Contextual control of skin immunity and inflammation by Corynebacterium

The skin is the body’s most exposed interface with the environment and acts as a first line of physical and immunological defense. This organ is also a complex and dynamic ecosystem inhabited by a multitude of microorganisms (Belkaid and Segre, 2014). These microbes play a fundamental role in the control of skin physiology, including skin immunity and inflammatory processes (Lai et al., 2009; Naik et al., 2012). However, despite the formidable diversity of skin microbes, thus far only a handful of specific microbes and microbe-associated molecules have been linked to defined immunological or inflammatory processes. Although little is known about the mechanisms by which skin microbes influence the skin immune system at steady state, even less is known about how this dialog is altered under conditions of inflammation. Identifying dominant microbe-derived immune modulators and the context controlling the impact of these microbes on the immune system may help us understand the association between defined members of the skin microbiota and the skin immune system under both steady-state and disease settings.

Here, we demonstrate that members of a dominant bacterial genus of the skin, Corynebacterium, promote IL-23–dependent responses under steady-state and inflammatory conditions and reveal that expression of a major component of the Corynebacterium cell wall, mycolic acid, is required to mediate these responses. Further, we show that the impact of microbial determinants on tissue immunity can be highly controlled by the inflammatory and metabolic status of the host.

To uncover novel microbial species or microbiota-derived molecules that engage the skin immune system, we developed a generalizable culturing approach to isolate microbial taxa from the skin of WT mice, from the skin of mice with defined immune deficiencies, or from skin swabs collected from healthy human volunteers. We used both a classical (αβ TCR⁺) and nonclassical (γδ TCR⁺) skin lymphocyte cytokine potential profile as the read-out of an in vivo screen. Specific pathogen-free (SPF) animals, raised under conventional settings (with an endogenous microbiota), were topically associated with distinct bacteria. At 14 d after the initial microbial application, skin T cell subset frequency and cytokine potential profiles were assessed (Fig. 1 A and Fig. S1, A and B).

Notably, C. accolens had a particularly strong impact on the accumulation of IL-17A–producing γδ TCR^low T cells (Fig. 1, A–E; and Fig. S1 B), a population of migratory γδ T cells (γδ TCR^low) found in the mouse dermis (Cai et al., 2011). Corynebacterium is one of the three most abundant bacterial genera on human skin, found especially in moist sites (Grice et al., 2009). Corynebacterium species are also common members of the mouse skin microbiota (Grice et al., 2009; Belheouane et al., 2017). Given their prevalence, remarkably little is known about the effects of Corynebacterium on host immunity. After C. accolens association of mice previously devoid of Corynebacterium (Fig. 1 H), the frequency and absolute number of γδ TCR^low cells as well as their potential to produce IL-17A (γδ T17) were significantly increased in the skin and skin draining lymph node compared with unassociated controls (Fig. 1, B–E). Addition of brefeldin A (BFA) during tissue digestion revealed that, in contrast to cells from control mice, γδ TCR^low cells actively released IL-17A after C. accolens association (Fig. 1 E). The impact on the γδ T17 cell number was detectable by 7 d after association and was durable, lasting up to 90 d after association (Fig. 1, F and G). The increase of γδ T17 cells in the skin of associated mice was uncoupled from inflammation as indicated by a lack of neutrophils and inflammatory monocyte recruitment and a lack of keratinocyte hyperplasia (Fig. S1, C–F). Such a response occurred even when bacteria were applied at doses as low as 10⁶ CFU/cm² (Fig. S1, H and I) corresponding to the proposed biomass of skin commensals in humans (Grice and Segre, 2011). 16S ribosomal RNA (rRNA) gene sequencing of the skin-resident microbiota of associated and control mice confirmed that C. accolens was able to colonize the mouse skin in a durable manner and was associated with some changes to the structure of the resulting skin-associated microbiota, namely, an increase in Proteobacteria and a corresponding decrease in Bacteroidetes (Fig. 1 H). Colony-forming unit assessment confirmed that a low number of C. accolens was found on the skin of associated mice (Fig. S1, G and H). We next addressed if the impact of C. accolens was direct or dependent on Corynebacterium-induced changes in the endogenous community. As previously shown (Naik et al., 2015), mice raised in the absence of live microbes (germ-free [GF]) had a severe defect in IL-17A production by γδ T cells (Fig. 1 I). Monocolonization of GF animals with C. accolens resulted in a significant increase in γδ T17 cells within the skin revealing that the effect of C. accolens on γδ T cells was dominant and independent of other members of the cutaneous microbiota (Fig. 1 I).

At steady state, approximately half of the IL-17A–expressing γδ TCR^low cells also express Vγ4, whereas the remaining fraction is enriched in Vγ6 usage (Heilig and Tonegawa, 1986). Skin association with C. accolens resulted in the specific increase and activation of the Vγ4⁺ subset of γδ T cells (Fig. 2, A–C). This contrasted with Staphylococcus epidermidis, which preferentially enhanced Vγ4-negative T cells (Fig. 2, A–C). Mice lacking the Vγ4⁺ subset of γδ T cells (Gray et al., 2013) did not mount a compensatory Vγ4^neg response (Fig. 2 D). IL-17A–producing γδ T cells within the skin also expressed the chemokine receptor CCR6, allowing us to use this marker to isolate Vγ4⁺ γδ T17 cells (Fig. 2 A). We sorted Vγ4⁺ γδ TCR^low CCR6⁺ cells from the skin of mice that had been associated for 14 d with either C. accolens or S. epidermidis or from unassociated controls and performed gene expression analysis using a NanoString chip designed to capture activation-associated lymphocyte gene expression (Table S1). Of note, gene-level expression for the IL-23 receptor was high in Vγ4⁺ γδ TCR^low CCR6⁺ cells under steady-state conditions (Fig. 2 E). Furthermore, expression of CCR4, a chemokine receptor linked to skin homing, was up-regulated in Vγ4⁺ γδ TCR^low CCR6⁺ cells collected from C. accolens–associated mice when compared with unassociated controls. Additionally, and in contrast to cells isolated from S. epidermidis–associated mice, gene-level expression of markers of T cell activation, such as TNFRSF4 (OX40), CD279 (PD-1), and STAT5, were also up-regulated (Fig. 2 E). Collectively, these results support the idea that C. accolens preferentially promoted the activation of Vγ4⁺ γδ T cells.

