Advertisement
Journal of Investigative Dermatology Home

Deciphering the Role of Skin Surface Microbiome in Skin Health: An Integrative Multiomics Approach Reveals Three Distinct Metabolite‒Microbe Clusters

Open AccessPublished:August 02, 2021DOI:https://doi.org/10.1016/j.jid.2021.07.159
      The advent of 16S RNA profiling and shotgun metagenomics has enabled a holistic approach to the study of the skin microbiome composition. Despite the interesting findings in this rapidly developing scientific area, the big question remains: What role does the microbiome play in skin physiology? To begin answering this question, we employed an integrative methodology for microbiome and metabolome analysis of skin surface samples collected from the volar forearm of healthy infants aged 3–6-months. Whereas the infant skin metabolome was dominated by amino acids, lipids, and xenobiotics, the primary phyla of the microbiome were Firmicutes, Actinobacteria, and Proteobacteria. Zooming in on the species level revealed a large contribution of commensals belonging to the Cutibacterium and Staphylococcus genera, including Cutibacterium acnes, Staphylococcus epidermidis, and S. aureus. This heterogeneity was further highlighted when combining the microbiome with metabolome data. Integrative analyses delineated the coexistence of three distinct metabolite‒microbe clusters: one dominated by Cutibacterium linked to hydrophobic elements of the skin barrier, one associating Staphylococcus genus with amino acids relevant to the water holding capacity and pH regulation of the skin surface, and one characterized by Streptococcus and independent of any particular metabolomic profile.

      Abbreviations:

      ASV (amplicon sequence variant), LAR (legally authorized representative), SSH (skin surface hydration)

      Introduction

      Skin is the body’s first line of defense against infections and environmental stressors. It acts as a major physical and immunological protective barrier but also plays a critical role in temperature regulation, water holding, vitamin D production, and sensory perception. The outermost surface of the skin consists of a lipid- and protein-laden cornified layer dotted with hair follicles and eccrine glands that secrete lipids, antimicrobial peptides, enzymes, salts, etc. It harbors microbial communities living in a range of physiologically and anatomically distinct niches. Overall, this constitutes a highly heterogeneous and complex system.
      The skin surface is colonized immediately after parturition and is dynamically evolving during the first years of life. Although the long-term impact of delivery mode on the future composition of the skin microbiome remains unclear, it appears that the skin surface of infants born through cesarean section is predominantly colonized by commensal skin bacteria (Streptococcus, Staphylococcus, Propionibacterium), whereas the skin surface of vaginally delivered newborns is mostly colonized by microorganisms common to the female urogenital tract (Lactobacillus, Prevotella, Candida) (
      • Chu D.M.
      • Ma J.
      • Prince A.L.
      • Antony K.M.
      • Seferovic M.D.
      • Aagaard K.M.
      Maturation of the infant microbiome community structure and function across multiple body sites and in relation to mode of delivery.
      ;
      • Dominguez-Bello M.G.
      • De Jesus-Laboy K.M.
      • Shen N.
      • Cox L.M.
      • Amir A.
      • Gonzalez A.
      • et al.
      Partial restoration of the microbiota of cesarean-born infants via vaginal microbial transfer.
      ,
      • Dominguez-Bello M.G.
      • Costello E.K.
      • Contreras M.
      • Magris M.
      • Hidalgo G.
      • Fierer N.
      • et al.
      Delivery mode shapes the acquisition and structure of the initial microbiota across multiple body habitats in newborns.
      ;
      • Ward T.L.
      • Dominguez-Bello M.G.
      • Heisel T.
      • Al-Ghalith G.
      • Knights D.
      • Gale C.A.
      Development of the human mycobiome over the first month of life and across body sites.
      ). In the first weeks of life, microbial communities start developing site specificity (depending on dry, moist, or lipid-rich niches) while increasing in diversity (
      • Bouslimani A.
      • Porto C.
      • Rath C.M.
      • Wang M.
      • Guo Y.
      • Gonzalez A.
      • et al.
      Molecular cartography of the human skin surface in 3D.
      ;
      • Grice E.A.
      • Kong H.H.
      • Conlan S.
      • Deming C.B.
      • Davis J.
      • Young A.C.
      • et al.
      Topographical and temporal diversity of the human skin microbiome.
      ;
      • Mukherjee S.
      • Mitra R.
      • Maitra A.
      • Gupta S.
      • Kumaran S.
      • Chakrabortty A.
      • et al.
      Sebum and hydration levels in specific regions of human face significantly predict the nature and diversity of facial skin microbiome.
      ). At puberty, the stimulation of sebaceous gland secretion by hormones markedly shifts the physicochemical properties of the skin surface and favors the development of lipophilic taxa (Corynebacterium and Propionibacterium) (
      • Mukherjee S.
      • Mitra R.
      • Maitra A.
      • Gupta S.
      • Kumaran S.
      • Chakrabortty A.
      • et al.
      Sebum and hydration levels in specific regions of human face significantly predict the nature and diversity of facial skin microbiome.
      ). During adulthood and in the absence of shifts in external factors, the individual skin microbiome remains relatively stable (
      • Oh J.
      • Byrd A.L.
      • Park M.
      • Kong H.H.
      • Segre J.A.
      NISC Comparative Sequencing Program
      Temporal stability of the human skin microbiome.
      ), despite the large interindividual variability (
      • Bouslimani A.
      • Porto C.
      • Rath C.M.
      • Wang M.
      • Guo Y.
      • Gonzalez A.
      • et al.
      Molecular cartography of the human skin surface in 3D.
      ), suggesting that mutualistic and commensal interactions exist among microbes and between microbes and host, even for bacterial species often considered as opportunistic pathogens. Under healthy skin conditions, most of the microbes living on the skin behave as commensal or mutualistic organisms. Such microbes inhibit the spread of opportunistic parasites employing various mechanisms, including the stimulation of secretion of innate immunity factors secretion (e.g., IL-1⍺) (
      • Naik S.
      • Bouladoux N.
      • Wilhelm C.
      • Molloy M.J.
      • Salcedo R.
      • Kastenmuller W.
      • et al.
      Compartmentalized control of skin immunity by resident commensals.
      ) or antimicrobial peptides by keratinocytes (
      • Nakatsuji T.
      • Chen T.H.
      • Narala S.
      • Chun K.A.
      • Two A.M.
      • Yun T.
      • et al.
      Antimicrobials from human skin commensal bacteria protect against Staphylococcus aureus and are deficient in atopic dermatitis.
      ). Moreover, commensal microbes contribute to the education of the immune system and to healthy skin barrier homeostasis. In case of skin barrier breach or immunosuppression, these carefully balanced relationships may transition from commensalism to pathogenicity, a transition referred to as dysbiosis (
      • Chen Y.E.
      • Fischbach M.A.
      • Belkaid Y.
      Skin microbiota-host interactions [published correction appears in Nature 2018;555:543].
      ), enabling the overgrowth of pathogenic species, common in skin conditions such as acne (
      • Agak G.W.
      • Qin M.
      • Nobe J.
      • Kim M.H.
      • Krutzik S.R.
      • Tristan G.R.
      • et al.
      Propionibacterium acnes induces an IL-17 response in acne vulgaris that is regulated by vitamin A and vitamin D.
      ;
      • Barnard E.
      • Shi B.
      • Kang D.
      • Craft N.
      • Li H.
      The balance of metagenomic elements shapes the skin microbiome in acne and health [published correction appears in Sci Rep 2020;10:6037].
      ;
      • Fitz-Gibbon S.
      • Tomida S.
      • Chiu B.H.
      • Nguyen L.
      • Du C.
      • Liu M.
      • et al.
      Propionibacterium acnes strain populations in the human skin microbiome associated with acne.
      ), psoriasis (
      • Gao Z.
      • Tseng C.H.
      • Strober B.E.
      • Pei Z.
      • Blaser M.J.
      Substantial alterations of the cutaneous bacterial biota in psoriatic lesions.
      ), ulcer (
      • van Rensburg J.J.
      • Lin H.
      • Gao X.
      • Toh E.
      • Fortney K.R.
      • Ellinger S.
      • et al.
      The human skin microbiome associates with the outcome of and is influenced by bacterial infection.
      ), and atopic dermatitis (
      • Byrd A.L.
      • Deming C.
      • Cassidy S.K.B.
      • Harrison O.J.
      • Ng W.I.
      • Conlan S.
      • et al.
      Staphylococcus aureus and Staphylococcus epidermidis strain diversity underlying pediatric atopic dermatitis.
      ).
      Since the early 1950s, studies involving microbial cultures were undertaken, aiming to understand the role of the skin microbiome in physiology and disease (
      • Kong H.H.
      • Segre J.A.
      Skin microbiome: looking back to move forward.
      ;
      • Roth R.R.
      • James W.D.
      Microbial ecology of the skin.
      ). The systematic survey of the human microbiome has gained significant momentum over the past decade with the advent of 16S RNA profiling and shotgun metagenomic approaches coupled with second-generation sequencing technologies. Such methods enable the identification of potential causal relationships between microbial communities and clinical outcomes (
      • Schmidt T.S.B.
      • Raes J.
      • Bork P.
      The human gut microbiome: from association to modulation.
      ). Studies focusing on the role of individual species in skin physiology have followed a reductionistic approach. More recently, the metabolome has emerged as the Rosetta stone warranting the understanding of the molecular bases of microbial influence on host physiology through the production, modification, or degradation of bioactive metabolites (
      • Shaffer M.
      • Armstrong A.J.S.
      • Phelan V.V.
      • Reisdorph N.
      • Lozupone C.A.
      Microbiome and metabolome data integration provides insight into health and disease.
      ) in diseases ranging from obesity (
      • Maruvada P.
      • Leone V.
      • Kaplan L.M.
      • Chang E.B.
      The human microbiome and obesity: moving beyond associations.
      ), depression (
      • Valles-Colomer M.
      • Falony G.
      • Darzi Y.
      • Tigchelaar E.F.
      • Wang J.
      • Tito R.Y.
      • et al.
      The neuroactive potential of the human gut microbiota in quality of life and depression.
      ), autism (
      • Sharon G.
      • Cruz N.J.
      • Kang D.W.
      • Gandal M.J.
      • Wang B.
      • Kim Y.M.
      • et al.
      Human gut microbiota from autism spectrum disorder promote behavioral symptoms in mice.
      ), inflammatory bowel disease (
      • Lavelle A.
      • Sokol H.
      Gut microbiota-derived metabolites as key actors in inflammatory bowel disease.
      ), diabetes (
      • Liu Y.
      • Wang Y.
      • Ni Y.
      • Cheung C.K.Y.
      • Lam K.S.L.
      • Wang Y.
      • et al.
      Gut microbiome fermentation determines the efficacy of exercise for diabetes prevention.
      ), neurological conditions (
      • Hertel J.
      • Harms A.C.
      • Heinken A.
      • Baldini F.
      • Thinnes C.C.
      • Glaab E.
      • et al.
      Integrated analyses of microbiome and longitudinal metabolome data reveal microbial-host interactions on sulfur metabolism in Parkinson’s disease.
      ), as well as heart conditions (
      • Liu H.
      • Chen X.
      • Hu X.
      • Niu H.
      • Tian R.
      • Wang H.
      • et al.
      Alterations in the gut microbiome and metabolism with coronary artery disease severity.
      ;
      • Vojinovic D.
      • Radjabzadeh D.
      • Kurilshikov A.
      • Amin N.
      • Wijmenga C.
      • Franke L.
      • et al.
      Relationship between gut microbiota and circulating metabolites in population-based cohorts.
      ). Despite being successful in identifying metabolic pathways and bacterial targets to improve health, these more holistic, integrative approaches have so far been limited to the study of the gut microbiome (
      • Chen M.X.
      • Wang S.Y.
      • Kuo C.H.
      • Tsai I.L.
      Metabolome analysis for investigating host-gut microbiota interactions.
      ).
      Following these examples and applying the approaches used in the study of the gut microbiome to the skin, we aim to understand the role of the skin microbiome in skin physiology. This study constitutes an integrative analysis of skin microbiome and metabolome, examining the case of healthy infants.

