Disruption of gut microbiome and metabolome in treatment-naïve children with attention deficit hyperactivity disorder

Abstract

Background

Attention deficit hyperactivity disorder (ADHD) is a neurodevelopmental disorder with an increasing prevalence in children. Recent studies have suggested that the gut microbiota may play a significant role in the development of ADHD. However, the specific relationship between changes in intestinal bacteria and related metabolites in children with ADHD remains poorly understood.

Results

In this study, we illustrated the fecal microbiome, metabolome and lipidome, as well as plasma metabolome using 16S rRNA gene sequencing and LC–MS in 15 pairs of children with ADHD and healthy controls. Our results revealed imbalance of gut microbiota and dysregulation of metabolites in individuals with ADHD. Specifically, children with ADHD exhibited significantly lower abundance of the Actinobacteria phylum, particularly Bifidobacterium, Corynebacterium and Actinomyces, while Veillonella in the Negativicutes class showed significant high level. No children with ADHD were classified under enterotype 1, which was composed solely of healthy children. Integration of multi-omics data suggested that the Bifidobacterium genus, which is positively correlated with various neurotransmitter precursor amino acid metabolites, may contribute to ADHD by downregulating pathways involving dopaminergic, serotonergic and glutamatergic systems.

Conclusions

These findings highlight the crucial regulatory impact of gut microbiota in the development of ADHD through metabolic pathways, and provide a potential avenue to the diagnosis and intervention of ADHD.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12866-025-04048-7.

Introduction

Attention deficit hyperactivity disorder (ADHD) is a neurodevelopmental condition characterized by persistent inattention, hyperactivity and impulsivity, ultimately resulting in varying degrees of functional impairment [1, 2]. Previous research has indicated that the prevalence of ADHD is as high as 5.3% in school-age children globally [3], with around 65% of ADHD-related impairments persisting into adulthood [4]. In China, the overall pooled prevalence of ADHD among children and adolescents is reported to be 6.3% [5]. However, the pathogenesis of ADHD remains unclear.

In addition to polygenic cause [6], the role of gut microbiota is increasingly recognized in influencing the risk and severity of ADHD. We thank the referee for these suggestions. An unbalanced gut microbiome contributes to the dysfunction of the intestinal barrier. Harmful bacteria can produce enzymes or toxins that thin the mucus layer and degrade the proteins that form the tight junctions, such as occludin and zonula occludens (ZO-1). This degradation further weakens the barrier, making it more permeable, which was commonly referred to as"leaky gut"[7]. The leakage of harmful substances, such as lipopolysaccharides (LPS), through the compromised barrier triggers local and systemic immune responses, which have been associated with autoimmune diseases [8], metabolic diseases [9], gastrointestinal disorders [10], allergies [11], and neurological disorders [12, 13]. Alpha diversity refers to the overall richness and evenness of species within a particular environment, whereas beta diversity measures the differences in species composition between various habitats or communities. Preliminary studies have observed a notable decrease in alpha-diversity in the gut microbiota of individuals with ADHD and a correlation between Bacteroides levels and hyperactivity and impulsivity [14]. Conversely, Wang et al. later reported an increase in alpha-diversity in patients with ADHD, and the relative abundance of Bacteroides coprocola was negatively correlated between ADHD symptoms [15]. However, three other studies did not find any significant differences in diversity between children with ADHD and those without [16–18]. Despite this, distinct microbial compositions have been noted in individuals with ADHD. For example, the Bifidobacterium genus was found to be elevated in late adolescence [19], while significantly increased levels of Bacteroidaceae and Neisseria families were observed during early puberty [14]. These conflicting findings regarding the association between ADHD and gut microbiota may be attributed to the inclusion criteria for patients, which may or may not have considered influencing factors such as antibiotics, probiotics, medication treatments, delivery modes, breastfeeding practices, and dietary patterns. Furthermore, reproducible patterns of variation in the microbial community are categorized into clusters known as'enterotypes,'which stratify human gut microbiomes. Several associations between human disorders and enterotypes have been reported [20]. However, the enterotype patterns in neurodevelopmental disorders remains unclear. This study also investigates the relationship between ADHD and enterotypes.

Although the gut microbiota in children is closely related to ADHD pathogenesis, the mechanism is still far from being fully understood. Recent evidence suggests that the microbiome regulates the gut-brain axis primarily through three pathways: the vagus or enteric nerves, the neuroendocrine system and the immune system. This regulation is mediated by neurotransmitters such as dopamine, serotonin, glutamate and gamma-aminobutyric acid (GABA), as well as short-chain fatty acids (SCFAs) like butyrate, propionic acid and acetate [21]. While metabolite alterations in the brain, reflecting biological processes, have been identified as key features and therapeutic targets of ADHD, their role in the regulatory function of the gut microbiota is still unclear. Additionally, a distinct metabolomic signature has been observed in ADHD twins [22]. However, there is currently no research that integrates the gut microbiome and metabolome to reveal the communication between gut microbes and the brain in children with ADHD. Therefore, in this study, we enrolled ADHD patients taking into account influencing factors such as medications, delivery modes, breastfeeding practices, and dietary patterns. We conducted 16S ribosomal RNA (rRNA) gene sequencing on fecal samples and liquid chromatography mass spectrometry (LC–MS) on fecal and plasma samples. For the first time, our study integrated these multi-omics findings to characterize the synergistic relationship between intestinal microbiota and various metabolites in children with ADHD.