Numerous Corynebacterium species are common members of both mouse and human skin microbiota (Grice and Segre, 2011; Belheouane et al., 2017). To test if the ability to induce γδ T17 cells was a general feature of the Corynebacterium genus, we associated WT mice with nine different species of Corynebacterium. Notably, eight of the nine species induced a robust Vγ4⁺ γδ T17 response (Fig. 3, A and B), supporting the idea that such a response was likely the result of a conserved feature among the species tested. One characteristic that distinguishes Corynebacteria from other gram-positive bacteria is an external lipid bilayer of all-carbon, branched fatty acids called mycolic acids, which forms a structure similar in size and location to the gram-negative outer membrane (Fig. 3 C) (Dörner et al., 2009). Interestingly, C. amycolatum, a rare Corynebacterium species that does not contain mycolic acids (Barreau et al., 1993), was the only Corynebacterium tested that failed to induce γδ T17 cells (Fig. 3, A and B).

To assess if mycolic acids or elements from the characteristic Corynebacterium cell envelope contributed to the Vγ4⁺ γδ T17 expansion, we generated a mutant strain of C. accolens with a deletion in the mycolic acid synthase gene (Δ0503). In contrast to WT C. accolens, association of SPF mice with the mycolic acid–deficient strain Δ0503 did not induce Vγ4⁺ γδ T17 within the skin compartment (Fig. 3, D and E). Because a defect in the cell envelope could potentially lead to defects in colonization efficacy, we monocolonized GF mice using the mutant or WT strains of C. accolens. Under these conditions, C. accolens Δ0503 also failed to induce a γδ T17 response compared with the WT strain (Fig. 3 F). To further confirm the involvement of the Corynebacterium cell envelope, we used an in vitro approach in which purified Vγ4⁺ CCR6⁺ T cells from the skin and regional LN of C. accolens–associated animals were restimulated ex vivo. Vγ4⁺ CCR6⁺ T cells from C. accolens–associated mice did not respond to heat-killed (hk) C. accolens bacteria alone, supporting the idea that this response was not associated with direct engagement of the γδ T cells by a microbially derived element (Fig. 3 G). On the other hand, and in contrast to unloaded splenic DCs, DCs loaded with hk C. accolens cells were able to promote the production of IL-17A by Vγ4⁺ CCR6⁺ T cells (Fig. 3 G).

Further confirming a role for mycolic acid expression, DCs loaded with hk Δ0503 C. accolens or hk C. amycolatum failed to promote IL-17A production from purified Vγ4⁺ CCR6⁺ cells (Fig. 3 G). To further test the contribution of the elements of the Corynebacterium cell envelope to Vγ4⁺ γδ T17 cell induction/activation, we isolated total envelope extracts (TEEs) from WT C. accolens bacteria and included this fraction in liposomes. DCs loaded with liposomes carrying TEE but not empty liposomes were able to stimulate IL-17A production from purified Vγ4⁺ CCR6⁺ cells (Fig. 3 G). In contrast, DCs loaded with lipoarabinomannans (LAMs), which are also present in the outer surface material of these Corynebacteria (Burkovski, 2013), were unable to induce IL-17A production by Vγ4⁺ γδ T cells (Fig. 3 G). Accordingly, DCs loaded with hk cells from a mutant strain of C. striatum (Δ647) bearing a deletion in a key LAM synthesis enzyme were still able to promote the production of IL-17 by Vγ4⁺ CCR6⁺ T cells (Fig. 3 G). Thus, the ability of Corynebacterium to activate γδ T cells in vitro is dependent on a cell envelope containing mycolic acid but not LAMs. Together, these results support the idea that mycolic acid (directly or via its ability to anchor a defined molecular determinant) plays a crucial role in the ability of Corynebacterium to activate Vγ4⁺ γδ T cells.

γδ T cells can be activated via distinct mechanisms, including antigen-specific stimulation and various innate mechanisms, such as direct responses to inflammatory cytokines (Sutton et al., 2009; Vantourout and Hayday, 2013). An in vivo screen of mice with defects in dominant classical and lipid antigen presentation (β2M and CD1d) revealed that neither of these pathways was required for the effect of C. accolens on γδ T17 (Fig. S2, A and B). Further, MINCLE (CLEC4E), a receptor previously shown to be involved in the recognition of bacterial fatty acid derived from Mycobacterium tuberculosis (Matsunaga and Moody, 2009); CARD9, a downstream signaling mediator for several C-type lectin receptors; or MyD88, a signal transducer downstream of most Toll-like receptors, was not required to mount a response to C. accolens in vivo (Fig. S2, C and D).