      Results

      To characterize the infant skin metabolomic profile and microbiome composition, we analyzed skin swab and tape samples collected on the dorsal forearm of 16 healthy subjects (9 females, 7 males, aged 118 ± 29 days) (Figure 1 and Supplementary Table S1, see Materials and Methods for an overview of inclusion and exclusion criteria). Skin surface pH and skin surface hydration (SSH) values were also recorded. Matched swab samples (left and right arms) were subjected to 16S ribosomal RNA sequencing followed by profiling of microbial community taxonomic composition defining amplicon sequence variants (ASVs). Skin tapes were analyzed using a combination of ultrahigh-performance liquid chromatography-tandem mass spectrometry and gas chromatography-tandem mass spectrometry. The profiling was carried out using sensitive, high-resolution mass spectrometers in nontargeted mode, capturing a large number of known and uncharacterized metabolites. In addition, parents were asked to provide information on the delivery mode of the infant's birth.
      Figure thumbnail gr1
      Figure 1Experimental design and analytical strategy. Skin swabs and tapes were collected from the skin surface of the dorsal forearm of 16 healthy subjects. Each swab sample was subjected to 16S rRNA sequencing followed by profiling of microbial community taxonomic composition and imputation of functional potential. Each tape sample was analyzed by a combination of UHPLC/MS/MS and GC/MS/MS. Parents were asked to provide information on delivery mode at birth. Skin surface pH and SSH were also recorded on the same individuals on opposite arms. ASV, amplicon sequence variant; FAMD, factor analysis of mixed data; GC/MS/MS, gas chromatography with tandem mass spectrometry; RGCCA, regularized generalized canonical correlation analysis; rRNA, ribosomal RNA; sPLS, sparse partial least square; SSH, skin surface hydration; UHPLC/MS/MS, ultrahigh performance liquid chromatography-tandem mass spectrometry.

       Overview of the healthy skin surface microbiome and metabolome

      The composition and heterogeneity of the skin microbiome and metabolome in this cohort were analyzed first by estimating the relative contribution of each metabolic pathway and bacterial taxum and then grouped into superpathways and phyla, respectively. Overall, from the metabolome perspective, the leading superpathways were amino acids (28.2% of total metabolites), lipids (17.6%), and xenobiotics (16.8%), and from the microbiome perspective, the leading phyla were Firmicutes (68.9%), Proteobacteria (15.2%), and Actinobacteria (13.6%) (Figure 2).
      Figure thumbnail gr2
      Figure 2Overview of the healthy surface skin microbiome and metabolome at the high taxonomic level. Firmicutes are the dominating microbial phyla, whereas amino acids and lipids are the most prevalent metabolites. Barplots depicting the weight of (a) each superpathway and (b) genus in each sample. Areas are color coded according to superpathways (metabolome) or phylum (microbiome). The bars on the left show the average distribution across samples. Blacklines delineate (a) individual pathways and (b) genera.
      The core metabolome present in all the samples at ≥1.4% relative abundance consisted of 24 metabolites and contained fatty acid derivatives (2-hydroxyarachidate, eicosanoylsphingosine, phytosphingosine), amino acids and derivatives (asparagine, hydroxyproline, methionine, N-acetylglycine, dimethylaminoethanol), nucleosides (N6-carbamoylthreonyladenosine), carboxylic acids (1-methyl-4-imidazoleacetic acid), as well as uncharacterized compounds in even proportion across all subjects (Supplementary Figure S1a). Lowering the prevalence threshold to ≥8 samples while increasing the abundance threshold to ≥3% revealed that amino acids (N-acethyltrheonine, phenylalanine, arginine, histidine, gamma-gluthamylhistidine, gamma-glutamylleucine, etc.) were largely contributing to the core metabolome, together with tricarboxylic acid cycle and (an)aerobic cellular respiration byproducts (alpha-ketoglutarate, pyruvate, lactate), alpha-tocopherol, and lactose (Supplementary Figure S1b). When focusing only on metabolites that were on average contributing the most to the overall skin metabolome without putting any restriction in terms of prevalence, we found that among the most abundant compounds, a significant proportion belonged to xenobiotics (salicylate, propyl 4-hydroxybenzoate, 4-acetamidophenol, triethanolamine, bicine, dexpanthenol), likely originating from skincare routines (Supplementary Figure S1c).
      The core skin microbiome consisted of 14 genera present in ≥8 samples at ≥1% relative abundance and was dominated by Streptococcus (52.8%), Cutibacterium (11.8%), and Staphylococcus (8.1%) (Supplementary Figure S1d). This overall contribution of major genera was highly heterogeneous across samples: for example, the microbiome from sample 1101 was dominated by Cutibacterium (≈75% of the core microbiome), whereas the one from sample 1111 by Moraxella (≈50% of the core microbiome).

