Autism Spectrum Disorder and the Gut Microbiota in Children: A Systematic Review

Growing evidence suggests that the gut microbiota has a role in the pathophysiology of autism spectrum disorder (ASD) [1]. Differences in composition of the gastrointestinal (GI) microbiota in children with ASD compared to unaffected siblings and/or healthy unrelated controls have been reported in various studies. The rates of diagnosis of ASD have increased dramatically in the past few decades [2]. Although changes to diagnostic criteria and greater awareness of the condition may be contributing to the rise, it is thought that environmental factors are also important [3]. Non-genetic risk factors, including maternal and pregnancy-related factors such as intrauterine infections and exposure to medications have gained interest as ASD incidence has continued to increase at a rate that cannot be explained by genetics [4]. In addition, there is an increasing interest in the gut microbiota in relation to ASD [1].

The GI tract is home to one of the most complex ecosystems and contains around 100 trillion microbes [5]. “Microbiota” is a collective term for this microbial community which includes bacteria, archaea, eukaryotes and viruses. There is a degree of variation in the adult composition of gut microbiota and this can be influenced by diet, antibiotic use, lifestyle and genetics [5, 6]. The gut microbiota is crucial for health in humans, with several important metabolic, protective and trophic functions and has often been referred to as the “forgotten organ” in the literature [7]. With such an impressive metabolic capacity and contribution to host health, it is no surprise, that the gut microbiota has also been implicated in disease. Characterising and understanding the gut microbiota in health and disease is a promising avenue that may lead to therapeutic benefits through its manipulation via so-called microbial therapeutics.

Changes in the gut microbiota seen in ASD may have a causative role and perpetuate GI symptoms or may simply be a confounder driven by dietary restriction. Children with autism often suffer from a range of GI symptoms, including diarrhoea, abdominal pain, constipation and gastroesophageal reflux. Estimates of the prevalence of such symptoms vary from 9 to 91% across studies [8]. A meta-analysis by McElhanon et al. [9] concluded that there was a three-fold higher risk of GI symptoms in children with ASD than in those without. Recent studies have suggested that alterations in the gut microbiota composition in children with ASD may contribute to both GI and neurological symptoms. Findings appear to be inconsistent across studies. If the gut microbiota plays a role in pathophysiology, there may be scope for novel treatment through its manipulation by microbial therapeutics [10-13]. The aim of this article is to systematically review the existing literature to evaluate variations in the gut microbiota and understand its significance in ASD.

An extensive literature search was conducted using MEDLINE and EMBASE databases by 2 independent researchers (N.B. and G.L.H.). All studies published between 1966 and July 2019 were included. The following Medical Subject Heading [MeSH] terms were used, which included both the root term and text word: GI; gut; microbiota; autism; and ASD. The studies were evaluated for sample size, age range of children included, study methodology and composition of the gut microbiota. Manual searching of reference lists from potentially relevant articles was also undertaken to identify additional studies.

Randomised controlled trials, cohort studies and observational studies were included. Studies which reported duplicate results were excluded. Those where data could not be extracted were also excluded.

Studies were included if they compared the intestinal microbiota analysis of autistic children with those of healthy children and provided information on specific bacterial taxa.

Studies were excluded if they did not report on patterns of individual bacterial taxa differences. General reviews, studies based on adult subjects or animal models and in vitro studies were also excluded, along with conference abstracts and texts not in English. Clinical trials with an intervention were not included, except if the microbiota was assessed in both groups at baseline before intervention. Solely culture-based studies were excluded from this -review. We also excluded case reports and studies of fewer than 12 patients.

The Newcastle and Ottawa scale for case-control studies was used to assess the quality of the studies and a quality score [1-9] was allocated to each [14]. Data collection and quality assessment were conducted independently by 2 researchers (N.B. and T.H.P.). Any disagreements were resolved through discussion until a consensus was reached.

The initial search identified 898 records from -MEDLINE and EMBASE. 595 articles remained following removal of duplicates. Of these, 61 studies were identified for full text eligibility after title and abstract screening. After full text screening, 26 studies were included in the review. Manual searching identified 2 further eligible articles, and therefore a total of 28 papers were included in the review (Fig. 1).

The quality scores of the 28 studies were assessed according to the Newcastle-Ottawa Scale and given a score (Table 1).

