate, glutamine, and thymine were lower.

Differences in SCFA have also been described between subjects with ASD and NT controls (Figure 2; Tables 1, 2) Children with ASD showed significantly lower levels of total SCFA compared to NT, except for propionic acid which was increased (De Angelis et al., 2013). The proportion of SCFA correlated with levels of Faecalibacterium, Ruminococcus and Bifidobacterium (De Angelis et al., 2013). Liu et al. (2019) found correlations between microbial taxa and decreased fecal SCFA in ASD. Ruminococcaceae, Peptostreptococcaceae, Lactobacillales, Streptococcaceae, Eubacterium and Lachnospiraceae-NC2004 all correlated with decreased fecal concentrations of butyrate. A correlation was found between relative Acidobacteria levels and increased levels of fecal valerate (Spearman correlation, rs = 0.349, p = 0.013), Desulfovibrionaceae and increased fecal acetate (r = 0.360, p = 0.01), and Streptococcaceae and increased propionate (r = 0.368, p = 0.009) (Liu et al., 2019). Subjects in both cohorts were thought to have similar dietary contributions of butyrate and valerate, suggesting that differences in microbial taxa were the primary determinant of SCFA changes (Liu et al., 2019). Kang et al. (2018) found that fecal concentrations of propionate and butyrate were not different between ASD and NT controls. Other carbohydrate fermentation products of intestinal bacteria, formate and lactate, were detected at relatively lower concentrations in children with ASD, but their differences were not significant after multiple testing corrections (Table 1).

The study by Zhai et al. (2019) described high levels of Pb, As, Cu, Zn, Mg, Ca and Hg in children with ASD (Figure 2; Table 1). In particular, As and Hg were positively correlated with Oscillospira and Parabacteroides in ASD. Luna et al. (2017) assessed tryptophan and serotonin metabolite concentrations in supernatants of rectal biopsies obtain from ASD patients (all displaying FGID) and correlated these data with bacterial taxa that were abundant in biopsy preparations from an ASD cohort. Decreased tryptophan and increased 5-hydroxyindoleacetic acid (5-HIAA) levels were found in ASD rectal biopsies compared to NT control rectal biopsy preparations (Luna et al., 2017). Erysipelotrichaceae, C lituseburense, and Terrisporobacter species correlated with tryptophan concentrations from the rectal biopsies of the ASD cohort. L. bolteae, L. hathewayi, and F. plautii correlated with levels of serotonin in ASD tissue biopsy specimens (Luna et al., 2017; Table 1; Supplementary Table S5).

Microbial-derived metabolites were not directly investigated in ADHD studies.

Among RETT subjects, Borghi et al. (2017) found fecal butyrate and propionate to be significantly increased, as well as iso-butyrate and iso-valerate. This was found in conjunction with a greater abundance of Bacteroidetes and Veillonaceae, major propionate producers (Borghi et al., 2017; Table 4). Similar findings were described in a study by Strati et al. (2016), who found that the fecal content of SCFAs in RETT subjects was significantly enriched in propionate, isobutyrate and 2- methylbutyl isovalerate (Table 4). Concentrations of selected protein end-products of the gut bacterial metabolome, including GABA, tyrosine, and glutamate, were reduced in RETT subjects as compared to NT (Thapa et al., 2021; Table 4).

Overall, these studies show changes in microbial-host metabolites, neurotransmitters and trace elements in stool and intestinal mucosa of ASD and RETT patients that correlated with bacterial taxa altered in ASD patients. These results require further validation to infer a causal relationship between microbiota changes and neurodevelopmental outcomes.

