- Academic Editors
†These authors contributed equally.
Background: Although the characteristics of the gut microbiota of children with autism spectrum disorder (ASD) have been well studied, those of young adults with ASD have seldom been reported. Methods: Using 16S rRNA gene sequencing, we characterized the gut microbiota of 19 young adults with ASD and compared them with that of 19 healthy adults. A random forest prediction model was used to distinguish between the two groups at the genus level. Results: The abundance levels of one phylum, seven families, and 18 genera in adults with ASD were significantly different from those of controls. The genus Phascolarctobacterium was significantly enriched in adults with ASD, which might elicit ASD-like behavior through production of propionate. In addition, a random forest model identified 15 genera that could distinguish adults with ASD from healthy controls with areas under the receiver operating curve of 92.86%, and ten of them were biomarkers identified by LEfSe. Conclusions: Our results identified specific gut bacteria associated with ASD, and the successful application of certain genera in the prediction model further supports the association between gut microbiota and ASD.
Autism spectrum disorder (ASD) is clinically manifested as impairments in language development and social interaction, as well as behavioral stereotypes and narrow interests [1]. The onset of ASD generally begins at infancy and persists into adolescence and adulthood, with a worldwide prevalence in children of approximately 6.25 per 1000. Additionally, the incidence of ASD varies from country to country. For instance, it is about 18.5 per 1000 in the United States [2], and approximately 2.9 per 1000 in China [3].
To date, the exact etiology and pathogenesis of ASD remain unclear [4, 5], but this condition is generally regarded as the result of genetic and environmental interaction [6]. Though hundreds of genes have been associated with ASD, they only account for 10–20% of ASD cases [7]. The occurrence of ASD may also be attributed to maternal exposure to toxic chemicals in the environment during pregnancy [8]. Furthermore, delivery and breastfeeding patterns may be associated with ASD [9, 10, 11].
Advances in research into gut microbial ecology have linked gut microbiota with the health of the host. Once an ecological imbalance in the gut microbiota occurs, various diseases, including ASD, can develop [12]. In this process, the brain-gut axis plays an important role [9, 13, 14, 15]. Accordingly, fecal microbiota transplants (FMT) have demonstrated long-term efficacy in improving gastrointestinal (GI) and behavioral symptoms of ASD patients by modifying gut microbiota [16, 17]. However, previous studies have also had conflicting conclusions. For example, Strati et al. [18] reported that Bacteroidetes is significantly less abundant in children with ASD, whereas Zhang et al. [19] suggested that the Bacteroidetes level in Chinese children with ASD is significantly higher. Luna et al. [20] indicated that Blautia is less abundant in children with autism, whereas Berding et al. [21] reported that levels of this bacterium are significantly higher in children with autism.
The accumulating evidence indicates that the gut microbiota changes with the host’s age [22, 23] and dietary alterations [24]. Previous studies have focused on the characteristics of the gut microbiota in children with ASD, although the microbiota characteristics in adult patients with ASD have rarely been reported. Also, adults with ASD tend to be more obese than the general adult population, for both young [25] and old subjects [26]. In a previous study, we compared the gut microbiota of adults with ASD to non-ASD obese adults and found that the propionate-producing species Phascolarctobacterium succinatutens is enriched in adults with ASD, whereas Dialister succinatiphilus is enriched in adults with obesity [27]. In our samples, only 2 of 21 adults with ASD were obese, and we did not compare the gut microbiota of autistic and non-autistic normal-weight adults.
In the present study, we compared the gut microbiota of 19 young autistic Chinese adults to those of 19 healthy Chinese adults. All subjects lived in Jinan, Shandong Province, and the dietary habits of this region are characterized by high intake of meat and seafood, with both wheat and rice as staple foods [28, 29].
Nineteen patients diagnosed with ASD and 19 healthy adults were recruited from Jinan, Shandong Province, China. The patients with ASD were diagnosed in childhood by clinicians according to the diagnostic criteria for childhood autism in the International Classification of Diseases and Related Health Problems, 10th Revision (ICD-10) (World Health Organization, 1993). Patients with schizophrenia or other psychoses were excluded. The gender-matched, healthy controls that did not suffer from ASD, other neurodevelopmental disorders, or neuropsychiatric disease, were recruited from community residents of Jinan. All the subjects were of Han nationality, and had not taken antibiotics or probiotics for one month prior to fecal-sample collection.