Although we could not at this stage identify the receptor or potential alternative mode of antigen presentation involved in these responses, we assessed to which extent C. accolens could influence the cytokine milieu in a way that promoted γδ T cell responses. In the skin, we previously showed that S. epidermidis was able to promote IL-17A production from the Vγ4^neg subset via a local increase of IL-1 (Naik et al., 2012). In contrast, based on response observed in Il1r1^−/− mice, γδ T cell activation after association with C. accolens was IL-1 independent (Fig. 4 A). On the other hand, and consistent with the high Il23r gene expression in Vγ4⁺ IL-17A⁺ T cells (Riol-Blanco et al., 2010; Liang et al., 2013; Fig. 2 E), mice deficient in either IL-23R or the IL-12p40 subunit mounted a reduced response after C. accolens association compared with controls (Fig. 4, B and C; and data not depicted). Thus IL-23 contributes significantly to the immune impact of C. accolens under homeostatic conditions within the skin compartment.

To further confirm that cytokines promoted the γδ T cell response to C. accolens, we assessed if such a response could be reproduced in vitro. To this end, Vγ4⁺ CCR6⁺ T cells purified from the skin and regional LN of naive or C. accolens–associated animals were exposed to DCs loaded with hk C. accolens or to the filtered supernatant of these cells. DCs loaded with hk C. accolens or supernatants from hk C. accolens–loaded DCs induced a potent production of IL-17A by Vγ4⁺ γδ T cells in vitro (Fig. 4 D). Of interest and in contrast to cells isolated from preassociated mice, Vγ4⁺ γδ T cells from naive mice failed to respond when exposed to the filtered supernatant of hk C. accolens–loaded DCs (Fig. 4 D).

Purified DCs stimulated in vitro are likely to produce a broader range of cytokines than the relevant skin resident population, and as such we next used a more targeted approach to assess responsiveness to IL-23 and/or IL-1. Vγ4⁺ γδ T cells from C. accolens–associated mice were purified from the skin and submandibular LN and exposed to IL-23 or IL-1 alone. As previously shown (Holderness et al., 2013), both cytokines promoted IL-17 production by these cells when purified from naive or C. accolens–associated mice (Fig. 4, E and F; and Fig. S2 E). However, IL-23 was more potent than IL-1 in maximizing IL-17 production by γδ T cells (from both naive and associated mice; Fig. 4, E and F; and Fig. S2 E). This is consistent with the high level of IL-23R expression on Vγ4⁺ γδ T cells (Fig. 2 E; Cai et al., 2011; Riol-Blanco et al., 2014), a feature that may predispose them to preferentially respond to a minute increase in the IL-23 level. Of interest, Vγ4⁺ γδ T cells from C. accolens–associated mice had a heightened response to cytokine in vitro compared with Vγ4⁺ γδ T cells isolated from naive mice, further supporting the idea of a potential education or memory-like phenotype after association (Fig. 4, E and F; and Fig. S2 E). Such a phenomenon has been previously reported in the context of experimental psoriasis in which Vγ4⁺ γδ T cells can acquire a memory-like phenotype characterized by a heightened and more rapid response to inflammatory cytokines (Ramírez-Valle et al., 2015). Together, these results support a dominant role for IL-23 in mediating the impact of Corynebacterium on skin immunity.

Although under homeostatic conditions, the C. accolens targeted effect on Vγ4⁺ γδ T cells is uncoupled from inflammation, this microbe may condition the tissue for subsequent inflammatory disorders. Supporting this hypothesis, in mice with defective barrier function, C. bovis was shown to contribute to the inflammatory process (Kobayashi et al., 2015), and in humans, the cutaneous microbiota of psoriatic lesions is characterized by an abundance of isolated Corynebacterium species (Gao et al., 2008; Alekseyenko et al., 2013). In a model of experimental psoriasis induced by topical application of a TLR7 agonist (imiquimod [IMQ]), part of the inflammatory processes is mediated by Vγ4⁺ γδ T cells (Gray et al., 2013; Ha et al., 2014). To evaluate if, under this setting, C. accolens–induced Vγ4⁺ γδ T17 cells could exacerbate skin inflammation, mice were preassociated with the bacteria 2 mo before topical treatment with IMQ. At 2 mo after association, the representation of C. accolens was minimal compared with other members of the skin microbiota (Fig. 1 H). Although C. accolens alone had no inflammatory impact on the skin, animals preassociated and subsequently treated with IMQ had significantly increased skin thickness and cellular infiltrate compared with unassociated controls (Fig. 5 A and Fig. S3, A, C, and D).

Increased inflammation was associated with higher IL-17A production from Vγ4⁺ γδ T cells in the skin of preassociated animals treated with IMQ compared with IMQ alone (Fig. 5 B) and was Vγ4⁺ subset dependent (Fig. S3 B). Thus C. accolens, even when present at low levels, can promote activation of Vγ4⁺ γδ T17 cells, which in turn can increase skin inflammatory responses. Increased inflammation in the context of C. accolens was also associated with a selective increase in inflammatory monocytes but not in neutrophils, eosinophils, or TCRβ⁺ cells (Fig. S3 D). In contrast to C. accolens, we found that C. amycolatum did not exacerbate IMQ-induced inflammation, supporting the idea that cell envelope components are required for this effect (Fig. 5 C and Fig. S3, C and D).