       The skin surface metabolome shapes bacterial communities and impacts microbiome diversity

      To visualize in a graph the relationship between demographics (age and sex), mode of delivery, skin physicochemical properties, and microbial richness, we employed factor analysis of mixed data, a principal component method specifically designed to explore data from both continuous and categorical variables (Figure 3a). This analysis revealed an association between skin pH, microbiome diversity (Chao1), and SSH. Looking at individual pairwise correlations, we confirmed a weak positive correlation between SSH and microbial richness (Figure 3b). Whereas birth mode appeared to be associated with skin surface pH and SSH, no association was detected with skin microbial diversity in this cohort (Supplementary Figure S2a–c). Furthermore, we found no correlation between metabolite diversity (Shannon index) and pH or SSH (Supplementary Figure S3).
      Figure thumbnail gr3
      Figure 3Skin surface microbiome and metabolome correlate with pH and hydration. (a) Biplot for a FAMD. Variables indicated with an outlined triangle are well-projected in the reduced dimensional plan (cos2 > 0.5). (b) Dotplot depicting the correlation between SSH and Chao1 alpha diversity index for ASV. (c) Dotplots depicting the Pearson’s correlation coefficient between bacterial genus abundance and skin pH and bacterial genus abundance and SSH. Bacterial genera are color coded according to the phylum they belong to. A more positive correlation with SSH reflects an association between the phylum and a relatively better-hydrated environment, and the opposite holds for a negative correlation. A more positive correlation with pH reflects an association between the phylum and a relatively alkali environment, and the opposite holds for a negative correlation. (d) Dotplots depicting the Pearson’s correlation coefficient between metabolic pathways weight and skin pH and metabolic pathways weight and SSH. Metabolic pathways are color coded according to the superpathway they belong to. A more positive correlation with SSH reflects an association between the species and a relatively better-hydrated environment, and the opposite holds for a negative correlation. A more positive correlation with pH reflects an association between the species and a relatively alkali environment, and the opposite holds for a negative correlation. ASV, amplicon sequence variant; FAMD, factor analysis of mixed data; SSH, skin surface hydration.
      To explore the association between the skin microenvironment of the individual subject (skin pH and SSH) and bacterial communities, we computed the pairwise Pearson’s correlation coefficient between skin pH and bacterial genera abundance and between SSH and bacterial genera abundance. By combining in a single graph the coefficient values of the two correlations (genera abundance–pH vs. genera abundance–SHH), we were able to determine the affinity of each genus for distinct skin niches in terms of acidity and moisturization (Figures 3c and Supplementary Figure S2d). Whereas Pseudomonas, Ruminococcus, Atopobium, Schaalia, and Lactobacillus were present on subjects with relatively acidic and hydrated skin, Cutibacterium was more abundant on subjects with relatively basic and slightly dry skin, and Moraxella, Agrobacterium, and Acinetobacter were more abundant in those with slightly acid and slightly dry skin. This analysis also revealed that the genera inside a given phylum were settling in heterogeneous niches, hence the significance to study microbiome at the finest possible grain.
      We then applied the same method to explore the potential associations between the skin microenvironment and the measured metabolite concentrations (Figures 3d and Supplementary Figure S2e). As expected, amino acids and tricarboxylic acid‒ and urea cycle‒derived metabolites were mostly associated with more acidic and more hydrated skin. We also noticed a broad distribution of lipid-related metabolites across niches, reflecting the broad spectra of chemical properties of these metabolite classes. Whereas long-chain unsaturated fatty acids tended to be associated with slightly more acidic and drier skin, phospholipids were found at higher concentrations in relatively basic and more hydrated sites. Relatively basic and drier niches were more enriched in ceramides.

       Skin microbiome aggregates around three distinct communities characterized by their metabolite microenvironment

      To investigate the complex relationships between the skin microbiome and metabolome, we applied a regularized Canonical Correlation Analysis at different taxonomic levels: (i) bacterial phyla versus metabolic superpathways, (ii) bacterial genera versus metabolic pathways, and (iii) bacterial species versus metabolites. At the higher taxonomic level, this analysis revealed a strong positive correlation between the abundance of xenobiotics, cofactors, and vitamins and the relative abundance of Actinobacteria as well as a strong anticorrelation between the metabolic superpathways mentioned earlier and Firmicutes (Figure 4a). Zooming in on the genus and metabolic pathway levels (Figure 4b and Supplementary Figure S4a and b) revealed three major clusters: (i) the first one built on the association between Cutibacterium, Acinetobacter, and Corynebacterium in a niche enriched in fatty acid (free fatty acids, monounsaturated fatty acids, and saturated fatty acids), benzoate, tocopherol, and dihydroceramides (Supplementary Figure S4c); (ii) the second one associating Dermacoccus, Agrobacterium, Moraxella, Schaalia, Clostridium, and Staphylococcus with sugars (fructose, mannose), amino acids (leucine, isoleucine), peptides, and vitamin B6 (Supplementary Figure S4d); and (iii) the last one dominated by Streptococcus in a niche independent of any particular correlation with the metabolic pathways mentioned earlier (Figure 4b). The composition of these three communities was characterized in more detail by examining the microbiome and metabolome data at the species and individual metabolite levels (Figure 4c).
      Figure thumbnail gr4
      Figure 4Skin surface microbiome and metabolome are highly entangled. Heatmaps (right) and correlation circles (left) depicting canonical correlations—as defined with regularized generalized canonical correlation analysis—between (a) bacteria phyla and metabolic superpathways, (b) bacteria genera and metabolic pathways, and (c) ASVs and metabolites. For b and c, only correlations above R2 = 0.5 are shown. ASV, amplicon sequence variant; MUFA, monounsaturated fatty acid; SFA, saturated fatty acid.
      To validate the existence of the three clusters, we applied multiomic sparse partial least square unsupervised analysis, integrating microbiome genera abundance data together with metabolome abundance data (Figure 5). Retaining 15 variables in each -omic bloc was sufficient to properly discriminate three clusters of metabolomic and microbe variables (Figure 5a–c and Supplementary Figure S5). The first group of samples (violet cluster) was characterized by an association between fatty acid metabolites and ceramides with Cutibacterium, Actinobacterium, and Bergeyella and is less rich from the microbiome perspective (Figure 5d). The second group of samples (turquoise cluster) was driven by the association between Streptococcus, Porphyroimona, Propionibacterium, Dermacoccus, and Trueperella in a niche mostly independent of the presence of fatty acids, ceramides, sugars, and pyrimidine and is richer from the microbiome perspective (Figure 5d). The third group (green) was characterized by a richer microbiome associating Schaalia, Corynebacterium, Atopobium, Lactobacillus, Clostridium, Escherichia, and Staphylococcus with an environment rich in lysine, sugar, and tricarboxylic acid byproducts. Subjects born vaginally tended to host more frequently the first and the third cluster (Figure 5e). We did not find any further association with other parameters.
      Figure thumbnail gr5
      Figure 5Unsupervised multiblock sPLS analysis on metabolome and microbiome data highlights three distinct clusters. (a) Biclustering of metabolome and microbiome data (row Z-score) with the k-means clustering results overplotted for both individuals and variables and delineating three metabolite‒microbe clusters. (b) Sample plot from the metabolome perspective. (c) Sample plot from the microbiome perspective. (d) Boxplots showing the distribution for pH, SSH, and Chao1 microbiome diversity in the three metabolite‒microbe clusters. (e) Contingency heatmaps showing the association between the three metabolite‒microbe clusters and birth delivery mode. (f) Heatmap depicting top-correlated predicted microbial metabolic activities (Pearson's correlation > 0.8) with Cutibacterium, Staphylococcus, and Streptococcus abundance. NS, not significant; sPLS, sparse partial least square; SSH, skin surface hydration; TCA, tricarboxylic acid.
      Finally, we correlated the abundance of Cutibacterium, Streptococcus, and Staphylococcus with the predicted microbial metabolic pathway potential (Figure 5f and Supplementary Figure S6). Streptococcus contributed mainly to acetylene degradation, galactose degradation, and nucleosides catabolism. Interestingly, the abundance of Staphylococcus was highly correlated with the predicted metabolic activities involved in amino acid degradation (L-arginine and L-glutamate catabolism), whereas the abundance of Cutibacterium was highly correlated with oxides and pyruvate metabolism. Both oxides and pyruvate are important intermediates from the serine synthesis pathway, which is used in combination with palmitoyl-CoA during sphingosine synthesis, the precursor of ceramides.