Few studies assessed correlations between immune dysregulation and changes in microbial taxa associated with ASD (three studies) or RETT (one study). No studies assessed these relationships in individuals with ADHD. Hughes and Ashwood (2018) reported that children with ASD have elevations in IgG against Candida albicans with concomitant elevations of interleukin (IL)-17 (Table 1). Rose et al. (2018) also found that peripheral blood mononuclear cells (PBMC), isolated from the blood of subjects with ASD and GI symptoms (ASD-GI), produced elevated levels of interleukin (IL)-17, as well as IL-5 and IL-15, after exposure to the toll-like receptor (TLR)-4 agonist lipopolysaccharide (LPS), a component of gram-negative bacteria cellular walls. This elevation specifically occurred in PBMC isolated from ASD-GI subjects, while those with ASD-no-GI showed increased levels of IL-1α, IL-1β and Tumor Necrosis Factor (TNF)-α after TLR-4 stimulation of PBMC with LPS (Rose et al., 2018). Moreover, production of Transforming Growth Factor (TGF)-β1 was consistently decreased in ASD-GI subjects, as compared to subjects with ASD-no-GI and NT controls. The study by Luna et al. (2017) measured cytokines levels in the blood and biopsy supernatants from ASD-GI and NT controls, with or without GI symptoms (Figure 2; Table 1). In serum, decreased growth related oncogene alpha (GROα) and Interferon (IFN)-a2, and increased monocyte chemoattractant protein (MCP) and eotaxin levels were observed in ASD-GI subjects compared with NT-GI and NT controls. Moreover, increases in MCP-1 and eotaxin were found in ASD-GI subjects with abdominal pain. Luna et al. (2017) reported positive correlations between serum/mucosal cytokines and bacterial species associated with ASD-GI group (Table 1) and also between cytokines and bacterial species associated with pain in the ASD-GI group. C. disporicum correlated with macrophage inflammatory protein 1 alpha (MIP-1a) and beta (MIP-1β), vascular endothelial growth factor (VEGF), IFN-γ, IL12p70, IL17A, IL5, IL6 and IP-10. C. tertium correlated with IL1RA, IFN-γ, IL12p70, IL17A, IL1a, IL5, IL6, MIP-1a, and MIP-1b; P. excrementihominis correlated with GROα, and Tyzzerella nexilis with soluble CD40 ligand (Luna et al., 2017). The study by Tomova et al. (2015), which assessed probiotic supplementation in ASD subjects, identified a significant decrease in relative abundance of Firmicutes, Bifidobacterium and Desulfovibrio that correlated with decreased fecal TNFα levels in probiotic-treated ASD subjects (Table 5; Supplementary Table S9). Oxytocin was significantly lower in the plasma of subjects with ASD and NT first-degree sibling controls compared to unrelated NT controls. A positive correlation was found between plasma oxytocin levels and ASD severity (ADI) (r = 0.71, p < 0.05) (Tomova et al., 2015). Plasma levels of dehydroepiandrosterone sulfate (DHEA-S) were significantly lower in children with ASD, compared to NT first-degree sibling controls or unrelated NT controls. Testosterone levels did not show significant differences between groups but were positively correlated with ASD severity (as measured by ADI scores) (Tomova et al., 2015; Supplementary Table S9).

Inflammatory markers were not investigated in ADHD studies.

Strati et al. (2016) found that subjects with RETT had elevated levels of fecal calprotectin and serum erythrocyte sedimentation rate (ESR), which correlated with levels of serum IgA antibodies (Table 4). These increased inflammatory markers were hypothesized to be associated with RETT subjects’ lower intestinal bacterial microbiota richness and diversity as compared to NT controls (Strati et al., 2016).

GI symptoms were assessed using the Questionnaire on Pediatric Gastrointestinal Symptoms-Rome III (QPGS-RIII) in five studies (Son et al., 2015; Luna et al., 2017; Strati et al., 2017; Hughes and Ashwood, 2018; Rose et al., 2018) and using the 6-item Gastrointestinal Severity Index (6-GSI) in four studies (Kang et al., 2013, 2018; Liu et al., 2019; Ahmed et al., 2020; Supplementary Tables S5, S6). In two studies, GI symptoms were evaluated based on parental questionnaires (Gondalia et al., 2012; Tomova et al., 2015; Supplementary Tables S5, S6), and four studies did not assess the presence of GI symptoms in their patient cohorts (De Angelis et al., 2013; Pulikkan et al., 2018; Zhang et al., 2018; Zhai et al., 2019). Many ASD subjects included in these studies manifested GI symptoms including constipation, diarrhea, abdominal pain, and others.

