As for sample size calculation, the estimated proportion of participants effectively treated with FMT is 90% (19). Using PASS 11 software, with α = 0.05 (two-sided) and power = 0.90, the calculated sample size is 35. Assuming a dropout rate of 10% for the study participants, a sample size of 38 is required. Finally, this study included 38 children with ASD (age 3–14 years) who were diagnosed by the criteria from the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5) by the American Psychiatric Association, and was conducted as an open-label clinical trial. Among these children with ASD, 31 children had GI symptoms (such as abdominal pain, reflux, indigestion, diarrhea and constipation) and 7 children had no GI symptoms. During the study, every child received a 12-week FMT treatment, followed by a 8-week observation period after the treatment ended. The entire study period lasted for 20 weeks. To serve as a healthy control (HC) group, 30 healthy children without GI symptoms were recruited. The healthy children were not treated and the FMT treatment was administered orally to all children with ASD via lyophilized capsules. Prior to FMT, neither vancomycin nor proton pump inhibitors were administered to any of the children. The researchers were aware of the group allocation and outcome evaluation. The study design is illustrated (Figure 1).

This clinical study was approved by the Ethics Review Committee at the Shanghai Children's Hospital, Shanghai Jiao Tong University. Before enrolling in the study, written informed consent was obtained from the guardian of each participant. Furthermore, the study was registered at the Chinese Clinical Trial Registry (www.chictr.org.cn). The trial protocol is accessible through the registration number ChiCTR2200055943.

Children with ASD were primarily recruited from patients at the FMT multidisciplinary clinic of the Shanghai Children's Hospital. Healthy children were recruited from families of acquaintances of the study researchers. The inclusion criteria for the subjects include being under 18 years old and able to swallow capsules; meeting the diagnostic criteria for ASD in DSM-5 and commonly used diagnostic scales; having a legal guardian who fully understands and signs the informed consent form, agreeing to the researchers collecting clinical data and samples; and being able to participate in the study continuously and follow-up. Exclusion criteria included a definite diagnosis of Reye's syndrome or other mental illness; combined with a clear history of brain trauma, cerebral palsy, encephalitis and other organic brain diseases; the use of probiotics, antibiotics, proton pump inhibitors or other medications that affect the gut microbiota in the past 3 months; the presence of other serious GI diseases or organic lesions in the intestine such as congenital megacolon, intestinal obstruction, or intussusception; and poor compliance. The healthy children did not have any mental disorders, such as ASD, attention-deficit hyperactivity disorder, depression or anxiety, and none of them had a first-degree relative with ASD. Samples of stool, urine, saliva and blood were collected from children with ASD, while only initial samples of stool, urine and saliva were collected from healthy children. All participants' guardians were capable of comprehending and consenting to the informed consent, and all participants were capable of receiving follow-up evaluations for 8 weeks.

The medical histories of the volunteers who underwent stool donor screening were reviewed and the laboratory test results were gathered. To be included as stool donors, individuals had to meet the following criteria: age between 15 and 30 years; overall good health; body mass index (BMI) ranging from 18.5 kg/m² to 23.9 kg/m²; and regular bowel movements occurring once or twice per day with a Bristol stool score of type 4. Exclusion criteria included a previous history of GI or chronic systemic disease, malignant neoplasm or had received radiochemotherapy; combined with any ongoing diseases; the use of antibiotics, probiotics, prebiotics, proton pump inhibitors, immunosuppressant agents or blood products during the past 3 months; positive serum results for hepatitis A, B, or C, human immunodeficiency virus, syphilis, Epstein-Barr virus and cytomegalovirus; positive results of feces testing for bacteria, viruses and parasites, and detection of multidrug-resistant genes such as extended-spectrum beta-lactamases, carbapenemases and resistance to colistin (21). We recruited two donors in our study after screening.

A disposable bottle was employed to collect the fresh donor stool samples. 100 g of feces was mixed with 500 ml of saline to achieve homogeneity within a timeframe of 6 h, using automated microfiltration equipment (GenFMTer, Nanjing, Jiangsu Province, China). Afterwards, the fecal suspension was processed for filtration within the equipment according to the predetermined schedule (22). Following microfiltration, the suspension was gathered into 50 ml tubes and centrifuged at 1,500 g for 3 min. The supernatant was removed and the precipitate was dissolved in normal saline to form fecal bacterial solution. The lyophilized protective agent was added to the fecal bacterial solution, which was then freeze-dried into powder by low temperature freeze-drying machine. The ultimate lyophilized powder was double-encapsulated in size 00 hypromellose capsules (Anhui Huangshan Capsule, Xuancheng, Anhui Province, China) and kept at a temperature of −80°C for storage. The fecal capsules were transferred to a temperature of −20°C approximately 2 h before the oral administration.