      Discussion

      Since the late 19th century, the presence of microbes has been associated with diseases. However, mostly through a better understanding of the gastrointestinal tract, we have come to realize that there are commensal and mutualistic species living inside and on us. The particular anatomic location and function of the skin as the interface between the organism and the environment, where microbes are ubiquitous, makes it suitable for microbial colonization. We now understand the skin microbiome as an integral part of the organism's interface with the environment, which among others, restrains potential colonization by opportunistic pathogens. However, the actual mechanisms of microbe‒host interactions and the role of the microbiome in skin physiology remain obscure.
      As is the case for the whole human organism, skin is undergoing dramatic changes after birth. At parturition, the newborn starts its journey experiencing a drastic change from a constant-temperature, wet, and sheltered environment to a dry and highly variable surrounding, potentiating water loss, mechanical trauma, and infections. Development of the epidermal epithelium starts early in utero during the first pregnancy trimester in preparation for the subsequent formation of a functional stratum corneum (
      • Hardman M.J.
      • Moore L.
      • Ferguson M.W.
      • Byrne C.
      Barrier formation in the human fetus is patterned.
      ,
      • Hardman M.J.
      • Sisi P.
      • Banbury D.N.
      • Byrne C.
      Patterned acquisition of skin barrier function during development.
      ). At birth, neonatal skin is still immature relative to adult and gradually follows a maturation process during the first years of life (
      • Mack M.C.
      • Chu M.R.
      • Tierney N.K.
      • Ruvolo Jr., E.
      • Stamatas G.N.
      • Kollias N.
      • et al.
      Water-holding and transport properties of skin stratum corneum of infants and toddlers are different from those of adults: studies in three geographical regions and four ethnic groups.
      ;
      • Nikolovski J.
      • Stamatas G.N.
      • Kollias N.
      • Wiegand B.C.
      Barrier function and water-holding and transport properties of infant stratum corneum are different from adult and continue to develop through the first year of life.
      ;
      • Stamatas G.N.
      • Nikolovski J.
      • Luedtke M.A.
      • Kollias N.
      • Wiegand B.C.
      Infant skin microstructure assessed in vivo differs from adult skin in organization and at the cellular level.
      ). It is now established that infant stratum corneum is thinner (
      • Stamatas G.N.
      • Nikolovski J.
      • Luedtke M.A.
      • Kollias N.
      • Wiegand B.C.
      Infant skin microstructure assessed in vivo differs from adult skin in organization and at the cellular level.
      ;
      • Walters R.M.
      • Khanna P.
      • Chu M.
      • Mack M.C.
      Developmental changes in skin barrier and structure during the first 5 years of life.
      ) and dryer (
      • Hoeger P.H.
      • Enzmann C.C.
      Skin physiology of the neonate and young infant: a prospective study of functional skin parameters during early infancy.
      ;
      • Saijo S.
      • Tagami H.
      Dry skin of newborn infants: functional analysis of the stratum corneum.
      ;
      • Visscher M.O.
      • Chatterjee R.
      • Munson K.A.
      • Pickens W.L.
      • Hoath S.B.
      Changes in diapered and nondiapered infant skin over the first month of life.
      ;
      • Walters R.M.
      • Khanna P.
      • Chu M.
      • Mack M.C.
      Developmental changes in skin barrier and structure during the first 5 years of life.
      ;
      • Yosipovitch G.
      • Maayan-Metzger A.
      • Merlob P.
      • Sirota L.
      Skin barrier properties in different body areas in neonates.
      ) at birth but rapidly getting more hydrated at age 3 months than adult stratum corneum (
      • Mack M.C.
      • Chu M.R.
      • Tierney N.K.
      • Ruvolo Jr., E.
      • Stamatas G.N.
      • Kollias N.
      • et al.
      Water-holding and transport properties of skin stratum corneum of infants and toddlers are different from those of adults: studies in three geographical regions and four ethnic groups.
      ;
      • Nikolovski J.
      • Stamatas G.N.
      • Kollias N.
      • Wiegand B.C.
      Barrier function and water-holding and transport properties of infant stratum corneum are different from adult and continue to develop through the first year of life.
      ;
      • Stamatas G.N.
      • Nikolovski J.
      • Luedtke M.A.
      • Kollias N.
      • Wiegand B.C.
      Infant skin microstructure assessed in vivo differs from adult skin in organization and at the cellular level.
      ). Infant corneocytes are smaller (
      • Stamatas G.N.
      • Nikolovski J.
      • Luedtke M.A.
      • Kollias N.
      • Wiegand B.C.
      Infant skin microstructure assessed in vivo differs from adult skin in organization and at the cellular level.
      ), their collagen fibers are less dense (
      • Stamatas G.N.
      • Nikolovski J.
      • Luedtke M.A.
      • Kollias N.
      • Wiegand B.C.
      Infant skin microstructure assessed in vivo differs from adult skin in organization and at the cellular level.
      ), and there is a lower concentration of natural moisturizing factors (
      • Nikolovski J.
      • Stamatas G.N.
      • Kollias N.
      • Wiegand B.C.
      Barrier function and water-holding and transport properties of infant stratum corneum are different from adult and continue to develop through the first year of life.
      ) and lipids in the infant stratum corneum than in the adult. These parameters directly impact the skin barrier function and the physicochemical properties of the skin.
      Exploiting the skin microbiome to develop innovative topical treatments to treat skin conditions requires detailed knowledge of the crosstalk between the microbial community and the host, which is currently lacking. To fill this gap, we used a multidimensional approach combining 16S RNA sequencing and untargeted metabolomics in samples taken from the healthy infant skin surface. We further applied state-of-the-art dimension reduction methodologies to better understand how the microbiome shapes and is being shaped by the skin microenvironment in healthy conditions.
      Despite a relatively homogeneous distribution of the major microbial phyla and the metabolic superpathways, a more granular analysis of these two components revealed a substantial heterogeneity between samples. Whereas amino acids, lipids, and xenobiotics were dominating together with Firmicutes, Actinobacteria, and Proteobacteria, as already has been shown in neonates (
      • Chu D.M.
      • Ma J.
      • Prince A.L.
      • Antony K.M.
      • Seferovic M.D.
      • Aagaard K.M.
      Maturation of the infant microbiome community structure and function across multiple body sites and in relation to mode of delivery.
      ), zooming in to lower taxonomic levels revealed a large contribution of commensals belonging to the Cutibacterium and Staphylococcus genera, including species such as C. acnes, Sta. epidermidis, Sta. aureus, Sta. hominis, and Str. pneumoniae. As reported in other works, our data showed that species commonly driving dysbiosis exist even in healthy skin (
      • Byrd A.L.
      • Deming C.
      • Cassidy S.K.B.
      • Harrison O.J.
      • Ng W.I.
      • Conlan S.
      • et al.
      Staphylococcus aureus and Staphylococcus epidermidis strain diversity underlying pediatric atopic dermatitis.
      ;
      • Chen Y.E.
      • Fischbach M.A.
      • Belkaid Y.
      Skin microbiota-host interactions [published correction appears in Nature 2018;555:543].
      ;
      • Nakatsuji T.
      • Chen T.H.
      • Narala S.
      • Chun K.A.
      • Two A.M.
      • Yun T.
      • et al.
      Antimicrobials from human skin commensal bacteria protect against Staphylococcus aureus and are deficient in atopic dermatitis.
      ).
      This heterogeneity is further reflected in the association between the microbiome and the metabolome at the skin surface. Integrative analyses indeed enabled us to delineate the existence of three distinct metabolite‒microbe clusters at the skin surface in infants: (i) one built on the association between Cutibacterium, Actinomyces, and Bergeyella in individuals with ceramide- and lipid-rich, relatively drier and basic skin surface; (ii) one consisting of the association of multiple commensals such as Corynebacterium, Lactobacillus, Clostridium, Escherichia, Pseudomonas, and Staphylococcus in individuals with a lysine- and sugar-rich, relatively moistened and more acidic skin surface; and (iii) one anticorrelated or independent of a particular metabolite microenvironment.
      C. acnes is a major skin commensal and is the dominating species of the pilosebaceous gland, accounting for up to 90% of the total microbiome in sebum-rich sites such as the scalp or the face (
      • Grice E.A.
      • Kong H.H.
      • Conlan S.
      • Deming C.B.
      • Davis J.
      • Young A.C.
      • et al.
      Topographical and temporal diversity of the human skin microbiome.
      ). Although accumulating evidence shows its role in enhancing sebaceous gland lipogenesis and triglycerides synthesis in vitro and in vivo (
      • Iinuma K.
      • Sato T.
      • Akimoto N.
      • Noguchi N.
      • Sasatsu M.
      • Nishijima S.
      • et al.
      Involvement of Propionibacterium acnes in the augmentation of lipogenesis in hamster sebaceous glands in vivo and in vitro.
      ), its interplay with stratum corneum lipid metabolism remains elusive. Our data showed that C. acnes had a greater affinity for lipid-rich skin surface and accumulated at sites with greater amounts of fatty acids (2-hydroxystearate, 2-hydroxypalmitate, myristoleate, arachidate, palmitoleate), cholesterol, and ceramides (N-palmitoyl-sphinganine, N-palmitoyl-sphingosine, N-2-hydroxypalmitoyl-sphingosine, N-stearoyl-D-sphingosine, N-arachidoyl-D-sphingosine). Whether organized into broad bilayers in the intercorneocyte spaces or covalently bound to the corneocyte envelope in the stratum corneum, lipids are essential constituents of the human epidermis, supporting skin barrier function, cell signaling, and antimicrobial defense (
      • van Smeden J.
      • Janssens M.
      • Gooris G.S.
      • Bouwstra J.A.
      The important role of stratum corneum lipids for the cutaneous barrier function.
      ). Considering the importance of lipids in skin barrier function as well as the role of C. acnes in acne vulgaris, these results are of utmost relevance.
      Sta. aureus is known to be involved in the pathology of atopic dermatitis (
      • Leyden J.J.
      • Marples R.R.
      • Kligman A.M.
      Staphylococcus aureus in the lesions of atopic dermatitis.
      ). In fact, Sta. aureus typically dominates the microbiome composition on atopic lesions and is responsible for the observed decline in the overall microbiome diversity (
      • Kong H.H.
      • Oh J.
      • Deming C.
      • Conlan S.
      • Grice E.A.
      • Beatson M.A.
      • et al.
      Temporal shifts in the skin microbiome associated with disease flares and treatment in children with atopic dermatitis.
      ). This species relies on the branched-chain amino acids (isoleucine, leucine, valine) for the synthesis of proteins and membrane branched-chain fatty acids. These amino acids are therefore crucial for the species metabolism, adaptation, and virulence (
      • Kaiser J.C.
      • King A.N.
      • Grigg J.C.
      • Sheldon J.R.
      • Edgell D.R.
      • Murphy M.E.P.
      • et al.
      Repression of branched-chain amino acid synthesis in Staphylococcus aureus is mediated by isoleucine via CodY, and by a leucine-rich attenuator peptide.
      ).
      Untargeted metabolomic profiling of skin samples remains challenging and expensive and compelled us to focus on a restricted number of samples. To overcome this limitation, we designed relevant processing and analytical strategies, mostly relying on the nonparametric statistical framework. Furthermore, we examined the datasets at the pathway/superpathway and genus/phylum levels, avoiding potential problems arising from simple pairwise correlations computed on isolated and sparse variables.
      The microbial biomass of the skin is drastically lower than that in the gut, leading to a high host-to-microbe DNA content ratio in the collected samples. To cope with the host DNA contamination that represents up to 90% in skin swab samples (
      • Bjerre R.D.
      • Hugerth L.W.
      • Boulund F.
      • Seifert M.
      • Johansen J.D.
      • Engstrand L.
      Effects of sampling strategy and DNA extraction on human skin microbiome investigations.
      ), shotgun metagenomic sequencing should be performed at appropriate depth—which remains costly to date—to provide useful data on the microbial community. Despite being less resolutive from the phylogenic standpoint, 16S ribosomal RNA sequencing offers a more affordable alternative, enabling accurate characterization of the microbial community at the phylum and genus level and avoiding issues of host DNA contamination. We therefore opted for 16S ribosomal RNA profiling in this study.
      Metabolomic analysis of skin surface samples has been used in previous works to supplement skin microbiome data (
      • Bouslimani A.
      • da Silva R.
      • Kosciolek T.
      • Janssen S.
      • Callewaert C.
      • Amir A.
      • et al.
      The impact of skin care products on skin chemistry and microbiome dynamics.
      ,
      • Bouslimani A.
      • Porto C.
      • Rath C.M.
      • Wang M.
      • Guo Y.
      • Gonzalez A.
      • et al.
      Molecular cartography of the human skin surface in 3D.
      ). In this work, we aimed to integrate the statistical analysis of the two -omics datasets to unlock insights focusing on the cross-talk between metabolome and microbiome on healthy infant skin.
      This study provides insights on the interplay between metabolome and microbiome on healthy skin and opens new directions in research focusing on the association of ceramides, fatty acids, and Cutibacterium sp. but also relating to longitudinal studies focusing on the evolution of the diverse metabolite‒microbe clusters as characterized in this report. It remains to be seen whether the three skin microbial communities identified in this work persist in puberty and adulthood and whether they are predictive of—or preclusive to—pathophysiological outcomes later in life.