Among the 28 studies, 19 studied the difference between ASD and unaffected, unrelated children [15-33], 4 studies looked at the difference between ASD and unaffected siblings [34-36] or blood relatives [37] and 5 studies considered all 3 groups [38-42] (Table 2). The number of enrolled patients in a single study ranged from 12 to 104 participants and all were aged between 2 and 18 years. Of the included studies, the majority (n = 11) were from the United States, 6 from Europe [15, 20, 22, 35, 39, 42] and 3 from Australia [34, 40, 41]. The remaining 8 studies were from Asia [30, 37]. Study design including sampling strategies and microbiome analysis differed between studies. Twenty-four studies assessed faecal samples and 4 used mucosal biopsies (Table 2). Biopsy site varied for mucosa-based studies: 2 sampled ileum and caecum [21, 31], 1 sampled duodenum [17], and 1 sampled rectum [18]. The methods of collection of faecal samples also varied: from a single sample; 3 separate samples; cumulative samples over 48 h; freshly evacuated faeces; and faecal samples following an overnight fast. In 4 studies, microbial analysis was conducted using quantitative real-time amplification of bacterial DNA (qPCR), 4 used both qPCR and 16S rRNA sequencing techniques, one study used fluorescent in-situ hybridisation and the majority (18 studies) solely used 16S rRNA sequencing. One study used shotgun metagenomic sequencing [28].

The majority of studies referred to the Diagnostic and Statistical Manual of Mental Disorders, editions IV or V [43, 44] for the diagnosis of ASD. Other diagnostic tools were also used including Autism Diagnostic Interview-Revised [45], Autistic Diagnostic Observation Schedule [46], Childhood Autism Rating Scale [47], Pervasive Developmental Disorders Autism Society Japan Rating Scale [48], Modified Checklist for Autism in Toddlers [49], Autism Treatment Evaluation Checklist [50], Pervasive Developmental Disorder Behaviour Inventory [51], INCLEN Diagnostic Tool for ASD [52], Indian Scale for Assessment of Autism [53] and ICD-10 [54]. Some studies simply stated that the formal diagnosis was made by a psychiatrist, psychologist or a multidisciplinary team through history and observation. In one study, families were recruited through a registry called the Simons Simplex Community through the Interactive Autism Network [36]. As ASD is an umbrella term, some studies specified certain subtypes of children in their ASD groups, for example, Asperger’s syndrome and pervasive developmental disorder not otherwise specified. For control patients, numerous studies did not specifically exclude behavioural and developmental characteristics of ASD to ensure they were truly unaffected controls.

The heterogeneity in methodology within the included studies, specifically the diagnosis of ASD; gut site sampled; and laboratory methods used made meta-analysis inappropriate within this systematic review, and hence it has not been attempted.

Due to the dramatic impact of antibiotics on the gut microbiota [55] antibiotic usage was specifically interrogated. Reported antibiotic use by participants varied across studies (Table 2). More than half of the studies excluded subjects who reported the use of any antibiotics prior to sample collection. The duration of not using antibiotics ranged from 15 days to 3 months. The majority used 1 month as a cut-off. Eight studies did not address antibiotic use [15, 17, 19, 21, 25, 28, 39, 41]. The 3 remaining studies did not exclude participants based on antibiotics but did collect data [31, 40, 42]. Parracho et al. [42] was the only paper to report on the effect of antibiotics, finding no relationship between exposure and the microbiota.

The vast majority of studies considered and collected information on GI symptoms. In most cases, parents or carers were asked to complete questionnaires (Table 2). Only a few studies specified a standardised assessment such as ROME III [56] or IV criteria [57], the 6 item GI Severity Index [58] and the CHARGE GI history survey [59]. Two studies excluded all subjects with gut problems [30, 35]. In contrast, in 4 studies, all participants had GI symptoms and all used mucosa-based analyses [17, 18, 21, 31]. This is unsurprising as it would be unethical to perform an endoscopy without any clinical indication. The studies that assessed GI symptoms consistently reported a significantly increased burden in autistic children. The analyses of the gut microbiota did not however always take symptoms into account. Five studies described distinct microbiota profiles in ASD with and without GI symptoms [20, 23, 33, 40, 41]. Two studies found no association [26, 28]. The presence of GI symptoms is an important confounding factor in the study of gut microbiota and ASD.

Dietary restriction is a fundamental component of ASD throughout life, and diet is also a potent driver of the individual microbiota [29, 60, 61]. Therefore, good dietary data is fundamental in understanding microbial diversity in ASD. Despite this, data on dietary patterns were collected in less than half the studies (Table 2). Most used individual questionnaires or non-validated surveys to acquire information about diet and primarily collected information on whether participants were on restricted diets (Table 2). Son et al. [36] used the Stony Brook University Medical Centre Department of Paediatrics Food Diary/Calorie Count Sheet and asked parents to record intake for 7 days prior to stool collection. They also used nutrient analysis as well as reporting the number of children on special diets [36]. Kang et al. [16] gathered information on diet patterns such as gluten-free/casein-free diets as well as information on the use of probiotics, vitamins and seafood consumption. Rose et al. [33] asked parents to report allergies, if patients were on restricted diets and if they had any specific food dislikes and foods that made symptoms worse.

