Skip to main content
Download PDF

Gut microbiota and metabolic changes in children with idiopathic short stature

BMC Pediatrics volume 24, Article number: 468 (2024) Cite this article

Abstract

Background

Idiopathic short stature (ISS) is characterized by short stature with unknown causes. Recent studies showed different gut microbiota flora and reduced fecal short-chain fatty acids in ISS children. However, the roles of the microbiome and metabolites in the pathogenesis of ISS remains largely unknown.

Methods

We recruited 51 Chinese subjects, comprising 26 ISS children and 25 normal-height control individuals. Untargeted metabolomics was performed to explore the fecal metabolic profiles between groups. A shotgun metagenomic sequencing approach was used to investigate the microbiome at the strains level. Mediation analyses were done to reveal correlations between the height standard deviation (SD) value, the gut microbiome and metabolites.

Results

We detected marked differences in the composition of fecal metabolites in the ISS group, particularly a significant increase in erucic acid and a decrease in spermidine, adenosine and L-5-Hydroxytryptophan, when compared to those of controls. We further identified specific groups of bacterial strains to be associated with the different metabolic profile. Through mediation analysis, 50 linkages were established. KEGG pathway analysis of microbiota and metabolites indicated nutritional disturbances. 13 selected features were able to accurately distinguish the ISS children from the controls (AUC = 0.933 [95%CI, 79.9–100%]) by receiver operating characteristic (ROC) analysis.

Conclusion

Our study suggests that the microbiome and the microbial-derived metabolites play certain roles in children’s growth. These findings provide a new research direction for better understanding the mechanism(s) underlying ISS.

Peer Review reports

Introduction

Short stature is defined as a height more than 2 Standard Deviation (SD) below the mean height for a given age, sex, and population. Idiopathic short stature (ISS) is the most common cause of short stature in children and excludes all possible causes such as undernourishment, endocrine conditions, skeletal dysplasia, genetic abnormalities and other systemic diseases [1]. At present, the etiology and pathogenesis of ISS remain unclear [2]. While recombinant human growth hormone (rhGH) therapy has been approved to treat children with ISS [3], the growth response to rhGH varies widely reflecting the heterogeneity of ISS [4]. Understanding the pathogenesis and disease mechanisms are important steps to identify new therapeutic targets and treatment.

Accumulating data has raised the possibility that gut microbiota may play an important role in body growth. Germ-free infant mice grew slower than the wild-type mice suggesting that the microbiota can interact with the somatotropic hormone axis to promote systemic growth [5]. In stunted undernourished children, a causal relationship was found between growth stunting and the components of the small intestinal microbiota [6]. Moreover, transplantation of immature microbiota from undernourished children into germ-free mice revealed that the immature microbiota impaired the growth phenotype [7]. Recently, a study using 16S rDNA sequencing reported that the composition of gut microbiota in ISS was very different from controls, with a significant decrease in the butyrate-producing genera including Faecalibacterium [8]. Targeted metabolomic analysis of fecal short-chain fatty acids from ISS children also revealed marked differences compared to healthy children [8].

Based on these pieces of evidence, we propose that the gut microbiota and resulting metabolic products may play an important role in the development of ISS. To test this hypothesis, we used shotgun metagenomic sequencing and untargeted metabolomics to analyze the gut flora in ISS and control individuals. We identified 50 associations between the height standard deviation (SD) value, microbiome and their derived metabolites by mediation analysis, which revealed that adenosine, erucic acid, spermidine, and L-5-Hydroxytryptophan could mediate the association of bacterial strains and the height SD value.

Materials and methods

Study design

From September 2019 to April 2021, we recruited 25 normal-height healthy children as controls from the general community and 26 children with ISS hospitalized in either the Xinhua Hospital Affiliated to Shanghai Jiao Tong University School of Medicine, Taizhou central Hospital, Shengjing Hospital of China Medical University or Zhejiang Xiaoshan Hospital. The study was approved by the Ethics Committee of Xinhua hospital (XHEC-C-2021–041-2), and has been registered in the Chinese Clinical Trial Registry (ChiCTR2100047810). Each hospital pediatric center strictly followed the same standard procedures for subject enrollment and for collection of clinical information and fecal samples. All subjects were of Hans Chinese ethnicity.

The inclusion criteria for children with ISS were as follows: (1) Height more than 2 SD below the mean height for a given age, sex, and population; (2) Normal response to growth hormone during stimulation tests (> 10 ng/mL); (3) Normal body weight and length at birth; (4) Absence of all other possible causes of short stature including undernourishment, intrauterine growth retardation, endocrine conditions, skeletal dysplasia, chromosomal disease, inherited metabolic disease, mental illness and other chronic systemic diseases; (5) Normal or slow growth rate; (6) Normal or delayed bone age; (7) Age 3–14 years. The inclusion criteria for the controls were as follows: (1) Normal height for a given age, sex, and population; (2) Age 3–14 years. For all subjects, the exclusion criteria were as follows: (1) Autoimmune disorders such as inflammatory bowel disease, autoimmune thyroid disease and irritable bowel syndrome; (2) Gastrointestinal diseases such as cholelithiasis; (3) History of gastrointestinal surgeries such as gastrectomy, colectomy, ileectomy, cholecystectomy and appendectomy; (4) Neuropsychiatric disorders including autism, epilepsy and depression; (5) History of malignant tumors; (6) Other diseases including obesity, hypertension and diabetes mellitus; (7) Use of antibiotics, probiotics, hormones, proton pump inhibitors, insulin sensitizers or growth-promoting herbs in the previous three months; (8) Familial short stature. To reduce patient bias in the two groups, subjects were matched as closely as possible in terms of sex, age and living area by province.

