Alterations in gut microbiota and urine metabolomics in infants with yin-deficiency constitution aged 0–2 years

Abstract

Background

Based on the constitution theroy, infants are classified into balanced constitution (BC) and unbalanced constitution. Yin-deficiency constitution (YINDC) is a common type of unbalanced constitutions in Chinese infants. An infant's gut microbiota directly affects the child's health and has long-term effects on the maturation of the immune and endocrine systems throughout life. However, the gut microbiota of infants with YINDC remains unknown. Herein, we aimed to evaluate the intestinal flora profiles and urinary metabolites in infant with YINDC, find biomarkers to identify YINDC, and promote our understanding of infant constitution classification.

Methods

Constitutional Medicine Questionnaires were used to assess the infants’ constitution types. 47 infants with 21 cases of YINDC and 26 cases of BC were included, and a cross-sectional sampling of stool and urine was conducted. Fecal microbiota was characterized using 16S rRNA sequencing, and urinary metabolomics was profiled using UPLC-Q-TOF/MS method. YINDC markers with high accuracy were identified using receiver operating characteristic (ROC) analysis.

Results

The diversity and composition of intestinal flora and urinary metabolites differed significantly between the YINDC and BC groups. A total of 13 obviously different genera and 55 altered metabolites were identified. Stool microbiome shifts were associated with urine metabolite changes. A combined marker comprising two genera may have a high potential to identify YINDC with an AUC of 0.845.

Conclusions

Infants with YINDC had a unique gut microbiota and metabolomic profile resulting in a constitutional microclassification. The altered gut microbiome in YINDC may account for the higher risk of cardiovascular diseases. Metabolomic analysis of urine showed that metabolic pathways, including histidine metabolism, proximal tubule bicarbonate reclamation, arginine biosynthesis, and steroid hormone biosynthesis, were altered in infants with YINDC. Additionally, the combined bacterial biomarker had the ability to identify YINDC. Identifying YINDC in infancy and intervening at an early stage is crucial for preventing cardiovascular diseases.

1. Introduction

The TCM constitution (TCMC) is a comparatively consistent internal feature formed on the basis of inborn endowments and later characters, combined with one's lifetime morphological structure, physiological function and psychological state [1]. TCMC divides infants into nine constitutions: balanced constitution (BC), which indicates a comprehensive state of physical, mental and social well-being and less susceptibility toward illness, and eight unbalanced constitutions with their own unique biological characteristics and predisposition to different diseases [2,3]. For example, constipation is one of the most common diseases with a reported frequency of up to 30% [4], and 96% of cases are due to functional causes [5]. However, current diagnostic methods for functional constipation often fail to achieve cooperation among children or are harmful to their health. Furthermore, the evidence supporting the main therapeutic approaches is insufficient [6]. According to TCMC, constipation is a sign of yin-deficiency constitution (YINDC), which is a very common constitution in infants that is influenced by both congenital and acquired aspects. From the congenital aspect, the yin and yang bias of the parents' constitutions will affect the constitution of the offspring. Additionally, an improper diet during the mother's nurturing process will also cause YINDC in the offspring. Acquired aspects such as excessive consumption of fried foods, uncontrolled sexual life, over-intense emotions, specific natural and social environments, disease effects, and medical negligence can lead to YINDC [7]. YINDC is characterized by a deficiency of fluid (such as water) and the consequent presence of internal heat. Lack of fluid in the body will manifest as dry skin, lips, and eyes and constipation; internal heat presents as hot palms and soles, redness of the face and a preference for cold water. When the stomach yin is insufficient to fumigate the tongue, there is an uneven distribution of the tongue coating. At night, the deficient yin qi cannot restrain the yang qi, which forces sweating from the skin [8]. These YINDC manifestations can be treated with psychological adjustments, physical exercise, food therapy, herbal medicine, acupuncture, moxibustion, or massage [9].

Constitution can be identified by screening significantly different microbiota and metabolites and subsequently finding specific biomarkers. Increasing evidence shows that different constitutions have their own gut microbiota and metabolic characteristics [[10], [11], [12], [13]]; nevertheless, relevant researches have been performed mainly on adults. Recently, the gut microbiota of infants has drawn considerable attention because it directly affects the health of children at different developmental stages and has a long-term effects on the maturation of the immune and endocrine systems [[14], [15], [16]]. In particular, the first 2 years of life are a pivotal period of early childhood growing and developing when numerous microbiota gather and become stabilized, and the microbial community of the gastrointestinal tract serves as a key player in multiple pathways [17]. Accordingly, we focused on evaluating the correlation between 0 and 2 years old YINDC infants and intestinal microbiota in this study. Moreover, the metabolites generated by intestinal flora-host interactions are closely related to the phenotype and can directly reflect what happened in the body. Drawing blood samples from infants is inconvenient, easily infected, and does not indicate the total condition of the body [18], and stool samples are not as abundant as urine samples. Thus, we selected urine samples for metabolite detection. In this study, high-throughput 16S rRNA sequencing and untargeted metabolomics were used to gain insight into differential gut flora and metabolite characteristics in infants with YINDC aged 0–2 years old. This could help identify biomarkers and treat the corresponding diseases in infants with YINDC.