Son et al. [36] reported no significant differences in microbial composition with respect to daily intake of micronutrients. No relationship was found between diet and microbial populations in the study by Parracho et al. [42]. Kang et al. [16] performed multivariate analysis and reported no significant associations between dietary intake and genus abundances. Five studies highlighted the issue of diet but did not specifically collect data [20, 22, 26, 35, 37]. De Angelis et al. [35] suggested their subjects came from the same families and so differences in diet could be excluded. Similarly, Strati et al. [20] and Li et al. [26] stated their participants were consuming a Mediterranean and Chinese diet respectively and did not comment further [22]. Pulikkan et al. [37] reported all subjects were consuming an omnivore native diet but did not provide more information. Berding et al. [29] investigated the microbiota composition in relation to feeding behaviour, nutrient and food group intake as well as dietary patterns. They found that dietary fibre negatively correlated with abundance of Clostridiales. Faecalibacterium abundance was positively correlated with fried food and negatively with fruit. They also identified that 2 distinct dietary patterns were associated with unique microbial profiles in children with ASD [29]. Similarly, differential patterns in food and microbiota were also seen in children with ASD in the study by Plaza-Diaz et al. [15]. The vast majority of studies did not collect details on diet or assess dietary impact on the gut microbiota. As diet is an important confounding factor, the extent to which diet affects the microbial composition in the ASD population remains unclear.

When comparing microbial diversity of the gut microbiota in children with ASD against controls, 3 studies reported an increase in ASD [27, 35, 38], 2 reported a reduction [16, 32], and 3 described no difference [17, 36, 37] (Table 3). Firmicutes, Bacteroidetes and Proteobacteria were the most abundant phyla reported in all studies (Table 3). Bacteroidetes were increased in ASD children in 5 studies [22, 25, 27, 35, 38]. Three studies noted a significant increase in the Firmicutes/Bacteroides ratio due to a decrease in Bacteroidetes in ASD children [20, 31, 37] and 2 studies showed no difference in Bacteroidetes and Firmicutes levels [24, 26].

Within the Firmicutes phylum, increased Clostridiales levels have been reported in ASD patients in several 16S rRNA pyrosequencing studies [15, 18, 24, 28, 29, 31, 35]. This has also been seen with qPCR studies: 3 Clostridium clusters (I, XI, and XIVab) and one specific Clostridium species, C. bolteae, were statistically significantly higher in ASD in Song et al. [19]. A further qPCR study reported increased Clostridium cluster I in ASD [39]. Using fluorescent in-situ hybridisation, Parracho et al. [42] identified an increase in Clostridium histolyticum in stools from ASD compared to healthy controls, although unaffected siblings of ASD children also had higher levels of C. histolyticum than healthy controls. Strati et al. [20] showed that GI problems in ASD children were associated with high levels of Clostridia. Williams et al. [31] found that the increase in Clostridiales was largely attributable to increases in Lachnospiraceae and Ruminococcaceae. Increased Lachnospiraceae and Ruminococcaceae were also seen in ASD children with GI symptoms when compared to control children with similar symptoms in Rose et al. [33]. Pulikkan et al. [37] also noted an increase in Ruminococcaceae in their analysis whereas Liu et al. [23] found Ruminococcaceae to be significantly reduced in ASD.

Greater abundance of Sutterella has been reported in ASD children [21, 27, 35, 41]. Williams et al. [31] found elevated levels of Betaproteobacteria in ASD, reflecting the presence of Alcaligenaceae, and later described these as Sutterella species [21]. They confirmed Sutterella species in over 50% of ASD children and their complete absence in control children. Sutterella species have been identified in individuals with conditions such as IBD but also in healthy individuals [62, 63]. It is currently unclear whether they are normal commensals or pathobionts. Two other studies observed Sutterella was less abundant in ASD [16, 18].

Desulfovibrio species were shown to be higher in ASD in 2 studies [38, 39]. Moreover, a strong correlation of Desulfovibrio and the severity of autism manifestations were noted [39]. Desulfovibrio species are associated with increased propionic acid (PPA) production, thought to be associated with ASD pathogenesis [64] Lactobacillus were also reported to be significantly higher in ASD in 3 studies [20, 37, 39]. Most notably, Lactobacillus genus was observed to be 32-fold higher in ASD children compared to healthy children in Pulikkan et al. [37].

