Alterations in gut microbiome associated with severity of atopic dermatitis in infants

Atopic dermatitis (AD) often arises in infancy, and gut microbial dysbiosis is associated with the development of AD. However, less is known about specific changes in early‐life gut microbiome associated with AD and AD severity. This study aims to reveal the gut microbial composition and function profiles associated with the severity of AD in infants.


INTRODUCTION
Atopic dermatitis (AD) is the most common chronic inflammatory skin disease and usually appears below 2 years and affects up to 30% of Chinese infants aged 1-12 months. 1 Early-life AD patients are more likely to develop subsequent food allergies, asthma and rhinitis. 2oreover, AD has a significant impact on the quality of life (QoL) of patients and their families, including itchiness and sleep loss. 3However, available data to examine the correlation between the severity of AD and QoL lack in Asian countries.The causes of AD are not fully understood and are supposed to be a combination of genetic and environmental factors, including skin barrier dysfunction, systemic immune dysregulation and microbiota dysbiosis. 4ncreasing evidence has demonstrated an intimate, bidirectional connection between the gut and skin, linking gastrointestinal health to skin homeostasis. 5However, the results are inconsistent, and observational studies have reported significant gut microbial aberrancy in infants with AD than healthy infants, including Clostridia, Lactobacilli, or Short-chain fatty acids (SCFAs) producing bacteria. 6,7SCFAs are colonic bacterial metabolites produced by fermentation of non-digestible carbohydrates, primarily acetate, propionate and butyrate. 8SCFAproducing bacteria protect against allergic diseases by exerting anti-inflammatory properties and tolerogenic immune responses. 9One study showed that the severity of AD in infants correlated inversely with bacterial diversity and with the abundance of butyrate-producing bacteria. 10owever, butyrate and valerate levels were found to be lower in transient AD infants than in healthy controls and those in persistent AD infants, 6 suggesting that infant ADassociated compositional and functional alteration of specific gut microbiota (including SCFA-producing bacteria) in AD progression remains inconclusive, and more studies are needed.This study aimed to identify the gut microbial composition and functional pathways in Chinese infants with AD, and the correlations between the severity of AD and the gut microbiota.

Study population
Study infants were recruited at the Department of Dermatology, Shanghai Children's Medical Center, and are the baseline data of a clinical trial with the probiotic intervention, which was approved by the Ethics Committee of Shanghai Children's Medical Center, China (SCMCIRB-K2020038-1). Written informed consent for this study was obtained from the patients' parents or guardians.This study was registered at Chinese Clinical Trial Registry (Identifier: ChiCTR2100045308).All procedures were performed in compliance with the Declaration of Helsinki.
Atopic dermatitis was diagnosed by an experienced paediatric dermatologist using the UK Working Party criteria, 11 and the severity of AD was assessed using the SCORAD values.A diagnosed AD infant was included if he/she (1) was 30 days to 12 months of age; (2) was born between 38 ≤gestational age <42 weeks with birth weight between 2500 and 4000 g; (3) had no gastrointestinal tract disorders or metabolic diseases; (4) had not received antibiotics or probiotics and remained free of diarrhoea 2 weeks before the study.Patients were divided into three groups based on the AD severity classified according to SCROAD values: mild group (SCORAD<25), moderate group (25 ≤ SCORAD <50) and severe group (51 ≤ SCORAD <103).Infants' Dermatitis Quality of Life (IDQoL) index and Dermatitis Family Impact (DFI) index were collected using published questionnaires 12 to evaluate the effect of AD on quality of life.Samples were collected and immediately stored at −80°C prior to further analysis.