Stool specimens were collected into fecal sample preservation solution of the MicroLockerT stool sample collector (Jiangsu YIMI Biotech Inc., Taizhou, China) and transferred to the laboratory within three hours at room temperature.
The genomic DNA of the stool samples was extracted with the QIAamp DNA Stool
Mini Kit (QIAGEN, Hilden, Germany). The V3–V4 region of the 16S rRNA gene was
amplified using primers 338F and 806R [30], then purified using an AxyPrep DNA
Gel Extraction kit (AXYGEN, Union City, CA, USA), and pooled. The 2
Raw FASTQ files were first imported into
QIIME2 (version 2022.2, https://qiime2.org/) [31]. The amplicon sequence variants (ASVs) were
constructed by Dada2 plugin of QIIME2 with 100% sequence identity [32]. The
alpha diversity was analyzed using the mothur software package (version 1.44.3, https://mothur.org/)
[33]. The differences in taxonomic abundance and microbiota functions between the
ASD and control groups were analyzed using LEfSe
(http://huttenhower.sph.harvard.edu/galaxy/)
[34], then p-values were adjusted using the Benjamini-Hochberg false
discovery rate (FDR) procedure (significant as FDR
The random forest (RF) model was used to identify specific microbial taxa that distinguished ASDs from controls using the “randomForest” package in R 4.2.2. Seventy percent of samples were randomly selected to train the model, whereas the remaining 30% were used to verify the model’s performance. “MeanDecreaseAccuracy” and “MeanDecreaseGini” indicators were used to evaluate the relative importance of each genus in the prediction model. The rfcv function implemented in the “RF” package was applied more than five times for ten-fold cross-validation in order to obtain the minimum number of top-ranking genera for prediction. The predictive power of biomarkers selected from the model was assessed by receiver operating characteristic (ROC) curve analysis followed by calculation of the area under the curve (AUC) using the “pROC” package.
Nineteen young adults with ASD (mean age = 21 years, range = 17–32 years, mean body mass index (BMI) = 21.9; sex = 5 females, 14 males) and 19 healthy adults (mean age = 29 years, range = 19–37 years, mean BMI = 24.3; sex = 5 females, 14 males) were involved in the study, with most of the ASD patients having GI symptoms. A total of 1,290,959 (24,290–51,389) sequences from the 38 samples were obtained. Of these sequences, a total of 24,290 sequences from each sample were analyzed by QIIME2, which identified a total of 1681 ASVs (Table 1).
Group | Sample | ASV | Coverage | Richness | Evenness | Diversity | |
Chao | ACE | Simpsoneven | Shannon | ||||
ASD | 19 | 1210 |
0.99996 | 1214.75 |
1215.27 |
0.0589 |
5.23 |
Control | 19 | 923 |
0.99999 | 923.77 |
923.96 |
0.0842 |
5.20 |
p-value | 0.0058 | 0.0057 | 0.0056 | 0.0156 | 0.6283 |
Note: data for ASV, Richness and Evenness were presented as Mean
Approximately 99.88%, 97.71%, and 86.80% of the ASVs were assigned to 11
phyla, 53 families, and 165 genera, respectively. Of the phyla,
Firmicutes (mean = 46.71%), Bacteroidetes
(mean = 45.45%), and Proteobacteria
(mean = 3.86%) were the most prevalent in the microbiota of all subjects.