Methods

Subject recruitment

This study was approved by the Ethics Committee of Shanghai Children’s Hospital. All participants signed informed consent. Fifteen children diagnosed ADHD were recruited from the Department of Child Health Care. Fecal samples from 15 pediatric patients newly diagnosed with ADHD and 15 healthy controls (HCs) were collected at the Department of Clinical Laboratory, Shanghai Children’s Hospital, Shanghai Jiao Tong University, China, between December 2020 and May 2021. The fresh fecal samples were immediately stored at −80 °C for DNA extraction and sequencing. Children with ADHD were diagnosed according to DSM-5 and ICD-11 [23]. The inclusion criteria were a) first-time diagnosis of ADHD without any neuro-psychotherapeutic medication treatment, including methylphenidate and atomoxetine, or non-pharmacologic treatments, such as behavioral therapy; b) children aged between 6 and 12 years old; c) duration of illness exceeding one year; d) non-vegetarians. The exclusion criteria included: a) incomplete basic clinical information; b) presence of other underlying diseases such as cancers, gastrointestinal diseases, infectious diseases, as well as inflammatory and immune diseases; c) recent use of medications including but not limited to antibiotics, probiotics, prebiotics, or anti-inflammatory drugs used within the past four weeks; d) recent surgical treatment within the past three months; and e) family history of neuropsychiatric diseases; f) lifetime history of head injury or surgery, as well as any neurological disorders, including but not limited to intellectual disabilities, concussion, autism, chronic tic disorder, bipolar disorder, psychosis, major depression or panic disorder. The recruited healthy individuals met the same inclusion and exclusion criteria.

16S rRNA gene sequencing

Following the manufacturer’s guidelines, genomic DNA was extracted from fecal samples using the QIAamp DNA Microbiome Kit (Qiagen, Germany). The extracted DNAs were then amplified to target the high-variant V3-V4 region of the 16S rRNA gene using primers 338 F (5′-ACTCCTACGGGAGGCAGCAG-3′) and 806R (5′-GGACTACHVGGGTWTCTAAT-3′). Each PCR reaction, with a total volume of 50 μL, included 5 μL of template genomic DNA, 2 μL of each forward/reverse primer (10 μM), 16 μL of ddH₂O, and 25 μL of 2 × Phanta Max Master Mix (Vazyme, China). The PCR reactions were initiated with denaturation at 95 °C for 3 min, followed by 35 cycles of denaturation at 95 °C for 15 s, annealing at 60 °C for 15 s, elongation at 72 °C for 30 s, and a final extension step at 72 °C for 5 min. Subsequently, the amplicons were barcoded, pooled and sequenced on an MGISEQ-2000 platform (MGI Tech, China) in PE300 mode.

Sequencing data analysis

Raw reads underwent preprocessing by cutadapt v2.6 to remove the adaptor contamination. Low-quality reads were filtered using a sliding window of 30 bp. Then, reads containing ambiguous bases and exhibiting low complexity were excluded from the analysis [24]. The resulting clean reads were used to create a consensus tag sequence through FLASH (Fast Length Adjustment of Short reads, v1.2.11) [25]. Tags sharing more than 97% sequence similarity were grouped into Operational Taxonomic Unit (OTU) using USEARCH v7.0.1090. Chimeric sequences were identified and removed with UCHIME v4.2.40 against the gold database (v20110519) for 16S rDNA sequences. OTUs were taxonomically classified using the RDP classifier v2.2 at a confidence level of 0.6. Prior to analyzing differential abundance, the read counts for each OTU were normalized by the total number of reads in each sample [26]. The rarefaction curve was utilized to visually represent the alternation of observed species or coverage along with the read counts increased. The data were subsampled to 60,000 read counts, with increments of 500 reads. Alpha diversity metrics, including Chao, Ace, Shannon, and Simpson indices, were calculated using Mothur v1.31.2 [27] with a sampling depth of 60,000 reads for rarefaction. Beta diversity was evaluated using QIIME v1.80 based on UniFrac distance [28]. The differences of beta-diversity between ADHD and HC in weighted UniFrac distance were assessed using permutational multivariate analysis of variance (PERMANOVA). Bacterial taxonomy was analyzed for differential abundance using the Mann–Whitney U test. The relative abundance of gut microbes at the genus level was clustered using Jensen-Shannon Distance and Partitioning Around Medoids. The optimal K value for clustering was determined using the Calinski-Harabasz index. The linear discriminant analysis (LDA) effect size (LEfSe) [29] was employed to identify bacterial communities that significantly contributed to the observed differences, which was conducted using Galaxy v1.0 with a cutoff of 2. Significance was assessed using the Kruskal–Wallis sum-rank test. PICRUST2 (Phylogenetic Investigation of Communities by Reconstruction of Unobserved States, v2.3.0) was utilized to predict the functional differences of microbial communities.