2. Materials and methods

2.1. Participants

0–2 years old infants were recruited via posters and social-media posts from Beijing and Guangdong, China. Constitutional Medicine Questionnaires were completed by the infants’ guardians, and only infants who met the criteria for conversion scores were screened for the next set of inclusion and exclusion criteria. The questionnaire comprised 43 items with 9 subscales, each measuring one constitution, namely, BC, items 1–3 and 43; qi–deficiency constitution, items 4–9; yang–deficiency constitution, items 10–13; YINDC, items 14–20; phlegm–dampness constitution, items 21–24; damp–heat constitution, items 25–29; blood stasis constitution, items 30–33; qi stagnation constitution, items 34–37; and inherited special constitution, items 38–42 (Supplemental file 1). Then, the conversion scores of the different constitutions corresponding to these items were calculated [19]. Individuals with BC score ≥60 and other conversions score <30 were considered to be BC. Individuals with YINDC were the ones with the highest YINDC score of ≥40 among all unbalanced constitutions.

YINDC and BC individuals were enrolled in fecal flora and urine metabolomics studies. The infants' guardians, generally their parents, were thoroughly informed of the trial procedure and signed informed consent forms. Ethical approval was acquired from the Ethics Committee of Beijing University of Chinese Medicine (No. 2020BZYLL122).

2.2. Inclusion and exclusion criteria

Participants were required to meet the following criteria: (1) infants aged 0–2 years old; (2) parents who had cared for the infants for a long time and had a detailed knowledge of the infant's basic condition; (3) infants with YINDC or BC diagnosed using the Constitutional Medicine Questionnaires.

Participants meeting one or more of the following criteria were excluded: (1) infants who received antibiotics and probiotics within 3 months prior to sample collection; (2) infants with a mental illness or severe disease that were unable to participate in and complete clinical studies in a standardized manner; (3) infants living far from the hospital where stool and urine samples were not readily available; (4) infants with incomplete clinical data.

2.3. Instrumentations

The following instrumentations were used: Qubit@ 2.0 Fluorometer (Thermo Scientific), Agilent Bioanalyzer 2100 system (Agilent), Bio-rad T100 thermal cycler (Bio-rad), NovaSeq6000 (Illumin), 1290 Infinity LC ultra-high pressure liquid chromatograph (Agilent Technologies), Triple TOF 6600+ Mass Spectrometer (AB SCIEX), Centrifuges (Eppendorf 5430R), 2.1 mm × 100 mm ACQUIY UPLC BEH 1.7 μm column (Waters, Ireland), Vortex (QiTe QT-1, Shanghai), MP Fastprep-24 Automated Homogenizer (MP Biomedicals), Ultrasonic Liquid Processors (Scientz JY92-II，Ningbo), Concentrator plus/Vacufuge (Eppendorf Concentrator Plus), −80 °C ultra-low temperature refrigerator (Thermo Scientific), Electronic balance (METLER TOLED AL104).

2.4. Chemicals

The following chemicals were used: Magnetic Soil And Stool DNA Kit (TIANGEN), Phusion®High-Fidelity PCR Master Mix (New England Biolabs), AxyPrepDNA Gel Extraction Kit (AXYGEN), TruSeq® DNA PCR-Free Sample Preparation Kit (TruSeq), Acetonirile (CAN, I592230123, Merck), NH3·H2O (221228, Sigma-Aldrich), Formic Acid (FA, 06450, Fluka), Ammonium acetate (70221, Sigma-Aldrich), MeOH (A456-4, Fisher Chemical).

2.5. Stool sample collection and sequencing

Freshly produced stool samples that did not contain urine were collected at home; >3 g (one-third the volume of the stool tube or two scoops) from the middle of the stool was scooped into a sterile stool cup. Then, the stool samples were transported with ice packs to a hospital within 1 h and stored at −80 °C after rapid freezing. Bacterial DNA was obtained by CTAB/SDS, and primers (341 F-806 R) were designed to amplify the 16S rRNA V3–V4 genes. Amplicons were mixed and purified, and amplicon libraries were generated. Lastly, the sequencing of the library yielded 250/300 bp paired-end reads [20].

