Cobamide Sharing Is Predicted in the Human Skin Microbiome

Abstract

The skin microbiome is a key player in human health, with diverse functions ranging from defense against pathogens to education of the immune system. While recent studies have begun to shed light on the valuable role that skin microorganisms have in maintaining the skin barrier, a detailed understanding of the complex interactions that shape healthy skin microbial communities is limited. Cobamides, the vitamin B₁₂ class of cofactor, are essential for organisms across the tree of life. Because this vitamin is only produced by a limited fraction of prokaryotes, cobamide sharing is predicted to mediate community dynamics within microbial communities. Here, we provide the first large-scale metagenomic assessment of cobamide biosynthesis and utilization in the skin microbiome. We show that while numerous and diverse taxa across the major bacterial phyla on the skin encode cobamide-dependent enzymes, relatively few species encode de novo cobamide biosynthesis. We show that cobamide producers and users are integrated into the network structure of microbial communities across the different microenvironments of the skin and that changes in microbiome community structure and diversity are associated with the abundance of cobamide producers in the Corynebacterium genus, for both healthy and diseased skin states. Finally, we find that de novo cobamide biosynthesis is enriched only in Corynebacterium species associated with hosts, including those prevalent on human skin. We confirm that the cofactor is produced in excess through quantification of cobamide production by human skin-associated species isolated in the laboratory. Taken together, our results reveal the potential for cobamide sharing within skin microbial communities, which we hypothesize mediates microbiome community dynamics and host interactions. IMPORTANCE The skin microbiome is essential for maintaining skin health and function. However, the microbial interactions that dictate microbiome structure, stability, and function are not well understood. Here, we investigate the biosynthesis and use of cobamides, a cofactor needed by many organisms but only produced by select prokaryotes, within the human skin microbiome. We found that while a large proportion of skin taxa encode cobamide-dependent enzymes, only a select few encode de novo cobamide biosynthesis. Further, the abundance of cobamide-producing Corynebacterium species is associated with skin microbiome diversity and structure, and within this genus, de novo biosynthesis is enriched in host-associated species compared to environment-associated species. These findings identify cobamides as a potential mediator of skin microbiome dynamics and skin health.

Keywords: cobamides; ecology; genomics; metagenomics; skin microbiome.

FIG 1

De novo cobamide biosynthesis is limited to distinct taxa on the skin. (A) Simplified de novo cobamide biosynthesis pathway. Subsections of the pathway are indicated by color, with gene names and white boxes indicating each enzymatic step. Aerobic and anaerobic corrin ring synthesis pathways contain orthologous enzymes that are indicated with dashed lines. hemL in parentheses is required for synthesis from glutamate. (B) Taxon contribution reflects the proportion of normalized cobamide biosynthesis gene hits assigned to each taxon out of the total normalized cobamide biosynthesis gene hits within a sample. Normalization was performed by dividing hits to each gene by its profile HMM length. Taxon contributions are shown for the top 6 taxa, grouped by skin site. Color indicates microenvironment classification. (C) The top 40 most abundant bacterial species within the data set were determined by totaling the hits to single-copy gene rpoB for each species. The remaining species were grouped into “Other.” Species names that are not yet validly published under the International Code of Nomenclature of Prokaryotes (ICNP) are indicated in parentheses. Individual values in the heatmap represent the number of hits assigned to the species for a particular cobamide biosynthesis gene divided by the total number of hits to the gene. Gene hits were normalized by profile HMM length and sequencing depth prior to calculation. Black squares represent taxonomic abundance from 0 to 0.01%. The colored bar above cobamide biosynthesis genes indicates pathway subsection in panel A.

FIG 2

Phylogenetically diverse skin bacteria encode cobamide-dependent enzymes and transporters. The total normalized hits for cobamide-dependent enzymes, cobamide transport protein btuB, and single-copy conserved gene rpoB are shown (totals hits normalized to profile HMM length and sequence depth), with the taxonomic abundance of the hits expanded as relative proportions above. Hits to distinct B12-dependent radical SAM proteins are grouped together as “B12-dep radical SAM.”

FIG 3

Cobalamin riboswitch regulation varies across skin taxa. (A) The taxonomic abundance of hits for cobalamin riboswitches (Rfam clan CL00101) are shown, with an expanded view of low abundance hits to the right. Total cobalamin riboswitch hits within each microenvironment are indicated. (B to G) Cobalamin riboswitch-containing reads identified from INFERNAL analysis were aligned to Cutibacterium acnes KPA171202 (B), Veillonella parvula DSM 2008 (C), Pseudomonas putida KT2440 (D), Corynebacterium kroppenstedtii DSM 44385 (E), Corynebacterium amycolatum FDAARGOS 1107 (F), and Streptococcus sanguinis SK36 (G) genomes. Dark gray lines along the light gray genome track indicate the position of mapped INFERNAL hits within the genome. Genes upstream and downstream of the riboswitches are colored by their general functional annotation. White (other function) indicates genes not currently known to be associated with cobamides. Gray (hypothetical) indicates a hypothetical protein that has no functional annotation. Right-facing gene arrows and upright dark gray riboswitch icons indicate forward strand orientation, and left-facing gene arrows and inverted riboswitch icons indicate reverse strand orientation. Genomic regions are not to scale.

FIG 4

Corynebacterium amycolatum produces cobamides and constitutively expresses cobamide biosynthesis genes. (A) E. coli ATCC 14169 was used as a microbiological indicator for the detection of cobamide concentration in Corynebacterium amycolatum LK19 cell extracts. Growth of E. coli was measured in minimal media with cyanocobalamin standards between 0.1 and 1.5 ng/ml to generate a standard curve. (B) E. coli growth with cyanocobalamin standards or different dilutions of C. amycolatum LK19 cell extract diluted between 10,000- and 50,000-fold. OD₆₀₀ values from 6 biological replicates and at least 3 technical replicates are shown. (C) The expression levels of cobB, cobH, and cobT were measured after 48 h of C. amycolatum LK19 culture in minimal media with cyanocobalamin supplementation (0, 250, and 2,500 ng/ml). Each culture/condition was grown in triplicate. Bars represent the mean gene expression ratio, calculated using the Pfaffl method, and error bars represent standard deviation. Cob gene expression was normalized to the 16S rRNA gene. Significance testing between cyanocobalamin concentrations was performed for each cob gene using the Kruskal-Wallis test, with significance defined as P < 0.05.

FIG 5

Skin microbiome networks reveal microbial associations among cobamide producers, precursor salvagers, and users. (A) The SPIEC-EASI method was used to identify microbial associations within each microenvironment of three independent skin microbiome data sets. Consensus networks are shown, representing associations identified in at least 2 of the 3 data sets. Species are represented by nodes and colored by phylum. Green and pink edges represent positive and negative associations, respectively. Node shape represents cobamide biosynthesis category and node size reflects mean species relative abundance within each microenvironment. Cobamide-dependent species are outlined in black. (B to E) In each final network, the number of species classified to each cobamide biosynthesis category (B), the number of species that are cobamide dependent or independent (C), the percentage of total edges that fall into each cobamide biosynthesis edge category (D), and the percentage of total edges that exist between cobamide producers and cobamide-dependent species that are nonproducers or precursor salvagers (E) are shown. NP, nonproducer; P, producer; S, precursor salvager.

FIG 6

Cobamide-producing Corynebacterium abundance is associated with microbiome diversity and atopic dermatitis disease state. Within each metagenome, the cumulative relative abundance of cobamide-producing corynebacteria (CPC) was calculated. (A) NMDS plots based on Bray-Curtis indices for healthy adult samples within each skin microenvironment are shown. Points are colored by Corynebacterium cobamide producer relative abundance and sized by alpha diversity (Shannon). (B) The relative abundance of CPC in pediatric atopic dermatitis patient samples at baseline, flare, and postflare time points or in healthy control subjects. A Kruskal-Wallis test was performed followed by Dunn’s test with Bonferroni correction to determine statistical significance between groups (*, P < 0.05; **, P < 0.01). The sample size for each group are as follows: baseline n = 70, flare n = 144, postflare n = 144, control n = 57. (C) The relative abundance of CPC in each individual skin site sampled. Black lines connect time points for a given patient. Certain sites were sampled from both sides of the body, therefore each point represents the average abundance of for each individual at the specified skin site.

FIG 7

De novo cobamide biosynthesis is host-associated within the Corynebacterium genus. (A and B) Genome length (A) and number of gene clusters (B) for 71 Corynebacterium genomes from host- or environment-associated species were determined using anvi’o. (C and D) Significantly enriched COG functions in environment-associated (C) or host-associated (D) genomes were identified with anvi’o function “anvi-get-enriched-functions-per-pan-group.” The top 20 significantly enriched COG functions (q < 0.05) are shown, ordered by ascending significance. (E) A Corynebacterium phylogenetic tree based on comparison of 71 conserved single-copy genes was generated using FastTree within the anvi’o environment. The tree is rooted with Tsukamurella paurometabola, and bootstrapping values are indicated (*, 100% bootstrap support). Species are colored by host or environment association, and by genome length. KOfamScan was used to identify the presence (dark pink) or absence (light pink) of cobamide biosynthesis genes within each genome. Cobamide biosynthesis subsections are indicated and are differentially colored based on those found in Fig. 1A. Blue, host-associated; orange, environment-associated. *, P < 0.0001 as calculated by a two-sided Mann-Whitney U test.