Liu et al. (2019) observed that only constipation was significantly increased in ASD subjects compared to NT controls. ASD subjects with constipation displayed stool enrichments in Fusobacterium, Barnesiella, Coprobacter and valeric acid-associated taxa (Actinomycetaceae), and reduced butyrate-producing taxa (Liu et al., 2019). The study by Strati et al. (2017) found that taxa belonging to Clostridium cluster XVIII and pro-inflammatory bacteria Escherichia/Shigella positively correlated with ASD constipation status. Finegold et al. (2017) did not report their methodology for assessment of GI symptoms. This study found that samples from subjects with ASD have significantly higher counts of β2-toxin expressing-Clostridium perfringens compared to NT controls. 79% of the 33 ASD subjects showed the presence of β2- toxin gene in stool, compared to 38% of the 13 NT controls. Those subjects with ASD also showed more severe GI symptoms, although the correlation between β2- toxin gene levels and GI symptoms severity was not assessed. Bacterial richness was negatively correlated with GI symptoms in ASD subjects in the study by Kang et al. (2013) (Table 1). Almost 40% of anti-C. albicans⁺ IgG-or anti-C. albicans^-I IgG- ASD subjects had GI symptoms in the study by Hughes and Ashwood (2018). Son et al. (2015) observed increased abdominal pain and functional constipation in ASD subjects compared with their NT siblings, but no significant correlation was identified between ASD, FGIDs or gut bacteria composition (Son et al., 2015).

In a study evaluating 50 individuals with ASD, Rose et al. (2018) showed that the relative abundance of Bacteriodaceae, Lachnospiraceae, Ruminococcaceae and Prevocellaceae differed between ASD-GI subjects and NT-GI controls. No differences were found among these taxa between ASD-no-GI subjects, and NT-no-GI controls. ASD-GI subjects had more irritability and agitation, social withdrawal, lethargy, and hyperactivity compared to ASD-no-GI. Further analyses showed that microbiota-related biological pathways were differentially represented between groups and that alterations were related to the presence of ASD, and not GI symptoms independently (Rose et al., 2018). In contrast, the studies by Tomova et al. (2015) and Kang et al. (2018) reported positive correlations between GI symptoms and ASD severity in their cohort of patients. Luna et al. (2017) found positive correlations between bacterial taxa (from mucosal biopsies) and GI symptoms in both ASD and NT subjects. Six bacterial taxa were higher in ASD with abdominal pain compared to ASD without abdominal pain, as well as NT controls with and without abdominal pain (Luna et al., 2017; Table 1).

Ahmed et al. (2020) showed that the 6-GSI score was not correlated with changes in microbiota profiles among subjects with ASD. This was corroborated by Gondalia et al. (2012), who also found no differences in intestinal microbial profiles between ASD-GI and ASD-no-GI subjects.

In the study by Yap et al. (2021), reduced stool consistency, as measured on one stool sample using the Bristol Stool Chart, was associated with reduced dietary and microbiota diversity in ASD patients. Kushak et al. (2017) reported constipation and gastroesophageal reflux in ASD subjects assessed for GI symptoms through upper GI endoscopy. However, the authors did not observe any effects of GI symptoms on mucosal-associated microbiota or duodenal enzyme activity (Kushak et al., 2017). Overall, these findings confirm the presence of GI comorbidities in subjects with ASD. Nevertheless, relationships between gut microbes, GI symptoms and ASD-related neurobehavioral outcomes remain unclear.