Fecal sample collection and DNA extraction

Written informed consent was obtained from all subjects meeting inclusion and exclusion criteria. Fresh stool samples were collected using disposable fecal collection tubes (Orienter, China), and immediately preserved at -80 °C until further analyses. Genomic DNA was extracted from fecal samples using Magnetic Soil And Stool DNA Kit (Qiagen, US) (Batch 1) or E.Z.N.A.® Soil DNA Kit (Omega Bio-tek, US) (Batch 2). DNA integrity was assessed by agarose gel electrophoresis and DNA concentrations measured using NanoDrop2000 (Thermo Fisher Scientific, US).

Metagenomic sequencing and data analysis

Construction of sequencing libraries

Genomic DNA was fragmented to an average size of about 350 bp (Batch 1) or 400 bp (Batch 2) for library construction. The library was constructed using NEBNext® UltraTM DNA Library Prep Kit for Illumina (NEB, US) (Batch 1) or NEXTflexTM Rapid DNA-Seq (Bioo Scientific, US) (Batch 2). Adapters containing the full complement of sequencing primer hybridization sites were ligated to the blunt end of fragments.

Processing of sequencing data

Libraries were sequenced on the Illumina NovaSeq 6000 (Illumina, US). Raw sequencing reads were processed to acquire clean data using Trimmomatic v0.39 softwareand the following steps: (1) Removal of low-quality reads which contain more than 35 bp of ‘N’ bases (default quality threshold value ≤ 15); (2) Removal of reads which shared overlapping sequences above a certain portion with the adapter (default length of 10 bp).

Since the sequencing data for microbial DNA are polluted by host DNA and animal DNA derived from diet, a blast analysis of host databases using Bowtie2.4.1 software (http://bowtie-bio.sourceforge.net/bowtie2/index.shtml) was used to filter out contaminating reads, enriching for microbial sequences.

Metagenome assembly

High-quality paired-end reads from each sample were de novo assembled into at least 500bp scaffolds using SOAPdenovo software v2.04 (http://soap.genomics.org.cn/soapdenovo.html). Assembled scaffolds were interrupted from the N connections and scafftigs without N bases collected as a set for further analysis.

Gene prediction and abundance analysis

  1. (1)

    The ORF prediction for the assembled set of microbial scafftigs were performed using MetaGeneMark v2.10 software (http://topaz.gatech.edu/GeneMark/) and ORFs less than 100 nt were filtered out.

  2. (2)

    For ORF prediction, CD-HIT software v4.8.1 (http://www.bioinformatics.org/cd-hit) was used to remove the redundant sequences and obtain a non-redundant initial gene set (nucleic acid sequences encoded by non-redundant consecutive genes were called genes).

  3. (3)

    Genes in each sample were calculated by comparing the clean data to the initial gene set using SOAP2 software v2.21 (http://soap.genomics.org.cn/). The final gene set (Unigenes) for further analysis were obtained by filtering out the genes with less than two reads per sample.

  4. (4)

    The abundance of each gene in individual samples was then calculated by comparing the number of reads and considering gene length.

  5. (5)

    Batch normalization was performed using ComBat_seq (with batch and group information) from sva (R package).

Taxonomy annotation

  1. (1)

    Comparisons between the unigenes and bacterial, fungal, archaeabacteria and viral sequences extracted from NCBI NR databases (version 202,004, https://www.ncbi.nlm.nih.gov/) was performed using DIAMOND software v2.0.8.146 (https://github.com/bbuchfink/diamond/).

  2. (2)

    For sequence comparison, we selected those with an evalue less than the minimum evalue * ten. Because of the possibility of multiple outputs for each sequence, the Least Common Ancestors (LCA) algorithm (system classification applied to MEGAN, https://en.wikipedia.org/wiki/Lowest_common_ancestor) was used to determine taxonomy.

  3. (3)

    An abundance table with the number of genes for individual samples at each taxonomic rank (phyla, family, genus and species) was then generated. The abundance for a particular species in one sample was defined as the sum of gene abundance annotated as that species. The number of genes for a particular species in one sample was defined as the number of genes (not equal to zero) annotated as that species.

Functional database annotations

  1. (1)

    Using DIAMOND software v2.0.8.146 (https://github.com/bbuchfink/diamond/), we compared the unigenes with KEGG database version 2019.10 (http://www.kegg.jp/kegg/). For comparisons, we selected the best blast hits for further analysis.

  2. (2)

    From these comparisons, the number of genes for each sample at different taxonomic ranks was obtained. The number of genes for a particular function in different samples is equal to the number of genes functionally annotated whose abundance is not equal to zero.

Binning analysis

  1. (1)

    Software mmseq2 version 13.45111 (https://github.com/soedinglab/MMseqs2) was then applied to build unique scaftig collections. Next bowtie2 was used to build unique scaftig collection indexes and map the reads back to this collection.

  2. (2)

    Software metabat2 version 2.15 (https://bitbucket.org/berkeleylab/metabat) was used to run the binning process based on each sample scaftig depth data. A single binning process was first conducted using scaftigs depth data from all samples followed by a mixed binning process. Merged single binning and mix binning results were generated by dRep v3.2.0 (https://github.com/MrOlm/drep). CheckM v1.1.3 (https://github.com/Ecogenomics/CheckM) and finally used for bin genome completeness and contamination evaluation. Bins were retained with completeness > 75% and contamination < 10%.