DNA extraction and 16S rRNA sequencing analysis
DNA was extracted from faecal samples using the QIAamp Fast DNA Stool Mini Kit (Qiagen, Germany) according to the manufacturer's protocol.The V3-V4 region of the bacterial 16S ribosomal RNA(RNA) genes was amplified by PCR using barcoded primers 341F (5′-CCTACGGGRSGCAGCAG-3′) and 806R (5′-GGACTACVVGGGTATCTAATC-3′).All quantified amplicons were pooled to equalize the concentrations for sequencing using the Illumina MiSeq (Illumina, Inc., CA, USA).The paired-end reads of 425 bp were overlapped on their 3 ends for concatenation into original longer tags by using PANDAseq (https:// github.com/ neufe ld/ pandaseq, version 2.9).DNA extraction, library construction and sequencing of MiSeq platform were conducted at the Realbio Genomics Institute (Shanghai, China).Operational taxonomic units (OTUs) were clustered with 97% similarity using UPARSE (http:// drive5.com/ uparse/ ), and chimeric sequences were identified and removed using Usearch (version 7.0.1090).The taxonomy of each 16S rRNA gene sequence was analysed by the RDP Classifier (http:// rdp.cme.msu.edu/ ) against the SILVA (SSU 117/119) 16S RNA database using a confidence threshold of 0.8.Alpha diversity (Chao, Simpson, and Shannon index) analyses were also performed using the Python scripts of QIIME (version 1.9.1).Phylogenetic investigation of communities by reconstruction of unobserved states (PICRUSt) analysis was performed to predict the bacterial functionality using the PICRUSt package.

Statistical analysis
The demographic characteristics of the infants among the three groups were compared using the one-way ANOVA for continuous variables of normal distribution, the Chi-Square test for categorical variables and the Kruskal-Wallis test for bacterial alpha diversity.Generalized linear model (GLM) was then performed to compare the normalized data of bacterial composition, and functional pathways between groups with the adjustment of age, feeding mode and delivery mode.Principal coordinate analysis (PCoA) was performed to visualize the dissimilarity of microbial composition based on the normalized values and weighted UniFrac distances, unweighted UniFrac distances (R package, vegan).Linear discriminant analysis (LDA) effect size (LEfSe) method was performed to identify the characteristic bacterial taxa/functional pathways of different microbiome communities and effectively aid in explaining the biology underlying differences in microbial communities.The threshold on the logarithmic score of LDA analysis was set to 2.0.Spearman's analysis was performed to identify associations between the SCORAD values and specific bacterial genera and functional pathways (GraphPad Prism version 9, San Diego).Unless otherwise specified, all statistical analyses were performed using SPSS version 26.0 (Chicago, IL, USA).Statistical significance was set at p < 0.05.

Baseline characteristics of the infants with AD
Sixty-two AD infants (average age, 4.7 months) were included in the present study, divided into mild, moderate and severe groups based on the AD severity.There were no significant differences in the overall demographic characteristics, including sex, delivery mode, age, feeding mode and supplementation of VD between the three groups.There were significant differences in the SCORAD, IDQoL, and DFI scores among the three groups (p = 0, p = 0.002, and p = 0.01, respectively) (Table 1).

Differences of gut microbial diversity and composition in three AD groups
No significant differences in the microbial αdiversity were shown as the Chao, Shannon and Simpson indices among the three groups (Figure 1a-c).No significant differences were observed in PCoA analyses among the three groups (Figure 1d,e).Neither significant differences were shown in the PCoA analysis between the breastfeeding group and the non-breastfeeding group, nor between the vaginal delivery group and the caesarean group (Figure S1).Six bacterial families, including Enterobacteriaceae, Veillonellaceae, Bifidobacteriaceae, Clostridiaceae, Lachnospiraceae and Streptococcaceae, dominated in the three AD groups, covering 89.0%-96.0% of the total bacterial abundance in different groups (Figure 2a).The top six bacterial genera in each AD group included Escherichia/ Shigella, Bifidobacterium, Veillonella, Bacteroides, Klebsiella and Clostridium sensu stricto, ranging from 4.0% to 22.1% of the total bacterial relative abundance (RA) (Figure 2b).
The LDA method identified the characteristic taxa in each AD group.Clostridium sensu stricto, Clostridium XI, Negativicoccus, Proteus and Erysipelotrichaceae_incertae_ sedis were significantly higher in the mild group compared to moderate and severe groups.Streptococcus, Rothia and Staphylococcus were the characteristic bacterial genera in the moderate group while Bacteroides and Helicobacter were biomarker genera in the severe group (Figure 3a).We next explored the significantly bacterial genus with the adjustment of potential confounding factors (age, feeding mode and delivery mode).Compared to the mild group (reference), both the moderate and severe groups had increased level of Parabacteroides (p = 0.023, and p = 0.024, respectively) and decreased level of Collinsella (p = 0.034, and p = 0.024, respectively).Meanwhile, the severe group presented less Clostridium sensu stricto than mild group (p = 0.001), and the trend of higher Bacteroides (p = 0.051) (Table S1).