The study by Wan et al. (2020) was the only study included in our systematic review to describe GI comorbidities in ADHD patients. An association was reported between bacterial changes and constipation symptoms in the ADHD cohort (Supplementary Table S7).

All individuals with RETT suffered from constipation, although relationships between microbiota alterations or disease severity with GI symptoms were not investigated (Table 4; Supplementary Table S8).

Eight out of 17 studies on ASD documented dietary patterns of patients. The effect of diet in shaping the gut microbiota composition, however, was evaluated only by Yap et al. (2021) (Table 2; Supplementary Table S6). The investigators in this study performed stool metagenomics in a large (n = 247) cohort of ASD and NT controls from a closely observed population from the Australian Autism Biobank and Queensland Twin Adolescent Brain project. ASD subjects had greater restriction in food intake and a less-diverse diet, which was associated with lower microbial α-diversity (Yap et al., 2021). This study found little evidence for a direct association between intestinal microbiota and ASD; rather, any observed changes were found to be most associated with influences of stereotypic dietary intakes in ASD subjects compared to NT controls (Yap et al., 2021).

A strictly controlled cross-sectional cohort study was conducted by Prehn-Kristensen et al. (2018) where subjects with ADHD were matched with NT controls with similar BMI and food intake (meat, fruits/vegetables, yogurt, milk products, or fast food during the month prior to microbiota analysis; Table 3; Supplementary Table S7). Differences were observed in α-diversity between subjects with ADHD and NT, and authors speculated that microbial changes in ADHD individuals were due to the underlying disease and not dietary patterns. In a cohort study by Wang et al. (2020), diets of ADHD subjects and NT controls were compared. ADHD subjects were found to have higher intake of refined grains (p = 0.027) and a lower proportion of dairy (p = 0.020) and vitamin B2 (p = 0.033) intake (Wang et al., 2020). In this study, S. stercoricanis was significantly associated with dairy, nut, seed and legume intake, and B. uniformis correlated with fat and carbohydrate intake. S. stercoricanis and B. ovatus also positively correlated with ADHD symptoms. These findings correlate with findings from Aarts et al. (2017), who demonstrated greater abundance of B. uniformis and B. ovatus in ADHD patients. Bacteroides sp. have been associated with the development of the frontal lobe, cerebellum and hippocampus (Aarts et al., 2017; Table 3; Supplementary Table S7). Overall, these data suggest that changes in the microbiota may correlate with dietary intake and not the actual disease state.

The influence of diet on gut microbiota profiles among NT controls and RETT subjects was only investigated by Thapa et al. (2021) (Table 4; Supplementary Table S8). Differences in bacterial composition and increased α-diversity were observed in RETT subjects who consumed table foods alone or combined with formula compared with RETT subjects who received exclusive formula. However, the effect of the diet in influencing disease severity was not investigated (Thapa et al., 2021).

Two studies examined the effects of probiotic treatment on gut microbiota profiles, disease outcomes and GI symptoms in patients with ASD (Tomova et al., 2015; Shaaban et al., 2018; Table 5; Supplementary Table S9). Both studies used a mix of Lactobacillus and Bifidobacteria strains, orally administered once (Shaaban et al., 2018) or three times per day (Tomova et al., 2015) for one month (Table 5). In both studies, NT controls were siblings of ASD subjects. Shaaban et al. (2018) found that decreases in the amount of Bifidobacterium in stool from ASD subjects recovered to levels observed in NT controls following probiotic treatment (Table 5). Conversely, Tomova et al. (2015) showed higher amounts of Bifidobacterium and increased Clostridium, Desulfovibrio, and Lactobacillus in ASD subjects. After probiotic treatment, levels of Bifidobacterium and Lactobacillus of ASD subjects decreased, reaching levels observed in NT controls (Tomova et al., 2015).