2.6. Gut microbiota analysis

Paired-end reads were assembled and controlled for quality on 16S rRNA sequencing data. The sequences were performed with UPARSE. Sequences with ≥97% similarity were attributed to the same OTUs. Alpha diversity was used to evaluate the diversity of the microbial community within the sample and included the Chao 1, ACE, and Shannon. Beta diversity was applied to investigate the variation of biological groups between infants with YINDC and those with BC in species complexity. The linear discriminant analysis (LDA) of effect size (LEfSe) was able to discover taxa affecting the sample division. PICRUSt (version 1.1.1;) was used to predict the gene composition to analyze the functional differences in the KEGG database [21].

2.7. Urine sample collection and profiling

Middle part of urine were gathered in the morning in a sterile urine cup (>1.5 mL). Thereafter, the urine samples were transported to the hospital in ice bags in 1 h and centrifuged for 30 min (2000×g, 4 °C). The supernatant is then quickly frozen and stored at −80 °C. For analysis, the supernatant is thawed and mixed with pre-cooled methanol/acetonitrile solvent at a ratio of 1:4. The mixture was mixed by vortexing, sonicated at low temperature, held at −20 °C for 60 min, and centrifuged for 20 min (14000×g, 4 °C). After collecting, drying, redissolving the supernatant in acetonitrile/water solvent, it was again mediated by mixing for 1 min and centrifuged for 15 min. Finally, the supernatant was used for chromatographic analysis and subsequent Q-TOF mass analysis [22].

2.8. Urine metabolism analysis

ProteoWizard was used to convert the original data into mzXML format [23], and then XCMS was used for peak alignment, retention time correction and peak area extraction [24]. Metabolite structure identification was performed in a self-constructed laboratory database. 1 set of missing ion peak data >50% was removed. Data were preprocessed by Pareto scaling and subjected to multidimensional statistical analysis, represented as PCoA plots, and univariate analysis using Student 's t-test and fold change analysis, represented as volcano plots. The MetaboAnalyst was used for metabolic pathway analysis. The significance of the enrichment pathways was evaluated using Fisher's exact test.

2.9. Correlation analysis, ROC analysis, and statistical analysis

The correlation analysis was used to find the correlation for the differential flora and metabolites by Spearman. Heatmap and Network diagram analyses were performed using R and Cytoscape. The markers was measured by the area under the receiver operating characteristic (ROC) curve (AUC). For statistical analysis, SPSS and GraphPad Prism were used. P < 0.05 was considered a statistically significant difference.

3. Results

3.1. Basic information of infants

Among the 52 subjects recruited for this study, one subject without available samples, two whose parents were too busy to collect samples, and two whose samples failed sequencing or quality control were excluded. This left 47 infants, including 26 with BC and 21 with YINDC. Gut microbiota is well acknowledged to be affected by age, gender, delivery mode, and feeding pattern [[25], [26], [27], [28], [29]], so the YINDC and BC groups were compared using binary logistic regression analysis.No significant differences were observed (Table 1).

Table 1.

Influencing factors of the YINDC and BC groups.

Influencing factors	YINDC	BC	P value
Age (months)	12.48 ± 4.68	10.42 ± 6.67	0.104
Gender
Boys	16 (76.2%)	14 (53.8%)	0.113
Girls	5 (23.8%)	12 (46.2%)
Delivery Mode
Vaginal Delivery	13 (61.9%)	14 (53.8%)	0.579
Caesarean Delivery	8 (38.1%)	12 (46.2%)
Feeding Pattern
Breast Feeding	11 (52.4%)	18 (69.2%)	0.237
Non-Breast Feeding	10 (47.6%)	8 (30.8%)

Note: Age values are shown as the mean ± standard deviation.

3.2. Alteration of gut microbiota diversity in infants with YINDC

The stool samples yielded 110,207 and 124,236 reads of the YINDC and BC groups, respectively. They were classified into 193 and 157 unique OTUs in the YINDC and BC groups, respectively (Fig. 1A). Although the species evenness index Shannon was similar in two groups (p = 0.297), the species richness indices ACE and Chao 1 were higher in the YINDC group than in the BC group (all p < 0.01), which suggested that YINDC had a higher intestinal flora diversity than BC did (Fig. 1B). To investigate the clustering of microbial taxa in both groups, PCoA was constructed (Fig. 1C). Results revealed inconsistent YINDC and BC clusters based on unweighted (p = 0.00059) and weighted (p = 0.00068) unifrac distances, indicating that the microbial community structure significantly differed in both groups.