The dosage and treatment duration of FMT for ASD are still under research, with variations in dosage and duration reported in previous literatures. The FMT dosages of each course were determined based on the weight of the donor stool and the body weight of the ASD children, with a ratio of 1 g of donor stool per 1 kg of recipient body weight. Based on our research on FMT treatment for pediatric diseases, the dosage of FMT is generally in line with the current ratio and has shown efficacy. Additionally, through the analysis of the dynamic changes in patients before and after FMT, the duration of microbiota transplantation is approximately 4 weeks. Therefore, we chose to use 3 treatment courses to enhance the colonization of gut microbiota (23–25). Each course of capsules were taken on an empty stomach before meals, completed within 3 days, with a frequency of one course every 4 weeks, totaling 3 courses over 12 weeks.

Participants with ASD underwent the physical examination at week 0 of the study. Samples of stool were collected from ASD participants during week 0, 12 and 20 of the study for the purpose of analyzing the composition and abundance of gut microbiota through 16s rRNA and fungal ITS sequencing. Urine and saliva samples were collected at week 0, 12 and 20. Blood samples were collected at week 0 and 12. Urine and blood samples were collected for subsequent research on the mechanisms related to neurotransmitter metabolism and saliva samples for subsequent oral microbiota detection. Stool samples were collected from healthy children at the beginning of the study for 16s rRNA sequencing. The collected samples were processed immediately and stored at −80°C. After the FMT treatment, each participant was interviewed by phone. If any patients had questions or experienced adverse symptoms following the treatment, additional interviews or consultations would be conducted.

In order to evaluate the gastrointestinal symptoms of each participant, the parents were requested to fill out the Gastrointestinal Symptoms Rating Scale (GSRS). The GSRS is a self-reported questionnaire that uses a 7-point Likert scale with the following descriptive indicators: 1 (asymptomatic), 2 (slight), 3 (mild), 4 (moderate), 5 (moderate to severe), 6 (severe), and 7 (very severe) to assess the severity of gastrointestinal symptoms. It was initially designed as a hierarchical scale for face-to-face questioning to evaluate common gastrointestinal symptoms. It was later modified to create a self-management questionnaire. The GSRS is a 15-item questionnaire that assesses 5 dimensions of GI symptoms (abdominal pain, reflux, indigestion, diarrhea and constipation) experienced by participants in the last 2 weeks. The score for each dimension is calculated as the average score of all items within the dimension, ranging from 1 to 7.

To evaluate symptoms related to ASD, the following three scales were used: Autism Behavior Checklist (ABC), Childhood Autism Rating Scale (CARS) and Social Responsiveness Scale (SRS). The ABC includes 57 items that evaluates problem behaviors in five areas that are frequently observed in children with ASD. These areas are sensory, relating, body and object use, language, social and self-help. The CARS is a 15-item scale that can be utilized for both diagnosing ASD and evaluating the overall severity of the symptoms. The SRS is a 65-item questionnaire completed by parents or caregivers, and it assesses social communication, social cognition, social motivation and autistic mannerisms. The Sleep Disturbance Scale for Children (SDSC) is a questionnaire used to assess sleep problems in children. The SDSC consists of 26 items and evaluates six different domains of sleep disturbance. The ABC, CARS, SRS and SDSC were all given to participants at three different time points: baseline (week 0), 4 weeks after treatment (week 12) and at the final time of the evaluation period (week 20).