Untargeted metabolomics analysis

Fresh fecal tissues frozen in liquid nitrogen were ground and metabolites extracted with 80% methanol. The supernatant was injected into an LC–MS/MS system and analysis performed using the Vanquish UHPLC system (ThermoFisher, Germany) coupled with an Orbitrap Q ExactiveTM HF mass spectrometer (Thermo Fisher) in both positive and negative ionization modes(Some metabolites are more prone to adhering to hydrogen cations, resulting in the formation of molecular ion peaks easily in the positive mode, while others are more likely to lose hydrogen cations, facilitating the formation of molecular ion peaks in the negative mode). The raw data from each sample was then analyzed by Compound Discoverer 3.1 software (ThermoFisher) to obtain a list of peaks with retention time, m/z, and integrated peak area. The peak intensities were normalized and matched with a database to obtain accurate qualitative and quantitative measurements which were annotated using the KEGG (https://www.genome.jp/kegg/pathway.html), HMDB (https://hmdb.ca/metabolites), and LIPIDMaps (http://www.lipidmaps.org/) databases. After filling the gaps for missing values, different peak intensities of samples were then normalized to the total spectral intensity. Metabolite features with coefficient of variation (CV) ≥ 30% were then excluded after normalization.

Statistical analysis

Untargeted metabolomics analysis

The differential metabolites were screened by Partial least squares discriminant analysis (PLS-DA) using R package ropes. Screening criteria for key metabolites were as follows: (1) Variable Importance in Projection (VIP) > 1; (2) The absolute value of the logarithm base 2 of the fold change was > 1. Metabolic pathways with P-values < 0.05 were considered as significantly enriched.

Distance matrix-based variance estimation

Applied feature selection based on the permutational multivariate analysis of variance (PERMANOVA) using euclidean distance was used to estimate the contributions of clinical factors to different omics data (metabolomics and microbiome).

Microbial diversity analysis

For α-diversity, microbiome diversity was evaluated by the Shannon index through R package vegan. Microbiome richness was evaluated by the number of features and microbiome evenness calculated by the Shannon index divided by the logarithm base 2 of richness. For analysis of the microbiome β-diversity, R package vegan was also used to perform principal coordinate analysis and PERMANOVA (Jaccard distance).

Linear discriminant analysis effect size (LEfSe)

LEfSe was used to determine the differential microbiota characteristics (strain level and KEGG modules). A Linear Discriminant Analysis (LDA) score > 2 in the LEfSe model was used as the screening criteria.

Associations with clinical parameters

Associations between clinical parameters and metabolomics and gut microbiome (strain level) were assessed by Spearman correlation and the false detection rate (FDR) calculated using the Benjamini–Hochberg procedure.

Prediction model

Random forest (RF) models were built using the R package RandomForest to distinguish key features associated with the ISS and control cohorts. Randomly, 70% of individuals were selected as a training dataset and 30% of individuals were selected as a testing dataset. Five repeats of tenfold cross-validation were then used to estimate model performance. A ROC curve and plot of the test model was generated using the R package pROC procedure.

Biomarkers identification through recursive feature elimination

We used a recursive feature(RF) elimination approach to identify the smallest subset of features that could produce an effective model with good prediction accuracy. In brief, this involved the iterative fitting of RF, whereupon each iteration, a specified proportion of variables, with the smallest variable importance progressively discarded. The R package Caret process was applied recursively until only a single variable remains available as input. At each iteration, the model performance is assessed in terms of the out-of-bag error, where RF is used in a classification capacity or mean squared error is used for regression forests. Following this sequence, a set of variables with the smallest number predictive features was identified.

Bi-directional mediation analysis

By Spearman correlation (FDR < 0.05), we first checked whether any clinical parameters were associated with the differential metabolites. Next, we carried out a bi-directional mediation analysis with interactions between mediator and outcome, using the mediation function from mediation software version 4.5.0 to infer the mediation effect of metabolites and microbiome with clinical parameters. The FDR was calculated based on the Benjamini–Hochberg procedure.

Results

Study groups

Based on strict inclusion and exclusion criteria, a total of 51 participants were enrolled comprising 26 subjects with ISS and 25 controls. Variables such as age and gender in the two groups were generally matched to reduce complications. The clinical characteristics of children with ISS and controls are summarized in Table 1 and Supplementary files. Spearman’s correlation analysis between clinical characteristics and the height SD value identified a significantly positive correlation with BMI (p < 0.05). For all 51 fecal samples analysed, we obtained meaningful data on gut microbiome and metabolic derivatives for statistical analysis.

Table 1 Characteristics of the study groups
Full size table

Metabolic profiling of children with ISS

We used untargeted LC–MS to detect the composition of key fecal metabolome features. In total, 1987 compounds were observed in the positive and negative modes of LC–MS. PLS-DA score plots separated the ISS and control groups in both positive and negative modes, suggesting the presence of metabolic disturbances in the ISS children. Here, the R2 and Q2 scores were 0.97 and 0.56 in the positive mode (Fig. 1A) and 0.99 and 0.50 in the negative (Fig. 1C), respectively. The PERMANOVA test showed that the PLS-DA model was not overfitting and thus was valid for this study (positive mode: Fig. 1B, negative mode: Fig. 1D).

Fig. 1
figure 1

Differential metabolic profiles in individuals with ISS and Controls. A PLS-DA score plot in positive mode (R2 = 0.97, and Q2 = 0.56). B PLS-DA significance test in positive mode(pR2Y = 0.65, pQ2 = 0.03). C PLS-DA score plot in negative mode (R2 = 0.99, and Q2 = 0.5). D PLS-DA significance test in negative mode (pR2Y = 0.27, pQ2 = 0.04)

Full size image

Under the set cut-off conditions for the PLS-DA model (VIP > 1 and log2(FoldChange) > 1), we identified differential metabolite derivatives between the ISS and control groups in positive mode (Fig. 2A). Compared with controls, ISS exhibited an increase in organic acid, phosphatidyl choline, esters and fatty acids and, a decrease in amino acids, nucleotides, polyamines and lysophosphatidyl ethanolamine. A total of 52 metabolic pathways were enriched in the positive mode, among which 14 pathways exhibited a significant difference (p < 0.05, Fig. 2B). We observed that most of these pathways were concentrated in lipid and amino acid metabolism, such as tryptophan, glycerophospholipid, arginine and proline metabolism and, in the biosynthesis of unsaturated fatty acids and arachidonic acid metabolism. The altered metabolic profiles suggested a disturbance in nutrient metabolism in ISS children. Among these, four metabolites in particular had significant change, including increased level of erucic acid and decreased levels of spermidine, adenosine, and L-5-Hydroxytryptophan in ISS (Fig. 2C-F).

