Gut microbiota and urine metabolomics signature in autism spectrum disorder children from Southern China

Abstract

Background and aim

Autism spectrum disorder (ASD) is a neurodevelopmental disorder that may have long-term effects on individual development, family functioning, and social integration. This study aimed to determine the gut microbiota and urine metabolomics signature and identify the regional characteristics in ASD from Southern China.

Methods

We conducted a cohort study of 88 well-characterized participants from Guangxi Zhuang Autonomous Region in Southern China. Gut microbiota and urine metabolomics signature was explored by 16 S rRNA sequences and untargeted metabolomic profiles respectively.

Results

The gut microbial α-diversity of ASD were significantly lower than healthy controls. The β-diversity analysis indicated that the community structure in ASD group was obviously distinctive. Significant microbiota enriched in 5 sensitive species, Faecalibacterium prausnitzii, Bifidobacterium catenulatum, Blautia obeum, Lachnoclostridium sp., and Blautia sp. in ASD children. In addition, functional analysis of the gut microbiota revealed that the ATP-binding cassette and ABC-2 type transport system ATP-binding protein were closely associated with ASD. Notably, microbiota showing a positive correlation with Androstenedione, Stearamide, Oleamide, Cadaverine, Hexadecanamide, Orotic acid, Linoleic acid, Palmitoleic acid, Lauric acid, suggesting a potential association with the Arginine and proline metabolism pathway.

Conclusion

This study found lower α-diversity, unique β-diversity, enriched species, and positive correlations between microbiota and Arginine/Proline metabolis, which demonstrated typical signature of microbiota and metabolites discriminated Zhuang ethnic group ASD children of regional characteristics.

Impact

• Studied gut microbiota of ASD children in Guangxi Zhuang Autonomous Region, found lower α-diversity, unique β-diversity, enriched species, and positive correlations with Arginine/Proline metabolism links.

• Contributed to ASD regional specificity research, offering new insights into diversity and complexity, focusing on Guangxi Zhuang Autonomous Region.

• Insights advance understanding of ASD pathogenesis, especially regional/ethnic aspects, informing targeted interventions for precision medicine.

Introduction

Autism Spectrum Disorder (ASD) is a diverse neuro-developmental condition marked by heterogeneity and multifaceted etiology. Prevalence rates of ASD vary widely across regions, with notable increased in the US from 6.7‰ in 2000 to 23.0‰ in 2018 [1, 2], particularly prevalent in Korea (26.4‰) [3]. Conversely, low prevalence rates have been reported, such as 1.1‰ in Quito, Ecuador [4], and 2.9‰ across 8 provinces in Chinese from 2014 to 2016 [5], underscoring geographical disparities. Baio J et al. [6] also reported that the ASD prevalence in children of non-Hispanic whites was higher than non-Hispanic blacks. These heterogeneities may stem from complex genetic and environmental factors, as researches have revealed no single gene accounts for more than 1% of cases, and definitive biomarkers still remain elusive [7–9].

Yatsunenko et al. [10] documented that the gut microbiota of participants from the Amazon of Venezuela, rural Malawi, and metropolitan areas of the United States was compared, revealing significant variations in both composition and functionality across different age groups and geographical populations. Additionally, significant differences in gut microbiota structure were observed among different ethnic groups, while samples from the same ethnic group clustered closely together [11]. Furthermore, they highlight the importance of identifying regional specific gut microbiota for the diagnosis and treatment of regional specific diseases. For example, Wan et al. [12] reported that chronological age was the main factor associated with the disturbances of gut microbiota genera in the ASD cohort. They also identified neurotransmitter biosynthesis pathways were decreased with age, which could affect the functionality of the gut microbiota in ASD children.

Metabolic abnormalities have been reported in the pathogenesis of ASD. Accumulated evidence has shown the disturbances in metabolism associate with amino acids, mitochondrial dysfunction, oxidative stress, purine intermediates, and gut microbiota in ASD [13, 14]. Urine sample, compared to other biofluids, it can be readily obtained in large quantities non-invasive and presented low risk of infection to participants. Concentrations of metabolites in the urine are often higher than that of plasma, providing a richer matrix for analysis [15]. Thus, it is an attractive option for the discovery of metabolism biomarkers. A clinical study by Milagros et al. [16] has shown that Homocysteine (Hcy) is altered in the urine (also in blood) of children with ASD, and the overproduction of Hcy levels is significantly associated with the severity of the disorder. Wan et al. [12] reported that increased levels of N-methylnicotinic acid and N-methylnicotinamide were observer in the urine of children with ASD, which may influence tryptophan nicotinic acid metabolism. Both clinical trails and experimental animal studies have shown that an elevation in urine hydroxyl-polycyclic aromatic hydrocarbons (PAHs) were positively correlated with the severity of ASD [17, 18].

Therefore, alteration of gut microbial composition and urine metabolomics may contribute to the pathogenesis of ASD. However, the specific change in individual were inconsistent across different cohort demography and geography. To gain a deeper understanding of the pathogenesis of ASD among ethnic minority children in the Guangxi Zhuang Autonomous Region, we conducted a comprehensive research that focused on detecting the signature and inter-correlations in gut microbiota and urinary metabolites of ASD children in this region.

Materials and methods

Participants

58 ASD children (ASD group) were recruited between June 2022 and June 2023 in the Departments of Pediatrics of the First Affiliated Hospital of Guangxi Medical University. ASD group based on the following inclusion criteria: (1) typical autistic features; (2) an Autism Behavior Scale (ABC) score of ≥ 53 or a Modified Checklist for Autism in Toddlers score of (M-CHAT) ≥ 7, and an Autism Rating Scale for Children (CARS) score of ≥ 30 at screening and baseline. Exclusion criteria included: (1) schizophrenia, mood disorders and other psychosis; (2) intellectual disability; (3) the use of probiotics/prebiotics, or antibiotics within 4 weeks prior to the collection of fecal samples; (4) immune deficiency; (5) organic illnesses that affected their GI function, including digestive tract malformations, Hirschsprung’s disease chronic gastrointestinal diseases, spina bifida, and hypothyroidism. 30 healthy controls (CON group) were recruited from local kindergartens. The same set of exclusion criteria that was applied to the ASD group was similarly applied to the CON group. They were excluded if they had history of speech deficits, autistic features and other neurological/psychiatric disorders.

All participants were from the Guangxi Zhuang Autonomous Region and without infections, malnutrition, immunodeficiency, metabolic disorders, chronic inflammatory disease. All participants did not take any antibiotics, probiotics, or prebiotics within 1 month. Fecal and urine samples were collected. The demographics information, clinical information, obstetric histories, and imaging of participants were collected. This study was approved by the Ethics Committee of the First Affiliated Hospital of Guangxi Medical University (NO.2021-KT-003) and parents of participants have provided written informed consent.