Fig. 1.

Intestinal microbial diversity analysis. (A) Venn diagram demonstrating the existence of different OTUs in the YINDC and BC groups. (B) Species alpha diversity differences between the two groups based on the ACE, Chao 1, and Shannon indices. (C) The overall microbial structure of the two groups as represented by PCoA of unweighted (left) and weighted (right) unifrac distances.

3.3. Alteration of gut microbiota composition in infants with YINDC

The relative proportions of dominant taxa at different levels were assessed, and specific taxa with statistically different proportions were identified in each group (Fig. 2A). At the phylum level, the YINDC and BC groups primarily comprised Actinobacteria, Firmicutes, Bacteroidetes, and Proteobacteria. Compared to the BC group, the relative abundance of Bacteroidetes (BC, 8.83%; YINDC, 20.54%) significantly increased (p = 0.0233), whereas the relative abundance of Proteobacteria (BC, 15.29%; YINDC, 9.04%) decreased (p = 0.0443) in the YINDC group. At the family level, Lachnospiraceae, Ruminococcaceae, and Bacteroidaceae were abundant in the YINDC and BC groups, consistent with the flora characteristic of infants on complementary foods [30]; At the genus level, Bifidobacterium and Bacteroides were abundant in both groups, and the proportion of Faecalibacterium (BC, 0.81%; YINDC, 5.27%) was higher in the YINDC group than in the BC group (p = 0.0220).

Fig. 2.

Alterations in gut microbiota associated with YINDC. (A) Relative abundance of the most prevalent bacterial phyla, families, genera and Faecalibacterium in the YINDC and BC groups. (B) Cladogram showing the distribution of intestinal microbial phylogeny. (C) Histograms of different taxa. (D) KEGG pathways of YINDC-associated intestinal flora.

To identify the significantly different intestinal flora between the YINDC and BC groups, LEfSe analysis was performed according to LDA score (log10) > 2.0, with p < 0.05. As expected, discriminative bacteria were identified in both groups. The YINDC group had a predominantly of phylum Bacteroidetes, and genera Ruminococcus__torques_group, Anaerostipes, Bradyrhizobium, Candidatus_Solibacter, Delftia, Faecalibacterium, Nitrospira, Pseudarthrobacter, Rikenellaceae_RC9_gut_group, Ruminococcaceae_UCG_013. Meanwhile, the microbiome in the BC group was characterized by the predominance of phylum Proteobacteria and genera Rothia, Romboutsia, and Lachnoclostridium (Fig. 2B and C).

3.4. Functional analysis of the microbiome in infants with YINDC

PICRUSt analyses were applied to evaluate the gut microbiota function involved in YINDC. Genes associated with cardiovascular disease (p = 0.003) and circulatory system (p = 0.003) were increased and genes associated with membrane transport (p = 0.029) were decreased in the YINDC group (Fig. 2D).

3.5. Altered urine metabolic profile in infants with YINDC

To explore the effect of YINDC on metabolism, urine samples were collected and assessed using untargeted metabonomics The samples of two groups were distinguished on different sides in two ion modes (Fig. 3A), suggesting that the metabolic profile of infants with YINDC was significantly altered.

Fig. 3.

Alterations in urine metabolites associated with YINDC. (A) OPLS-DA score plot in the positive- (above) and negative-ion (below) modes with VIP >1.0 and p < 0.05. (B) Volcano map of the significantly different metabolites in the positive (left) and negative (right) ion modes. Upregulated metabolites were detected by FC > 1.5 and p < 0.05, whereas downregulated metabolites were detected by FC < 0.67 and p < 0.05. (C) KEGG pathways of YINDC-associated metabolites.

3.6. Significantly different urine metabolites between the YINDC and BC groups

Univariate statistical analysis based on volcano plots was used to visualize the overall distribution difference of metabolites between the YINDC and BC groups. The upregulated and downregulated metabolites were shown in Fig. 3B. To further screen YINDC-related metabolites, OPLS-DA analysis was used. Finally, 55 metabolites were obtained and considered to be significantly different metabolites (Table 2). And the structural illustrations of the differential metabolites were shown in Supplemental file 2. Collectively, these results demonstrated that YINDC infants had a unique metabolic profile.

Table 2.

Differential metabolites between the YINDC and BC groups.