Microbial DNA was extracted from fecal samples by using the QIAamp DNA Stool Mini Kit (Qiagen, Hilden, Germany). Bacterial 16S rRNA V3-V4 hypervariable regions were amplified using polymerase chain reaction (PCR) primers 338F (5′-ACTCCTACGGGAGGCAGCAG-3′) and 806R (5′-GGACTACHVGGGTWTCTAAT-3′). Fungal internal transcribed spacer (ITS) regions were amplified using primers ITS1F (5′-CTTGGTCATTTAGAGGAAGTAA-3′) and ITS2R (5′-GCTGCGTTCTTCATCGATGC-3′). The PCR products were purified by using the AxyPrep DNA Gel Extraction Kit (Axygen Biosciences, Union City, CA, USA). Paired-end sequencing of purified amplicons were performed on an Illumina MiSeq PE300 platform (Illumina, Inc., San Diego, CA, USA) at Majorbio Bio-Pharm Technology Co., Ltd. (Shanghai, China).

FLASH (v1.2.11) was utilized to merge the raw 16S rRNA and ITS gene sequences and quality filtered with fastp (0.19.6) (26, 27). The high-quality sequences were then denoised using DADA2 to obtain unique Amplicon Sequence Variant (ASV) (28). The Qiime2 (version 2020.2) pipeline with recommended parameters was used for further analysis. Taxonomic assignment of ASV was conducted by using the Blast consensus taxonomy classifier implemented in Qiime2, the SILVA 16S rRNA database (v138) and the UNITE v8 database (29). The bacterial and fungal community composition data was analyzed by using the Majorbio Cloud Platform. The alpha diversity of the fecal microbiota was assessed using the observed richness (Sobs), Chao, abundance-based coverage estimator (ACE), Shannon and Simpson indexes. Moreover, the beta diversity was measured through principal coordinate analysis (PCoA) to Bray-Curtis distance and weighted UniFrac metric. The Wilcoxon rank-sum test was utilized to analyze the taxa difference in relative abundance between different groups.

In this study, the statistical analysis was conducted using SPSS 22.0 statistical software (SPSS Inc., Chicago, IL, USA). Continuous variables were represented as median and interquartile range (IQR). Normality tests were performed to guide the selection of statistical tests. The measurement data were compared using t-test, nonparametric Mann-Whitney U-test, Kruskal-Wallis H-test and Friedman test. In addition, the categorical data were compared using the chi-square test. All p-values were corrected with FDR, and if not stated otherwise, p-value less than 0.05 is regarded as statistically significant.

This study included 38 children who met the critical inclusion criteria for ASD, as well as 30 healthy children from separate families. The children with ASD in the study underwent FMT treatment for 12 consecutive weeks and were then monitored for an additional 8 weeks. The healthy children did not receive any treatment during the entire trial period (Figure 1). At the beginning of the trial, all participants were compared in terms of their demographic characteristics, such as gender, age distribution and body mass index (BMI) (Table 1). In addition, we assessed the severity of autism and GI symptoms in the children with ASD. Dietary assessment revealed that 47.4% (18 of 38) of ASD children had varying degrees of imbalanced dietary intake.

In the follow-up phase, we used the GSRS to measure the overall GI symptoms of participants after FMT, and found significant improvement in symptoms such as indigestion, constipation, and diarrhea. The average GSRS score decreased by 51% after the cessation of FMT treatment but reversed (32% decrease from baseline) 8 weeks after treatment stopped (Figure 2A), indicating that GI symptoms were obviously improved after FMT. We also observed improvements in symptoms related to ASD and sleep after FMT treatment. After stopping FMT treatment, the ABC scores decreased by 20% and remained decreased (23% decrease from baseline) after 8 weeks of stopping treatment (Figure 2B). At the end of the follow-up phase, there was a 10% reduction in scores on the CARS (Figure 2C) and a 6% reduction in scores on the SRS (Figure 2D). After 8 weeks of stopping treatment, scores on the SDSC, which evaluates sleep symptoms, decreased by 10% (Figure 2E). Changes in each scale scores before and after FMT treatment were measured (Table 2). There were no dropouts during the trial. In addition, short-term adverse events observed during follow-up included vomiting and fever in 2 participants, which were self-limiting and resolved within 24 h, and no long-term adverse events were observed, indicating that FMT therapy is generally safe. Therefore, our data suggests that oral lyophilized FMT may improve GI, ASD related and sleep symptoms without causing any serious adverse events.