The outcome of C. accolens encounters with the skin may also depend on the physiological status of the host. Obesity has been linked to an increased prevalence of skin inflammatory disorders (Padhi and Garima, 2013). To test if host metabolic alterations could impact host responses to the commensal, we fed SPF mice with either a control or synthetic high-fat diet for 1 mo, before cutaneous association with C. accolens. Of note, we associated mice at an early time point after the diet change, before the development of metabolic syndrome and tissue inflammation (Zhang et al., 2015). Unassociated animals fed a high-fat diet had a modest increase in skin thickness compared with their counterparts fed a control diet (Fig. S3 E). On the other hand, and in contrast with control-diet-fed animals, mice fed a high-fat diet and associated with C. accolens had a significant increase in skin thickness within 2 d of association (Fig. 5 D). Interestingly, other populations of γδ T cells (Vγ4^neg γδ T) were also increased under these settings (Fig. 5 E); however, depleting Vγ4⁺ cells (Fig. S3 F) abrogated this inflammatory response (Fig. 5 D). 16S rRNA gene profiling revealed that the colonization efficiency of C. accolens was lower in animals fed a high-fat diet compared with control-diet counterparts, supporting the idea that enhanced inflammation was not caused by aberrant C. accolens colonization (Fig. S3 G).

In agreement with the targeted effect of C. accolens in mice fed a control diet, a small number of transcripts (48 up-regulated and 31 down-regulated) modestly changed (twofold average) after association (Fig. 5 F). In contrast, application of C. accolens to the skin of a high-fat-fed animal was associated with a broad change in gene expression (599 up-regulated and 511 down-regulated; Fig. 5 F). Pathway analysis revealed that the majority of highly up-regulated gene ontology (GO) pathways were linked to keratinocyte/hair follicle responses (Fig. 5 G). Histological assessment revealed hyperkeratinization in C. accolens/high-fat animals compared with unassociated controls and, as assessed by flow cytometry, enhanced keratinocyte proliferation (Fig. 5, H and I). Expression of numerous immune-related genes was also up-regulated, including those known to contribute to skin inflammation, such as genes encoding for several members of the IL-1 family as well as IL-23R (Fig. 5 J). Nonetheless, no changes in cellular infiltrate (neutrophils, eosinophils, and monocytes) were observed in the context of C. accolens/high-fat diet compared with high-fat diet alone, supporting the idea that changes in skin physiology result dominantly from a targeted impact on keratinocytes (Fig. S3 H). Thus, in the context of a high-fat diet, the addition of an innocuous skin microbe can be sufficient to trigger inflammation. The rapid inflammatory impact of C. accolens on the skin of high-fat-diet–fed mice was significantly reduced in IL-23R–deficient mice, further implicating this cytokine in mediating the effect of C. accolens in both steady state and inflammation (Fig. 5 K).

We next tested if the expression of mycolic acid could mediate the pathogenic impact of C. accolens. To this end, high-fat-diet–fed mice were associated with WT or Δ0503 C. accolens or C. amycolatum. To normalize colonization by the Δ0503 mutant, mice were associated daily starting 4 wk after introduction of the high-fat diet. As previously shown, C. accolens alone induced skin inflammation as early as 2 d after association. In contrast, the impact of C. accolens Δ0503 or C. amycolatum was significantly reduced compared with high-fat-diet–fed mice associated with control bacteria (Fig. 5 L). Thus, mycolic acid expression is associated with the pathogenic effect of Corynebacterium. These results also reveal that host metabolic status can have a profound impact on the sensing and interpretation of microbial signals by the host’s immune system and may predispose to inflammatory responses to an otherwise commensalistic or mutualistic symbiont.

Despite the growing understanding that microbes are critical to the promotion of skin immunity, the microbial determinants responsible for mediating these interactions remain undefined. Our data demonstrate that Corynebacterium activates the skin immune system and that this property depends on the expression of mycolic acid, a highly conserved surface molecule that distinguishes Corynebacteria from other gram-positive skin microbes. Together, our results also reveal that host metabolic status can have a profound impact on the sensing and interpretation of microbial signals by the host’s immune system and may profoundly impact the outcome of the host–immune system encounter. Our data show that skin microbes can, in a predisposed skin state, be sufficient to trigger inflammation. Moreover, they reveal that a change in diet alone can be sufficient to transform a mutualistic member of the skin microbiota into an inflammatory trigger. In contrast to the gut, the skin is characterized by a scarcity of nutrients available to fuel the microbiota (Scharschmidt and Fischbach, 2013). In these arid conditions, subtle alterations in nutrient availability can potentially have a dramatic impact on the impact of the skin microbiota. Obesity has been linked to the increased severity and prevalence of many skin inflammatory disorders (Padhi and Garima, 2013). Based on our present findings, it is intriguing to consider that these effects may be, in part, mediated by altered responses to canonical microbial components. Further, based on the prevalence of Corynebacteria on mammalian skin and the ability of this genus to selectively engage IL-23–dependent responses, our work may have identified a dominant mediator of skin immunity and inflammation. Clinical observations have implicated IL-23 signaling in inflammatory diseases, and this cytokine is now considered a major therapeutic target in immune-based pathologies such as psoriasis (Puig, 2017). Furthermore, in the context of inflammation-related cancer, IL-23 is thought to contribute to tumorigenesis and progression to metastatic disease (Langowski et al., 2006; Wang and Karin, 2015). The ability of Corynebacteria to promote IL-23 signaling, an effect that can potentially target numerous immune populations within the skin environment, including classical T cells, could be highly relevant to the etiology of a variety of skin disorders in humans.