      Materials and Methods

       Clinical study, measurements, and sample collection

      A single-center, randomized, evaluator-blind, 5-week trial (NCT03457857) was conducted to assess the effects of two skincare regimens on the cutaneous microbiome, metabolome, and skin physiology of healthy infants aged between 3 and 6 months in general good health on the basis of medical history and without any skin conditions or family history of known allergies. In this report, we used only the baseline data to assess the cross-talk between the microbiome, metabolome, and skin physiology. An institutional review board (IntegReview, Austin, TX) approved the study, and parents/legally authorized representatives (LARs) of study participants provided written informed consent. Parents/LARs of prospective participants were screened for eligibility criteria using an institutional review board‒approved screener. Parents/LARs were required to be aged at least 18 years. Participant eligibility was assessed at an initial screening visit by the primary investigator, and the study physician confirmed the eligibility of each participant before enrollment. All eligible study participants entered a 7-day washout period, during which parents/LARs were instructed to use a marketed gentle baby cleanser (Johnson’s Head-To-Toe Wash & Shampoo: Johnson & Johnson Consumer, Skillman, NJ) in place of their infant’s normal body cleanser, at least three times during the week, and to refrain from the use of any type of moisturizer or lotion. They were also instructed not to bathe or cleanse the children for at least 12 hours before the scheduled visit. Sample collection from the left or right dorsal forearm was determined by randomization, with one arm used for skin swabs for microbiome analysis and skin tape samples for metabolomic analysis and the opposite arm used for SSH and skin pH readings. SSH was assessed using a Corneometer CM 825 (Courage+Khazaka Electronic, Cologne, Germany) using three consecutive readings from the subject’s dorsal forearm. Skin pH measurements were obtained from five consecutive readings within each test site on the subject’s dorsal forearm using a Skin-pH-Meter PH 905 (Courage+Khazaka Electronic). For microbiome collection, a 1 cm × 4 cm skin area was sampled by swabbing with a polyester-tipped sterile swab, which was first dipped into deionized water and wrung of excess liquid. The lateral edge of the swab was rubbed back and forth in a crosswise manner in the defined area for 30 seconds. The head of each swab was placed into a sterile microcentrifuge tube and aseptically cut from the handle before closing the tube lid. Samples were frozen at –80 °C until shipment on dry ice for DNA extraction and high throughput sequencing. Gloves were worn at all times when handling the swabs and when sampling the microflora to prevent possible contamination. Skin swab samples together with blank swabs (used as negative controls) were sent to an independent laboratory (RTL Genomics, Lubbock, TX) for DNA extraction and sequencing of the skin microflora. Two consecutive skin tape samples were collected from the dorsal forearm, adjacent to the site used for microbial sample collection. Samples were collected using D-Squame Standard Sampling Discs (CuDerm, Dallas, TX) applying 30 seconds of constant pressure. The tape was then removed with forceps and placed into a scintillation vial (adhesive side in) and immediately stored at –80 oC. Metabolomic analysis of the collected samples and blank tapes (negative controls) was performed by an independent laboratory (Metabolon, Morrisville, NC).