Fig. 2
figure 2

Identification of the fecal differential metabolites and pathways between the ISS and Control group in positive mode. A Visualization of differential metabolites. B The bubble plot displayed significantly enriched pathways (p < 0.05), C The relative intensity of the adenosine was significantly decreased in ISS. D The relative intensity of the L-5-Hydroxytryptophan was significantly decreased in ISS. E The relative intensity of the spermidine was significantly decreased in ISS. F The relative intensity of the erucic acid was significantly increased in ISS. Red and blue colours denote ISS and Control groups, respectively. ** p-value < 0.01; *** p-value < 0.001

Full size image

Microbial community analysis

We investigated gut microbial alterations in ISS children using metagenomic sequencing, and the clean reads across the 51 samples ranged from 38835568 to 53845134 (44417337 ± 3355970, mean ± SD) after QC of the raw sequencing data. After correcting batch effects (Supplementary files), we performed data analysis. The microbial community compositions between the two groups were similar at the rank of phylum with Firmicutes, Bacteroidetes, Actinobacteria, and Proteobacteria being the most abundant phylum (Supplementary files).

To identify any differences in bacterial diversity and richness between the two groups, we performed alpha diversity analysis, and initial results showed that the differences were not significant (pShannon = 0.594) (Fig. 3A). Next, we evaluated the beta diversity to compare the composition similarity between the microbiota communities and observed the gut microbiota composition. From this analysis, we found that ISS children were significantly different from controls (pAdonis = 0.001, R2 = 0.033) (Fig. 3B). Comparing the relative percentages of bacterial genera between ISS and control children, we identified 7 genera with an increased abundance and 7 genera that had a decreased abundance in ISS children (Supplementary files). At the species level, the relative percentages of bacterial species in ISS were significantly different from those of controls, including 21 species with an increased abundance and 34 species with a decreased abundance (Supplementary files). We further evaluated microbial communities at the strain level. Using a total of 5309009 contigs across all 51 samples to bin the metagenome-assembled genomes (MAGs) and 333 high-quality MAGs (completeness > 75%; contamination < 5%) were generated. These MAGs were annotated to the database and taxonomy classifications at the strain level were obtained. A total of 66 differentially abundant microbial strains exhibited significant changes in abundance, of which 23 strains (33.3% were Lachnospirales, 19% were Actinomycetes, 19% were Oscillospirales, and 9.5% were Bacteroidales) were significantly increased in ISS and 43 strains (26.6% were Bacteroidales, 26.6% were Oscillospirales, 15.5% were Coriobacteriales, and 6.6% were Lachnospirales) were significantly decreased in ISS (Fig. 3C). Most MAGs were assigned to Oscillospirales (24.2%), followed by Bacteroidales (21.2%), Lachnospirales (15.2%), Coriobacteriales (10.6%) and Actinomycetes (7.6%).

Fig. 3
figure 3

The alteration of the gut microbiota composition in individuals with ISS and Controls. A Shannon index showed no difference in α diversity (pShannon = 0.594). B Principal coordinate analysis (PCoA) analysis showed that the gut microbiota of individuals with ISS significantly differed from Controls (pAdonis = 0.001). C Cladogram indicated the phylogenetic distribution of strains correlated with the ISS or Control groups

Full size image

A total of 31 differential KEGG modules were identified between the two groups, mainly involved in nutrient metabolisms, such as amino acid, vitamin, nucleotide and organic acid metabolism (refer to Supplementary files). We explored the roles of the gut microbiome in modulating erucic acid, spermidine, adenosine, and L-5-Hydroxytryptophan metabolism by mapping the differential enzymes involved in the production of these metabolites (refer to Supplementary files). We observed that adenosine deaminase involved in adenosine metabolism to inosine [9] was up-regulated in ISS. There was a higher abundance of 21 gut bacterial strains expressing adenosine deaminase in ISS children (Fig. 4A). In addition, we observed that arginine decarboxylase involved in the arginine and spermidine metabolism [10] was decreased in the ISS group relative to the control group. There was a lower abundance of 25 bacterial strains expressing arginine decarboxylase in ISS children (Fig. 4B). Furthermore, we found that the abundance of tryptophan decarboxylase was enriched in ISS relative to control subjects, which may result in increased L-5-Hydroxytryptophan metabolism to serotonin [11] (Fig. 4C). Unfortunately, we were unable to identify any strains harboring tryptophan decarboxylase.

Fig. 4
figure 4

Key metabolic pathways mapped by microbial enzyme and fecal metabolism. A Microbial adenosine deaminase that converts adenosine to inosine was up-regulated in ISS. 21 bacterial strains that expressed adenosine deaminase were up-regulated in ISS. B Microbial arginine decarboxylase involved in the arginine metabolism was down-regulated in ISS. 25 bacterial strains containing arginine decarboxylase were down-regulated in ISS. C Microbial tryptophan decarboxylase involved in L-5-Hydroxytryptophan metabolism was up-regulated in ISS. * p-value < 0.05; ** p-value < 0.01; *** p-value < 0.001

Full size image

Multi-omics analysis

To compare the relative importance of fecal microbiome and metabolites in explaining the variability of inter-individual clinical indicators(age, height, weight, height SD and BMI), the proportion of variance explained by these two factors for the individual clinical indicators was calculated separately. We found that the type and abundance of gut bacteria and the composition of the metabolites could explain 49% and 64% of the variance, respectively (Fig. 5). Together, these two factors combined explained 72% of the variance in the clinical indicators.

Fig. 5
figure 5

Inter-individual variation in clinical indicators explained by bacteria and metabolites

Full size image

We performed Spearman's correlation analysis to investigate whether the differential metabolites were correlated with the clinical parameters. There were 37 differential metabolites significantly associated with the height SD value (Fig. 6), of which 22 were positively correlated and 15 were negatively correlated. Spermidine, adenosine, and L-5-Hydroxytryptophan, enriched in the control group, were positively correlated with the height SD value. In addition, erucic acid was enriched in the ISS group and showed a significant association with the height SD value. We further assessed the association between the metagenomic abundance changes and clinical indicators by Spearman’s correlation analysis. There were 35 bacterial strains significantly associated with the height SD value (Fig. 6), of which 23 were positively correlated and 12 were negatively correlated. We observed Longicatena caecimuris_bin.4400, Anaerostipes hadrus_bin.4374, and Streptococcus thermophilus_bin.2251 were negatively correlated with the height SD value and, Christensenellales sp000437595_bin.1962 and Bacilli sp001916775_bin.6360 were positively correlated with the height SD value. Moreover, we assessed associations between the metabolites and bacterial strains across all subjects. Given an FDR of 5%, 66 bacterial strains were correlated with 50 metabolites, presenting 461 significant associations (Fig. 6). Among them, we found Longicatena caecimuris_bin.4400 was negatively correlated with adenosine, L-5-Hydroxytryptophan, and spermidine. Anaerostipes hadrus_bin.4374 was negatively correlated with spermidine. Streptococcus thermophilus_bin.2251 was negatively correlated with L-5-Hydroxytryptophan. Christensenellales sp000437595_bin.1962 and Bacilli sp001916775_bin.6360 were negatively correlated with erucic acid.

Fig. 6
figure 6

Spearman correlation analysis of clinical indicators, bacterial strains, and metabolities. * p-value < 0.05; ** p-value < 0.01; *** p-value < 0.001

Full size image

A mediation analysis was also performed to investigate the links between clinical information, the microbiome, and metabolites. The analysis identified 126 mediation linkages (pmediation ≤ 0.05, pinverse-mediation ≥ 0.05) (Fig. 7) including 14 for the microbiome impact on the BMI through metabolites, 31 for the microbiome impact on the weight through metabolites, 50 for the microbiome impact on the height SD value through metabolites and 31 for the microbiome impact on the height through metabolites. We observed adenosine mediated a strong association of Longicatena caecimuris_bin.4400 with the the height SD value (Fig. 8A). In addition, we found erucic acid could mediate the association of Christensenellales sp000437595_bin.1962 and Bacilli sp001916775_bin.6360 with the height SD value (Fig. 8B-C). Moreover, spermidine could mediate the association of Anaerostipes hadrus_bin.4374 with the height SD value (Fig. 8D), and L-5-Hydroxytryptophan could mediate the association of Streptococcus thermophilus_bin.2251 with the height SD value (Fig. 8E).

Fig. 7
figure 7

Sankey plots showed the correlation network between the gut microbiome and the clinical indicators was mediated by fecal metabolites. Red connections indicate positive correlations (FDR < 0.05), whereas blue connections indicate negative correlations (FDR < 0.05)

Full size image
Fig. 8
figure 8

Mediation analysis and identification of interrelationships between the microbiome, metabolities, and the height SD. A Analysis of the effect of Longicatena caecimuris_bin. 4400 on the height SD of children as mediated by adenosine. B Analysis of the effect of Christensenellales sp000437595_bin.1962 on the height SD of children as mediated by erucic acid. C Analysis of the effect of Bacilli sp001916775_bin.6360 on the height SD of children as mediated by erucic acid. D Analysis of the effect of Anaerostipes hadrus_bin.4374 on the height SD of children as mediated by spermidine. E Analysis of the effect of Streptococcus thermophilus_bin.2251 on the height SD of children as mediated by L-5-Hydroxytryptophan. The gray lines showed the Spearman associations between the two factors. Red and blue arrows indicate direct mediation and reverse mediation, respectively

Full size image

Based on the results of mediation analysis, we included metabolites and bacterial strains of mediation linkages in the recursive feature elimination (RFE) model (Fig. 9A). Finally, 13 features were selected and used to perform receiver operating characteristic (ROC) analyses. The model had a significant discrimination for diagnostic accuracy [AUC(Test) = 0.933 (95%CI, 79.9–100%)] (Fig. 9B). Among the discriminatory features included in the classifier, 2-(tert-butyl)-6,7-dimethoxy-4H-3,1-benzoxazin-4-one had the greatest impact. The mean decrease in accuracy of 13 features was calculated in Fig. 9C.

Fig. 9
figure 9

Metabolites and microbiome classifies ISS from Controls. A Recursive feature elimination (RFE) identified the 13 most important features. B ROC analysis was performed to evaluate these 13 features, and the AUC of the training set and test set were 100% and 93.3%, respectively. C The mean decrease in accuracy of 13 features was calculated. Red and blue shading denotes metabolites and microbiomes, respectively

Full size image

Discussion

In this study, using untargeted metabolomics and metagenomic sequencing, we identified significant differences in fecal metabolite profiles and intestinal microbiota in ISS children compared with healthy children. Our findings revealed that specific microbial taxa were significantly correlated with metabolites and the height SD value suggesting that the composition of the gut microbiota may influence children’s growth by regulating intestinal metabolic processes. In the clinic, we observe that some ISS children are thinner and shorter than healthy children, even though they have a balanced diet, adequate sleep and regular exercise. Through the correlation analysis between clinical indicators, we found the height SD value was positively related to BMI, supporting the notion that ISS children may have intestinal malabsorption.

Intestinal flora interacts with the human body by regulating metabolites. We identified 37 height SD-associated metabolites through untargeted metabolomics. Correlation analysis showed some fecal metabolites decreased in ISS, such as spermidine, adenosine, and L-5-Hydroxytryptophan, whereas erucic acid was enriched. Spermidine, as a polyamine, is known to play an important role in bone remodeling and fetal growth [12]. We also found decreased amounts of arginine in the gut of ISS children, a precursor for spermidine synthesis [13]. Adenosine is generated from the hydrolysis of adenine nucleotides and participates in energy metabolism. A study reported that chondrocytes could secrete adenosine to regulate the homeostasis and function of cartilage [14]. L-5-Hydroxytryptophan is produced from tryptophan, and further transformed into melatonin [11]. A study observed that L-5-Hydroxytryptophan could increase rapid eye movement sleep amount [15], suggesting that it may promote the secretion of growth hormone. Erucic acid is present in rapeseed oil and has been reported to inhibit animal growth [16]. A study showed that high erucic acid feeding decreased the absorption of nutrient substances, such as lipids and protein, inhibited growth, and disrupted the intestinal development of fish [16]. Among the differential metabolites, we observed that a number of lipids increased in ISS, while some amino acids and nucleotides decreased in ISS. KEGG pathway analysis showed that most metabolic pathways were enriched in amino acid and lipid metabolic pathways, indicating that ISS may have problems in the absorption and utilization of nutrients.

In this study, significant alterations in the abundance of gut microbiota were found in ISS children. The relative abundance percentages of butyrate-producing bacteria, such as Faecalibacterium and Eubacterium has been reported to be significantly reduced in ISS children [8]. However, our study revealed different butyrate-producing bacteria in the gut, Butyricicoccus, Coprococcus, Fusobacterium_mortiferum, Alistipes_putredinis, and Coprococcus_comes, were reduced in ISS children. About 95% of butyrate is absorbed and utilized within the colon, whereas only a small amount enters the circulation [17]. Butyrate plays a role in the regulation of the intestinal homeostasis and in repairing the intestinal barrier function [18]. We hypothesise that reduced amounts of butyrate in ISS children may have an influence on normal functions, such as digestion and absorption of nutrients. Moreover, we identified a total of 31 differential KEGG modules between the two groups. These differential modules were also mainly involved in the nutrient metabolism pathways such as amino acid metabolism and nucleotide metabolism. We speculate that disturbance of nutrient metabolism was of relevance to the gut ecosystem of ISS. Further, we explored the roles of the gut microbiota in modulating fecal metabolism by mapping the differential enzymes associated with the disturbed metabolisms. We found that the level of fecal spermidine and its relevant metabolic enzyme (arginine decarboxylase) were consistently decreased in ISS children. We also found a group of bacterial strains expressing arginine decarboxylase was also decreased in ISS children. In addition, the microbial metabolic enzyme (adenosine deaminase) involved in the metabolism of adenosine was up-regulated and a group of bacterial strains containing adenosine deaminase exhibited the same trend. Taken together, these findings suggest that the fecal levels of spermidine and adenosine in children with ISS may be collectively modulated by groups of gut bacterial strains. Bi-directional mediation analysis found 5 highly significant causal relationships. Longicatena caecimuris_bin.4400 may affect children’s growth by regulating the metabolism of adenosine. Correlation analysis showed Longicatena caecimuris_bin.4400 was negatively related to the height SD value and adenosine. We analyzed the enzymes expressed by the Longicatena caecimuris_bin.4400 and found two enzymes (purine nucleoside phosphorylase and adenosine deaminase) involved in adenosine metabolism. Purine nucleoside phosphorylase catalyzes adenosine to the adenine [19] whereas adenosine deaminase catalyzes the hydrolytic deamination of adenosine to produce inosine [9]. The significantly enriched presence of Longicatena caecimuris_bin.4400 in the gut of ISS children may thus play a role in the reduction of adenosine. The mechanism of how Christensenellales sp000437595_bin.1962 and Bacilli sp001916775_bin.6360 regulate the metabolism of erucic acid to affect children's growth remains to be further explored. A study reported that big broiler chicks had a higher abundance of Christensenellales compared to small chicks [20], which suggests that Christensenellales may promote growth. Bacilli sp001916775_bin.6360 belongs to the class Bacilli. Several studies reported that a dietary supplementation with Bacillus strains help improved the growth performance of mice and broilers [21, 22].

A combination of 13 fecal metabolites and bacteria enabled construction of a disease classifier. The model discriminated ISS children from healthy children with a relatively high degree of accuracy (AUC of 0.953 in the test set). It is suggested that this identified metabolite signature has the potential of assisting clinical diagnosis and treatment in the future. In the clinic, we have observed that some ISS children are picky eaters. There is mounting evidence that dietary intake modulates the composition and function of the microbiome and metabolites [23]. In this study, we found erucic acid was increased in the ISS. Rapeseed and mustard oils have a high content of erucic acid [24], suggesting that a lower intake may be beneficial to ISS children. Further, based on the finding of low amounts of adenosine it may be desirable for ISS children to change to a diet with a high content of adenosine and arginine. From other clinical observations, some ISS children have insufficient sleep or poor sleep quality, and adequate sleep is important for normal growth and development. Since L-5-Hydroxytryptophan was also decreased in ISS children, and L-5-Hydroxytryptophan is mainly produced from tryptophan, supplementing their diet with food rich in tryptophan may also help to improve sleep quality. A study reported that children with stunting had lower serum concentrations of tryptophan [25]. Future research can explore whether children with ISS have lower serum tryptophan or deficits in conversion of tryptophan to L-5-Hydroxytryptophan.

Despite our efforts to correct batch effects using the Combat-seq method, this remains a limitation of our current study. Furthermore, the relatively small sample size constitutes another limitation. In our future research, we will collect data in a more systematic manner and expand the sample size in order to enhance the reliability of our findings. In conclusion, we identified gut microbiota and metabolite disturbances in children with ISS that may affect their growth and development. Through mediation analysis, we found several fecal metabolites closely related to the abundance of specific gut bacterial strains and found significant correlations with height SD value in ISS children. On this basis, our study points to dietary recommendations that may help promote normal growth and development of these children.

Availability of data and materials

We uploaded the raw sequence data to the National Center for Biotechnology Information (NCBI) database (SUB12933851).

References

  1. Cohen P, Rogol AD, Deal CL, Saenger P, Reiter EO, Ross JL, et al. Consensus statement on the diagnosis and treatment of children with idiopathic short stature: a summary of the Growth Hormone Research Society, the Lawson Wilkins Pediatric Endocrine Society, and the European Society for Paediatric Endocrinology Workshop. J Clin Endocrinol Metab. 2008;93(11):4210–7. https://doi.org/10.1210/jc.2008-0509.

    Article  CAS  PubMed  Google Scholar 

  2. Yuan J, Du Z, Wu Z, Yang Y, Cheng X, Liu X, et al. A novel diagnostic predictive model for idiopathic short stature in children. Front Endocrinol (Lausanne). 2021;12:721812. https://doi.org/10.3389/fendo.2021.721812.

    Article  PubMed  Google Scholar 

  3. Inzaghi E, Reiter E, Cianfarani S. The challenge of defining and investigating the causes of idiopathic short stature and finding an effective therapy. Horm Res Paediatr. 2019;92(2):71–83. https://doi.org/10.1159/000502901.

    Article  CAS  PubMed  Google Scholar 

  4. Cutfield WS, Albert BB. Growth hormone treatment for idiopathic short stature. Pediatr Endocrinol Rev. 2018;16(Suppl 1):113–22. https://doi.org/10.17458/per.vol16.2018.ca.ghidiopathicshortstature.

    Article  PubMed  Google Scholar 

  5. Schwarzer M, Makki K, Storelli G, Machuca-Gayet I, Srutkova D, Hermanova P, et al. Lactobacillus plantarum strain maintains growth of infant mice during chronic undernutrition. Science. 2016;351(6275):854–7. https://doi.org/10.1126/science.aad8588.

    Article  CAS  PubMed  Google Scholar 

  6. Chen RY, Kung VL, Das S, Hossain MS, Hibberd MC, Guruge J, et al. Duodenal microbiota in stunted undernourished children with enteropathy. N Engl J Med. 2020;383(4):321–33. https://doi.org/10.1056/NEJMoa1916004.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  7. Blanton LV, Charbonneau MR, Salih T, Barratt MJ, Venkatesh S, Ilkaveya O, et al. Gut bacteria that prevent growth impairments transmitted by microbiota from malnourished children. Science. 2016;351(6275):aad3311. https://doi.org/10.1126/science.aad3311.

    Article  CAS  PubMed  Google Scholar 

  8. Li L, Chen L, Yang Y, Wang J, Guo L, An J, et al. Characteristics of gut microbiome and its metabolites, short-chain fatty acids, in children with idiopathic short stature. Front Endocrinol (Lausanne). 2022;13:890200. https://doi.org/10.3389/fendo.2022.890200.

    Article  PubMed  PubMed Central  Google Scholar 

  9. da Cunha JG. A adenosinadesaminase. Uma enzima pluridisciplinar [Adenosine deaminase. A pluridisciplinary enzyme]. Acta Med Port. 1991;4(6):315–23.

    PubMed  Google Scholar 

  10. Wu G, Morris SM Jr. Arginine metabolism: nitric oxide and beyond. Biochem J. 1998;336(Pt 1):1–17. https://doi.org/10.1042/bj3360001.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  11. Maffei ME. 5-Hydroxytryptophan (5-HTP): natural occurrence, analysis, biosynthesis, biotechnology, physiology and toxicology. Int J Mol Sci. 2020;22(1):181. https://doi.org/10.3390/ijms22010181.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  12. Pegg AE. The function of spermine. IUBMB Life. 2014;66(1):8–18. https://doi.org/10.1002/iub.1237.

    Article  CAS  PubMed  Google Scholar 

  13. Halloran KM, Stenhouse C, Wu G, Bazer FW. Arginine, agmatine, and polyamines: key regulators of conceptus development in mammals. Adv Exp Med Biol. 2021;1332:85–105. https://doi.org/10.1007/978-3-030-74180-8_6.

    Article  CAS  PubMed  Google Scholar 

  14. Tesch AM, MacDonald MH, Kollias-Baker C, Benton HP. Effects of an adenosine kinase inhibitor and an adenosine deaminase inhibitor on accumulation of extracellular adenosine by equine articular chondrocytes. Am J Vet Res. 2002;63(11):1512–9. https://doi.org/10.2460/ajvr.2002.63.1512.

    Article  CAS  PubMed  Google Scholar 

  15. Wyatt RJ, Zarcone V, Engelman K, Dement WC, Snyder F, Sjoerdsma A. Effects of 5-hydroxytryptophan on the sleep of normal human subjects. Electroencephalogr Clin Neurophysiol. 1971;30(6):505–9. https://doi.org/10.1016/0013-4694(71)90147-7.

    Article  CAS  PubMed  Google Scholar 

  16. Gan L, Wu P, Feng L, Jiang WD, Liu Y, Jiang J, Kuang SY, et al. Erucic acid inhibits growth performance and disrupts intestinal structural integrity of on-growing grass carp (Ctenopharyngodon idella). Aquaculture. 2019;513:734437.

    Article  CAS  Google Scholar 

  17. Zhang L, Liu C, Jiang Q, Yin Y. Butyrate in energy metabolism: there is still more to learn. Trends Endocrinol Metab. 2021;32(3):159–69. https://doi.org/10.1016/j.tem.2020.12.003.