Liquid chromatography mass spectrometry

Fresh peripheral blood from all participants was collected into heparin-anticoagulant tubes, and immediately centrifuged at 2000 × g for 30 min. The supernatant was then stored at −80 ℃ before use. LC–MS was performed to identify the composition of metabolites from gut microbiota and peripheral blood plasma in an untargeted manner, following previously established methods [30]. Briefly, a mixture of ice-cold methanol and water (4:1, v/v) containing 0.224 mM phenylhydrazine hydrochloride was used to extract polar metabolites from 67 mg fecal samples or 50 µL of plasma. The homogenized samples were incubated at −20 °C for 30 min for derivatization of alpha-keto acids [31]. The resulting supernatant was subjected to a repeat extraction process. The combined extracts were then dried using a SpeedVac under H₂O mode. Following reconstitution in 2% acetonitrile in water, the extracts were analyzed using an Agilent 1290 II UPLC coupled to Sciex 5600 + quadrupole-TOF MS system. Separation of polar metabolites was achieved using a Waters ACQUITY HSS-T3 column (3.0 × 100 mm, 1.8 μm). The MS parameters included an ESI source voltage of − 4.5 kV in negative ion mode, vaporizer temperature of 500 °C, drying gas pressure of 50 psi, nebulizer gas pressure of 50 psi, curtain gas pressure of 35 psi, and a scan range of 60–800 m/z. The collision energy was maintained at (-) 35 ± 15 eV.

Lipids were extracted from stool samples using a mixture of acetonitrile in water (1:1, v/v), then derivatized with 3-nitrophenylhdyrazine and analyzed using Japer™ HPLC coupled to SCIEX Triple Quad 4500 MD system [32]. The lipids were separated on a Phenomenex Kinetex C18 column (100 × 2.1 mm, 2.6 μm) with 0.1% formic acid in water and acetonitrile as mobile phases A and B, respectively. Internal standards used were octanoic acid-1–13 C1 (Sigma-Aldrich) and butyric-2,2-d2 (CDN Isotopes).

Data processing for untargeted LC–MS

Analyst® TF 1.7.1 Software (AB Sciex, ON, Canada) was utilized for data acquisition and processing. The information of all detected ions including mass to charge ratio (m/z), retention time, peak areas and isotopic peaks was extracted using MarkerView 1.3 (AB Sciex, Canada). This data was then compared with the Metabolites database (AB Sciex, Canada), HMDB and standard references to annotate ion identities by PeakView 2.2 (AB Sciex, Canada) [33]. The quantification of metabolites was performed based on spike-in isotopically labeled internal standards (IS) such as _L-Tryptophan-D8, _L-lactate-13 C3, _L-Tyrosine-D7, and others. Peak areas of endogenous metabolites were normalized to their corresponding isotopically labeled structural analogues for quantitation. In cases where endogenous metabolites lacked labeled structural analogues, an automated algorithm, guided by the rule of minimal coefficients of variations (COVs), selected the optimal internal standard for quantitation [33]. Orthogonal projections to latent structures discriminant analysis (OPLS-DA) were conducted to identify differential metabolites using the R package ‘ropls’. Sparse partial least squares analysis (sPLS) [34] was employed to determine the correlation between gut microbes and fecal metabolites utilizing the R package ‘mixOmics'.

Statistical analysis of demographic data

The student's t-test was used to analyze continuous variables and the results were presented as means ± SD. The Chi-square test was employed to analyze categorical data. Benjamini–Hochberg adjustment was used for multiple testing correction. All statistical analyzes were conducted using IBM SPSS Statistics v26.

Results

Participant recruitment

In this study, 30 participants were recruited, consisting of 15 pediatric patients with ADHD and 15 healthy controls (HCs). There were no notable differences in sex, age, height, weight, and body mass index (BMI) between the two groups, while significant more children with ADHD were born via caesarean section compared to HCs (Table 1, Supplementary Table 1). However, there is no significant variations between the two groups in terms of breastfeeding practices after birth or the duration of breastfeeding beyond six months. Additionally, dietary patterns including consumption of vegetables, fruits, meat, grains, dairy products, fried foods, soft drinks and desserts did not show any significant differences between children with ADHD and healthy children (Table 2).

Table 1.

Characteristics of the recruited participants

Clinical indices	Recruited participants (n = 30)		P value
	ADHD (n = 15)	HC (n = 15)
Sex (male/female)	11/4	11/4	0.659
Delivery mode (vaginal birth/caesarean section)	6/9	9/6	0.007
Age (year, mean ± SD)	7.80 ± 1.21	7.47 ± 1.36	0.582
Height (m, mean ± SD)	1.31 ± 0.09	1.29 ± 0.09	0.870
Weight (kg, mean ± SD)	28.17 ± 7.53	28.87 ± 6.47	0.545
BMI (kg/m2, mean ± SD)	16.15 ± 2.83	17.24 ± 3.41	0.871

Table 2.