       Microbiome profiling

      To profile skin microbiota, sequencing was conducted by RTL Genomics. Briefly, DNA was extracted using MagAttract PowerSoil DNA Isolation (Qiagen, Hilden, Germany) on the KingFisher 96-well extraction robot (Thermo Fisher Scientific, Waltham, MA) following the manufacturer’s instructions. Sample amplification for sequencing was conducted using primers encompassing variable regions 1 through 3 (28 forward: GAGTTTGATCNTGGCTCAG, 519 reverse: GTNTTACNGCGGCKGCTG) of the 16S ribosomal RNA gene as previously described (
      • Phillips C.D.
      • Hanson J.
      • Wilkinson J.E.
      • Koenig L.
      • Rees E.
      • Webala P.
      • et al.
      Microbiome structural and functional interactions across host dietary niche space.
      ). Sequencing was conducted on the Illumina MiSeq platform (Illumina, San Diego, CA) according to the manufacture instructions and targeting a minimum depth of 10,000 taxonomically classified reads per sample, a threshold defined by rarefaction analysis (Supplementary Figure S7). Raw paired-end sequencing reads were first merged using custom R script, and PCR primers were removed from the obtained sequences. These sequences were further quality trimmed, filtered and denoised using DADA2 framework (
      • Callahan B.J.
      • McMurdie P.J.
      • Rosen M.J.
      • Han A.W.
      • Johnson A.J.
      • Holmes S.P.
      DADA2: high-resolution sample inference from Illumina amplicon data.
      ) to infer ASVs. Among the 1,647,259 read pairs generated, 1,071,553 were kept. Taxonomy was assigned using the HiMAP National Center for Biotechnology Information‒derived database (Segota and Long, 2019
      Segota I, Long T. A high-resolution pipeline for 16S-sequencing identifies bacterial strains in human microbiome. bioRxiv 2019.
      ). ASV abundance matrix, sample metainformation, and taxonomy were finally stored as a phyloseq object (
      • McMurdie P.J.
      • Holmes S.
      phyloseq: an R package for reproducible interactive analysis and graphics of microbiome census data.
      ). ASVs detected in less than two samples were excluded from the analysis. Microbiome analysis of the negative control (blank swab) showed no relevant contamination (only one ASV detected from the blank swab, among the 1,328 detected in the global experimental design). The ASV count matrix was finally used as an input for PICRUSt2 to define microbiome metabolic pathway potential (Douglas et al., 2019
      Douglas GM, Maffei VJ, Zaneveld J, Yurgel SN, Brown JR, Taylor CM, et al. PICRUSt2: an improved and extensible approach for metagenome inference. bioRxiv 2019.
      ).

       Metabolomics

      Untargeted metabolomic analysis of skin tape samples was performed by Metabolon (Durham, NC), as previously described (
      • Evans A.M.
      • DeHaven C.D.
      • Barrett T.
      • Mitchell M.
      • Milgram E.
      Integrated, nontargeted ultrahigh performance liquid chromatography/electrospray ionization tandem mass spectrometry platform for the identification and relative quantification of the small-molecule complement of biological systems.
      ). Individual compounds were identified by comparing the mass spectroscopy data of the samples with a library of >4,500 authenticated purified standards. The library includes information on the retention-time index, mass-to-charge ratio, and chromatographic data (including tandem mass spectrometry data) on all molecular entries. In a given sample, the peak intensities corresponding to each metabolite were normalized to the total intensity count for that sample.

       Statistical analyses

      All statistical analyses were performed in R, version 4.0.0, and rely on the packages mixOmics (
      • Rohart F.
      • Gautier B.
      • Singh A.
      • Lê Cao K.A.
      mixOmics: an R package for ‘omics feature selection and multiple data integration.
      ), FactorMineR (
      • Lê S.
      • Josse J.
      • Husson F.
      FactoMineR: an R package for multivariate analysis.
      ), vegan, and phyloseq (
      • McMurdie P.J.
      • Holmes S.
      phyloseq: an R package for reproducible interactive analysis and graphics of microbiome census data.
      ). Factorial analysis of mixed data was applied on a matrix containing pH, SSH, microbiome Chao1 index, as well as sex and mode of birth information for each sample. Regularized Canonical Correlation Analysis was performed on the combination of the metabolomic abundance matrix and the microbiome relative abundance matrix after regularization through Ridge regression (ℓ2 penalties) of parameters λ1 and λ2 using a leave-one-out cross-validation procedure. To define metabolite‒microbe clusters, a block sparse Partial Least Square analysis was applied on the combination of the metabolomic abundance matrix (pathway level) and the microbiome relative abundance matrix (genera level) after fine tuning the numbers of dimensions and variables to select using a k-fold cross-validation procedure. The samples and the selected variables were then clustered using k-means biclustering. The optimal number of sample clusters was defined using the gap statistic. When relevant, group comparisons were performed using nonparametric Kruskal‒Wallis and Wilcoxon‒Mann‒Whitney rank-sum tests, and a P-value threshold cutoff at 0.05 was considered. Group associations with descriptive parameters were evaluated using the Chi-square test. Correlations were evaluated using Pearson’s correlation and Spearman’s rank method.

       Ethical statements

      An institutional review board (IntegReview) approved the study (NCT03457857), and parents/LARs of study participants provided written informed consent.

       Data availability statement

      Raw data are available at the Sequence Read Archive (Bioproject: PRJNA707369). Processed data are available in Supplementary Table S2 (metabolome) and Supplementary Table S3 (microbiome).

      ORCIDs

      Conflict of Interest

      All authors are employees of Johnson & Johnson Santé Beauté France.

      Acknowledgments

      We thank the volunteers who participated in this study as well as Kimberly Capone, Lorena Telofski, Janeta Nikolovski, and Diana Friscia for managing the clinical study.