Name	RT (s)	m/z	Adduction	VIP	p value
Mitoxantrone	53.0615	443.17175	[M − H]-	10.08	0.0020
d-Quinovose	212.492	223.08057	(M + CH3COO)-	1.95	0.0025
Amoxicillin	53.131	729.18481	[2M − H]-	1.14	0.0030
Arctigenin	39.984	371.15127	[M − H]-	1.31	0.0032
Cortisol 21-sulfate	41.122	441.15545	[M − H]-	3.06	0.0050
1,5-anhydro-d-sorbitol	214.764	163.05991	[M − H]-	1.84	0.0062
Estrone glucuronide	39.314	427.17627	[M-H-H2O]-	4.15	0.0063
3-dehydroepiandrosterone sulfate	56.666	367.15666	[M − H]-	2.35	0.0081
Alpha-d-Glucose	433.422	179.05481	(M − H)-	1.20	0.0082
[glu1] trh	39.226	379.19233	[M − H]-	1.44	0.0096
2′-fucosyllactose	431.907	487.16385	[M − H]-	2.03	0.0108
Boldenone sulfate	38.019	365.14084	[M − H]-	1.04	0.0113
1h-indazole-1-pentanoic acid,3-[[[1-(aminocarbonyl)-2,2-dimethylpropyl]amino]carbonyl]-	57.347	373.16638	[M − H]-	1.82	0.0113
Osajin	86.01	403.17668	[M − H]-	1.53	0.0132
Propionylglycine	391.0055	130.0497	[M − H]-	1.11	0.0136
Cortisone acetate	79.548	401.19847	(M − H)-	1.93	0.0136
3,4-dihydroxyhydrocinnamic acid	352.812	181.00802	[M − H]-	1.59	0.0142
Acetic acid, 2-[4-[3-(4-acetyl-3-hydroxy-2-propylphenoxy)propoxy]phenoxy]-	113.211	401.16014	[M − H]-	2.46	0.0169
L-Galactono-1,4-lactone	348.0265	215.0912	(M + K–2H)-	1.89	0.0170
Alpha-ketoglutarate	369.937	145.01292	[M − H]-	2.01	0.0170
(2r,3s,4s,5r,6r)-2-[[(2s,3r,4r)-3,4-dihydroxy-4-(hydroxymethyl)oxolan-2-yl]oxymethyl]-6-[4-(4-hydroxyphenyl)butan-2-yloxy]oxane-3,4,5-triol	114.254	459.16622	[M − H]-	4.39	0.0177
pregnenolone sulfate	70.595	395.18675	(M − H)-	3.64	0.0214
Propylpyrazoletriol	57.292	385.16626	[M − H]-	4.64	0.0216
3-(n-maleimidopropionyl)biocytin	83.069	425.19754	[M-H-C4H3NO2]-	2.06	0.0240
4-pyridoxic acid	52.365	182.04462	[M − H]-	9.70	0.0279
Allantoin	188.214	157.03556	[M − H]-	1.55	0.0286
Cis,cis-muconic acid	351.4185	141.01667	[M − H]-	11.68	0.0316
Hydroxysebacic acid	354.2715	217.10703	[M − H]-	3.34	0.0357
(1-acetyloxy-3-hydroxy-6,8a-dimethyl-7-oxo-3-propan-2-yl-2,3a,4,8-tetrahydro-1h-azulen-4-yl) 4-hydroxybenzoate	59.415	429.19212	[M − H]-	6.34	0.0394
N-acetyl-l-glutamate	385.099	188.05591	[M − H]-	1.85	0.0497
4-methyl-1h-pyrazole	365.0525	83.05883	[M+H]+	1.77	0.0011
Erucamide	40.5455	338.34094	[M+H]+	6.62	0.0014
4-ketopimelic acid	369.5635	129.06416	[M + H–CH2O2]+	4.66	0.0030
5-methyl-2′-deoxycytidine	205.012	126.06452	[M + H–C5H8O3]+	2.31	0.0031
Benzeneethanamine, 3,5-dimethoxy-.alpha.-methyl-4-propoxy-	379.45	254.16003	[M+H]+	2.36	0.0033
Pro-Asn-Arg	357.508	386.21833	[M+H]+	1.44	0.0048
15-deoxy-goyazensolide	139.501	284.33021	[M+K]+	1.18	0.0049
Ephedrine	182.737	166.12144	[M+H]+	3.68	0.0058
Lumichrome	72.1785	243.08667	[M+H]+	4.38	0.0069
(2r)-3-hydroxyisovaleroylcarnitine	302.943	262.16418	[M+H]+	2.91	0.0103
1-methylhistamine	364.3375	126.08971	[M+H]+	1.22	0.0121
Didodecyl 3,3′-thiodipropionate oxide	38.2345	531.40674	[M+H]+	9.13	0.0148
1,5-pentanediamine	264.4155	86.09527	[M + H–NH3]+	4.33	0.0151
N8-Acetylspermidine	440.246	188.17435	(M + H)+	1.65	0.0154
Anserine	301.8185	241.15358	[M+H]+	6.24	0.0157
2-naphthalenol, 1-[(2,4-dimethylphenyl)azo]-	40.715	277.1425	[M+H]+	1.30	0.0171
Asp-Asp-Arg	431.97	203.06519	[M+2H]2+	3.67	0.0209
Adipoyl-l-carnitine	390.888	290.15906	[M+H]+	11.85	0.0263
Gly-Arg	374.997	232.15034	[M+H]+	1.03	0.0303
Asp-Asp-Lys	373.823	189.07474	[M+2H]2+	1.26	0.0311
Vanylglycol	50.5185	167.06886	[M + H–H2O]+	1.49	0.0315
Diethyl 2,4-dimethylpyrrole-3,5-dicarboxylate	221.678	240.10067	[M+H]+	1.13	0.0403
Fusarenone-x	131.117	377.1041	[M+Na]+	1.18	0.0420
N-acetylputrescine	339.865	131.11631	[M+H]+	2.84	0.0488
5-aminovaleric acid betaine	387.571	160.13156	[M+H]+	17.11	0.0497