Tomova et al. (2015) did not assess the effect of probiotic treatment on disease severity. Shaaban et al. (2018) observed a significant decrease in ASD severity in probiotic-treated patients, represented by improvements in ATEC scores (Table 5).

In the study by Pärtty et al. (2015), Lactobacillus rhamnosus GG was administered to mothers daily, from four weeks before delivery up to six months of life (Table 5). Subjects were followed for 13 years to assess whether probiotic administration in early life reduced the risk of developing AS and ADHD in adolescence (Pärtty et al., 2015). ADHD or AS was diagnosed in 17.1% of subjects in the placebo group vs. 0% of the probiotic-supplemented group by age 13. There were no differences in bacterial abundance between ADHD subjects and NT controls, by 3 months or 13 years. Decreases in Bifidobacterium, Clostridium histolyticum, Bacteroides, Lactobacillus, and Enterococcus observed in 2-year-old children treated with placebo were associated with the development of AS or ADHD at 13 years old. These taxa differences did not persist at age 13 years (Pärtty et al., 2015; Table 5).

In an open-label study, Kang et al. (2017) showed that fecal microbiota transfer (FMT) to patients with ASD, using stool derived from healthy controls (Standardized Human Gut Microbiota) (Hamilton et al., 2012) (Table 5; Supplementary Table S9) increased bacterial diversity compared to NT controls and remained higher than baseline bacterial diversity 8 weeks after final administration of FMT (Kang et al., 2017). In ASD patients, the relative abundance of Bifidobacterium, Desulfovibrio, and Prevotella increased following FMT. Median microbial richness at week 18 was not statistically different between ASD and NT control groups (p > 0.05). A positive engraftment of phageome from the healthy donor to ASD subjects and increased abundance of phage was also observed after FMT (Table 5). Similar to the probiotic treatment described by Shaaban et al., FMT significantly improved ASD symptoms by improving social skill deficits, stereotypy, communication, daily living skills and socialization (Kang et al., 2017) (Table 5).

Probiotic and FMT intervention have demonstrated beneficial effects in improving GI symptoms in ASD patients (Table 5). Among recipients of probiotic supplementation, improvements in GI symptoms (assessed by the 6-GSI) were also strongly correlated with improvements in autism severity (assessed by the ATEC) (r = 0.674, p = 0.0001; Shaaban et al., 2018). Specifically, these subjects had a significant reduction in constipation scores (p < 0.01), stool consistency (p < 0.023), flatulence (p < 0.037) and abdominal pain (p < 0.002). Further, a significant correlation was found between probiotic administration, stool colony counts of Bifidobacteria, and reduction of constipation severity score (r = −0.441, p < 0.015; Shaaban et al., 2018). In the FMT study by Kang et al. (2017), 89% of ASD subjects had improvement in GI symptoms as reflected by an 80% reduction in the Gastrointestinal Symptom Rating Scale (GSRS). Further, improvements persisted up to 8 weeks after FMT (follow up) though similar changes in microbial richness were also observed in one of two non-responders (subjects whose GI symptoms improved less than 50% on the GSRS; Kang et al., 2017).

Stevens et al. (2019) assessed the effects of dietary supplementation on the modulation of ADHD symptom expression and gut microbiota composition in a subset of participants from a large RCT (Table 5; Supplementary Table S9). Participants ages seven to 12 years with a diagnosis of ADHD were randomized to receive a 10-week course of broad-spectrum micronutrients containing a mixture of vitamins, minerals, amino acids, and antioxidants, or inert placebo control. Subjects receiving the micronutrient supplementation had improved symptoms of attention, emotional regulation, and aggression relative to placebo. Eighteen children from this cohort (n = 8 placebo, n = 10 micronutrient group) had their fecal microbiota assessed and showed a decrease in Bifidobacterium genera by 25% in the micronutrient treatment group while Coriobacteriales increased though these changes were not significant (Stevens et al., 2019).