Together, our present results support the idea that a dominant genus of the skin microbiota has a profound impact on both skin immunity and inflammation and that canonical cell envelope components from these microbes may represent a dominant skin microbiota signal. Further, our results support the idea that conserved microbial determinants from an innocuous colonist can have a highly context-specific effect on the skin immune system and that in the skin a simple diet alteration can have a dramatic effect on the ability of the host to integrate microbe-derived signals. Although understanding the mechanism of interaction between the immune system and the microbiota remains in its infancy, this line of investigation is required not only to understand the impact of defined microbial communities on skin physiology but also as a means to uncover novel classes of immunological adjuvants.

Human skin swabs were obtained from anonymous healthy volunteers. All volunteers gave written informed consent, and collection of skin swabs was conducted according to the principles of the declaration of Helsinki after approval by the Institutional Review Board of National Institute of Allergy and Infectious Diseases (NIAID), National Institutes of Health (NIH).

C57BL/6 SPF mice were purchased from Taconic Farms. GF C57BL/6 mice were bred and maintained in the NIAID Microbiome Program gnotobiotic animal facility. B6.SJL-Cd45a(Ly5a)/Nai (B6.SJL), C57BL/6-[KO]IL17A (Il17a^−/−), C57BL/6NTac-[KO]β2m (B2m^−/−), and C57BL/6-[KO]IL1r1 (Il1r1^−/−) mice were obtained through the NIAID-Taconic exchange program. B6.129S6-Del(3Cd1d2-Cd1d1)1Sbp/J (Cd1d^−/−) mice and their C57BL/6J WT controls were purchased from the Jackson Laboratory. Il22^−/− mice were generated by Pfizer Inc. Il23r^−/− mice (Awasthi et al., 2009) were obtained by breeding Il23r^eGFP+/− reporter mice (a gift from M. Oukka, Seattle Children’s Research Institute, Seattle, WA) to generate Il23r^eGFP+/eGFP+ (Il23r^−/−) and Il23r^eGFP+/− (littermate control) mice. B6.129P2-Myd88^tmAki (Myd88^−/−), C57BL/6 Clec4e^−/− (Yamasaki et al., 2009; Mincle^−/−, originally generated by S. Yamasaki, Kyushu University, Fukuoka, Japan) and C57BL/6 Card9^−/− (Hsu et al., 2007) mice were gifts from G. Trinchieri (National Cancer Institute/NIH, Bethesda, MD), A. Sher (NIAID/NIH, Bethesda, MD), and M. Lionakis (NIAID/NIH, Bethesda, MD), respectively, and backcrossed to C57BL/6 for at least 10 generations. For experiments with knockout mice, littermate controls were used as WT controls, and/or WT (littermate or C57BL/6) and knockout mice were cohoused at 3–4 wk of age for 2–3 wk in the same cage, and then separated before the start of experimental manipulations. All mice were bred and maintained under pathogen-free conditions at an American Association for the Accreditation of Laboratory Animal Care–accredited animal facility at the NIAID and housed in accordance with the procedures outlined in the Guide for the Care and Use of Laboratory Animals. All experiments were performed at the NIAID under an animal study proposal approved by the NIAID Animal Care and Use Committee. Sex- and age-matched mice between 6 and 12 wk of age were used for each experiment. When possible, preliminary experiments were performed to determine requirements for samples size, taking in account resources available and ethical, reductionist animal use. In general, each mouse of the different experimental groups is reported. Exclusion criteria, such as inadequate staining or low cell yield due to technical problems, were predetermined. Animals were assigned randomly to experimental groups.

Bacillus lincheniformis and Escherichia coli (17KO) were isolated from the skin draining lymph nodes of Il17a^−/− mice. Candida albicans, E. coli (22KO), and Enterococcus faecalis were isolated from the skin of the ear pinnae of Il22^−/− mice. S. epidermidis strain NIHLM087 (Conlan et al., 2012) and Micrococcus luteus strain HV1032 were isolated from the skin of healthy human volunteers. All those bacteria were cultured for 18 h in tryptic soy broth at 37°C. C. accolens strain ATCC 49725, C. amycolatum strain SK46, C. glutamicum strain DSM 20300, C. jeikeium strain DSM 7171, C. propinquum strain DSM 44285, C. pseudodiphtheriticum strain DSM 44287, C. striatum strain DSM 20668, C. tuberculostearicum strain ATCC 35692, and C. urealyticum strain ATCC 43042 were cultured for 18 h in a brain–heart infusion broth supplemented with 1% Tween 80 (Sigma-Aldrich) at 37°C. Bacteria were enumerated before topical application by assessing CFUs using traditional bacteriology techniques and by measuring OD at 600 nm with the use of a spectrophotometer.

For topical association of bacteria, each mouse was associated with bacteria by applying bacterial suspension (5 ml, ∼10⁹ CFU/ml for most experiments, or 10⁸ or 10⁷ CFU/ml for Fig. S1, H and I, as indicated) across the entire skin surface (∼36 cm²) using a sterile cotton swab. Application of bacterial suspension was repeated every other day for a total of four times (or for Fig. S1, H and I, performed only once as indicated). In experiments involving the topical application of various bacterial species or strains, 18-h cultures were normalized using OD₆₀₀ to achieve similar bacterial density (∼10⁹ CFU/ml).

The induction of psoriasis-like inflammation on ear skin was done as previously described (Gray et al., 2013). Mice were treated daily for up to 5 d with 5 mg 5% IMQ cream (Aldera cream; Valeant Pharmaceuticals) topically applied on each ear. Ear-skin thickness was measured using a digital caliper (Mitutoyo). For each mouse, a daily ear-skin-thickness value was calculated by averaging the thickness of both ears. The change in ear-skin thickness over time was reported as the difference related to the first day of IMQ application.