3.7. Pathway analysis of the significantly different metabolites in infants with YINDC

To further investigate metabolic mechanisms of YINDC, the changed metabolites were mapped to physiological pathways. Differential abundance analysis scores showed that YINDC affected histidine metabolism, proximal tubule bicarbonate reclamation, arginine biosynthesis, and steroid hormone biosynthesis, with negative scores corresponding to the excretory system, amino acid metabolism, and lipid metabolism in the hierarchy 2 pathway (Fig. 3C).

3.8. Correlation between the significantly different flora and metabolites

To determine if metabolic alterations related to intestinal flora, spearman analysis was performed. Several bacteria were very closely related to metabolites (Fig. 4A). We further narrowed the correlation coefficient to 0.5 < |r| < 1.26 pairs of significantly correlated bacteria and metabolites were screened, of which five pairs were positively correlated and 21 pairs were negatively correlated (Fig. 4B). Genera Romboutsia, Ruminococcus__torques_group, Anaerostipes, and Faecalibacterium were closely associated with the altered metabolites. Moreover, the correlation coefficient between Romboutsia and 3-dehydroepiandrosterone sulfate was >0.6, indicating that they may have a larger impact on YINDC.

Fig. 4.

Correlation analysis conducted by Spearman. (A) Heatmap of different microbiota and metabolites with |r| < 1. (B) Network diagram of significantly related microbiota and metabolites with 0.5 < |r| < 1. The red line and blue line indicate positive and negative correlation respectively. A darker color indicated a stronger correlation. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

3.9. Microbial and metabolic markers of YINDC

As an analytical method widely used to evaluate the predictive ability of models, ROC was used for the marker identification of YINDC. The top three metabolites were 3-dehydroepiandrosterone sulfate (AUC = 0.795, 95% CI = 0.652–0.899), 2′-fucosyllactose (AUC = 0.788, 95% CI = 0.644–0.893), and 4-pyridoxic acid (AUC = 0.780, 95% CI = 0.636–0.888). Genera with AUC >0.8 included Bradyrhizobium (AUC = 0.807, 95% CI = 0.665–0.907) and Delftia (AUC = 0.811, 95% CI = 0.671–0.911) (all p < 0.001) (Fig. 5A–E). Furthermore, the combination of two genus markers whose AUCs were >0.8 was found to have a higher diagnostic accuracy for YINDC than the markers alone (AUC = 0.845, 95% CI = 0.710–0.934, p < 0.001) (Fig. 5F). The combined marker had a sensitivity and specificity of 71.43% and 80.77%.

Fig. 5.

ROC curves for differential genera and metabolites of YINDC. (A) Genus Bradyrhizobium. (B) Genus Delftia. (C) Metabolite 3-dehydroepiandrosterone sulfate. (D) Metabolite 2′-fucosyllactose. (E) Metabolite 4-pyridoxic acid. (F) Combined marker composed of the two genus markers.

4. Discussion

Due to the fast development of modern detection methods and TCMC theory, more and more researchers are focusing on the microscopic study of YINDC. Four years ago, adults with YINDC were found to be associated with cardiovascular diseases (CVDs), infectious diseases, and certain cancers through salivary RNA [18]. In the current study, we found an increase in genes related to CVDs and the circulatory system in infants with YINDC via the intestinal microbiota. This suggests that YINDC may be associated with cardiovascular pathology, which remains a major contributor to global morbidity as well as mortality [31]. In a recently published meta-analysis, YINDC accounted for 14.8% of all patients with high normal blood pressure, which is a known risk factor for CVDs [32]. Therefore, identifying YINDC in infancy and intervening early are of critical importance in preventing CVDs. In this study, different gut microbiota of YINDC infants was identified, some of which were closely related to various CVDs.