The dietary patterns of the recruited participants. (Chi-squared test)

		ADHD (n = 15)	HC (n = 15)	P-value
Feeding mode				0.583
Exclusive breastfeeding		66.7%	60.0%
Formula feeding		6.7	6.7%
Mixed milk feeding		26.7%	33.3%
Duration of breastfeeding				0.341
< six months		20.0%	26.7%
≥ six months		80.0%	73.3%
Dietary Pattern
Vegetables	≤ 1 day/week	0.0%	0.0%	0.193
	2–6 days/week	13.3%	20.0%
	daily	86.7%	80.0%
Fruits	≤ 1 day/week	0.0%	0.0%	0.329
	2–6 days/week	33.3%	40.0%
	daily	66.7%	60.0%
Meat	≤ 1 day/week	6.7%	13.3%	0.126
	2–6 days/week	13.3	20.0%
	daily	80%	73.3%
Grains and nuts	≤ 1 day/week	26.7%	33.3%	0.125
	2–6 days/week	60.0%	46.7%
	daily	13.3%	20.0%
Dairy products	≤ 1 day/week	20.0%	13.3%	0.372
	2–6 days/week	33.3%	33.3%
	daily	46.7%	53.3%
Fried foods	≤ 1 day/month	46.7%	53.3%	0.529
	2–3 days/month	26.7%	26.7%
	≥ 1 day/week	26.7%	20.0%
Soft drinks	≤ 1 day/month	46.7%	53.3%	0.085
	2–3 days/month	40.0%	26.7%
	≥ 1 day/week	13.3%	20%
Desserts	≤ 1 day/week	26.7%	20.0%	0.166
	1–2 days/week	33.3%	46.7%
	≥ 3 day/week	40.0%	33.3%

Diversity of gut microbiota in children with ADHD

To investigate the composition of gut microbiota in children with ADHD, a total of 2,625,491 raw reads were obtained from 16S rRNA gene sequencing of fecal samples. After filtering, the clean reads with high quality were clustered into 5075 OTUs (Supplementary Table 2). The species accumulation curves showed a plateau, suggesting that the sample size in this study was sufficient to characterize the overall community structure (Supplementary Fig. 1 A). The species compositions approached saturation and high-evenness, as depicted by the OTU rank curve (Supplementary Fig. 1B). In addition, the rarefaction curve indicated that the sequencing depth and coverage were adequate for the representing bacterial species diversity (Supplementary Fig. 1 C and 1D). Then, the bacteria were annotated at various taxonomic levels (Supplementary Fig. 2).

The diversity of the gut microbiota within individuals, referred to as α-diversity, is often linked to health status. The α-diversity could be estimated with the community richness using Ace and Chao index, or with the community diversity using Shannon and Simpson index. However, no significant differences were found in the four indices of α-diversity between children with ADHD and HCs (Fig. 1A, Supplementary Table 3). Subsequently, an analysis of β-diversity was conducted to compare the overall microbial communities between the samples. Weighted UniFrac distance was utilized to account for both phylogenetic relationship and species abundance. Samples from children with ADHD were observed to be more closely related to each other and distinct from those of the HCs, which is implied by the shorter distance between samples within ADHD group (Fig. 1B) (Mann–Whitney U test, p < 0.001). Furthermore, as shown by analyses including principal coordinate analysis (PCoA), non-metric multidimensional scaling analysis (NMDS), principal component analysis (PCA) and Partial least squares discrimination analysis (PLS-DA) (Fig. 1C, Supplementary Fig. 3), the gut microbiome was found to be significantly different between the two groups, which is determined by PERMANOVA (p = 0.0003).

Fig. 1.

Gut microbial diversity in children with ADHD and HCs. A The boxplot showing the α-diversity indices including the Ace index, the Chao index, the Shannon index, and the Simpson index. B The heterogeneity estimated with weighted UniFrac distance was significantly different between children with ADHD and HCs. **, p < 0.01. Each point represents the weighted UniFrac distance between each pair of samples. C The PCoA analysis showing the distinct composition of gut microbiota between children with ADHD and HCs

Altered gut microbial compositions in children with ADHD

The composition of gut microbes was compared between individuals with ADHD and HCs at various taxonomic levels. While Firmicutes, Bacteroidetes and Proteobacteria phyla remained stable, the Actinobacteria phylum exhibited a significant low abundance of less than one-tenth in children with ADHD (Fig. 2A, Supplementary Fig. 4, Supplementary Table 4). At the class level, in addition to the low level of Actinobacteria, Negativicutes and Betaproteobacteria showed more than a 20-fold and tenfold higher level in patients with ADHD, respectively (Fig. 2B). Similarly, at the order level, there was a lower relative abundance of Bifidobacteriales and Coriobacteriales from the Actinobacteria class, and a higher relative abundance of Selenomonadales from the Negativicutes class, and Burkholderiales from the Betaproteobacteria class (Fig. 2C). Further analysis revealed significant low abundance in Bifidobacteriaceae, Coriobacteriaceae, and Peptostreptococcacea, along with a marked high abundance in Veillonellaceae at the family level (Fig. 2D). Lastly, at the genus level, children with ADHD displayed notably lower levels of Bifidobacterium, Actinomyces, Romboutsia and Saccharibacteria genera, while Lachnospiracea incertae sedis and Veillonella genus were significantly more abundant compared to HCs (Fig. 2E, Supplementary Table 4).

Fig. 2.