ASD diagnostic

The diagnosis of ASD was made by two specialists (neurologist and psychiatrists) on the basis of the 4th edition of Diagnostic and Statistical Manual of Mental Disorders (DSM-IV) criteria [19], in line with China’s national conditions and conformed to the “Expert consensus on early diagnosis of autism spectrum disorder in Chinese young children" [20]. According to the above expert consensus, the ABC and M-CHAT were used for parents interviews and screening, and the CARS was adopted for diagnostic assessment. The Rome IV diagnostic criterias of Childhood Functional Gastrointestinal Disorders were used for assess Gastrointestinal (GI) status [21, 22], including stomach pain and hurt, discomfort when eating, and nausea/vomiting.

Sample preparation

Fecal samples were collected in the morning and stored at −80 °C with GUHE Flora Storage buffer (20190682, GUHE Laboratories, Hangzhou, China) for 16 S rRNA sequencing. Urine samples were collected by sterilized instruments on the same day and stored at −80 °C. Before Liquid Chromatography coupled with Mass Spectrometry (LC-MS) analysis, 100 µL urine sample was thawed and mixed with 300 µL methanol by vortexed for 30s. The mixture was centrifuged at 4 °C at 12,000 rpm for 15 min. 200 µL supernatant was then mixed with 5 µL internal standard (1 mg/mL, DL-o-Chlorophenylalanine) as samples pretreatment.

DNA extraction and 16 S rRNA sequencing

Total DNA were extracted from fecal samples using the GHFDE100 DNA isolation kit (20190952, GUHE Laboratories, Hangzhou, China). The quantity and quality of extracted DNAs were assessed using a NanoDrop ND-1000 spectrophotometer (Thermo Fisher, USA). PCR amplification of the V4 region of 16S rRNA gene was performed using the forward primer 515 F (5’-GTGCCAGCMGCCGCGGTAA-3’) and the reverse primer 806 R (5’-GGACTACHVGGGTWTCTAAT-3’). The PCR amplicons were purified using Agencourt AMPure XP Beads (Beckman Coulter, Indianapolis, IN) and then sequenced at the Illlumina NovaSeq6000 platform (Hangzhou, China). 16 S rRNA sequencing raw data have been deposited to National Center for Biotechnology Information (NCBI) under the BioProject number PRJNA1277165.

Gut microbiota analysis

Paired-end reads were obtained and merged using Vsearch V2.18.0, followed by Operational Taxonomic Unit (OTU) picking [23]. The OTUs and sequencing data were process by the Quantitative Insights Into Microbial Ecology (QIIME2, v2024.2) [24]. OTU taxonomic classification was performed by aligning the representative sequences set against the SILVA138 database [25]. And then OTUs were assigned for bioinformatics and statistical analysis utilizing QIIME2 and R packages (v3.2.0). The key OTUs were analyzed by 10-fold cross-validation.

We standardized the sequencing depth by rarefaction. The rarefaction was applied uniformly across all samples. α-diversity was assessed by species richness indices (chao and ace) as well as species diversity indices (sobs) via t-tests. β-diversity was estimated through unifrac distance metrics-across samples. Principal component analysis (PCA), non-metric multidimensional scaling (NMDS), principal-coordinate analysis (PCoA), and partial least-squares discriminant analysis (PLS-DA), all with 7-fold cross validation, and a PERMANOVA test were employed to visually assess the overall dissimilarity and similarity of communities. Species differences inter-group was identified by Kruskal-Wallis test.

DESeq2 (DESeq 1.26.0) was the primary method performed to analysis species differences between groups, then Linear discriminant analysis effect size (LEfSe, v1.1.0) and Welch’s t-test as supplementary. The P-values from DESeq2 and Welch’s t-test were FDR-adjusted to Q-values. The cladogram was graphed with linear discriminant analysis (LDA) value > 2 and P < 0.05 [26]. PICRUSt 1.1.4 was utilized to predict the pathways of gut microbiota and the activity of gut-brain modules (GBMs) [27]. And MetaCYC database was used to identify metabolic pathways, mostly involved in the biosynthetic pathway.

LC-MS analysis

The gradient elution method based on acetonitrile-water as the mobile phase was used for separation at ACQUITY HSS T3 ultra-performance liquid chromatography (UPLC) and a mass spectrometer (Thermo Scientific, USA). All samples were both analyzed in reverse phase chromatography with positive ionization methods (POS) and negative ionization conditions (NEG) for compounds. Quality control (QC) samples were prepared by mixing equal amounts of the test samples, and were analyzed before, during, and after the LC-MS injection of the test samples.

LC-MS analysis raw data have been deposited to Metabolights datebase (MTBLS12605). Raw data de-noising and normalization were performed to correct variation and further obtained the three-dimensional matrices, which included the peak number, sample name, and normalized peak area. LC-MS data were then analyzed by SIMCA14.1 (Sartorius Stedim Data Analytics AB, Umea, Sweden).

HMDB database was applied for the qualitative analysis of metabolites. PCA and Orthogonal projections for latent structures-discriminant analysis (OPLS-DA) were applied to obtain a higher level of separation and classification between groups. OPLS-DA model first principal component Variable Importance in the Projection (VIP) value was used to identify the differential metabolites between groups. The metabolites with VIP > 1, t test P < 0.05, and Q < 0.05 (multiple hypothesis test correction by False discovery rate (FDR)-adjusted Q-values), and |log2 fold change|>1.5 were considered as significant features.

Simultaneously, MS2 score and Standard score (z-score) were used to evaluate these differential metabolites. MS2 score [0, 100] is the score for material secondary matching, the larger MS2 score, the better the differentiation. Z-score was based on the mean and standard deviation of Control group, and was used to measure the relative abundance of metabolites at the same level in various samples.

MetaboAnalyst 4.0 and the Kyoto Encyclopedia of Genes and Genomes (KEGG) metabolomics reference library were utilized to conduct metabolite pathway analysis. We used the P value (Holm adjust) after multiple hypothesis test correction by Holm-Bonferroni method, and the P value after multiple hypothesis test correction by FDR method to reduce false positives.

Statistical methods

Power analysis of sample sizes was conducted using G*Power 3.1.9.7., and power > 0.8 suggested adequate statistical power. Clinical information was statistically analyzed using SPSS 26.0. Quantitative data was presented using median (interquartile range). T-tests was used for comparison between groups when the indices pass normality test and had adequate statistical power, if not, used Wilcoxon -Mann-Whitney tests. Count data was represented by frequency (%), and chi-square test analysis was used for comparison between groups when the indices had sufficient sample size and adequate statistical power, if not, used Fisher’s exact test. Spearman correlations corrected by FDR, db-RDA and CCA were used to link microbes and metabolites. All analyses showed statistically significant differences with P < 0.05 or Q < 0.05.