      Author Contributions

      Conceptualization: TO, GS; Data Curation: PFR; Formal Analysis: PFR; Funding Acquisition: TO, GS; Investigation: TO, GS, PFR; Methodology: GS, PFR; Supervision: TO, GS; Writing - Original Draft Preparation: TO, GS, PFR; Writing - Review and Editing: TO, GS, PFR

      Supplementary Material

      Supplementary Materials

      Figure thumbnail fx1
      Supplementary Figure S1Overview of the healthy surface skin microbiome and metabolome. Barplots depicting the weight of (a) core metabolites with RA > 1.4% in 16 samples, (b) core metabolites with RA > 3% in eight samples, (c) top 20 contributing metabolites, and (d) core microbial genus with RA > 1% in eight samples. The bars on the left of each graph show the average distribution across samples. RA, relative abundance.
      Figure thumbnail fx2
      Supplementary Figure S2Skin surface microbiome and metabolome are highly entangled. (a–c) Boxplots highlighting the relationships between birth delivery mode and (a) Chao1 diversity, (b) pH, and (c) SSH. (d) Dotplots depicting the correlation between SSH (green) and pH (red) and Pseudomonas, Granulicatella, and Cutibacterium abundance. The red and green lines correspond to the linear regression for pH (red) and SSH (green). (e) Dotplots depicting the correlation between SSH (green) and pH (red) and urea cycle‒related metabolites, ceramides, and long-chain PUFA. The red and green lines correspond to the linear regression for pH (red) and SSH (green). PUFA, polyunsaturated fatty acid; SSH, surface skin hydration.
      Figure thumbnail fx3
      Supplementary Figure S3Metabolomic diversity versus SSH and pH. Scatter plots showing the distribution of metabolomic Shannon diversity index as a function of the SSH (left) and pH (right). RA, relative abundance; SSH, skin surface hydration.
      Figure thumbnail fx4
      Supplementary Figure S4Top correlated metabolites from the lipid category for Cutibacterium and from the amino acid category for Staphylococcus. (a–b) Hierarchical biclustering based on pairwise (a) Pearson’s correlation or (b) Spearman’s correlation between microbial and metabolite abundances summarized at the genus and the pathway level. Red boxes delineate the outstanding associations that are conserved across methods. (c–d) Dotplots showing the top correlated metabolites with (c) Cutibacterium RA and (d) Staphylococcus RA. RA, relative abundance.
      Figure thumbnail fx5
      Supplementary Figure S5Unsupervised multiblock sparse partial least square analysis on metabolome and microbiome data highlights three distinct clusters. Hierarchical biclustering of metabolome and microbiome data (row Z-score) with the k-means clustering results from a overplotted for both individuals and variables. ASV, amplicon sequence variant.
      Figure thumbnail fx6
      Supplementary Figure S6Sparse partial least sPLS-DA on PICRUSt2 metabolic pathway potential data. Biclustering of PICRUSt2 metabolic pathway potential data (row Z-score) with the k-means clustering results overplotted for individuals. The shortlist of MetaCyc biological pathways was selected using an sPLS-DA. sPLS-DA, sparse partial least square discriminant analysis.
      Figure thumbnail fx7
      Supplementary Figure S7Rarefaction analysis on microbiome data. Rarefaction curves showing the number of expected ASVs detected as a function of the subsampled reads. ASV, amplicon sequence variant; SSH, skin surface hydration.