Taxonomic composition of gut microbiome in ADHD and HC. Relative abundance of the gut microbial community in both groups at the levels of phylum (A), class (B), order (C), family (D) and genus (E). ***, p < 0.001. F The enterotypes of gut microbiota in children with ADHD and HCs. G Different relative abundance of the gut microbiota between AHs and HCs. LEfSe (linear discriminant analysis effect size) analysis based on LDA (linear discriminant analysis) effect size demonstrated differences in taxonomic composition of ADHD children compared to HCs showed by cladogram (G) and histogram (H)

Based on the relative abundance of gut microbiota at the genus level, children were categorized into three enterotypes (Fig. 2F). Interestingly, no individuals with ADHD were designated to enterotype 1 featured by Bacteroides, which consisted of HCs only. Both groups had individuals in enterotype 2 driven by Prevotella. Among the 15 individuals with ADHD, three were classified into enterotype 3, characterized by prominent Ruminococcus [35]. However, host properties such as sex, age or BMI were not found to be significantly correlated with the enterotypes, which is consistent with previous reports [35]. In addition, there is no significant difference between individuals with ADHD in Enterotype 2 and Enterotype 3, both in terms of physiological features and neurological parameters (Supplementary Table 5). Furthermore, using LEfSe, key discriminative microbial features in children with ADHD were identified (Supplementary Table 6). With a cutoff of LDA > 2.5, the most distinguishing features of the ADHD gut microbiota were the downregulated Actinobacteria-Bifidobacterium lineage as well as the upregulated Negativicutes-Veillonellaceae lineage (Fig. 2G and H). These findings suggest that children with ADHD exhibit distinct microbial signatures.

The gut microbial lipidome in children with ADHD remains unchanged

The regulation of neural function by gut microbes primarily occurs through their products, with SCFAs playing a crucial role in mediating the microbiota-gut-brain axis (mGBA) [36]. In our study, we utilized HPLC–MS to analyze the lipidome in fecal samples obtained from children with ADHD and HCs (Supplementary Table 7). The individuals from both groups did not exhibit clear separation based on PCA analysis (Supplementary Fig. 5 A). The abundant patterns of the lipids, including nine SCFAs and five medium-chain fatty acids (MCFAs), were found to be similar between the two groups (Supplementary Fig. 5B). Furthermore, no significant differences were observed in the abundance of any lipids between the ADHD and HC groups (Supplementary Fig. 5 C and 5D, Supplementary Table 8).

Fig. 5.

The associations between gut microbiota and metabolites. A The heatmap of the correlation between gut microbiome and metabolome at the OTU level. B The concentric circle plot indicating the correlation between gut microbiome and metabolome at the OTU level. The distance between points reflects the correlation. The acute angle between two points represents positive correlation, while the obtuse angle represents negative correlation. C The co-occurrence network between gut microbiota and metabolites. The red dash lines indicate positive correlation, while the grey dash lines indicate negative correlation

Disturbed gut microbial metabolome in the feces of children with ADHD

To assess the impact of gut microbiota on the metabolome, we conducted untargeted UPLC-MS analysis on fecal samples obtained from children with ADHD and HCs (Supplementary Fig. 6 A, Supplementary Table 9). The intestinal metabolite profiles of in children with ADHD were strikingly distinguished from those of the healthy controls (Fig. 3A, Supplementary Fig. 6B). Out of the 219 metabolites detected, 56 showed significant differential abundance in children with ADHD (Student’s t-test, BH-adjusted p < 0.05) (Fig. 3B, Supplementary Table 10). Among them, five metabolites were up-regulated by more than twofold, while 50 metabolites were down-regulated. In comparison to HCs, amino acid metabolites (_L-tryptophan, _L-phenylalanine, _L-serine, threonine _L-valine, _L-tyrosine, _L-histidine, tyrosyl-glutamate) were significantly less in fecal samples from children with ADHD (Fig. 3C).

Fig. 3.

The composition of fecal metabolome in children with ADHDs and HCs. A The PCA of fecal metabolome from children with ADHD and HCs. B Volcano plot showing fecal metabolites with different abundance between children with ADHDs and HCs. C The boxplot showing the abundance of representative fecal metabolites down-regulated in children with ADHD. D The OPLS-DA indicating distinct fecal metabolomic pattern between children with ADHDs and HCs. *, p < 0.05; **, p < 0.01; ***, p < 0.001

Next, OPLS-DA were utilized to profile the metabolic differences between individuals from both groups. The analysis identified samples of ADHD22 and HC320 as potential outliers (Supplementary Fig. 6 C). The OPLS-DA score plot clearly showed a distinct separation between samples from the ADHD and HC groups along the x-axis, indicating inter-group differences (Fig. 3D). On the y-axis, the dispersed HC samples implied relatively high intra-group variability compared to the more clustered ADHD samples. Following validation through 200 permutations, the original R2Y and Q2Y values were notably higher than those obtained in the permutation test, suggesting absence of overfitting (Supplementary Fig. 6D). Metabolites with VIP values exceeding one were selected for intersection with the differential metabolites for subsequent pathway enrichment analysis (Supplementary Table 11). However, these metabolites did not show significant enrichment in specific functional pathways (Supplementary Fig. 6E-F, Supplementary Table 12).

Changes of the metabolome in the plasma of children with ADHD

Metabolites from the gut can enter the brain through the blood–brain barrier, serving as precursors of neurotransmitters or other neural structures, thereby influencing the brain-gut axis. To ascertain the corresponding plasma metabolome changes in response to altered gut metabolites, we conducted further analysis on the plasma metabolome of children with ADHD and HCs using non-targeted UPLC-MS (Supplementary Fig. 7 A, Supplementary Table 13). A distinct metabolite profile was also found in the plasma of children with ADHD (Fig. 4A). Among the 508 detected metabolites, 95 showed significant differential abundance in children with ADHD (Student’s t-test, BH-adjusted p < 0.05) (Fig. 4B, Supplementary Table 14). Specifically, the amino acid metabolites such as _L-tyrosine, _L-threonine, _L-phenylalanine, _L-serine, _L-tryptophan and tyrosyl-glutamate, which were significantly low abundant in fecal samples, were also found to be at a low level in the plasma of children with ADHD (Fig. 4C). A total of 32 metabolites were commonly downregulated in both sample types, which accounted for 39% and 24% of the significantly low abundant metabolites in feces and plasma, respectively (Fig. 4D).