Fifteen families (
Family | ASD | Control | Enriched in | FDR |
Acidaminococcaceae | 1.83% |
0.37% |
ASD | 0.002006 |
Akkermansiaceae | 1.27% |
0.04% |
||
Bacteroidaceae | 11.12% |
34.73% |
Control | 0.00105 |
Bifidobacteriaceae | 2.63% |
0.5% |
||
Enterobacteriaceae | 0.69% |
2.56% |
Control | 0.002006 |
Erysipelotrichaceae | 3.49% |
0.26% |
ASD | 0.000024 |
Fusobacteriaceae | 0.14% |
2.47% |
||
Lachnospiraceae | 23.24% |
20.53% |
||
Peptostreptococcaceae | 1.08% |
0.39% |
ASD | 0.041808 |
Prevotellaceae | 29.47% |
8.98% |
ASD | 0.039061 |
Rikenellaceae | 2.44% |
0.92% |
||
Ruminococcaceae | 13.37% |
16.16% |
||
Selenomonadaceae | 0.64% |
3.83% |
||
Sutterellaceae | 0.79% |
2.28% |
Control | 0.005931 |
Veillonellaceae | 1.34% |
0.54% |
Note: *, data for ASD and Control group were presented as Mean
Genus | ASD | Control | Enriched in | FDR |
Agathobacter | 3.93% |
1.47% |
||
Akkermansia | 1.27% |
0.04% |
||
Alistipes | 2.44% |
0.92% |
||
Anaerobutyricum | 0.22% |
0.8% |
Control | 0.007219 |
Anaerostipes | 0.16% |
0.55% |
||
Bacteroides | 2.69% |
9.13% |
Control | 0.008436 |
Bifidobacterium | 2.63% |
0.5% |
||
Blautia | 3.45% |
2.36% |
ASD | 0.043184 |
Catenibacterium | 0.5% |
0.00% |
ASD | 0.004553 |
Clostridium IV | 0.71% |
0.14% |
||
Clostridium sensu stricto | 0.81% |
0.22% |
ASD | 0.043184 |
Clostridium XlVb | 0.28% |
0.57% |
Control | 0.045836 |
Coprococcus | 0.89% |
0.35% |
ASD | 0.049717 |
Dialister | 0.7% |
0.40% |
||
Dorea | 0.6% |
1.00% |
||
Escherichia/Shigella | 0.59% |
1.14% |
Control | 0.011537 |
Faecalibacillus | 1.37% |
0.17% |
ASD | 0.003047 |
Faecalibacterium | 2.87% |
8.62% |
Control | 0.019944 |
Fusicatenibacter | 0.55% |
0.41% |
||
Fusobacterium | 0.14% |
2.47% |
||
Gemmiger | 1.29% |
0.92% |
||
Holdemanella | 1.38% |
0.04% |
ASD | 0.002805 |
Klebsiella | 0.10% |
1.11% |
Control | 0.002826 |
Lachnospira | 3.23% |
3.09% |
||
Lachnospiracea_incertae_sedis | 1.40% |
0.64% |
||
Mediterraneibacter | 1.11% |
0.13% |
ASD | 0.000748 |
Megamonas | 0.56% |
3.83% |
||
Megasphaera | 0.61% |
0.03% |
||
Parabacteroides | 0.67% |
0.94% |
||
Parasutterella | 0.14% |
1.45% |
Control | 0.002254 |
Phascolarctobacterium | 1.83% |
0.37% |
ASD | 0.002006 |
Phocaeicola | 8.43% |
25.6% |
Control | 0.002006 |
Prevotella | 28.60% |
8.59% |
ASD | 0.043184 |
Prevotellamassilia | 0.64% |
0.00% |
ASD | 0.007618 |
Romboutsia | 0.62% |
0.36% |
||
Roseburia | 1.52% |
2.59% |
||
Ruminococcus | 1.48% |
1.73% |
||
Streptococcus | 0.61% |
0.43% |
||
Sutterella | 0.54% |
0.64% |
Note: *, data for ASD and Control group were presented as Mean
The results of PCoA and PerMANOVA analysis revealed a significant difference in bacterial diversity (F = 3.642, p = 0.0001) between the ASD and control groups (Fig. 1).
Principal coordinate analysis (PCoA) based on Bray-Curtis dissimilarity among ASVs of ASD and control groups. Each point represents samples, and the circles surrounding the samples represent 80% confidence interval.