      References

        • Agak G.W.
        • Qin M.
        • Nobe J.
        • Kim M.H.
        • Krutzik S.R.
        • Tristan G.R.
        • et al.
        Propionibacterium acnes induces an IL-17 response in acne vulgaris that is regulated by vitamin A and vitamin D.
        J Invest Dermatol. 2014; 134: 366-373
        • Barnard E.
        • Shi B.
        • Kang D.
        • Craft N.
        • Li H.
        The balance of metagenomic elements shapes the skin microbiome in acne and health [published correction appears in Sci Rep 2020;10:6037].
        Sci Rep. 2016; 6: 39491
        • Bjerre R.D.
        • Hugerth L.W.
        • Boulund F.
        • Seifert M.
        • Johansen J.D.
        • Engstrand L.
        Effects of sampling strategy and DNA extraction on human skin microbiome investigations.
        Sci Rep. 2019; 9: 17287
        • Bouslimani A.
        • da Silva R.
        • Kosciolek T.
        • Janssen S.
        • Callewaert C.
        • Amir A.
        • et al.
        The impact of skin care products on skin chemistry and microbiome dynamics.
        BMC Biol. 2019; 17: 47
        • Bouslimani A.
        • Porto C.
        • Rath C.M.
        • Wang M.
        • Guo Y.
        • Gonzalez A.
        • et al.
        Molecular cartography of the human skin surface in 3D.
        Proc Natl Acad Sci USA. 2015; 112: E2120-E2129
        • Byrd A.L.
        • Deming C.
        • Cassidy S.K.B.
        • Harrison O.J.
        • Ng W.I.
        • Conlan S.
        • et al.
        Staphylococcus aureus and Staphylococcus epidermidis strain diversity underlying pediatric atopic dermatitis.
        Sci Transl Med. 2017; 9eaal4651
        • Callahan B.J.
        • McMurdie P.J.
        • Rosen M.J.
        • Han A.W.
        • Johnson A.J.
        • Holmes S.P.
        DADA2: high-resolution sample inference from Illumina amplicon data.
        Nat Methods. 2016; 13: 581-583
        • Chen M.X.
        • Wang S.Y.
        • Kuo C.H.
        • Tsai I.L.
        Metabolome analysis for investigating host-gut microbiota interactions.
        J Formos Med Assoc. 2019; 118: S10-S22
        • Chen Y.E.
        • Fischbach M.A.
        • Belkaid Y.
        Skin microbiota-host interactions [published correction appears in Nature 2018;555:543].
        Nature. 2018; 553: 427-436
        • Chu D.M.
        • Ma J.
        • Prince A.L.
        • Antony K.M.
        • Seferovic M.D.
        • Aagaard K.M.
        Maturation of the infant microbiome community structure and function across multiple body sites and in relation to mode of delivery.
        Nat Med. 2017; 23: 314-326
        • Dominguez-Bello M.G.
        • Costello E.K.
        • Contreras M.
        • Magris M.
        • Hidalgo G.
        • Fierer N.
        • et al.
        Delivery mode shapes the acquisition and structure of the initial microbiota across multiple body habitats in newborns.
        Proc Natl Acad Sci USA. 2010; 107: 11971-11975
        • Dominguez-Bello M.G.
        • De Jesus-Laboy K.M.
        • Shen N.
        • Cox L.M.
        • Amir A.
        • Gonzalez A.
        • et al.
        Partial restoration of the microbiota of cesarean-born infants via vaginal microbial transfer.
        Nat Med. 2016; 22: 250-253
        • Evans A.M.
        • DeHaven C.D.
        • Barrett T.
        • Mitchell M.
        • Milgram E.
        Integrated, nontargeted ultrahigh performance liquid chromatography/electrospray ionization tandem mass spectrometry platform for the identification and relative quantification of the small-molecule complement of biological systems.
        Anal Chem. 2009; 81: 6656-6667
        • Fitz-Gibbon S.
        • Tomida S.
        • Chiu B.H.
        • Nguyen L.
        • Du C.
        • Liu M.
        • et al.
        Propionibacterium acnes strain populations in the human skin microbiome associated with acne.
        J Invest Dermatol. 2013; 133: 2152-2160
        • Gao Z.
        • Tseng C.H.
        • Strober B.E.
        • Pei Z.
        • Blaser M.J.
        Substantial alterations of the cutaneous bacterial biota in psoriatic lesions.
        PLoS One. 2008; 3: e2719
        • Grice E.A.
        • Kong H.H.
        • Conlan S.
        • Deming C.B.
        • Davis J.
        • Young A.C.
        • et al.
        Topographical and temporal diversity of the human skin microbiome.
        Science. 2009; 324: 1190-1192
        • Hardman M.J.
        • Moore L.
        • Ferguson M.W.
        • Byrne C.
        Barrier formation in the human fetus is patterned.
        J Invest Dermatol. 1999; 113: 1106-1113
        • Hardman M.J.
        • Sisi P.
        • Banbury D.N.
        • Byrne C.
        Patterned acquisition of skin barrier function during development.
        Development. 1998; 125: 1541-1552
        • Hertel J.
        • Harms A.C.
        • Heinken A.
        • Baldini F.
        • Thinnes C.C.
        • Glaab E.
        • et al.
        Integrated analyses of microbiome and longitudinal metabolome data reveal microbial-host interactions on sulfur metabolism in Parkinson’s disease.
        Cell Rep. 2019; 29: 1767-1777.e8
        • Hoeger P.H.
        • Enzmann C.C.
        Skin physiology of the neonate and young infant: a prospective study of functional skin parameters during early infancy.
        Pediatr Dermatol. 2002; 19: 256-262
        • Iinuma K.
        • Sato T.
        • Akimoto N.
        • Noguchi N.
        • Sasatsu M.
        • Nishijima S.
        • et al.
        Involvement of Propionibacterium acnes in the augmentation of lipogenesis in hamster sebaceous glands in vivo and in vitro.
        J Invest Dermatol. 2009; 129: 2113-2119
        • Kaiser J.C.
        • King A.N.
        • Grigg J.C.
        • Sheldon J.R.
        • Edgell D.R.
        • Murphy M.E.P.
        • et al.
        Repression of branched-chain amino acid synthesis in Staphylococcus aureus is mediated by isoleucine via CodY, and by a leucine-rich attenuator peptide.
        PLoS Genet. 2018; 14e1007159
        • Kong H.H.
        • Oh J.
        • Deming C.
        • Conlan S.
        • Grice E.A.
        • Beatson M.A.
        • et al.
        Temporal shifts in the skin microbiome associated with disease flares and treatment in children with atopic dermatitis.
        Genome Res. 2012; 22: 850-859
        • Kong H.H.
        • Segre J.A.
        Skin microbiome: looking back to move forward.
        J Invest Dermatol. 2012; 132: 933-939
        • Lavelle A.
        • Sokol H.
        Gut microbiota-derived metabolites as key actors in inflammatory bowel disease.
        Nat Rev Gastroenterol Hepatol. 2020; 17: 223-237
        • Lê S.
        • Josse J.
        • Husson F.
        FactoMineR: an R package for multivariate analysis.
        J Stat Softw. 2008; 25: 1-18
        • Leyden J.J.
        • Marples R.R.
        • Kligman A.M.
        Staphylococcus aureus in the lesions of atopic dermatitis.
        Br J Dermatol. 1974; 90: 525-530
        • Liu H.
        • Chen X.
        • Hu X.
        • Niu H.
        • Tian R.
        • Wang H.
        • et al.
        Alterations in the gut microbiome and metabolism with coronary artery disease severity.
        Microbiome. 2019; 7: 68
        • Liu Y.
        • Wang Y.
        • Ni Y.
        • Cheung C.K.Y.
        • Lam K.S.L.
        • Wang Y.
        • et al.
        Gut microbiome fermentation determines the efficacy of exercise for diabetes prevention.
        Cell Metab. 2020; 31: 77-91.e5
        • Mack M.C.
        • Chu M.R.
        • Tierney N.K.
        • Ruvolo Jr., E.
        • Stamatas G.N.
        • Kollias N.
        • et al.
        Water-holding and transport properties of skin stratum corneum of infants and toddlers are different from those of adults: studies in three geographical regions and four ethnic groups.
        Pediatr Dermatol. 2016; 33: 275-282
        • Maruvada P.
        • Leone V.
        • Kaplan L.M.
        • Chang E.B.
        The human microbiome and obesity: moving beyond associations.
        Cell Host Microbe. 2017; 22: 589-599
        • McMurdie P.J.
        • Holmes S.
        phyloseq: an R package for reproducible interactive analysis and graphics of microbiome census data.
        PLoS One. 2013; 8: e61217
        • Mukherjee S.
        • Mitra R.
        • Maitra A.
        • Gupta S.
        • Kumaran S.
        • Chakrabortty A.
        • et al.
        Sebum and hydration levels in specific regions of human face significantly predict the nature and diversity of facial skin microbiome.
        Sci Rep. 2016; 6: 36062
        • Naik S.
        • Bouladoux N.
        • Wilhelm C.
        • Molloy M.J.
        • Salcedo R.
        • Kastenmuller W.
        • et al.
        Compartmentalized control of skin immunity by resident commensals.
        Science. 2012; 337: 1115-1119
        • Nakatsuji T.
        • Chen T.H.
        • Narala S.
        • Chun K.A.
        • Two A.M.
        • Yun T.
        • et al.
        Antimicrobials from human skin commensal bacteria protect against Staphylococcus aureus and are deficient in atopic dermatitis.
        Sci Transl Med. 2017; 9eaah4680
        • Nikolovski J.
        • Stamatas G.N.
        • Kollias N.
        • Wiegand B.C.
        Barrier function and water-holding and transport properties of infant stratum corneum are different from adult and continue to develop through the first year of life.
        J Invest Dermatol. 2008; 128: 1728-1736
        • Oh J.
        • Byrd A.L.
        • Park M.
        • Kong H.H.
        • Segre J.A.
        • NISC Comparative Sequencing Program
        Temporal stability of the human skin microbiome.
        Cell. 2016; 165: 854-866
        • Phillips C.D.
        • Hanson J.
        • Wilkinson J.E.
        • Koenig L.
        • Rees E.
        • Webala P.
        • et al.
        Microbiome structural and functional interactions across host dietary niche space.
        Integr Comp Biol. 2017; 57: 743-755
        • Rohart F.
        • Gautier B.
        • Singh A.
        • Lê Cao K.A.
        mixOmics: an R package for ‘omics feature selection and multiple data integration.
        PLoS Comput Biol. 2017; 13e1005752
        • Roth R.R.
        • James W.D.
        Microbial ecology of the skin.
        Annu Rev Microbiol. 1988; 42: 441-464
        • Saijo S.
        • Tagami H.
        Dry skin of newborn infants: functional analysis of the stratum corneum.
        Pediatr Dermatol. 1991; 8: 155-159
        • Schmidt T.S.B.
        • Raes J.
        • Bork P.
        The human gut microbiome: from association to modulation.
        Cell. 2018; 172: 1198-1215
        • Shaffer M.
        • Armstrong A.J.S.
        • Phelan V.V.
        • Reisdorph N.
        • Lozupone C.A.
        Microbiome and metabolome data integration provides insight into health and disease.
        Transl Res. 2017; 189: 51-64
        • Sharon G.
        • Cruz N.J.
        • Kang D.W.
        • Gandal M.J.
        • Wang B.
        • Kim Y.M.
        • et al.
        Human gut microbiota from autism spectrum disorder promote behavioral symptoms in mice.
        Cell. 2019; 177: 1600-1618.e17
        • Stamatas G.N.
        • Nikolovski J.
        • Luedtke M.A.
        • Kollias N.
        • Wiegand B.C.
        Infant skin microstructure assessed in vivo differs from adult skin in organization and at the cellular level.
        Pediatr Dermatol. 2010; 27: 125-131
        • Valles-Colomer M.
        • Falony G.
        • Darzi Y.
        • Tigchelaar E.F.
        • Wang J.
        • Tito R.Y.
        • et al.
        The neuroactive potential of the human gut microbiota in quality of life and depression.
        Nat Microbiol. 2019; 4: 623-632
        • van Rensburg J.J.
        • Lin H.
        • Gao X.
        • Toh E.
        • Fortney K.R.
        • Ellinger S.
        • et al.
        The human skin microbiome associates with the outcome of and is influenced by bacterial infection.
        mBio. 2015; 6 (e01315–15)
        • van Smeden J.
        • Janssens M.
        • Gooris G.S.
        • Bouwstra J.A.
        The important role of stratum corneum lipids for the cutaneous barrier function.
        Biochim Biophys Acta. 2014; 1841: 295-313
        • Visscher M.O.
        • Chatterjee R.
        • Munson K.A.
        • Pickens W.L.
        • Hoath S.B.
        Changes in diapered and nondiapered infant skin over the first month of life.
        Pediatr Dermatol. 2000; 17: 45-51
        • Vojinovic D.
        • Radjabzadeh D.
        • Kurilshikov A.
        • Amin N.
        • Wijmenga C.
        • Franke L.
        • et al.
        Relationship between gut microbiota and circulating metabolites in population-based cohorts.
        Nat Commun. 2019; 10: 5813
        • Walters R.M.
        • Khanna P.
        • Chu M.
        • Mack M.C.
        Developmental changes in skin barrier and structure during the first 5 years of life.
        Skin Pharmacol Physiol. 2016; 29: 111-118
        • Ward T.L.
        • Dominguez-Bello M.G.
        • Heisel T.
        • Al-Ghalith G.
        • Knights D.
        • Gale C.A.
        Development of the human mycobiome over the first month of life and across body sites.
        mSystems. 2018; 3 (e00140–17)
        • Yosipovitch G.
        • Maayan-Metzger A.
        • Merlob P.
        • Sirota L.
        Skin barrier properties in different body areas in neonates.
        Pediatrics. 2000; 106: 105-108