Wang et al. [40] reported a lower relative abundance of Bifidobacterium in ASD compared to unrelated controls and unaffected siblings. Decreased Bifidobacterium in ASD children was also reported in 3 other studies [29, 35, 38], with 2 reporting a significant increase [15, 37]. Both Lactobacillus and Bifidobacterium species are commonly used in probiotic supplements [65]. Probiotic use could therefore have contributed to increased numbers of these bacteria; however, the use of probiotics at baseline was not specified in the studies concerned.

Abundance of Akkermansia muciniphila was reported decreased in ASD relative to controls in Wang et al. [40]. A. muciniphila is a mucin-degrading bacterium, usually present in large amounts in the healthy gut, thus its absence may contribute to altered mucus barrier function. In contrast, increased levels of Akkermansia have also been reported in ASD [35]. Lower levels of Prevotella genera were also noted in ASD in 4 studies [16, 17, 20, 32], although this was significant only in 3 [16, 17, 32]. A significant reduction in Prevotella melaninogenica species was seen in the study by Li et al. [26]. Plaza-Diaz et al. [15] reported significantly increased levels of Prevotella in ASD.

The Brain-Gut-Microbiome axis is a concept that refers to the complex interactions between the central nervous system, GI system and the microorganisms of the GI tract. The significance of GI issues in ASD children has led to the investigation of gut microbial involvement in the disease. In this paper, we have systematically reviewed the literature and identified 28 studies assessing the microbiota in individuals with ASD compared to healthy individuals. Unfortunately, we were not able to complete a meta-analysis due to the heterogeneity of studies.

Gut microbiome alterations are implicated in ASD; species which have been reported to change include Clostridium, Desulfovibrio, Lactobacillus and Sutterella. Altered gut microbial composition may cause disruption of the gut barrier, potentially allowing translocation of bacteria and their antigens, toxins and metabolites [12]. Bacterial fermentation of dietary carbohydrates normally leads to the production of short chain fatty acids -(SCFAs). These are signalling molecules with a variety of functions: the most abundant SCFAs are acetic acid, butyric acid, and PPA. MacFabe [66] showed that when PPA or other SCFAs were injected into the cerebral ventricles of rats, the rats demonstrated biological, chemical, and pathologic changes that were characteristic of autism. Of note, many of the bacteria implicated in ASD, such as Clostridium and Desulfovibrio are PPA producers. Additionally, specific strains produce harmful toxins such as lipopolysaccharides, which can theoretically lead to the production of inflammatory cytokines and impair neurodevelopment [38, 67]. Increased circulating levels of pro-inflammatory cytokines have been reported in ASD [68].

Much focus has been given to Clostridium groups in ASD, although the causal relationship remains to be proven. Bolte et al. [69] first postulated that Clostridium tetani could induce autism in 1998, although offered no mechanistic insight. Clostridiaceae synthesise metabolic products (e.g., phenols, p-cresol, certain indole derivatives) that are potentially toxic for humans. One of the first studies to speculate that disruption of the gut microbiota might contribute to autistic symptomatology was published in 2000 by Sandler et al. [70]. This small study showed improvement of autistic symptoms after oral vancomycin. Symptoms relapsed following cessation and it has been suggested this may be because Clostridium are spore-forming organisms, promoting recurrence [70]. However, of note, this study used non-validated measures to assess symptoms of autism. These findings are also yet to be replicated in further studies.

Although not a target of this review, an altered gut microbiota has also been seen in rodent models of ASD [71]. Prenatal exposure to the anticonvulsant valproate (VPA) is a risk factor for ASD and exposure to VPA in rodent models results in behavioural impairments. Studies have been reported showing altered gut microbiota, altered GI morphology and CNS inflammation following prenatal VPA treatment [71]. A study by de Theije et al. [72] showed that VPA-treated 28-day offspring had decreased abundance of Bacteroidetes phyla, mainly consisting of Bacteroidales, and increased Firmicutes microbial taxa, mainly consisting of Clostridiales. This is in keeping with findings from some of the studies discussed in this review.