Differences of bacteria functional analysis in three AD groups
Predicted functional genes were annotated using the Kyoto Encyclopaedia of Genes and Genomes (KEGG) The Kruskal-Wallis test was performed to compare differences between the three AD infant groups, followed by the Wilcoxon rank-sum test to identify differences between any two groups, if there were significant differences.Mild, Mild group; Moderate, Moderate group; Severe, Severe group.pathway and functional pathways in the severe group included those responsible for the metabolism of sphingolipids, glycosphingolipids, arginine and proline metabolism, and carbon fixation in photosynthetic organisms (Figure 3b).After adjusting the confounding factors in the GLM analysis, similar pathways were significantly different in the severe group compared with the mild group (Table S2).

Association between the key bacterial genera and functional pathways
Spearman's correlation analysis revealed significant relationships between the key bacterial genera and KEGG pathways among the groups.The RA of Bacteroides was positively related to the pathways of sphingolipid metabolism and glycosphingolipid biosynthesis; however, it was negatively correlated with Staphylococcus aureus infection (p < 0.01) (Figure 4).Moreover, the RAs of Streptococcus and Staphylococcus were positively associated with the pathways responsible for S. aureus infection (p < 0.01) (Figure 4).

DISCUSSION
This study demonstrated evident discrepancies in gut microbial composition and functional pathways in Chinese infants with mild-to-severe atopic dermatitis.Furthermore, the strong positive correlation between IDQoL, DFI values and SCORAD values suggested that AD severity could significantly increase the negative effect on the quality of life for infants and their families.The adverse impact of AD on children and their families is worth more attention.
Escherichia/Shigella, Bifidobacterium, Veillonella, Bacteroides, Klebsiella and Clostridium sensu stricto were shown to be the predominant gut bacterial groups in the AD infants, similar to previous findings of infants less than 1 year old. 13However, the proportion of the predominant bacterial groups varied with the severity of AD, including those belonging to the SCFA-producing bacteria.The strong negative correlation between the SCORAD index and Clostridium sensu stricto was in line with a recent infant study that Clostridium was negatively correlated with the SCORAD index. 6High abundance of Clostridia in infancy was also shown to be negatively correlated with blood eosinophils, which played an important role in the development of AD by regulating a variety of pathways in immune homeostasis. 14A recent study indicated that initial butyrate producers during infant gut microbiota development were endospore formers, including Clostridiaceae and Erysipelotrichaceae. 15 Of the family Clostridiaceae, Clostridium sensu stricto was the pioneer butyrateproducing bacteria, often saccharolytic and proteolytic, to access a wide variety of substrates in the infant diets and employ the butyrate kinase (buk) pathway to form butyrate. 15,16 Erysipelotrichaceae is another endospore bacterial family with butyrate formation in infants during the first year. 15arabacteroides and Bacteroides enriched in the severe AD group, consistent with recent studies in AD children and adults. 17,18Bacteroides could mediate T-cell activation, and some species are significant clinical pathogens with potent virulence factors involved in skin and tissue infections. 19Parabacteroides and Bacteroides are sphingolipid producing bacteria.This is the first study to discover a significant positive correlation between the SCORAD index and pathways responsible for sphingolipid metabolism, glycosphingolipid biosynthesis in infants.Sphingolipids regulate inflammation and immunity in eukaryotes via activating natural killer T cells and rapid release of cytokine; and some pathogens rely on host sphingolipids to promote their virulence (e.g.botulism, cholera). 20Altered profiles of sphingolipids and their metabolite ceramides have been reported in patients with AD patients. 21arabacteroides and Bacteroides are also acetate producers. 22Moreover, Bacteroides utilizes sugars and polysaccharides forming propionate, and succinate as metabolic end products to mediate competitive interactions with other bacteria. 16,19The significant correlation between decreased Clostridium sensu stricto, increased Parabacteroides, Bacteroides and more severe AD in the present study suggests alterations in SCFA-producing bacteria associated with the severity of AD may represent imbalanced SCFA production towards the increase in acetate and propionate and the decrease in available butyrate.This may suggest a hyperactive immune system in AD infants as one earlier study indicated. 23Acetate, propionate, and butyrate, how the three SCFA products function with host in the gut-skin axis is intriguing and needs further investigation.
In addition, some factors could impact on gut microbiota in infants, such as feeding mode and delivery method. 24Studies showed that gut microbiota profile in vaginally delivered infants resembled the vaginal microbiota in their mothers, while C-section born infants harboured the gut microbiome more resemble as those of the maternal skin or the environment. 25,26Another system review indicated that certain bacterial genera (such as Bifidobacterium and Bacteroides) seems to be significantly associated to the mode of delivery in the first 3 months after birth; however, the observed significant differences disappeared after 6 months of age of the infants. 27The gut microbiome structure showed similarity in regard of different delivery modes and feeding methods in the present study.The pathogenesis of AD remains complicated, whether these factors influence gut microbiome in the AD infants needs further investigation.
This study has several limitations.First, the population of the mild group is limited, because it is difficult for the parents of the mild group to complete the regular visits, particularly during COVID-19 pandemic period.Second, this is a cross-sectional study to identify specific bacterial genus associated with AD severity of infants and cannot extrapolate to the cause effect conclusion.The altered functional pathways associated with the SCORAD index revealed in this study are intriguing and require further investigation.A combination of metabolomic and metagenomic studies with host immune biomarkers would better reveal how gut microbiota and metabolites function in the development and progression of AD.
In conclusion, discrepancies in butyrate-producing bacteria and sphingolipid-producing bacteria compositions and related functional pathways were associated with the SCORAD index in infants.Our study provides new information on the connection between specific gut bacterial groups, related functional pathways and AD severity.Further larger-scale studies with metagenomic and metabolomic techniques that confirm specific bacterial groups and the metabolites associated with the AD severity may help to further discover microbial targets for personalized treatment of AD.

E 1
Comparison of αdiversity of gut microbiota among the three groups.(a) Chao diversity (b) Shannon diversity (c) Simpson diversity (d) PCoA plot based on weighted UniFrac distances (e) PCoA plot based on unweighted UniFrac distances.

F I G U R E 2
Comparison of the gut microbiota among the three infant groups.(a) Mean relative abundance (RA) at family level (b) Mean RA at genus level.Other mean combinations of taxa <1% of mean RA across all samples.p < 0.05 was considered as significant difference.Mild, Mild group; Moderate, Moderate group; Severe, Severe group.F I G U R E 3 Linear discriminant analysis (LDA) was performed to distinguish the representative (a) taxa of the three infant groups (b) functional pathways at level 3. Bacterial taxa and functional pathways with LDA scores >2.0 were indicated as biomarkers of each AD group.Mild, Mild group; Moderate, Moderate group; Severe, Severe group.

F I G U R E 4
Relationships between the key bacterial genera and functional pathways with significant difference in three AD groups.The colour scale represents R values.Red colour indicates positive correlation and green colour for negative correlation.*p < 0.05; **p < 0.01.F I G U R E 5 Spearman correlation analysis between SCORAD values and (a) Clostridium sensu stricto (b) Bacteroides (c) Parabacteroides (d) Collinsella (e) IDQoL scores (f) DFI scores.p < 0.05 was considered as significant difference.Mild, Mild group; Moderate, Moderate group; Severe, Severe group.
Characteristics of the study population.
T A B L E 1Note: Data are presented as mean ± SD; The demographic characteristics of the infants were compared using the one-way ANOVA for continuous variables; if significant, paired comparisons were then adjusted with Bonferroni test.The Chi-Square test was used for categorical variables.Abbreviations: DFI, Dermatitis Family Impact; IDQoL, Infants' Dermatitis Quality of Life; n, number; SCORAD, scoring atopic dermatitis; V/CS, vaginal delivery/caesareancesarean delivery.*p adj <0.05, when compared with the mild group; **p adj <0.01, when compared with the mild group; ## p adj <0.01, when compared with the moderate group.1. seven missing data of the body height; 2. four unknown data of the VD supplementation.