Results

Clinical characterization

As shown in Table 1, age, gender, BMI, and antibiotic using in pregnancy resulted with power < 0.8, and other indices with power > 0.8. There were no statistically significant differences in age (Z=−0.795, P = 0.427), gender (P = 0.406), and BMI (Z=−1.585, P = 0.113) between the CON and ASD groups. As for GI problems, there was statistically significant difference in the incidence between two groups (χ² = 23.25, P < 0.0001), with higher proportion in the ASD group experiencing functional constipation and diarrhea.

Table 1.

Comparison of the characteristics of children in the CON group and the ASD group

Characteristics		CON(n = 30)	ASD(n = 58)	POWER	Z/χ2	P
Age(month)	/	48.00(13.25)	49.00(19.75)	0.183	−0.795	0.427
Gender	Male	27(90.00%)	55(94.83%)	0.137	/	0.406
	Female	3(10.00%)	3(5.17%)
BMI (kg/m2)	/	14.88(0.81)	14.46(3.53)	0.151	−1.585	0.113
GI problems	No	22(73.33%)	17(29.31%)	1.000	23.250	< 0.0001
	Functional abdominal pain ‒ not otherwise specified	5(16.67%)	5(8.62%)
	Functional constipation	2(6.67%)	29(50.00%)
	Functional diarrhea	1(3.33%)	7(12.07%)
Mode of delivery	Natural delivery	22(73.33%)	24(41.38%)	0.816	8.092	0.004
	Caesarean section	8(26.67%)	34(58.62%)
Time of delivery	Premature infant	7(23.33%)	31(53.45%)	0.847	8.856	0.006
	Term infant	23(76.67%)	25(43.10%)
	Overdue delivery	0(0.00%)	2(3.45%)
Antibiotic using in pregnancy	No	30(100.00%)	53(91.38%)	0.124	/	0.161
	Yes	0(0.00%)	5(8.62%)
Breadfed for month	≤ 12	16(53.33%)	56(96.55%)	1.000	24.827	< 0.0001
	>12	14(46.67%)	2(3.45%)
ABC Scales	/	21.50(2.75)	75.50(7.50)	1.000	−7.674	< 0.0001
CARS Scales	/	14.00(2.00)	36.00(3.00)	1.000	−7.723	< 0.0001
GDS score	Gross motor	98.00(18.00)	55.00(25.00)	1.000	−7.619	< 0.0001
	Fine motor	95.00(16.00)	56.00(25.25)	1.000	−7.670	< 0.0001
	Adaptive behavior	92.00(14.50)	49.50(20.50)	1.000	−7.620	< 0.0001
	Language function	88.00(16.00)	35.00(9.00)	1.000	−7.679	< 0.0001
	Personal/social function	90.00(17.50)	38.00(11.75)	1.000	−7.686	< 0.0001

There was statistical difference in the delivery methods between two groups (χ² = 8.092, P = 0.004). Most CON group children were giving birth naturally, but most ASD group children by caesarean section. Moreover, the distribution for time of delivery between the two groups also showed statistical differences (χ² = 8.856, P = 0.006). The majority of CON group children were term infants, but the majority of ASD group children were premature infants. There was a statistically significant difference in the distribution of breastfeeding time between the two groups (χ² = 24.827, P < 0.0001), and 96.55% of ASD group breastfed for ≤ 12 months. However, there was no statistically significant difference in whether the mothers of the two groups of children took antibiotics during pregnancy (P = 0.161).

The ABC and CARS scores of ASD children were significantly higher than those of the CON group (Z=−7.674, P < 0.0001, Z=−7.723, P < 0.0001). The GDS scores of ASD children were significantly lower than those of CON group (Z _{Gross moto}=−7.619, P < 0.0001, Z _{Fine motor}=−7.67, P < 0.0001, Z _{Adaptive behavior}=−7.62, P < 0.0001, Z _{Language function}=−7.679, P < 0.0001, Z _{Personal/social function} =−7.686, P < 0.0001).

Gut microbiota abundance and diversity

The abundance and diversity of the microbial communities were determined by 16 S rRNA sequencing. The Rarefaction Curve (Fig. 1A) was progressively flatter, suggested the adequate amount of sequencing in this study. Rank-Abundance reflected the richness and evenness of the microbiota. The richer the microbiota, the wider the curve in the direction of the horizontal coordinate; the more even the microbiota, the flatter the curve. Results suggested low evenness in ASD group when compared with CON group (Fig. 1B).

Fig. 1.

Sample diversity analysis. The Rarefaction curves (A) and Rank-Abundance curves (B) of Autism spectrum disorder group and control group. Comparison of the indices ace, chao, and sobs for calculating community richness between the Autism spectrum disorder group and the control group (C). The distance of the sample in the Principal Component Analysis (PCA) plot (D) reflects the similarity of the sample species composition. Non metric Multidimensional Scaling (NMDS) plot (E) reflects the degree of difference between different samples through the distance between points. PCoA (Principal Coordinates Analysis) plot (F) reflects the similarity of species composition and structure through inter group sample distance. Similarity analysis (Anosim) (G)

We used α-diversity indices to measure microbial abundance between groups, and found ASD group had a significant lower α-diversity than CON group (ace/chao/sobs Indices, P < 0.001, Fig. 1C). β-diversity analysis resulted with a clear separation that could discriminate ASD from CON group via PCA (Fig. 1D), NMDS (Fig. 1E), and PCOA (Fig. 1F) algorithms. These indicated the microbial composition of ASD group had been altered distinctively, as confirmed by Distances box plot (P = 0.020, Fig. 1G) and PERMANOVA test formally test group-level compositional differences significantly (F.Model = 7.318, R²=0.079, P = 0.001).

Gut microbiota composition

The Venn diagram provided a visual representation of the number of species shared between groups. As showed in Figs. 2A and 386 species were shared in both ASD and CON groups, while 311 and 147 species were unique to ASD and CON group respectively. Circos plot (Fig. 2B) revealed the abundance of Bacteroides and Faecalibacterium was beyond 5%, which were also showed visually as dominant genus in the abundance 3D bar chart (Fig. 3A). A Taxonomic Tree was drawn using GraPhlAn and results were shown in Fig. 3B, suggesting the dominant community structure included 28 genus, such as Bacteroides, Faecalibacterium and so on.

Fig. 2.

Community composition analysis. Venn diagram (A); Circos diagram of community composition (B)

Fig. 3.

Abundance analysis. The abundance 3D bar chart (A) provides a more three-dimensional observation of the dominant species or functional distribution in all samples; Taxonomic Tree (B) showed the classification of dominant species based on the taxonomy of the sample, combined with species abundance information, presented in a circular dendrogram

DESeq2 tool was preformed to detect differentially abundant taxa and then RF classifiers were used to identify a gut microbiota signature for ASD. In total, 5 species (Faecalibacterium prausnitzii, Bifidobacterium catenulatum, Blautia obeum, Lachnoclostridium sp., Blautia sp., all Q < 0.05) with low error rate plus standard deviation were selected as the most important biomarkers (Fig. 4A). Additionally, LEfSe analysis (Fig. 4B) and Welch’st-test (Fig. 5A) results suggested the proportions of these spices were significantly upregulated in ASD group when compared with CON group (all Q < 0.05). These species had significant value as the critical microbiota to distinguish the ASD from healthy control.

Fig. 4.

Microbial categories. DESeq2 (A) selects the most important microbial categories for sample classification. LEfSe (Linear Discriminant Analysis Effect Size) (B): the horizontal axis represents the significant marker LDA value, and the vertical axis represents the significant marker analyzed

Fig. 5.

Different species analysis and functional enrichment. STAMP difference analysis (Welch's t-test) (A): The left figure shows the abundance ratios of different species classifications in two groups of samples, and the middle figure shows the proportion of differences in species classification abundance within the 95% confidence interval. The rightmost value is the P-value and Q-value, where P-value and Q-value<0.05 indicates significant differences, *represents 0.01<P-value (Q-value)<0.05,**represents 0.001<P-value (Q-value)<=0.01, ***represents P-value (Q-value)<=0.001. Functional abundance heatmap (B): columns represent samples, rows represent functions, and color blocks represent functional abundance values. The redder the color, the higher the abundance, and vice versa

Gut microbiota functional and pathway

The FAPROTAX histogram (Fig. 5B) suggested that the ASD children were inactivation in chemoheterotrophy and fermentation microbiota infunction. The KEGG pathway analysis suggested that the microbiota in the ASD group were enriched in function K06147 (ABCB-BAC; ATP-binding cassette, subfamily B, bacterial), K01990 (ABC-2.A; ABC-2 type transport system ATP-binding protein) and K03088 (rpoE; RNA polymerase sigma-70 factor, ECF subfamily) (Table 2). The results of the predicted COG function (Fig. 6A) is enriched in E (Amino acid transport and metabolism), G (Carbohydrate transport and metabolism), K (Transcription) and J (Translation, ribosomal structure and biogenesis).

Table 2.

Enriched function

KO	Description
K06147	ABCB-BAC; ATP-binding cassette, subfamily B, bacterial
K01990	ABC-2.A; ABC-2 type transport system ATP-binding protein
K03088	rpoE; RNA polymerase sigma-70 factor, ECF subfamily
K02004	ABC.CD.P; putative ABC transport system permease protein
K02003	ABC.CD.A; putative ABC transport system ATP-binding protein
K01992	ABC-2.P; ABC-2 type transport system permease protein
K02529	lacI, galR; LacI family transcriptional regulator
K07024	SPP; sucrose-6-phosphatase [EC:3.1.3.24]
K02025	ABC.MS.P; multiple sugar transport system permease protein

Fig. 6.

COG functional enrichment and metabolic PLS-DA/OPLS-DA. COG functional abundance analysis (A): The horizontal axis represented COG functional categories, and the vertical axis represented the relative abundance values of functions. PLS-DA (B): two-dimensional plot ellipses representing 95% confidence intervals. OPLS-DA(C): two-dimensional plot ellipses representing 95% confidence intervals

Urine metabolic signatures

LC-MS of urine samples from 58 ASD children and 30 healthy controls and identified 97 metabolites (including 66 positive ions, ES⁺ and 31 negative ions, ES⁻) from the microbiota and host (Table 3). In total, we discovered 66 metabolites of positive ionic were responsible for discriminating between the two groups, of which 10 were up-regulated and 56 were down-regulated (VIP value > 1, t test P < 0.05, and Q < 0.05, Table 3). Z-score was used to measure the relative content of these 7 metabolites.

Table 3.

Urine metabolic signatures

MS2 name	MS2 Score	VIP	P-velue	Q-velue	Fold Change	LOG_Fold Change
3-Methylxanthine	95.80	1.03	0.02	0.03	0.20	−2.31
Theobromine	98.80	1.30	0.02	0.02	0.21	−2.25
9-Methyluric acid	92.30	1.42	0.02	0.02	0.23	−2.10
D-Proline	99.00	1.30	0.00	0.00	0.38	−1.38
4-GUANIDINOBUTANOATE	83.90	1.26	0.01	0.01	0.45	−1.15
LysoPC(18:0/0:0)	87.60	1.04	0.00	0.01	0.48	−1.05
Octanoyl-L-Carnitine	86.30	1.35	0.02	0.02	0.55	−0.86
Xanthurenic acid	95.30	1.38	0.00	0.00	0.55	−0.86
N-(2-Furoyl)glycine	89.10	1.51	0.04	0.04	0.56	−0.84
4-oxododecanedioic acid	79.30	1.95	0.01	0.02	0.57	−0.82
Decanoylcarnitine	94.80	1.33	0.02	0.03	0.57	−0.82
Glycyl-L-leucine	69.40	1.63	0.00	0.00	0.57	−0.82
Glycocholic Acid	89.90	1.13	0.01	0.02	0.58	−0.79
Trimethylamine N-oxide	87.80	1.40	0.01	0.01	0.59	−0.75
4-Hydroxyphenylacetic acid	96.70	1.37	0.00	0.01	0.60	−0.74
N6,N6,N6-Trimethyl-L-lysine	97.50	1.32	0.01	0.02	0.61	−0.71
Salicylic acid	65.20	1.47	0.00	0.00	0.62	−0.69
4-Pyridoxic acid	99.30	1.05	0.00	0.01	0.63	−0.68
α-Aspartylphenylalanine	86.20	1.06	0.01	0.02	0.63	−0.66
4-Hydroxybenzoic acid	86.40	1.10	0.03	0.04	0.64	−0.65
Hexanoylcarnitine	99.00	1.23	0.00	0.01	0.64	−0.63
Creatinine	99.90	1.47	0.00	0.00	0.64	−0.63
Pantothenic acid	94.80	1.05	0.01	0.02	0.65	−0.62
cis, cis-Muconic acid	77.60	1.24	0.00	0.00	0.65	−0.61
Urocanic acid	98.30	1.53	0.00	0.00	0.65	−0.61
DL-α-Aminocaprylic acid	98.70	1.29	0.01	0.02	0.66	−0.60
Kynurenic acid	99.40	1.16	0.00	0.00	0.66	−0.59
Pipecolic acid	68.00	1.40	0.00	0.00	0.66	−0.59
Prolylleucine	92.40	1.36	0.00	0.00	0.67	−0.57
2,8-Quinolinediol	79.80	1.09	0.00	0.01	0.68	−0.55
Suberic acid	77.40	1.82	0.03	0.03	0.68	−0.55
Leucylproline	99.20	1.10	0.00	0.01	0.70	−0.52
2-Pyrrolidineacetic acid	72.70	1.30	0.01	0.02	0.70	−0.52
Glycylproline	85.70	1.16	0.00	0.01	0.70	−0.51
Picolinic acid	98.00	1.20	0.00	0.01	0.71	−0.50
1-Methyladenosine	99.00	1.12	0.00	0.00	0.72	−0.48
gamma-Glutamylleucine	92.50	1.34	0.00	0.00	0.72	−0.47
N6-Methyladenine	93.80	1.10	0.00	0.00	0.72	−0.47
N6-Threonylcarbamoyladenosine	88.50	1.15	0.00	0.00	0.72	−0.47
N-Acetylornithine	61.20	1.22	0.02	0.03	0.72	−0.47
Nicotinuric acid	71.40	1.16	0.00	0.01	0.73	−0.46
3-Hydroxyanthranilic acid	92.20	1.10	0.00	0.01	0.73	−0.45
Thymine	69.50	1.10	0.00	0.00	0.74	−0.44
Proline betaine	61.40	1.32	0.01	0.01	0.75	−0.42
Hydroxyproline	82.00	1.30	0.00	0.00	0.75	−0.41
L-Pyroglutamic acid	96.30	1.35	0.00	0.00	0.76	−0.40
L-Valine	87.00	1.37	0.00	0.01	0.76	−0.39
Adenosine	99.90	1.20	0.01	0.01	0.77	−0.38
N2,N2-Dimethylguanosine	90.90	1.18	0.00	0.01	0.77	−0.38
Citrulline	62.30	1.14	0.00	0.01	0.77	−0.37
Imidazolelactic acid	96.20	1.38	0.00	0.00	0.78	−0.36
Adenine	89.20	1.12	0.01	0.01	0.78	−0.36
L-Homoserine	65.40	1.05	0.00	0.01	0.79	−0.35
2-Hydroxyhippuric acid	70.80	1.09	0.04	0.04	0.79	−0.34
Creatine	99.80	1.04	0.01	0.02	0.79	−0.33
4-Acetamidobutanoic acid	82.20	1.02	0.01	0.02	0.82	−0.29
Androstenedione	65.20	1.07	0.00	0.01	1.27	0.34
Arachidoyl Ethanolamide	70.40	1.14	0.02	0.02	1.40	0.48
Dihydrosphingosine	61.70	1.23	0.00	0.00	1.43	0.51
Stearamide	76.80	1.43	0.00	0.00	1.63	0.71
Oleamide	78.40	1.81	0.00	0.00	1.67	0.74
Cadaverine	70.70	1.92	0.00	0.00	1.76	0.81
Hexadecanamide	95.60	1.76	0.00	0.00	1.78	0.83
Erucamide	94.20	1.66	0.00	0.00	1.79	0.84
Monoethylhexyl phthalic acid	99.60	1.79	0.00	0.00	1.92	0.94
Phosphoric acid	96.80	1.42	0.00	0.00	2.03	1.02

Urine metabolic pathway

Pathway enrichment analysis was performed to understand the functions of these differential metabolites compared with healthy controls, using the KEGG metabolic pathway database. The KEGG mtabolic pathway histogram (Table 4) suggested that the microbiota in the ASD group were enriched in hsa01100 Metabolic pathway (33 metabolics), hsa00330 Arginine and proline metabolism (6 metabolics) and hsa00380 Tryptophan metabolism (4 metabolics). We further screened the pathways and identify the key pathways with the highest correlation with differences by enrichment analysis and topology analysis. Metabolite mapping results suggested urine metabolic of ASD group enriched in Arginine and proline metabolism, Arginine biosynthesis, Pantothenate and CoA biosynthesis pathway (Table 5).

Table 4.

Metabolites pathway enrichment

KEGG pathway	Compound
hsa01100 Metabolic pathways - Homo sapiens (human)(33)	cpd: C01104 Trimethylamine N-oxide; cpd: C01672 Cadaverine; cpd: C00791 Creatinine; cpd: C00763 D-Proline; cpd: C00183 L-Valine; cpd: C00263 L-Homoserine; cpd: C10164 Picolinic acid; cpd: C00178 Thymine; cpd: C01879 5-Oxoproline; cpd: C00408 L-Pipecolate; cpd: C01157 Hydroxyproline; cpd: C00300 Creatine; cpd: C00147 Adenine; cpd: C00805 Salicylate; cpd: C00156 4-Hydroxybenzoate; cpd: C00785 Urocanate; cpd: C02480 cis, cis-Muconate; cpd: C02946 4-Acetamidobutanoate; cpd: C01035 4-Guanidinobutanoate; cpd: C00642 4-Hydroxyphenylacetate; cpd: C00632 3-Hydroxyanthranilate; cpd: C16357 3-Methylxanthine; cpd: C00437 N-Acetylornithine; cpd: C00327 L-Citrulline; cpd: C07480 Theobromine; cpd: C00847 4-Pyridoxate; cpd: C03793 N6,N6,N6-Trimethyl-L-lysine; cpd: C01717 4-Hydroxy-2-quinolinecarboxylic acid; cpd: C00864 Pantothenate; cpd: C00212 Adenosine; cpd: C00280 Androstenedione; cpd: C00836 Sphinganine; cpd: C01921 Glycocholate
hsa00330 Arginine and proline metabolism - Homo sapiens (human)(6)	cpd: C00791 Creatinine; cpd: C00763 D-Proline; cpd: C01157 Hydroxyproline; cpd: C00300 Creatine; cpd: C02946 4-Acetamidobutanoate; cpd: C01035 4-Guanidinobutanoate
hsa00380 Tryptophan metabolism - Homo sapiens (human)(4)	cpd: C10164 Picolinic acid; cpd: C00632 3-Hydroxyanthranilate; cpd: C01717 4-Hydroxy-2-quinolinecarboxylic acid; cpd: C02470 Xanthurenic acid

Table 5.

Metabolite mapping results

Metabolic Pathway	Arginine and proline metabolism	Arginine biosynthesis	Pantothenate and CoA biosynthesis
Total	38	14	19
Hits	5	2	2
Raw P	0.0049	0.0650	0.1110
-ln(P)	5.3208	2.7339	2.1981
Holm adjust	0.4107	1	1
FDR	0.4107	1	1
Impact	0.0731	0.2284	0.0071
Hits Cpd	Creatine cpd: C00300; D-Proline cpd: C00763; Hydroxyproline cpd: C01157; 4-Guanidinobutanoate cpd: C01035; 4-Acetamidobutanoate cpd: C02946	N-Acetylornithine cpd: C00437; L-Citrulline cpd: C00327	Pantothenate cpd: C00864; L-Valine cpd: C00183

Associations of gut microbiota and urine metabolites

To obtain the metabolic potential of gut microbiota in the ASD group, we calculated Spearman correlation coefficient of altered gut microbiota and differentially accumulated urine metabolites at specie level. The permutation test of orthogonal projections to latent structures-discriminant analysis (PLS-DA) and OPLS-DA model for group ASD vs. CON revealed a separation of positive and negative urine metabolites representing 95% confidence intervals (Fig. 6B, C). Volcano plot of metabolites of ASD patients compared to healthy controls was showen in Fig. 7A. Z-score analysis showed metabolites expression in the ASD and controls (Fig. 7B).

Fig. 7.

Metabolic profiles in the positive ion mode of ASD patients and healthy controls. Volcano plot (A) of metabolites of ASD patients compared to healthy controls. The y-axis representsP-value converted to negative log₁₀ (Scale) and the x-axis represents log₂ (Fold change). Up regulated significant metabolites were highlighted in red. Down-regulated significant metabolites were highlighted in blue. Z-score analysis (B) showing metabolites expression in the ASD and controls

Results revealed that 11 metabolites had strong correlations with the microbes of ASD patients. These metabolites involved lipids and lipid-like molecules, organic acids and derivatives and polyketides, and organic oxygen compounds, among others (Fig. 8A-C). Specifically, we found the 9 metabolites (Androstenedione, Stearamide, Oleamide, Cadaverine, Hexadecanamide, Orotic acid, Linoleic acid, Palmitoleic acid, Lauric acid) positively correlated with microbiotas (Blautia obeum, Lachnoclostridium sp., Blautia sp., Roseburia inulinivorans, Bacteroidefragilis, Bacteroidethetaiotaomicron, Akkermansia sp., Erysipelotrichaceae UCG-003 sp., Bacteroidexylanisolvens, Collinsella stercoris, Tuzzerella sp.) and played promoted role in neurodevelopment dysfunction and up-regulated in ASD.

Fig. 8.

Correlations analysisi between metabolites and microbes. Spearman correlations corrected by FDR (A); db-RDA (B) and CCA (C) were used to link microbes and metabolites

Disscusion

Currently, there is still no specific diagnostic approach for diagnosing ASD. So, it is imperative to conduct a comprehensive evaluation during the diagnosis process. Though the ADI-R and ADOS-2 are internationally recognized standardized diagnostic tools for ASD [28], studies on the reliability of these two tools reveal that their specificity is limited. The diagnostic tools commonly employed are intricate, time-consuming, and lack specificity. Consequently, there is a need for simpler screening methods to address the challenge of delayed diagnosis.

In this cohort, a total of 88 participants were included, we examined gut microbiota composition and urine metabolomics to discover the signature for discriminating ASD children. Considering the regional host and environment, we focused on patients were from Guangxi Zhuang Autonomous Region. We found the frequency of gastrointestinal problems in ASD group was high to 70.69%, with a higher incidence of constipation and diarrhea. This was consistent with previous reports that have found ASD children have an increased rate of gastrointestinal disorders in general [29]. It is possible that ASD children is attributable to the behavioral features of ritualistic tendencies and insistence on sameness, which more likely to manifest gastrointestinal symptoms [30].

Most ASD children in our study were delivered by caesarean section and were premature infant. Curran EA et al. [31] found that children born by caesarean section are approximately 20% more likely to be diagnosed as having ASD. Chen M et al. [32] conducted a meta-analysis revealed caesarean section was a risk factor for ASD in offspring compared with vaginal delivery. A longitudinal study results suggested an estimated prevalence of ASD in the very premature population of 18.46% [33]. Brito A et al. [34] indicated that breastfeeding might act as a protective factor for ASD children whose mothers took antibiotics during pregnancy. A longer period of exclusive breastfeeding was associated with a subsequent reduced likelihood of ASD diagnosis [35]. However, 96.55% ASD children in our study were breastfed for ≤ 12 months, and most of them had higher ABC and CARS scores, lower GDS scores. Therefor, it is very necessary to pay close attention to care for premature infants who have undergone cesarean section, and to improve the breastfeeding period and nutrition.

The high prevalence of GI symptoms, C-section delivery, and prematurity in the ASD group were potential confounders that could independently affect microbiome and metabolome profiles. Limited by the sample size of sub-groups in this study, we couldn’t independently statistically analyzed these influence factors. Age, gender, BMI, and antibiotic using in pregnancy also might be considered insufficient evidence due to insufficient efficacy, they need to be verified in the future.

Here, we found that a distinct gut microbiota signature could identify ASD from Guangxi geographical cohorts, and discovered the alterations in the characteristics of urine metabolites in children with ASD. Our study showed the gut microbial α-diversity of ASD was significantly lower than healthy controls. It has been reported that α-diversity usually down-regulated in various diseases, such as overweight, cancer, neurological and psychiatric disorders [36–38]. The β-diversity analysis showed that a large proportion of the ASD group did not cluster with the CON group, indicating that the community structure was obviously distinctive.

ASD children showed a significantly increased specie abundance of Faecalibacterium prausnitzii, Bifidobacterium catenulatum, Blautia obeum, Lachnoclostridium sp., and Blautia sp.. Coretti L et al. also found an increase of Faecalibacterium prausnitzi in ASD children from Department of Translational Medical Science-Pediatric Section, University of Naples Federico II, Naples, Italy [39]. Guangxi’s climate is similar to that of Naples, being consistently warm and humid. Such climatic conditions might allow beneficial bacteria in the environment to provide Faecalibacterium prausnitzi with the nutrients needed for growth, which is conducive to the growth and reproduction of microorganisms, promoting the growth of Faecalibacterium prausnitzi and other microorganisms in the intestines of children with ASD.

In addition, Faecalibacterium prausnitzi is a late colonizer of human gut and a major butyrate producer [40]. High levels of Faecalibacterium prausnitzi were associated to increase of fecal butyrate levels within normal range [41]. Guangxi Zhuang Autonomous Regioni has many characteristic sour fermented food, such as sour bamboo shoots, kimchi, fermented bean products and so on. These fermented foods are often made using natural fermentation, which relies on microbial communities in the local environment to participate in the fermentation process. These might have a symbiotic or synergistic relationship with Faecalibacterium prausnitzi. Yap CX et al. [42] found that the limited dietary diversity and high preference for high-carbohydrate, high-fat foods (UPFs) among ASD children were associated with specific ASD characteristics. In Guangxi, where fermented foods were widely consumed, children with autism might frequently consume these fermented foods, which could alter the gut microbiome, stimulating the growth and proliferation of Faecalibacterium prausnitzi, leading to an increase in its population.