Fig. 4.

The composition of plasma metabolome in children with ADHDs and HCs. A The PCA of plasma metabolome from children with ADHD and HCs. B Volcano plot showing plasma metabolites with different abundance between children with ADHDs and HCs. C The boxplot showing the abundance of representative plasma metabolites down-regulated in children with ADHD. D Venn diagram showing commonly up- (left) and down- (right) regulated metabolites in feces and plasma. E The OPLS-DA indicating distinct plasma metabolomic pattern between children with ADHDs and HCs. *, p < 0.05; **, p < 0.01; ***, p < 0.001

Moreover, the samples from the ADHD and HC groups were clearly separated in the OPLS-DA score plot along the x-axis. Similar to the gut metabolome pattern, the HC samples were dispersed on the y-axis compared to the more condensed ADHD samples in the plasma (Fig. 4E, Supplementary Fig. 7B). Metabolites with VIP values exceeding one were selected (Supplementary Table 15). Plasma metabolites with differential abundances significantly enriched in pathway of aspartate metabolism (Supplementary Fig. 7 C-D, Supplementary Table 16).

Association between intestinal bacteria and metabolites

Next, we explored the potential relationship between gut microbes and fecal metabolites. Utilizing sPLS, we selected the key variants. We found that Actinobacteria exhibited positive correlation with various amino acids, such as _L-lactic acid, _L-isoleucine, 5-hydroxy-_L-tryptophan, _L-tryptophan, and _L-phenylalanine at the phylum level. Moving to lower taxonomic levels, the orders of Bifidobacteriales, Actinomycetales and Coriobacteriales within the Actinobacteria class showed positive correlations with these amino acids, which were in turn negatively correlated with the Selenomonadales order in the Negativicutes Class and the Burkholderiales order in the Betaproteobacteria Class. At the genus level, a striking positive correlation was observed between Bifidobacterium and various metabolites. Specifically, _L-tryptophan and 5-hydroxy-_L-tryptophan were positively correlated with Bifidobacterium longum and Actinomyces odontolyticus, while a negative correlation was found with Lachnospiraceae (Fig. 5A, Supplementary Fig. 8). The correlation circle plot highlighted strong associations primarily within genera belonging to Actinomycetes. Notably, while there was a general positive correlation between Actinobacteria and metabolites at the phylum level, at the genus level, Bifidobacterium, Actinomyces, and Collinsella were positively linked with differential metabolites, whereas Atopobium and Corynebacterium exhibited a negative correlation. Furthermore, tyrosyl-glutamate, 1H − Indole − 3 − carboxaldehyde and _L-tyrosine showed strong positive correlation with Collinsella aerofaciens (Fig. 5B, Supplementary Fig. 9).

Our study observed a complex co-occurrence relationship between intestinal bacteria and fecal metabolites in pediatric patients with ADHD (Fig. 5C, Supplementary Table 17). Within this network, microbes with differential abundance were grouped into three clusters. Cluster 1, comprised of Actinobacteria phylum, Actinobacteria class, Actinomycetales order, Bifidobacteriales order, Coriobacteriales order, Bifidobacteriaceae family, Coriobacteriaceae family, Atopobium genus, and Bifidobacterium genus, showed positive correlations with amino acid metabolism, aromatic metabolism and nucleotide metabolism. Cluster 2, composed of Negativicutes class, Selenomonadales order, Veillonellaceae family, and Dialister and Veillonella genus, displayed negative corrections with various metabolites including amino acid metabolism, phenolic metabolism, aromatic metabolism, and organic acid metabolism. Cluster 3, including Verrucomivrobiae class, Verrucomicrobiales order, Verrucomicrobiaceae family, and Akkermansia genus, exhibited a positive correlation with glycochenodeoxycholic acid. These results suggest that the unique gut microbiome composition in ADHD patients interacts with multiple metabolic pathways, potentially influencing the progression of the disorder.

Discussion

The gut microbiota communicates bidirectionally with the brain through the brain-gut axis by producing neurotransmitter precursors and metabolites, stimulating the vagus nerve, and modulating the systemic immune response. Various factors can substantially affect the composition of gut microbiota, such as diet, infections and medication. Consuming a high-fat or high-sugar diet is associated with an increase in Firmicutes and a decrease in Akkermansia, leading to inflammation triggered by cytokines and lipopolysaccharides. However, this negative effect can be mitigated by adopting a high-fiber diet, which promotes the growth of Prevotella and Bifidobacterium, thereby enhancing the metabolism of the SCFA and butyrate [37]. A recent study has also found that breastfeeding, antibiotic exposure, and probiotic intake can influence neurodevelopment in children with ADHD by changing the composition of intestinal microbiota [38]. Therefore, distinct from most of previous studies on the gut microbiota in ADHD, we gathered the meta data on common influencing factors and considered the history of diseases and medications as exclusion criteria. Interestingly, we did not observe significant differences in dietary habits or breastfeeding practices between the two groups. However, a higher percentage of children with ADHD were delivered via caesarean section. Research on the gut microbiota composition during the neonatal period has shown that the majority of early colonizing microbes in vaginally delivered infants come from the mother, including enriched commensal genera like Bifidobacterium [39]. In contrast, newborns delivered by caesarean section tend to have a gut microbiota dominated by bacteria linked to the hospital environment, whose mother-newborn transmission of Bifidobacterium was disrupted [40]. These findings suggest that the altered gut microbiome associated with ADHD observed in our study may be established very early in life and remain unaffected by normal dietary patterns. This underscores the importance of interventions aimed at promoting the colonization of beneficial bacteria during this critical developmental window.