The phylum Bacteroidetes, and its class Bacteroidia, and genus Bacteroides, which were abundant in the YINDC group, are considered to be the dominant flora in coronary heart disease. Bacteroidetes is the most numerous phylum of bacteria in the gut of the human body. It is engaged in polysaccharide fermentation, nitrogenous substances utilization, propionates production, and is often described as a beneficial bacterium in prior studies [33,34]. However, it is also participating in the process of protein hydrolysis with the release of toxic substances, thus promoting inflammation [35]. Shin et al. [36] found that Bacteroidetes was negatively related to HDL and acted as a counterpart in obesity. In this study, Bacteroidetes were abundant in YINDC where individuals are mostly thin, in agreement with the above study. Apart from the above bacteria, infants with YINDC had an increased abundance of Anaerostipes, Rikenellaceae_RC9_gut_group and Ruminococcus_torques_group and decreased abundance of Lachnoclostridium. Anaerostipes is reportedly associated with the development of CVDs [37]. Treatment with traditional Chinese medicine decreases its abundance in patients with chronic stable angina [38]. Some bacteria of this genus can reduce glucose levels and also promote propionate production through inositol or phytate, thereby reducing the risk of metabolic diseases [39,40]. Considerable changes in the Rikenellaceae_RC9_gut_group may be involved in acute myocardial ischemia through multiple pathways [41]. Ruminococcus torques is predictive of cardiac survival outcomes [42]. Lachnoclostridium counts are lower among those with higher maternal consumption of meat [43]. According to TCMC, YINDC is partially caused by excessive meat intake, which may be associated with reduced Lachnoclostridium level. Interestingly, the biomarker Bradyrhizobium, which can discriminate YINDC from BC, has been found to be negatively related to water intake and positively correlated with meat intake [44]. This could explained the preference for greasy, heavy-tasting food and the characteristics of water deficiency in individuals with YINDC.

In addition to being associated with CDVs, some genera are associated with the body shape of YINDC. For example, Ruminococcus_torques_group possibly modifies fat turnover. Low abundance of the group facilitates body fat control and enhances the contribution of resistant starch to abdominal obesity [45]. Conversely, an high abundance may exert adverse effects. Positive effect of probiotics on non-obese diabetic mice involved a reduction of this group [40]. Bacteria belonging to Ruminococcuss_torques_group probably consist of species that are involved in small intestinal permeability and immune responses during the first two years of life [46]. Specifically, it utilizes intestine mucin, whose increased mucin degradation allows luminal antigens to pass through the intestinal barrier, stimulating the immune system and causing intestinal diseases [47]. Infants with YINDC are usually underweight, possibly because Ruminococcus__torques_group increases intestinal permeability, thus nutrients cannot be completely digested and absorbed. A lower relative abundance of Anaerostipes is observed in obese patients and mice [48]. The consumption of lactic acid-producing bacteria increases the abundance of Anaerostipes, which is accompanied by weight loss [44]. The relative abundance of Lachnoclostridium is significantly higher in obese mice [49], which has been linked to the formation of visceral fat and is positively correlated with cholesterol levels [50,51]. Faecalibacterium is a butyrate producing bacterium with protective effects against obesity [52]. These changes in obesity-associated bacteria were also observed in the present study. Additionally, a high-fat diet reportedly results in the progression of primary Sjögren's syndrome [53], decreased Faecalibacterium is commonly observed among Sjogren's syndrome subjects [54]. The relationship between Faecalibacterium and dry symptoms requires further investigation.

“Was your child always excited” is an important personality item to determine whether an infant belongs to YINDC. If a child is always excited, he/she is more likely to behave as an individual with YINDC. The abundance of specific microbiota varies among people of different personalities. Participants with autism spectrum disorder have a decreased relative abundance of Ruminococcaceae UCG 013 [55], whereas excited YINDC individuals had an increased number of Ruminococcaceae UCG 013. Faecalibacterium, affiliated with the Ruminococcaceae family, is negatively correlated with the severity of depressive symptoms [56] and is more abundant in individuals with YINDC. Therefore, we speculated that the personality of YINDC infants may be correlated with the genera Ruminococcaceae UCG 013 and Faecalibacterium.