As for Blautia obeum, Li Y et al. [43] showed a reducing Blautia obeum level in ASD children. However, Blautia obeum showed significantly increased relative abundance in Parkinson’s disease and Colorectal Cancer [44, 45]. The studies of Yang C et al. [46] and Ding X et al. [47] exhibited elevated levels of Lachnoclostridium at the genus level in ASD children, while the studies of Ma B et al. [48] with a decrease in the relative abundance. There are few studies about Bifidobacterium catenulatum and Blautia sp in ASD. More in-depth research is needed in the future to better discover the patterns of specific microbiota communities in children with ASD.

It was confirmed that dietary preferences in different regions affects ASD-associated gut dysbiosis and metabolic capacity. Due to varying dietary patterns across regions, when compared with the reports from Hong Kong, China [49], and abroad [50], we found unique microbial communities in Guangxi region. In our future research, we will strengthen the detailed record and collection of data such as dietary habits and nutrient intake of ASD children and control groups in our region. Through maslin2 method, these factors will be added to study their effects on the results of associated microbiota.

Pathway enrichment analysis result suggested that ASD associated with Energy and Metabolism related pathways, such as ATP-binding cassette, PPP pathways and so on. Whole genome sequencing has shown that ATP-binding cassette pathway is associated with an increased incidence of gallstone disease, gallbladder and bile duct carcinoma or elevated liver function tests [51]. Deficiency of PPP rate-limiting enzyme restricts growth and predisposes to ASD.

As COG results showed that differential species were enriched in biological functions, such as Amino acid transport and metabolism, Carbohydrate transport and metabolism, Transcription and Translation, ribosomal structure and biogenesis. The alterations in these pathways could link to metabolic dysfunction and immune system changes that affect the brain-gut axis [52, 53]. Intestinal microbiome molecules, such as neurotransmitters, amino acids, short-chain fatty acids (SCFA), lipopolysaccharides (LPS), and microbiota-associated molecular patterns (MAMPs), interacted with the host immune system through circulation, influencing the host’s metabolism and nervous system [54, 55]. MAMPs included peptides, nucleotides, carbohydrates, and lipids. When the host failed to detect MAMPs, it might accompany by acute to chronic inflammation to alter brain development and function [56].

Decreased metabolism function of carbohydrate, amino acid and lipid in gut microbiota were found in the IFN-γ-high ASD children [57]. Aziz-Zadeh L reported that abnormal levels of tryptophan, a metabolite of gut flora, might exacerbate symptoms including social impairment, sensory sensitivity and aversion by affecting the brain function regions such as the insula and the putamen in ASD children [58]. CHD8 is a high-risk factor in ASD. Kawamura A et al. found that the activation of amino acid metabolism, and transport related pathways in neurons of mice with CHD8 mutations might lead to ASD core symptoms (social deficits, repetitive behaviors) by disrupting gene expression dependent on neuronal activity [59].

Mothers’ health problems during pregnancy, such as anxiety and depression, directly affect the neurodevelopment of the fetus through genes, immune system or other physiological mechanisms, leading to autism. Children with autism are also often accompanied by symptoms such as anxiety and depression [60]. Li Q et al. found the different microbiota between the prenatal stress depress pregnant female rats and control group were mainly involved in carbohydrate metabolism, and amino acid metabolism pathways [61]. Zhu J et al. have reported that gut microbiota mediated early life stress-induced social dysfunction and anxiety-like behaviors by impairing amino acid transport at the gut of maternally separated (MS) mice [62]. Tyrosine intestinal microbial metabolism of 4-ethylphenol sulfate into the mouse brain affected the activation and connectivity of specific brain regions, and disrupted the maturation and myelination patterns of oligodendrocytes in the brain, thus regulating brain activity and anxiety-like behavior in mice [63]. Glutamate could be converted to γ-aminobutyric acid (GABA) by the gut bacteria glutamate decarboxylase, and it has been observed to reduce symptoms of depression and anxiety [64].

Many metabolomics studies have revealed the altered metabolic were associated with ASD children in plasma and feces samples, but few studies in urine. Feces and urine are both excretions of the human body, which are important non-invasive examination methods to evaluate body health. Active substances produced by gut microbial communities through metabolic activities (such as short-chain fatty acids and amino acid metabolism) could be absorbed into the circulatory system through the intestinal tract and finally excreted through the urinary system [65]. Thus, we chose the untargeted metabolomics analysis to find that ASD children had unique urine metabolic characteristics compared to controls in this study.

In total, we discovered 10 metabolites were up-regulated and 56 metabolites were down-regulated mainly from lipid and amide, mainly involved in the Arginine and proline metabolism, with the highest statistical significance. Arginine biosynthesis, Pantothenate and CoA biosynthesis were shown higher pathway impact scores. Similar to the recent findings in ASD children, Arginine and proline metabolism was the significant pathway for discrimination both for plasma and urine samples [66, 67]. It has been confirmed that changes in the levels of amino acids could reflect the Autism Diagnostic Observation Schedule (ADOS) severity scores in children with autism [66]. Studies have shown that patients with abnormally elevated proline levels could develop seizures and intellectual disorders, which may be associated with a proline-induced reduction in glutamate release [68].

Specifically, we found the 9 metabolites (Androstenedione, Stearamide, Oleamide, Cadaverine, Hexadecanamide, Orotic acid, Linoleic acid, Palmitoleic acid, Lauric acid) positively correlated with microbiotas (Blautia obeum, Lachnoclostridium sp., Blautia sp., Roseburia inulinivorans, Bacteroidefragilis, Bacteroidethetaiotaomicron, Akkermansia sp., Erysipelotrichaceae UCG-003 sp., Bacteroidexylanisolvens, Collinsella stercoris, Tuzzerella sp.) and played promoted role in neurodevelopment dysfunction and up-regulated in ASD. The genus Blautia was abundant in ASD children, in agreement with a previous study on the gut microbiota by Liu et al. [69]. Our results showed higher abundance of fecal Lachnoclostridium sp. in ASD individuals, who presented poorer cognitive and GI symptoms. Lachnoclostridium sp. is a pro-inflammatory bacterium, which is related to tryptophan metabolism [70], consistent with our research, suggesting that Lachnoclostridium sp. might affect tryptophan metabolism through inflammatory pathways and participate in ASD.