The high-fat diet (TD.06414; 60% of total calories from fat) and the corresponding control diet (TD.150064; 10% of total calories from fat) were purchased from Envigo Teklad Diets. Mice were placed under those diet regimens at weaning (3 wk old) for 4 wk before topical association with bacteria. Ear thickness was measured with a digital caliper (Mitutoyo). For each mouse, the ear-skin thickness value at a defined time point was calculated by averaging the thickness of both ears measured on this time point. The change in ear-skin thickness over time was reported as the difference related to the first day the mice were first given specific diets.

Mice were treated i.p. with anti–TCR Vγ4 antibody (clone UC3-10A6; BioXCell) or Armenian Hamster IgG isotype control antibody (BioXCell). The treatment schedule was as follows: 0.5 mg of either antibody was injected 2 d before topical association with C. accolens; each mouse then received 0.5 mg of either antibody every 3 d starting on the day of the first topical association until analysis.

Cells from the ear pinnae were isolated as previously described (Naik et al., 2012). Single-cell suspensions from the ear draining lymph nodes were prepared by passing tissue through a 70-µm cell strainer after digestion with 0.1 mg/ml Liberase TL purified enzyme blend (Roche) and 0.5 mg/ml DNase I (Sigma-Aldrich) for 30 min at 37°C.

For detection of basal cytokine potential, single-cell suspensions from various tissues were cultured directly ex vivo in a 96-well U-bottom plate in complete medium (RPMI 1640 supplemented with 10% FBS, 2 mM l-glutamine, 1 mM sodium pyruvate and nonessential amino acids, 20 mM Hepes, 100 U/ml penicillin, 100 µg/ml streptomycin, and 50 mM β-mercaptoethanol) and stimulated with 50 ng/ml PMA (Sigma-Aldrich) and 5 µg/ml ionomycin (Sigma-Aldrich) in the presence of BFA (GolgiPlug; BD Biosciences) for 2.5 h at 37°C in 5% CO₂. After stimulation, cells were assessed for intracellular cytokine production as described in the next section.

Mouse single-cell suspensions were incubated with fluorochrome-conjugated antibodies against surface markers CCR6 (clone 29-2L17), CD4 (RM4-5), CD8β (eBioH35-17.2), CD11b (M1/70), CD11c (N418), CD19 (MB19-1), CD31 (MEC13.3), CD34 (RAM34), CD40 (IC10), CD45 (30-F11), CD45.2 (104), CD45R (RA3-6B2), CD49b (DX5), CD49f (eBioGoH3), CD64 (X54-5/7.1), CD90.2 (53–2.1), Ly6C (AL-21), Ly6G (1A8), MHC II (M5/114.15.2), NK1.1 (PK136), Sca-1 (D7), Siglec F (E50-2440), TCRβ (H57-597), TCR Vγ4 (UC3-10A6), and/or γδ TCR (eBioGL3) in HBSS for 20 min at 4°C and then washed. For CCR6 staining, cell suspensions were first incubated with antibody against CCR6 (29-2L17) for 45 min at room temperature before incubation with other fluorochrome-conjugated antibodies against surface markers for 20 min at 4°C. LIVE/DEAD Fixable Blue Dead Cell Stain kit (Invitrogen Life Technologies) was used to exclude dead cells. Cells were then fixed for 30 min at 4°C using the fixation/permeabilization buffer supplied with the BD Cytofix/Cytoperm kit (BD Biosciences) and washed twice with permeabilization buffer. For intracellular staining, cells were then stained with fluorochrome-conjugated antibody against IL-17A (eBio17B7 or TC11-18H10.1) and/or Ki-67 (SolA15) in permeabilization buffer supplied with the BD Cytofix/Cytoperm kit (BD Biosciences) for 1 h at 4°C. Each staining was performed in the presence of purified anti–mouse CD16/32 (93), 0.2 mg/ml purified rat IgG, and 1 mg/ml normal mouse serum (Jackson Immunoresearch). All antibodies were purchased from BD Biosciences, Biolegend, eBioscience, or R&D Systems. Cell acquisition was performed on an LSR Fortessa flow cytometer using FACSDiVa software (BD Biosciences), and data were analyzed using FlowJo software (TreeStar).

Mouse ear pinnae skin samples were sterilely obtained and processed using a protocol adapted from (Grice et al., 2010). For 16S rRNA amplicon Illumina MiSeq sequencing, the DNA from each sample was amplified using Accuprime High Fidelity Taq polymerase (Invitrogen Life Technologies) with universal primers flanking variable regions V1 (primer 27F; 5′-AGAGTTTGATCCTGGCTCAG-3′) and V3 (primer 534R; 5′-ATTACCGCGGCTGCTGG-3′). For each sample, the universal primers were tagged with unique sequences (“barcodes”) to allow for multiplexing/demultiplexing. PCR products were then purified using the Agencourt AmPure XP kit (Beckman Counter Genomics) and quantitated using the QuantIT dsDNA High-Sensitivity Assay kit (Invitrogen Life Technologies). Approximately equivalent amounts of each PCR product were then pooled and purified using Agencourt AMPure magnetic beads before sequencing on an MiSeq instrument (Illumina). Sequencing data were analyzed as previously described (Naik et al., 2012). Data deposition is with the Sequence Read Archive and all sequences can be accessed under BioProject ID no. PRJNA389472 (www.ncbi.nlm.nih.gov/bioproject/389472) and no. PRJNA384162 (www.ncbi.nlm.nih.gov/bioproject/384162).

The ear pinnae of topically associated or unassociated control mice were swabbed with a sterile cotton swab previously soaked in brain–heart infusion broth supplemented with 1% Tween 80. Swabs were streaked on Columbia blood agar plates. Plates were then placed at 37°C under aerobic conditions for 18 h. CFUs on each plate were enumerated, and the number of CFUs was reported per square centimeter of skin.

Vγ4⁺ CCR6⁺ γδ T cells (CD45⁺CD90.2⁺ γδ TCR^low TCR Vγ4⁺ CCR6⁺, >95% purity) were isolated from the skin of unassociated mice or mice associated with S. epidermidis or C. accolens strain ATCC 49725 (day 14 after association) by cell sorting using a FACSAria cell sorter (BD Biosciences). The nCounter analysis system (NanoString Technologies) was used to screen for the expression of signature genes associated with lymphocyte activation. In brief, RNA from each sample was obtained by lysing the sorted cells (10³ cells/µl) in RLT buffer (Qiagen) and then hybridized with the customized Reporter CodeSet and Capture ProbeSet of a custom NanoString chip (132 genes; Table S1), according to the manufacturer’s instructions. mRNA molecules were counted on a NanoString nCounter. Data analysis was performed according to NanoString Technologies recommendations. mRNA counts were processed to account for hybridization efficiency, background noise, and sample content using the R package NanoStringNorm with arguments: CodeCount = “geo.mean,” Background = “mean.2sd,” and SampleContent = “housekeeping.geo.mean.” Each sample profile was normalized to geometric mean of three housekeeping genes (Eef1g, Gusb, and Oaz1). After normalization, genes with mean counts less than two SDs above the mean count of the highest negative control (count > 4.34092) were disregarded, and differential expression of remaining genes was determined using a nonparametric Welch t test with correction for multiple testing using the Benjamini-Hochberg false discovery rate controlling procedure in the multtest package in R. By using those genes with a false discovery rate <0.05, a heatmap was rendered using the R package pheatmap.

For gene expression analysis by microarray of total tissue homogenate, ear pinnae were isolated from unassociated control mice or mice previously associated with C. accolens strain ATCC 49725 (day 14 after association) and stored in RNAlater until RNA extraction. In brief, tissues were lyzed in TRIzol reagent (Invitrogen/Life Technologies) using bead disruption, and total RNA was extracted using phenol/chloroform extraction followed by the miRNeasy kit (Qiagen) according to the manufacturer’s instructions. Total RNA was then used for GeneChip analysis using the Mouse Gene 2.0 ST Array (Affymetrix). Terminal-labeled cDNA synthesis, hybridization to genome-wide Mouse Gene 2.0 ST GeneChips, and scanning of the arrays were performed according to the manufacturer’s instructions (Affymetrix). The microarray data discussed in this publication have been deposited in the Gene Expression Omnibus (GEO) database under accession no. GSE96932. Pathway analysis was performed using Enrichr (http://amp.pharm.mssm.edu/Enrichr/; Kuleshov et al., 2016) based on all genes up-regulated between samples from unassociated and associated mice given the high-fat-diet treatment. Redundant gene lists obtained from the GO term database on the Enrichr site were collapsed. The top nine GO pathways are shown based on enrichment score (−log10 [adjusted P value]).

C. accolens Δ0503, a mutant devoid of mycolic acids, was constructed in the C. accolens background ATCC 49725 by replacing the gene encoding mycolic acid condensase (locus tag: HMPREF0276_0503) with a 21-bp in-frame scar. C. striatum Δ647, a mutant devoid of LAMs and lipomannans, was similarly constructed in the C. striatum background DSM 20668 by replacing the gene encoding the proximal mannan chain elongase (locus tag: HMPREF0276_0647) with a 21-bp in-frame scar. To make each deletion construct, ∼1.2- to 1.5-kb flanking regions up- and downstream of each gene were amplified with the following primer pairs, with the scar sequence underlined: JC_0503UP_F (5′-CAATTGCGGATCCGCCTTCTTGCAGTACACCTCCGGC-3′)/JC_0503UP_R (5′-C​C​C​A​T​C​C​A​C​T​A​A​A​C​T​T​A​A​A​C​A​C​A​T​A​A​A​C​C​G​T​C​T​T​T​C​C​T​A​G​T​G​G-3′) and JC_0503DOWN_F (5′-T​G​T​T​T​A​A​G​T​T​T​A​G​T​G​G​A​T​G​G​G​T​G​A​G​A​A​G​C​T​C​G​C​T​G​A​C​C​T​G-3′)/JC_0503DOWN_R (5′-CGAATTCTGCAGCTAGTTAATATCGCCGCTCAAGTTCTTCGACCC-3′); and JC_0647UP_F (5′-CAAGACAATTGCGGATCCAGCCATGTTCTCCGTGTTGATG-3′)/JC_0647UP_R (5′-C​C​C​A​T​C​C​A​C​T​A​A​A​C​T​T​A​A​A​C​A​C​A​G​T​G​A​G​A​C​G​C​T​G​T​T​T​C​A​T​A​A​A​G-3′) and JC_0647DOWN_F (5′-T​G​T​T​T​A​A​G​T​T​T​A​G​T​G​G​A​T​G​G​G​A​G​T​G​T​C​C​T​A​A​A​C​T​A​G​T​C​C​T​T​C​G​T​G-3′)/JC0647UP_R (5′-AAGCTTCGAATTCTGCAGCTCGACAACACCCATGAGAATG-3′).