We conducted an evaluation of alpha diversity and observed notable variations in alpha diversity between the HC and ASD groups during the initial assessment (Figure 3A). In order to investigate the dissimilarities in the gut microbiota of both groups, we computed beta diversity and detected significant disparities in beta diversity between the two groups (Figure 3B). When examining the microbiota composition at the phylum and genus levels, we noticed that there were dissimilarities between the microbiota of ASD patients and that of HC individuals (Figure 4A, Figure 3C). We used LEfSe analysis to identify two genera that showed significant differences in abundance between the ASD and HC group at baseline (LDA score greater than 4.0). Children in ASD group had a relatively higher abundance of Lachnospiraceae in family level, and Blautia in genus level, while HC group had a relatively higher abundance of Enterobacteriaceae in family level, and Escherichia-Shigella in genus level (Figure 4B).

After FMT treatment for 12 weeks, although there was no significant difference in alpha and beta diversity in children with ASD before and after FMT therapy (Figures 3A,B), FMT treatment resulted in alterations in the relative abundances of various bacterial genera in the samples of ASD patients. Specifically, there was an increase in the abundance of Eubacterium_hallii_group, Anaerostipes, Fusicatenibacter, Collinsella, Ruminococcus_torques_group and Dorea, and a decrease in the abundance of Blautia, Prevotella and Sellimonas (Figures 3C,D). The gut bacterial microbiota in ASD children with or without GI symptoms and changes after FMT were also analyzed, and we found that there may be significantly fewer individuals without GI symptoms, leading to no statistically significant differences in α and β diversity compared to the HC group (Supplementary Figures S1A,B). However, there are still differences in the abundance of specific bacterial genera (Supplementary Figures S1C,D).

Through amplicon-based sequencing of the fungal ITS region, we analyzed the gut mycobiota of our study cohort. We found high-quality fungal sequences in 31 out of 38 individuals with ASD and in 5 out of 30 HC. We observed a significant difference in fungal alpha diversity between the HC and individuals with ASD (Figure 5A). The analysis of beta diversity using PCoA showed that the gut mycobiota of individuals with ASD differed significantly from that of the HC (Figure 5B). At both the phylum and genus levels, we observed differences in mycobiota composition between individuals with ASD and the HC (Figure 4C, Figure 5C). At baseline, we utilized LEfSe analysis to identify seven genera that exhibited significant differences in abundance between the ASD and HC (LDA score greater than 4.0). In terms of family-level abundance, children in the ASD group exhibited relatively higher levels of Aspergillaceae, Saccharomycopsidaceae, Sporidiobolaceae and Cladosporiaceae. At the genus level, ASD children had relatively higher levels of Aspergillus, Saccharomycopsis, Rhodotorula and Cladosporium. In contrast, the HC group had relatively higher levels of Ascomycota and Saccharomyces at the genus level (Figure 4D).

Following 12 weeks of FMT treatment, we did not observe any significant differences in alpha and beta diversity between baseline and post-treatment samples from children with ASD (Figures 5A,B), the FMT treatment induced changes in the relative abundances of several fungal genera in samples from patients with ASD. Notably, we observed an increase in the abundance of Aspergillaceae and a decrease in the abundance of Saccharomycopsis, Cystobasidium and Chaetomium (Figures 5C,D). The analysis also included the gut fungal microbiota in ASD children with or without GI symptoms, as well as the changes after FMT. Due to the significantly lower number of individuals without GI symptoms, we observed no statistically significant differences in α and β diversity compared to the HC group (Supplementary Figures S2A,B). Nonetheless, distinct differences in the abundance of specific fungal genera were still present (Supplementary Figures S2C,D).

The associations of changed gut bacteria, fungi and clinical core symptoms in ASD patients were analyzed using Spearman's rank correlation coefficient. Intestinibacter, Terrisporobacter and Turicibacter were positively correlated with SRS score, including symptoms of social communication and cognition impairment. Ruminococcus_torques_group was positively correlated with ABC score, including five areas of sensory, relating, body and object use, language, social and self-help. On the other hand, increased abundances of Fusicatenibacter and Erysipelotrichaceae_UCG-003 were negatively correlated with the scores of ASD core symptoms (Figure 6A). The increased abundances of certain fungal genera in patients with ASD, such as Cladosporium and Auricularia, were positively correlated with the scores of core symptoms. Conversely, increased abundances of Saccharomyces, Rhodotorula, Cutaneotrichosporon and Zygosaccharomyces were negatively correlated with the scores of ASD core symptoms (Figure 6B).