Increased consumption of linoleic acid is associated with maternal intestinal dysbiosis and consequently starts the inflammatory process, harming myelinization [71]. Conjugated linoleic acid could significantly increase the abundance of Lachnoclostridium, Roseburia, etc., which are believed to be beneficial for health for their anti-inflammatory and anti-oxidative stress activities, SCFA production, polyunsaturated fatty acids transport, and other properties [72]. Gu X et al. reported that SCFAs produced by gut microbiota fermentation of dietary fiber regulate energy metabolism and played a crucial role in the microbiome-gut-brain axis [73].

Majewska MD et al. suggested that salivary levels of selected steroids (included androstenediol) might serve as biomarkers of autism pathology. Because they are neuroactive and modulate GABA, glutamate, and opioid neurotransmission, affecting brain development and functioning. These steroids may contribute to autism pathobiology and symptoms such as elevated anxiety, sleep disturbances, sensory deficits, and stereotypies among others [74]. Stearamide was found associated with lower birth weight [75]. Stearamide is one of the metabolism of Phthalates and polycyclic aromatic hydrocarbons. Prenatal exposure to PAEs was reported to contribute to the development of autism, attention deficit and hyperactivity disorder, mental retardation, and other developmental disorders [76].

Our data also showed an increase in the concentration of fatty acids in the autistic children, including Linoleic acid (LA), Palmitoleic acid (PA) and Lauric acid (LAA). In contrast to our study, Adams [77] reported lower fatty acids in children with ASD. This inconsistency might be due to the specific alteration of the metabolites in Guangxi populations and the intakes dietary change in the different studies. We also observed a higher level of Lachnoclostridium sp. and a positive correlation between Lachnoclostridium sp. and these fatty acids in ASD. Lachnoclostridium sp. is considered to be related with peripheral neuropathy in mice with a high-fat diet, which may have a potential correlation with cognitive level.

We observed increases in the abundance of taxa in the ASD children, including Akkermansia sp. Golubeva et al. [78] revealed that BTBR mice (p-Cresol exposure, idiopathic model for autism) displayed an increase in Akkermansia. Similarly, Altimiras et al. [79]. also identified an elevated abundance of Akkermansia genera in Fmr1 KO2 mice. We found that taxa were positively correlated with Hexadecanamide (palmitic amide, PA). In contrast to our study, Pan et al. [80] reported the levels of PA were negatively correlated with the abundance of Akkermansia in mice with PND (Perioperative neurocognitive disorders). Evidence by Pan et al. [80] has shown that PA significantly increased in aged mice than in young mice, but surgery further increased its levels. Moreover, they further observed that PA administration exacerbated surgically induced cognitive impairment and increased neuroinflammation, indicating that exacerbated neurocognitive disorders are explained partially by the underlying microbiota dysbiosis and its associated increase in PA levels.

We should acknowledge some limitations of our study. Firstly, the influence of environmental factors, diet, geography, and other factors on the microbiota and metabolomic results should also be considered. We cannot guarantee that our findings will apply to genetically diverse populations with different lifestyles and diets. Secondly, the functional role of identified microbiota and metabolites was not better defined as the observed results were not validated in animal models. We planed to validate our results in larger and geographically diverse cohorts, such as comparing ASD children from different nationalities (Zhuang nationality, Yao nationality, Han nationality, etc.) in larger Guangxi Zhuang Autonomous Region, or comparing ASD children between public databases (GEO database, etc.) and our cohort in the future study.

Conclusion

In conclusion, we performed a comprehensive study to explore ASD-related microbiota and metabolomics signature in Guangxi region. Discriminative microbiota and metabolites were identified between ASD subjects with healthy controls. 5 sensitive species Faecalibacterium prausnitzii, Bifidobacterium catenulatum, Blautia obeum, Lachnoclostridium sp., Blautia sp., and Bacteroides fragilis were observed significantly enriched in specifically in ASD children. Notably, bacterium showed a positive correlation with Androstenedione, Stearamide, Oleamide, Cadaverine, Hexadecanamide, Orotic acid, Linoleic acid, Palmitoleic acid, Lauric acid, suggesting a potential association with the Arginine and proline metabolism. Future investigations involving a larger cohort of ASD patients and focusing on the related microbiota and metabolomics signature are necessary to further validate these findings.

Abbreviations

ASD

Autism spectrum disorder

Hcy

Homocysteine

PAHs

Polycyclic aromatic hydrocarbons

ABC

Autism Behavior Scale

M-CHAT

Modified Checklist for Autism in Toddlers

CARS

Autism Rating Scale for Children

DSM-IV

The 4th edition of Diagnostic and Statistical Manual of Mental Disorders

GI

Gastrointestinal

LC-MS

Liquid Chromatography coupled with Mass Spectrometry

OTU

Operational Taxonomic Unit

QIIME

Quantitative Insights Into Microbial Ecology

PCA

Principal component analysis

NMDS

Non-metric multidimensional scaling

PCoA

Principal-coordinate analysis

PLS-DA

Partial least-squares discriminant analysis

LEfSe

Linear discriminant analysis effect size

LDA

Linear discriminant analysis

GBMs

Gut-brain modules

UPLC

Ultra-performance liquid chromatography

POS

Positive ionization methods

NEG

Negative ionization conditions

QC

Quality control

OPLS-DA

PCA and Orthogonal projections for latent structures-discriminant analysis

VIP

Variable Importance in the Projection

KEGG

Kyoto Encyclopedia of Genes and Genomes

SCFA

Short-chain fatty acids

LPS

Lipopolysaccharides

MAMPs

Microbiota-associated molecular patterns

MS

Maternally separated

GABA

γ-aminobutyric acid

ADOS

Autism Diagnostic Observation Schedule

LA

Linoleic acid

PA

Palmitoleic acid

LAA

Lauric acid

Authors’ contributions

X.L. conceived and supervised the project. Z.H., A.W., X.L. designed the research. Y.H. and S.H. collected the data. X.C., Y.H. and A.W. analyzed and interpreted the data. X.L., Z.H., A.W., and Y.H. wrote the manuscript and X.C., Y.H. and A.W. revised the paper with input from all the other authors. All authors read and approved the final manuscript.

Funding

This work was supported by the “Climbing” Program from the Youth Science and Technology Project of the First Affiliated Hospital of Guangxi Medical University (grant number YYZS2020019).

Data availability

Availability of data and materials16S rRNA sequencing raw data have been deposited to National Center for Biotechnology Information (NCBI) under the BioProject number PRJNA1277165. LC-MS analysis raw data have been deposited to Metabolights datebase (MTBLS12605). All data generated or analyzed during this study are included in this article. Further enquiries can be directed to the corresponding author.

Declarations

Ethics approval and consent to participate

The study was conducted in accordance with the Declaration of Helsinki. This study was approved by the Ethics Committee of the First Affiliated Hospital of Guangxi Medical University (NO.2021-KT-003). All parents of participants in this study have provided written informed consent.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.