Several metabolites associated with CVDs were also identified in infants with YINDC, including upregulated metabolites, such as 4-pyridoxic acid and downregulated metabolites, such as 3-dehydroepiandrosterone sulfate. The concentration of 4-Pyridoxic acid correlates with cardiovascular risk, and a high 4-pyridoxic acid/pyridoxine ratio is considered to be relevant to the global cardiovascular risk [57]. Conversely, dehydroepiandrosterone sulfate levels is negatively associated with CVDs [58,59]. The major metabolic pathways of YINDC included histidine metabolism, arginine biosynthesis, steroid hormone biosynthesis, and proximal tubule bicarbonate reclamation. Consistent with the present study, Tabrez et al. [30] found histidine metabolism and arginine metabolism alteration by untargeted metabolomics in patients with CVDs. Histidine metabolism and arginine biosynthesis belong to amino acid metabolism. Current evidence confirms that a defect in branched chain amino acids (BCAAs) catabolism, and increased BCAAs levels are significantly correlated with CVDs risk [[60], [61], [62]]. Thus, BCAAs and their derivatives are supported as the potential biomarkers of CVDs [63]. A relative deficiency in steroid hormones, such as dehydroepiandrosterone, may be causally correlated with the development of a series of diseases associated with aging, including CVDs [64]. Cardiovascular risk is well known to increase in obese patients, but some overweight and obese individuals remain cardiometabolically healthy [65]. Innovatively, our study found an association between YINDC and CVDs, suggesting that thin individuals are also at risk of CVDs.

The microbial community structure significantly differed between the YINDC and BC groups, but there was partial overlap between the two groups of samples, probably due to the small sample size and within-sample variation. Spearman's correlation analysis showed that the level of critical metabolite 3-dehydroepiandrosterone sulfate was negatively related to the number of genera Ruminococcaceae UCG 013, Faecalibacterium, Ruminococcus__torques_group and positively correlated with the abundance of genus Rothia, which was in line with their separate trends. On the basis of the significantly different bacteria and metabolites, we found a combined marker comprising two bacterial markers Bradyrhizobium and Delftia to highly identify YINDC. These findings could help in the early diagnosis of YINDC so that early initiation of treatment could improve the imbalance state in infancy and prevent CVDs as early as possible.

5. Study limitations

This study had several limitations: (1) The age range of 0–2 years is considered a critical period for the functioning of the gut flora. Identification of structural changes in the gut flora during this time and narrowing of the age range requires further studies. (2) The sample size was limited and potential errors may exist. The use of 16S rRNA sequencing to analyze the gut microbiota has low phylogenetic power at the species level and poor discriminatory power for some genera; thus, our study findings need to be validated in macro-genomic studies with larger samples. (3) This study found that infants with YINDC and BC were not proportionally equal in terms of gender and feeding pattern, and the differences were not statistically significant. This may be influenced by the characteristics of different constitutions and requires studies related to the factors associated with constitution formation.

6. Conclusions

In conclusion, the present study revealed disordered gut microbiota and urinary metabolome in infants with YINDC, providing critical clues to its pathogenesis. Infants with YINDC may be at risk for CVDs, and this study provided several new targets for intervention in these complex diseases. The main metabolic pathways in infants with YINDC included histidine metabolism and arginine biosynthesis, etc. Additionally, the combined presence of genera Bradyrhizobium and Delftia had high diagnostic accuracy for YINDC. Multi-center studies with large sample sizes, such as bacterial transplantation experiments, in infants with YINDC are required to further study the key flora, metabolites, and their interactions identified in this study.

Author contribution statement

Haihong Zhao: Conceived and designed the experiments; Performed the experiments; Analyzed and interpreted the data; Wrote the paper.

Qiqi Ren; Huai-yu Wang: Analyzed and interpreted the data.

Yuhan Zong: Performed the experiments.

Weibo Zhao; Yaqi Wang; Miao Qu: Contributed reagents, materials, analysis tools or data.

Ji Wang: Conceived and designed the experiments.

Funding statement

National Key Research and Development Program of China (2020YFC2003100, 2020YFC2003103), the Project Supported by the Foundation of National Infrastructures for Translational Medicine (Shanghai) (TMSK-2021-414), Innovation Team and Talents Cultivation Program of National Administration of Traditional Chinese Medicine (ZYYCXTD-C-202001), Research Fund of Chinese Nutrition Society: Feihe Constitution Nutrition and Health (CNS-Feihe2018A01).

Data availability statement

Data will be made available on request.

Declaration of interest’s statement

The authors declare no conflict of interest.

Appendix A. Supplementary data

The following are the Supplementary data to this article: