Strains to go: interactions of the skin microbiome beyond its species

Abstract

An extraordinary biodiversity of bacteria, fungi, viruses, and even small multicellular eukaryota inhabit the human skin. Genomic innovations have accelerated characterization of this biodiversity both at a species as well as the subspecies, or strain level, which further imparts a tremendous genetic diversity to an individual’s skin microbiome. In turn, these advances portend significant species- and strain-specificity in the skin microbiome’s functional impact on cutaneous immunity, barrier integrity, aging, and other skin physiologic processes. Future advances in defining strain diversity, spatial distribution, and metabolic diversity for major skin species will be foundational for understanding the microbiome’s essentiality to the skin ecosystem and for designing topical therapeutics that leverage or target the skin microbiome.

The skin harbors a diverse microbial community

Contrasted to the body’s other interfaces with the external environment, the human skin features a relatively low-nutrient barrier as its first line of defense against biotic and abiotic foreign matter. Despite a core function of ‘keep it out’, human skin is home to a diversity of microorganisms, including bacteria (primarily Cutibacterium, Staphylococcus, Corynebacterium, and Micrococcus spp. and other Actinobacteria, Proteobacteria, and Firmicutes, Figure 1A), fungi (primarily Malassezia sp.), viruses (both phage and human viruses, including papillomavirus and polyomavirus), and small eukaryotes such as mites (Demodex sp.)[1,2].

Figure 1. Overview of the skin microbiome.

A) The skin microbiome in healthy skin is considered to reside primarily in deeper structures such as pilosebaceous units (e.g., terminal or vellus hair follicles and sebaceous glands) and sweat glands, with relatively fewer microbes inhabiting the stratum corneum. Localization, density, and admixture of species within these structures remains little defined. These bacteria, fungi, and viruses have important roles in skin barrier homeostasis, cutaneous innate and adaptive immunity, and help condition the skin microenvironment. B) Representation of major skin microbiome changes over lifespan, where available. Piecharts are composite relative abundance data from the three indicated studies.

Considering the thick, scaly skin of the heel, the dry expanses of the forearm, the oily pores of the nose, or the hairy mucosal inner nares, the skin has marked physiologic differences over its breadth. In addition to differing numbers of pilosebaceous units, including different types of hair follicles such as terminal or vellus, sebaceous and apocrine glands, the thickness and composition of the skin’s major layers – the epidermis, the dermis, and the subcutaneous layer – vary based on location, features which collectively determine the skin’s numerous physiochemical variations in skin moisture, pH, salinity and oiliness[3]. Striking differences in the skin microbiome within a single individual reflect these physiologic differences, which dramatically remodel over lifespan (Figure 1B). In healthy individuals, the skin microbiome begins with seeding of the skin by the mother’s vaginal or skin microbiome during birth[4-6], then shifts towards a lipophile-dominated community following major hormonal changes during puberty, which increase skin oiliness[6,7]. Communities then remain relatively stable through adulthood[1], but will again remodel with age, diversifying in concurrence with major physiologic changes of thinning and drying skin[7]. Finally, the skin microbiome differs significantly between individuals and is influenced transiently or long-term by numerous additional intrinsic (e.g., genetics, immunocompetence, skin barrier status) and extrinsic factors (e.g., ethnicity, geography, hygiene and other personal habits, medication use, exposure to pathogens).

Different constituents of the skin microbiome have been linked to a wide range of cutaneous processes in health and disease, including modulation of innate and adaptive immunity through different stages of life[8-12], colonization resistance to pathogens[13,14], maintenance of the skin barrier and microenvironment[15,16], and wound healing[11,12,17], to name a few. Conversely, skin microbiome dysfunction and colonization of pathobionts like Staphylococcus aureus has been established to be associated with or to contribute to skin diseases such as atopic dermatitis[18-23], Netherton syndrome[24], and skin cancers[25]. As species-level interactions in skin health and disease have been reviewed excellently elsewhere [26-30], we will primarily discuss the contributions of two ubiquitous skin bacteria at sub-species level, which are the focus of the remainder of this review.

Deeper than species: host interactions are strain specific

Such studies have amply demonstrated the breadth of host processes that are regulated by different skin species. Adding to the challenges of investigating skin microbiome interactions is the extensive array of genetic diversity within each species. Indeed, microbial diversity and its phenotypic consequences are ultimately manifested at this finest taxonomic resolution, not necessarily at the species level. Infectious disease specialists are well-versed on the concept of this variation at the subspecies, or strain level – e.g., nosocomial vs. commensal, methicillin resistant vs. drug-susceptible S. aureus strains – thus a strain is defined as a genetic variant of a species, and “isolate”, a term often used interchangeably, is a heretofore uncharacterized strain. A lineage encompasses strains that are descended from a common ancestor, and strains can be very similar to their parent or differ significantly based on mutational rate or horizontal gene transfer introducing new genic elements.

Most studies to date have surveyed strain diversity in different individuals, or transmission from the environment in efforts to identify broad characteristic of disease-causing vs. healthy strains. However, two major recent efforts in skin, performing comparative genomics on libraries of Staphylococcus epidermidis and Cutibacterium acnes isolates, identified that striking strain diversity can exist within-individual; even within skin-site. In addition, these examples well exemplify that strain variation will take multiple manifestations depending on the species of interest.

Staphylococcal strain diversity

Zhou et al. surveyed 1,462 isolates of S. epidermidis isolated from 12 skin sites of 5 healthy individuals, targeting 10 isolates per skin sample[31]. While this depth is certainly not exhaustive, it was the first large-scale effort examining whether strains within a single skin site are clonal, or deriving from multiple lineages. Using a combination of tracing single nucleotide polymorphisms (SNPs) in the core genome (genic regions shared between all strains, which estimates evolutionary distance between genomes), and examining gene content differences (genic regions unique to only a subset of genomes, termed the accessory genome), they found that S. epidermidis strains within an individual and remarkably, even within a skin site, were far from clonal; nearly every isolate was a unique variant (Figure 2A).

Figure 2. Extensive within-individual strain variation of two major skin species.

A) An individual simultaneously harbors multiple lineages of S. epidermidis (left) and C. acnes (right) across their skin. For S. epidermidis (left), the phylogenetic tree was inferred based on whole genome assemblies of 1477 isolates. Lineages were defined as groups of isolates (n>2) with genetic distances not exceeding 0.05. Similarly, isolates with genetic distances not exceeding 0.15 were annotated to be of the same "strain type", which are arbitrarily named. For C. acnes (figure adapted from Conwill et al., 2022), lineages were defined as sets of colonies separated by <100 mutations. The 53 lineages generated from 947 isolate genomes are shown in the tree, with strain types named by single locus sequence type. The distribution of lineage richness of each subject is visualized in the heatmaps for each species. B) For S. epidermidis, two examples of functional diversity arising from strain-level genetic diversity are shown. Distribution of 18 different types of predicted bacteriocins, and their distribution across subjects (top row) and skin site (bottom row) is shown. An example of antibiotic resistance gene reservoirs differing between individuals but disseminated across skin sites, with red or green dots showing presence in at least one isolate from that skin site. For both bacteriocin distribution and antibiotic resistance, most are individual-specific but distributed across multiple skin sites within that individual (figure adapted from Zhou et al., 2020). C) A key unresolved question is to what degree does genetically diverse strains co-exist at a microscale, in isolated skin structures? From Conwill et al., C. acnes follows the “island” hypothesis, but each species may differ in their spatial distributions (Figure adapted from Kong, Oh, 2022).

We note several findings of particular interest. First, a within-individual analysis of Bacteroides fragilis found that healthy individuals possess a single lineage that diversified within the gut of an individual over time[32], whereas S. epidermidis strains within an individual derived from multiple founder lineages, rather than a single colonizer. Populations were subsequently shaped by site adaptation, particularly in the thick scaly skin of the foot as well as through transmission events between high-touch sites such as the hands. Strikingly, horizontal gene transfer could occur on short evolutionary timeframes. Sister strains, zero SNP differences in the core genome – that is, the genome that is shared between all strains, could possess different gene contents, and very closely related strains with few SNP differences could differ by hundreds of genes. Such genes were often associated with mobile elements and, notably, plasmid-borne antibiotic resistance genes, whose function they verified experimentally. Given the striking co-occurrence of genetically diverse strains, what might be a potential role in skin health? By making admixtures of S. epidermidis strains observed within a skin site and examining transcriptional response to admixture, the presence of multiple diverse strains was shown to suppress expression of virulence factors and modulate metabolism on a population-level. Thus, strain diversity might be one mechanism to suppress S. epidermidis’ potential transition to pathogenicity, an area of interest to pursue in in vivo and bloodstream infection models.

C. acnes strain diversity

C. acnes differs from S. epidermidis in its relatively closed accessory genome, with ~10% of the genome estimated to vary between strain types, vs. 20% for the latter[33]. Conwill et al. surveyed 947 isolates of C. acnes obtained from 300 samples from 16 healthy adults, obtaining 1–15 colonies per sample[34]. The unique aspect of their design was that 145 of these samples were isolated pore samples from 5 individuals, allowing them to assess strain diversity at a much finer geographic scale allowable than a bulk skin swab. Indeed, they found that a pore was effectively a genetic island, with most isolates derived from the same pore having very few SNP level differences. On a broader scale, similar (but not clonal) strains could co-exist within an individual (Figure 2A), suggesting that pores are monocolonized at random, i.e., a population bottleneck. They purport that there is limited competition between co-existing C. acnes strains – one would suspect because there are relatively few genic differences between strains that would beget a significant functional advantage, although the authors observed differences in in vitro growth rate. Indeed, a compelling segue to this experiment would be to understand if these island-like population structures persist in C. acnes-associated diseases (like acne), or if certain strains possess gene-level differences that would overcome the neutral processes in healthy skin. It may very well be a combination of dispersion as well as a potential for genetic specialization (together with host immunity and intrinsic states) that mediates C. acnes’ contribution to acne.

Functional consequences of strain diversity

Now, how might we contextualize the findings of studies of S. epidermidis’ or C. acnes’ contribution to skin health and disease? Most studies have not – or are not – able to systematically study a wide breadth of strains for each species’ study. Yet there is ample precedent that that these strain-level differences are impactful. Continuing with the example of acne, while certain phylotypes (genetically similar groups of strains, of 6 identified in C. acnes[35]), e.g., “phylotype I” strains, have greater associations in acne formation than those of phylotype II, which are more closely related to healthy skin as well as, interestingly, deep tissue infections, potentially on account of the opportunistic nature of this phylotype[36]. In addition to the suppression of population-level virulence by S. epidermidis as discussed, there is ample evidence of phenotypic diversity of strains with significant health consequences (Figure 2B). Several examples include: in atopic dermatitis, patients with less severe skin flares had distinct S. epidermidis strain diversity (predicted by metagenomic sequencing), with a loss of genetic groups, or clades, typical of healthy controls[19,22]. This is further bolstered by evidence in mice, in which skin barrier disruption was observed with specific strains of S. epidermidis that produce excessive amounts of a damaging protease[37]. Other strains of S. epidermidis can produce metabolites that are protective against skin cancer in mice[38]. Numerous cases have identified different strains that produce different antimicrobials or possess antibiotic resistance genes[13,39], which may be a risk factor for transfer to other microbes in the environment[40]. A particularly intriguing study recently demonstrated that S. epidermidis can regulate wound healing in skin via recruitment of mucosa-associated invariant T (MAIT) cells in mice[11], with implications that different strains present in different regions might then modulate wound healing in a skin-site specific manner.

However, it is important to note that few ‘smoking gun’ characteristics of disease-causing strains have been identified, and that most strains lie on a continuum of commensal to virulent[33]. This is likely because of the complexity of genetic variants observed in a given strain - numerous genes/variants may endow plasticity in health vs. disease environments, or genetic background can modify gene essentiality[41] or virulence[42]. For example, while acne-associated phylotype I strains innately produce higher levels of porphyrins, this production can be regulated by the availability of vitamin B12, while other strains are non-producers and also non-responsive to B12[43]. S. epidermidis strains can produce widely varying amounts of the metabolite of the riboflavin biosynthesis pathway that mediates MAIT cell activation, dependent on environmental growth conditions (unpublished data).

Challenges and alternative approaches to resolving strain diversity

We note several challenges to bringing strain diversity into the mainstream vis-à-vis their mechanistic role in host-microbiome interactions. While there are numerous efforts now being taken to create patient-specific isolate collections[13,44-48], a significant challenge remains in systematically defining strain diversity genetically and phenotypically because of the scale and scope required. Isolation by cultivation followed by whole genome sequencing is the gold standard for generating sufficiently high-quality genomes for differentiation at the SNP level. However, cultivation and isolate sequencing is laborious, especially for investigating within-population diversity, and can be further limiting for low abundance microbes, in the absence of methods for enrichment. At the other extreme, algorithms to infer strain diversity from bulk metagenomic data require significant depth for each species interrogated, but still can dramatically underestimate within-population strain diversity, e.g., by identifying a dominant strain type based on SNP variation in a set of conserved marker genes[49-53]. Culture-assisted metagenomics, in which a limited number of cultivated isolates are used to track strains over time in additional metagenomic samples, may prove a useful intermediate, however; this has similarly limited ability to resolve population-level diversity, as an individual’s specific set of strains for each species of interest must be characterized a priori for most accurate tracking[53]. A very recent innovation in droplet-based, single-cell microbial metagenomics has promise in reconstructing, with extraordinary throughput, individual genomes; however, due to genome incompleteness, genomes obtained from multiple droplets had to be merged to make a composite genome, losing some information on strain variation[54].

Additional tools to examine the functional consequences of genetic variation across multiple strain types will be useful to probe the accessory genome. Transposon mutagenesis (e.g., Tn-seq[55]) knockout/knockdown or CRISPRi tools[41,56-59] are promising approaches that can be deployed in multiple strains as opposed to more laborious gene knockout approaches, and can be used to profile fitness effect of genes in different environmental conditions. However, these approaches are still predicated on genetic transformability, which remains a major challenge in primary isolates, which possess numerous restriction modification systems and other barriers to efficient genetic transformation. Overexpression of pangenome regions in a genetically tractable strain is another possibility to evaluate a phenotype of interest. Ultimately, genomic data should also be leveraged to identify genetically diverse strain sets worthy of phenotyping in low- mid-throughput assays, or to identify genes and variants of interest for synthesis and testing. For example, a clever deployment of genome-wide association studies (GWAS) with 415 strains of S. epidermidis identified 61 gene variants associated with infection vs. commensalism, further enabling 80% accuracy in a random forests machine learning model in classifying disease-causing vs. commensal strains[60]. High throughput screens, such as those searching for antimicrobial production[13], immunomodulatory ability, or other specific host interaction (i.e., production of molecules sensed by G-protein coupled receptors[61]), will also help to refine candidates for more laborious mechanistic follow-ups.

Outlook

Probing the microbiome’s role in the skin with not only species, but also strain-level resolution is a major emerging frontier in microbiome research, and we envision that these ecological principles will be similarly investigated in other prevalent skin species – e.g., the many additional staphylococcal species in the skin, Corynebacterium, Micrococcus and others. Characterizing strain diversity within and between individuals in states of health and disease will dramatically refine the genetic blueprint of the microbiome first established by metagenomic sequencing. Studies of strain diversity will be further complicated by a species’ unique evolutionary trajectory in the human body, as it is likely that many, if not most inferences on a species’ genetic diversity at the strain level will need to be reconstructed on a species-by-species basis. Given the already substantial species-level diversity of the human microbiota, systematic studies to reconstruct strain diversity on a greater scope will require technological and algorithmic innovations, including recent efforts in massively parallelized single cell approaches[54], integrated cultivation-based and metagenomic analyses, and algorithms that can delineate standing genetic variation. To complement genomic reconstructions, analogous high throughput efforts in strain phenotyping will be needed to translate the functional consequences of this genetic diversity. Recent examples have included screening for antibiotic resistance[31], production of useful antimicrobials[13,62], and production of immunomodulatory metabolites[63-66].

Finally, innovations in spatial resolution of both species and strains[67-69], perhaps drawing on recent technologies in spatial transcriptomics[70-72], are critically needed to understand to what degree strains (and species) actually co-exist in the different skin structures, and which are influencing a given host cell type, and what is the host response. For example, while C. acnes and S. epidermidis are ubiquitous in skin, some strains of C. acnes produce cutimycin, an antimicrobial that can reduce the presence of S. epidermidis in the same hair follicle[73]. This antagonistic relationship appears of increasing importance in relation to skin health, as altered Cutibacterium:Staphylococcus ratios have been observed in skin cancer[25] and aging[7]. However, the degree to which different S. epidermidis strains co-localize or compete on a microscale with other genetically diverse strains, C. acnes, or other skin microbiota remains unknown (Figure 2C). In addition, it would be of significant value to pinpoint host interactions that result from microbial colonization – for example, recent efforts using single cell transcriptomics characterized immune cell populations in the follicular environment that appeared to influence the resident skin microbiome composition[74] (and potentially, vice versa). Key questions remain: where do specific microbes reside in proximity to host cells and to each other? Which host cells are responding to which microbes, and what is their response? Such forays will continue to transform our understanding of how the microbiome shapes skin health.

Highlights.

• The human skin microbiome differs between and within individuals

• Ubiquitous bacteria S. epidermidis and C. acnes have major roles in cutaneous immunity and physiology

• Strain-level differences can modulate interactions with the host

• Knowledge gaps remain: strain diversity at a microscale and between species