At the phylum level, the Proteobacteria were significantly less abundant in the ASD group than in the control group (Fig. 2). At the family level (Table 2), seven major families showed significant differences (48.47% in the ASD group and 49.58% in the control group).
Comparison of the bacterial taxa in the gut microbiota of ASD
and control groups. (A) A cladogram taxonomic representation representing
distinct bacterial taxa between the two groups. Red color indicates enrichment in
the ASD group, and green indicates enrichment in the control group. (B) A
histogram of the linear discriminant analysis (LDA) scores represents significant
differences in the abundance of the bacterial taxa between the two groups. Only
taxa with significant richness differences (FDR
At the genus level (Table 3, Fig. 2), 18 major genera differed significantly between the ASD (55.87%) and control groups (60.64%).
To distinguish the microbiota of young adults with ASD from that of healthy adults, we performed RF analysis with ten-fold cross-validation at the genus level. In general, the optimal model with the lowest error rate had an AUC value of 92.86% (95% confidence interval (CI): 78.86–100%) when 15 genera were included as biomarkers (Fig. 3A,B). Among them (Fig. 3C), ten genera showed a significant difference between the ASD and control groups, including Phocaeicola, Mediterraneibacter, Parasutterella, Phascolarctobacterium, Faecalibacterium, Klebsiella, Blautia, Anaerobutyricum, Bacteroides, and Holdemanella (Table 3).
Random forest model construction and performance analysis. (A) The error rate of random forest model was the lowest when 15 genera were combined through 10-fold cross-validation. (B) The random forest classifier was constructed at the genus level to classify ASDs and controls. (C) The 15 dominant genera were selected as biomarkers according to Gini index of RF model. The length of the bar indicates the importance of the genera. ASD, autism spectrum disorder; AUC, area under receiver operating characteristic curve; 95% CI, 95% confidence interval; RF, random forest; Mean Decrease in Gini Index, the influence of each genus on the heterogeneity of observed values at each node of the classification tree; the greater the value, the greater the importance of the genus.
The gut microbiota needs to be analyzed comprehensively to understand the differences in the bacterial communities in young adults with ASD and those of healthy controls. Our analysis indicated that four genera with an abundance of 14.96% in the ASD subjects and 37.37% in the control subjects comprise the core microbiota (i.e., Phocaeicola, Faecalibacterium, Blautia, and Anaerobutyricum are present in each of the 38 subjects). Overall, the microbiota of young adults with ASD differed significantly from that of the healthy controls in one phylum, seven families, and 18 genera.
At the phylum level, the abundance level of Proteobacteria was significantly lower in young adults with ASD than in controls (Fig. 2). Strati et al. [18] reported that the Firmicutes/Bacteroidetes ratio in the feces of individuals with ASD (mean age = 10 years) was significantly higher than that in children without ASD. In contrast, Zhang et al. [19] found that the Firmicutes/Bacteroidetes ratio in the feces of 35 Chinese ASD children (mean age = 4.9 years) was significantly lower than that of healthy children. In the present study, the Firmicutes/Bacteroidetes ratio in ASD patients (1.10) was slightly higher than that in healthy adults (0.96), suggesting that the difference in ages of the ASD subjects may help explain the conflicting results.
The families Acidaminococcaceae, Erysipelotrichaceae, Peptostreptococcaceae, and Prevotellaceae were enriched in ASD subjects, whereas the abundance levels of Bacteroidaceae, Enterobacteriaceae, and Sutterellaceae were significantly lower than those of healthy adults (Table 2). The enrichment of Acidaminococcaceae and Prevotellaceae and the depletion of Enterobacteriaceae were also observed in children with ASD [35], reflecting a common gut microbiota character of ASD, although the depletion of Bacteroidaceae showed a contrary trend in children with ASD [35].
The abundance levels of 10 genera (Blautia, Holdemanella,
Phascolarctobacterium, etc.) were significantly higher, whereas those of
eight genera (Klebsiella, Phocaeicola, Parasutterella,
etc.) were significantly lower, in young adults with ASD than in healthy adults
(Table 3). Berding et al. [21] reported that the levels of
Blautia are significantly higher in children with ASD (mean age = 4.1
Ding et al. [44] found that children with ASD have significantly lower levels of Parasutterella. In the present study, we also observed that the abundance of Parasutterella was significantly lower in young adults with ASD than in healthy subjects. A previous study reported that Parasutterella produces short-chain fatty acids (SCFA) that are beneficial for improving GI symptoms [45].