There are numerous challenges to studying how the gut microbiota may be implicated in ASD and this is evident from the conflicting and complex results of studies conducted so far. Importantly, there is a lack of a uniform definition of ASD and the diagnostic criteria have continued to evolve over the last 15 years. Moreover, autistic spectrum disorders encompass a heterogeneous range of conditions with varying severity. The numbers of participants in the studies are often small; therefore, the chances of type 2 statistical errors cannot be disregarded. There is significant variation in the sampling methods, sites and laboratory techniques used across studies which invariably makes comparison difficult. The gut microbiota ecosystem in mucosal surfaces is fundamentally different to that of faeces; hence, the 2 are not comparable [73]. The majority of studies reported here focused on faeces. Mucosal samples are understandably difficult to obtain in this age group due to logistic and ethical limitations; however, this is an important sampling issue to keep in mind when drawing conclusions from studies. Bacteria at the mucosal surface are in close contact with the host and may have a more potent pathophysiologic role than luminal bacteria. The studies discussed used various techniques to study microbial communities. Differences may arise from various aspects: sample handling; DNA extraction; sequencing of different regions of DNA; varying coverage achieved by PCR primers; sequencing depth; and bioinformatic/statistical methods.

Additionally, not all the studies have shared important baseline information about the participants. All results need to be considered with caution, especially given the potential impact of diet, antimicrobial use and GI symptoms on the gut microbiota. Individuals with ASD are known to have specific dietary interests: they are commonly restrictive with food choices by texture, smell and taste [74]. Some individuals with ASD also follow restrictions, such as gluten-free diet, in an attempt to improve symptoms. Although popular with parents and clinicians, there is little evidence to support the use of dietary therapies for the treatment of core symptoms of ASD [75]. It has been established that diet has a significant role in the modelling of the gut microbiota [76]. Protein, fats, digestible and non-digestible carbohydrates, probiotic and polyphenols can change the microbiome [60]. Several studies have shown that a Western diet (high in animal protein and fat, low in fibre) leads to a marked decrease in beneficial bacteria such as Bifidobacterium and Eubacterium species [60]. In contrast, the Mediterranean diet is highly regarded as a healthy balanced diet with positive effects on inflammation and lipids, mediated by increases in Lactobacillus, Bifidobacterium and Prevotella and decreases in Clostridium [60]. Furthermore, it has been shown that dietary alterations can induce large microbial shifts, even within 24 h of a change [77]. To clarify the relationship between diet, gut bacteria and autism, future studies need to be carefully designed with the inclusion of validated and objective food diaries and detailed dietary histories.

Studies have shown a higher incidence of antibiotic usage in individuals with ASD, particularly because of susceptibility to infections such as otitis media [78]. In a pyrosequencing study, healthy humans were exposed to ciprofloxacin and their microbiota was assessed before and after treatment. Ciprofloxacin was found to lower the diversity of bacteria [55]. The microbial composition largely returned to pre-antibiotic state after 4 weeks; however, a few species did not return to original numbers within 6 months [55]. It is clear that antibiotics can have a significant influence on the composition of the gut microbiota and therefore may be a confounding factor in studies assessing the gut microbiota in ASD.

It is widely accepted that children with ASD have an increased prevalence of GI symptoms. Although an altered gut microbiota has been seen in children with ASD with GI symptoms, whether these changes are also seen in ASD children without GI symptoms is not clear from studies thus far. Not all studies have addressed the presence and effect of GI symptoms on ASD and microbial profiles and very few have considered them in statistical analyses, adding to the complexity of interpreting microbial outcomes.

In conclusion, ASD is an increasingly common condition that affects millions of families across the world. Studies to date have been inconsistent in their findings; however, this in part can be explained by heterogeneous populations and methods. Future studies of the gut microbiota in autism need more objective clinical assessment, consistency of case ascertainment and description. Furthermore, details regarding antibiotic use, diet and GI symptoms are key to unravelling the link between autism and an altered gut microbiota. There are several noteworthy reasons to consider that the gut microbiota may be involved in autism: the significance of GI symptoms in ASD; the apparently distinct gut microbiome of these children; and the mechanistic plausibility of bacterial products causing neurobehavioral effects. Gaining a better understanding of this important gut-brain connection may offer insights into potential diagnostic or therapeutic options and help alleviate the rising burden of ASD.

None.

Ethical approval was not required for this work. Research was conducted ethically in accordance with the World Medical Association Declaration of Helsinki.

N.B. and T.H.P. have nothing to disclose. G.L.H. reports receiving personal fees from Ferring from outside the submitted work. R.H. reports receiving personal fees and non-financial support from Nutricia, personal fees and non-financial support from 4D Pharma from outside the submitted work.

This study did not require any funding. R.H. is supported by a personal fellowship from NHS Research Scotland.

N.B.: literature search, figures, study design, data collection, data analysis, data interpretation, writing. T.H.P.: literature search, figures, data collection, data analysis, data interpretation, writing. G.L.H.: literature search, study design, data interpretation, writing. R.H.: data analysis, study design, data interpretation, writing, conception of review.