Previous research on the gut microbiome of children with ADHD has yielded conflicting results. Wang et al. reported an increase in α-diversity, while Prehn et al. found a decrease in their study. In line with the majority of past studies, our research observed no significant changes in intestinal α-diversity in children with ADHD, but did observe less heterogenous compared to their healthy counterparts. The shifts in gut microbiota composition identified in this study are noteworthy, marked by a widespread reduction in the abundance of various bacteria within the Actinobacteria phylum, including the Bifidobacteriales and Coriobacteriales classes, as well as the Actinomyces genus, with a corresponding high level in the Veillonella genus. The study conducted by Wan et al. also noticed a change in the abundance of Veillonella, observing a decrease in individuals with ADHD [17]. In addition, they reported a reduction in Lachnospiracea, which is elevated in our current participants. These differing results could be attributed, at least in part, to the different delivery mode of the children, as all participating children in their study were born via vaginal delivery. Nevertheless, consistent with our findings, a notable increase in Veillonella [41] and Lachnospiracea [42] was observed in previous studies. Furthermore, our studies identified low abundance in the Romboutsia genus. Interestingly, a two-sample Mendelian randomization study showed that Romboutsia may have a protective effect against ADHD [43], while increased levels of Lachnospiracea incertae sedis have been implicated in autism [44]. Despite numerous studies demonstrating changes in fecal bacteria composition in ADHD patients across different ethnicities, dietary habits and delivery methods, it is still challenging to interpret the significance of gut microbiota without accompanying metabolome data.

Metabolites, as mediators in the interactions between gut microbiota and the host, play a crucial role in maintaining the normal physiological functions of the host. Typically, metabolites produced by gut microbes exert their effects in the way of neuronal signaling molecules and/or related precursors [45]. Therefore, an integrated analysis of the gut microbiome and metabolome can precisely reveal the functional microbiota changes specific for the disorders. Interestingly, while adult ADHD patients showed lower plasma levels of SCFAs compared to healthy individuals [46], there was no significant change of them was detected in fecal samples of children with ADHD. Surprisingly, we found an intensive disruption in the metabolome of children with ADHD. Gut microbiota plays pivotal role in modulating the amino acids homeostasis [47]. Our result showed that many of the changing metabolites are amino acids known to directly modulate neuronal function. ADHD is a disorder characterized by a dysregulated dopaminergic system. Phenylalanine can be converted to tyrosine, which serves as the precursor of dopamine. Dietary depletion of tryptophan, tyrosine, and phenylalanine in mice leads to reduced serum levels of these amino acids, along with decreases in both phenylalanine and tyrosine levels in the hippocampus, as well as diminished concentrations of serotonin and dopamine in a brain region-specific manner [48]. The deficiency of these monoamines correlates with hyperactive behavior in mice. In a double-blind clinical trial, the supplementation of phenylalanine to 19 patients with ADHD for two weeks led to a substantial improvement in mood before the development of subsequent drug tolerance [49]. Tryptophan acts as the precursor of the essential neurotransmitter serotonin, which plays a role in mood and cognition [50]. Imbalances in tryptophan metabolism not only affects the production of 5-hydroxytryptamine (5-HT) but also the accumulation of the downstream metabolites, interfering with neurofunctions [51]. The intestinal tract serves as a site to perform the conversion of tryptophan to 5-HT. Recent evidence suggested that gut-derived 5-HT is a key mediator of anxiety regulation [52]. Furthermore, lower levels of serum tryptophan have been reported in adult patients [53], dogs [54] and rats [55] that exhibit ADHD-like behaviors. Tyrosyl-Glutamate, a dipeptide composed of tyrosine and glutamate, also plays a role in neurotransmission. Tyrosine serves as a precursor of dopamine, while glutamate is a crucial neurotransmitter for excitatory impulses. Knock-out mice lacking various glutamate receptors have shown impaired spatial working memory and abnormal emotional and social behavior [56]. The involvement of glutamate in the etiology of ADHD is thought to revolve around the interaction between the dopaminergic and glutaminergic systems. In addition to the extensively studied ADHD-related metabolites, this study also found that serine and valine were slightly decreased in individuals with ADHD. The valine is classified as branched-chain amino acids (BCAAs), which are required for synthesis of succinyl-CoA and/or acetyl-CoA. Metabolic abnormalities in these amino acids can lead to defects in various organs, including neurodevelopmental disorders like ADHD. However, the detailed mechanism through which BCAAs induce ADHD remains to be fully understood [57]. Moreover, serine is implicated in neural diseases [58], but its involvement in ADHD requires further investigation. Interestingly, while the fecal metabolome of individuals with ADHD differed significantly from that of healthy controls, these altered metabolites were not enriched in any specific function or pathway. This could be due to the fact that, as previously mentioned, multiple rather than single metabolic pathways are affected in children with ADHD, reflecting the complexity of ADHD symptoms.

The conjoint analysis of the microbiome and metabolome implied the underlying regulatory relationships in ADHD. Main metabolites changed in ADHD, such as amino acids and aromatic metabolites, showed positive associations with the Actinobacteria phylum and negative associations with the Negativicutes class. Within the Actinobacteria phylum, Bifidobacterium has been utilized as potential therapeutic treatment of major depressive disorder by targeting serotonergic substances like tryptophan, 5-HTP, and 5-HT [30]. Animal studies have shown that supplementation with Bifidobacterium longum can reduce anxiety and depression-like behaviors [59]. Early microbial colonization, particularly with Bifidobacterium, has been found to influence the hypothalamic–pituitary–adrenal (HPA) axis response to stress [60]. Additionally, transplantation of human infant microbiota rich in Bifidobacterium has been linked to enhanced microglial reactivity and ramification, which aids in the synapse and neural circuit maturation [61]. Corynebacterium glutamicum has been identified as a producer of _L-isoleucine in ADHD [62] and 3,4-Dihydroxyphenyl-_L-alanine (_L-DOPA) in Parkinson’s disease [63]. Actinomyces has been shown to provide the major source of diverse secondary metabolites with antibiotic and antitumor properties [64]. Therefore, the regulation of brain function by Actinomyces through these metabolites may involve an indirect way of modulating host immunity or gut microecology. Veillonella, belonging to the Negativicutes class, was found to exhibit a strong negative correlation with the major pathways altered in ADHD in our study. Previous research has shown that Veillonella utilizes lactate as its sole carbon source to produce SCFAs [65]. However, this relationship was not directly observed in our current data. Thus, more research is needed to reveal their role in the regulation of metabolites in neurodevelopmental disorders. It is interesting that the gut microbiota of three children with ADHD belongs to enterotype 3, which is dominated by Ruminococcus. Additionally, the bile acid Chenodeoxycholic acid 3-glucuronide was five-fold higher in the plasma of children with ADHD. Given that Ruminococcus is capable to metabolize primary bile acids from the liver into secondary bile acids [66, 67], this observation might suggest a potential role for the microbiome-gut-liver axis in modulating ADHD symptoms in a subset of patients, presenting an interesting angle for further exploration.

Some limitations of this study should be acknowledged. First, although strict patient enrollment criteria were established, posing challenges in obtaining suitable samples, this recruitment strategy resulted in a relatively small cohort size (15 ADHD and 15 HCs), thereby limiting the statistical power and generalizability of the findings. Future validation through larger cohorts targeting ADHD-related bacteria and metabolites will enhance the significance of our current findings. An integrative model based on metagenomes and metabolomics of gut microbiota may aid in the diagnosis of ADHD. Second, the etiology of ADHD is influenced by an interplay of genetic causes and environmental factors including prenatal and perinatal insults, dietary components and psychosocial adversity [68]. Thus, the confounding factors considered and analyzed in this study remain limited. Third, while a comprehensive analysis of the altered gut microbiome, metabolome and lipidome, as well as plasma metabolome in pediatric ADHD patients has been conducted, further longitudinal research or functional validation is necessary to confirm causal relationships and elucidate molecular mechanisms, possibly through animal models or in vitro experiments.

Conclusions

In conclusion, we conducted a pioneering exploration into the multi-omics alterations observed in children with ADHD. Specifically, we found a notable low abundance in bacteria belonging to the Actinobacteria phylum, such as Bifidobacteriales, Coriobacteriales, and Actinomycetales order. Moreover, our results highlighted the significant role of amino acid metabolism disturbances in the pathogenesis of ADHD. The strong correlation between the altered intestinal bacteria and the metabolite dysregulation underscored the impact of disrupted homeostasis on ADHD development. These findings offer valuable insights into the pathogenesis of ADHD and have the potential to inform the development of innovative diagnostic tools and treatments for children with ADHD.

Authors’ contributions

D.H., H.Z. and J.C. conceived the project and designed the study and Y.Z. and R.W. recruited participants and W.L., J.H. and F.P. collected samples and F.P. and X.P. collected the data and D.H., Y.Z. and W.L. wrote the manuscript and analyzed the data and W.W., Y.W. and Z. M. revised the manuscript and J.C. and H.Z. supervised this investigation. All authors contributed to the article and approved the submitted version.

Funding

Natural Science Foundation of Shanghai,21ZR1452900,19ZR1477700,National Natural Science Foundation of China,82471882,Shanghai Municipal Key Clinical Specialty,shslczdzk06902,Three-Year Initiative Plan for Strengthening Public Health System Construction in Shanghai (2023-2025),GWVI-3,National Key Research and Development Program of China,2022YFC2705203.

Data availability

Sequence data that support the findings of this study have been deposited in the NCBI Sequence Read Archive database with the primary accession code “PRJNA1003257”.

Declarations

Ethics approval and consent to participate

This study was approved by the Shanghai Children’s Hospital Ethics Review Committee (No. 2021R040-E01) and was conducted in accordance with the Helsinki Declaration. Informed consent to participate was obtained from the parents or legal guardians of any participant under the age of 16.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.