Whether the gut microbiome is a cause or consequence of ASD remains controversial [46]. Yap et al. [24] reported that the unique gut microbiota character of children with ASD is the result of their dietary preference but not a contributor to ASD. Although the gut microbiota is important in the early stages of neural development [47], the study of Yap et al. [24] emphasized the importance of dietary data in ASD-microbiota relationship studies [48]. In the present study, we did not collect dietary data, so the potential effect of dietary preferences of the young adults with ASD might have contributed to the observed differences in gut microbiota between the two groups.
To date, the diagnosis of ASD mainly depends on physician observations and responses to questionnaires, such as the Autism Diagnostic Interview-Revised [49]. No biomarkers are currently available. However, because properties of the gut microbiota are associated with ASD, this raises the possibility that ASD can be predicted or diagnosed with the help of gut microbial markers. In the present study, a prediction model for ASD diagnosis using the RF analysis was established. The model identified 15 genera that enable distinguishing between autistic and healthy adults with 100% sensitivity and 85.7% specificity. However, the practical application of using gut microbial properties for the clinical diagnosis of ASD requires the evaluation of more samples and data.
There are some limitations in this study. First, the sample sizes in this study were small, which weakens our statistics and conclusions. Although the changes in the key microbiota revealed in this study were consistent with those of several previous studies, the evaluation of more adults with ASD and healthy controls is needed. Second, dietary data were not collected in our study. Although both the ASD and control groups lived in the same city, dietary differences between groups influence the gut microbiota [24]. Therefore, dietary data should be included in future studies. Third, we did not record intestinal transit time, which is a direct indicator of GI symptoms and plays a role in shaping gut microbiota [50]. Since intestinal transit time can be reflected by the Bristol Stool Scale [51], it would be useful to collect this information to help understand the association between gut microbiota and host health. Finally, the proportion of gram-positive bacteria might have been underestimated since we used a traditional stool DNA extraction method instead of bead-beating to lyse the cell wall of bacteria [52].
In summary, we observed that the gut microbiota in young adults with ASD was significantly different from that of healthy adults. Young adults with ASD were associated with more propionate-producing bacteria such as the genus Phascolarctobacterium and fewer SCFA-producing bacteria such as the genus Parasutterella, than were healthy controls. A prediction model based on RF analysis using 15 genera can distinguish the gut microbiota of autistic adults from that of healthy controls with high sensitivity and specificity, further indicating the association between gut microbiota and ASD.
The raw sequence data of gut microbiota have been deposited in the National Omics Data Encyclopedia (NODE, https://www.biosino.org/node/) under accession numbers OEX010410 (adults with ASD) and OEX014604 (healthy adults).
HZ and QL designed the project. YZ, BC and QZ collected samples. XP, QZ and MG performed experiments. XP, QZ and YW analyzed the data. XP, QZ and HZ drafted the manuscript with other authors involved in revision. All authors contributed to editorial changes in the manuscript. All authors read and approved the final manuscript. All authors have participated sufficiently in the work and agreed to be accountable for all aspects of the work.
All subjects gave their informed consent for inclusion before they participated in the study. The study was conducted in accordance with the Declaration of Helsinki, and the protocol was approved by the Medical Ethical Committee of the Shanghai Institute of Planned Parenthood Research (NO: PJ2019-17).
Not applicable.
This research was funded by Science and Technology Planning Project of Zunyi (grant number HZ-2021-52), and the Open Project Program of State Key Laboratory of Dairy Biotechnology (No. SKLDB2020-003).
The author Mr. Bin Chen is from the company. He is a volunteer caring for patients with ASD and helps us collecting fecal samples. All the authors, including Bin Chen, had no conflicts of interest.
Publisher’s Note: IMR Press stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.