    Article  CAS  PubMed  Google Scholar 

  18. Wang RX, Lee JS, Campbell EL, Colgan SP. Microbiota-derived butyrate dynamically regulates intestinal homeostasis through regulation of actin-associated protein synaptopodin. Proc Natl Acad Sci U S A. 2020;117(21):11648–57. https://doi.org/10.1073/pnas.1917597117.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  19. Bzowska A, Kulikowska E, Shugar D. Purine nucleoside phosphorylases: properties, functions, and clinical aspects. Pharmacol Ther. 2000;88(3):349–425. https://doi.org/10.1016/s0163-7258(00)00097-8.

    Article  CAS  PubMed  Google Scholar 

  20. Waters JL, Ley RE. The human gut bacteria Christensenellaceae are widespread, heritable, and associated with health. BMC Biol. 2019;17(1):83. https://doi.org/10.1186/s12915-019-0699-4.

    Article  PubMed  PubMed Central  Google Scholar 

  21. Xu Y, Yu Y, Shen Y, Li Q, Lan J, Wu Y, et al. Effects of Bacillus subtilis and Bacillus licheniformis on growth performance, immunity, short chain fatty acid production, antioxidant capacity, and cecal microflora in broilers. Poult Sci. 2021;100(9):101358. https://doi.org/10.1016/j.psj.2021.101358.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  22. Zhang N, Wang L, Wei Y. Effects of Bacillus pumilus on growth performance, immunological indicators and gut microbiota of mice. J Anim Physiol Anim Nutr (Berl). 2021;105(4):797–805. https://doi.org/10.1111/jpn.13505.

    Article  CAS  PubMed  Google Scholar 

  23. Zmora N, Suez J, Elinav E. You are what you eat: diet, health and the gut microbiota. Nat Rev Gastroenterol Hepatol. 2019;16(1):35–56. https://doi.org/10.1038/s41575-018-0061-2.

    Article  CAS  PubMed  Google Scholar 

  24. Wendlinger C, Hammann S, Vetter W. Various concentrations of erucic acid in mustard oil and mustard. Food Chem. 2014;153:393–7. https://doi.org/10.1016/j.foodchem.2013.12.073.

    Article  CAS  PubMed  Google Scholar 

  25. Semba RD, Shardell M, Sakr Ashour FA, et al. child stunting is associated with low circulating essential amino acids. EBioMedicine. 2016;6:246–52. https://doi.org/10.1016/j.ebiom.2016.02.030.

    Article  PubMed  PubMed Central  Google Scholar 

Download references

Acknowledgements

We thank Renfu Lai and Jiuben Biotechnology Company for support and data processing. We thank all the participants and families in this study.

Funding

This work was sponsored by the National Key R&D Program of China (No. 2022YFC2703400, to YGY), the National Natural Science Foundation of China (No. 82271904 and No. 82070914, to YGY), The Shanghai Municipal Commission of Health and Family Planning (No. 20204Y0451, to SSD), the Shanghai Scientific and Technological Innovation Action Plan (No.21YF1437800, to LS), the Shanghai Natural Science Foundation of China(21ZR1452700, to YKZ).

Author information

Author notes

  1. Luyan Yan, Bin Ye, Min Yang and Yongsheng Shan contributed equally to this work and share first authorship.

Authors and Affiliations

  1. Department of Pediatric Endocrinology and Genetic Metabolism, Shanghai Institute for Pediatric Research, Xinhua Hospital Affiliated to Shanghai Jiao Tong University School of Medicine, Shanghai, China

    Luyan Yan, Dan Yan, Kaichuang Zhang & Yongguo Yu

  2. Department of Pediatric Internal Medicine, Taizhou Central Hospital, Taizhou University Hospital, Taizhou, China

    Bin Ye & DanFeng Fang

  3. Department of Pediatrics, Shengjing Hospital of China Medical University, Shenyang, China

    Min Yang

  4. Department of Pediatrics, Xiaoshan Hospital Affiliated to Hangzhou Normal University, Hangzhou, China

    Yongsheng Shan

Authors
  1. Luyan Yan
    View author publications

    You can also search for this author in PubMed Google Scholar

  2. Bin Ye
    View author publications

    You can also search for this author in PubMed Google Scholar

  3. Min Yang
    View author publications

    You can also search for this author in PubMed Google Scholar

  4. Yongsheng Shan
    View author publications

    You can also search for this author in PubMed Google Scholar

  5. Dan Yan
    View author publications

    You can also search for this author in PubMed Google Scholar

  6. DanFeng Fang
    View author publications

    You can also search for this author in PubMed Google Scholar

  7. Kaichuang Zhang
    View author publications

    You can also search for this author in PubMed Google Scholar

  8. Yongguo Yu
    View author publications

    You can also search for this author in PubMed Google Scholar

Contributions

YLY contributed to the conception and design of the study, the acquisition and analysis of data, and drafted the manuscript. YB, YM, SYS, YD, FDF and ZKC were responsible for the collection of samples and analysis of data. YYG contributed to the conception and design of the study, and revised and commented on the draft. This study was completed with the assistance and approval of all authors.

Corresponding author

Correspondence to Yongguo Yu.

Ethics declarations

Ethics approval and consent to participate

Written informed consent was obtained from all subjects. The study was approved by the Ethics Committee of Xinhua hospital (XHEC-C-2021–041-2), and has been registered in the Chinese Clinical Trial Registry (ChiCTR2100047810, 26/06/2021).

Consent for publication

All subjects consent to this publication.

Competing interests

The authors declare no competing interests.

Additional information

Publisher’s Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary Information

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Yan, L., Ye, B., Yang, M. et al. Gut microbiota and metabolic changes in children with idiopathic short stature. BMC Pediatr 24, 468 (2024). https://doi.org/10.1186/s12887-024-04944-3

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: https://doi.org/10.1186/s12887-024-04944-3

Keywords

BMC Pediatrics

ISSN: 1471-2431

Contact us