The flanking fragments were then assembled by circular polymerase extension cloning (Quan and Tian, 2011) into pJSC232 (Converse and Cox, 2005), which was first amplified with primer pair JC_pJSC_F (5′-GCGGCGATATTAACTAGCTGCAGAATTCGAAGCTTATCG-3′)/JC_pJSC_R (5′-GGTGTACTGCAAGAAGGCGGATCCGCAATTGTCTTGGATCGG-3′) to generate ends compatible with the inserts. The circular polymerase extension cloning reaction of each resulting deletion construct was introduced into E. coli Top10 by electroporation, and both deletion constructs were verified by colony PCR and sequencing.

Competent C. accolens and C. striatum cells were generated as previously described for C. glutamicum (Eggeling and Bott, 2005) with the following adaptations. Cells were cultivated in brain–heart infusion sorbitol medium supplemented with 1% (wt/vol) of Tween 80 (BHIST), and cultures were incubated at 37°C. 2 µg of each deletion construct was introduced into a 150-µl aliquot of fresh, competent C. accolens or C. striatum by electroporation in a 0.2-cm cuvette (at 25 μF, 200 Ω, and 2.5 kV on a MicroPulser; Biorad) and immediately heat-shocked for 6 min at 46°C. The cells were allowed to recover for 90–120 min with shaking at 37°C before plating on BHIST plus 15 µg/ml kanamycin (Km). The plates were then incubated for 3–4 d at 37°C to obtain single crossover integrants, verified by colony PCR. Double-crossovers were obtained by plating single crossovers on brain–heart infusion plus 1% (wt/vol) Tween 80, supplemented with 20% sucrose to allow for counterselection of the sacB marker on the pJSC232 backbone. Deletion of the native gene was verified by colony PCR. We could not detect any extractable corynomycolic acids in C. accolens Δ0503 after whole-cell methanolysis (Minnikin et al., 1980). The colony morphology and growth defect are consistent with the equivalent mutant in C. glutamicum (Gande et al., 2004). In C. striatum Δ647, lipoglycan extraction and analysis by SDS-PAGE gel showed no bands corresponding to LAMs or lipomannans, similar to the equivalent mutant in C. glutamicum (Mishra et al., 2008).

TEEs were prepared from the cell pellets of 48-h cultures of C. accolens strain ATCC 49725. Cell pellets were first washed in distilled water and then subjected to extraction for 16 h with chloroform/methanol (1:2 vol/vol) at room temperature, followed by reextraction with chloroform/methanol (1:1 vol/vol) and finally with chloroform/methanol (2:1 vol/vol) for 16 h each (Bou Raad et al., 2010). The three organic phases resulting from the extraction procedures were pooled, concentrated by means of rotary evaporation, and then partitioned between aqueous and organic phases using a mixture of chloroform/water (1:1 vol/vol). The organic phase was retained and evaporated to dryness yielding the TEEs from C. accolens. Liposomes were prepared as previously described (Korf et al., 2005). 250 µg TEE or LAM from Mycobacterium smegmatis (Invivogen) was mixed with 45 µl l-α-phosphatidylcholine at 100 mg/ml in chloroform (Sigma-Aldrich). Empty liposomes were used as a negative control. Liposomes were dried down and dissolved with sonication in 1 ml PBS before use in mononuclear phagocytes–γδ T cell co-culture assays.

Vγ4⁺ CCR6⁺ γδ T cells (CD45⁺CD90.2⁺ γδ TCR^lowTCR Vγ4⁺ CCR6⁺, >95% purity) were sorted by flow cytometry from the ear-skin tissue and ear draining lymph nodes of C57BL/6 mice 2 wk after topical association with C. accolens strain ATCC 49725 using a FACSAria cell sorter. For splenic DC purification, single-cell suspensions from the spleen of congenic WT mice were magnetically enriched for CD11c⁺ cells by positive selection using CD11c MicroBeads and MACS separation columns (Miltenyi Biotec). Purified DCs and Vγ4⁺ CCR6⁺ γδ T cells were co-cultured at a 15:1 ratio (10⁴ Vγ4⁺ CCR6⁺ γδ T cells) in a 96-well U-bottom plate in complete medium for 16–18 h at 37°C in 5% CO₂. BFA (GolgiPlug) was added for the final 4 h of culture. DCs were previously incubated for 2 h with or without hk C. accolens strain ATCC 49725 (WT), hk corynomycolic acid synthase–deficient C. accolens strain Δ0503, hk C. amycolatum, hk C. striatum strain DSM 20668 (WT), or hk C. striatum mutant strain Δ647 (bacteria/DCs ratio, 500:1) and washed before co-culture with Vγ4⁺ CCR6⁺ γδ T cells. For some experiments, DCs were preincubated with 100 µl of one of the following liposome suspensions prepared as described in the previous section—empty liposomes or liposomes containing one of the following: C. accolens TEE, or LAM from M. smegmatis. In other experiments, Vγ4⁺ CCR6⁺ γδ T cells (10⁴) were cultured in a 96-well U-bottom plate for 16–18 h at 37°C in 5% CO₂ in presence of the filtered culture supernatant (dilution 1:3 in complete medium) of DCs incubated for 4 h with hk C. accolens strain ATCC 49725. IL-17A production by Vγ4⁺ CCR6⁺ γδ T cells after co-culture was assessed by flow cytometry after intracellular cytokine staining using the following antibodies: α-CD45.1 (clone A20), α-CD45.2, α-CD90.2, α-TCRβ, α-TCR Vγ4, α-γδ TCR, and α-IL-17A.

FACS-purified CD45⁺ CD90.2⁺ γδ TCR^low TCR Vγ4⁺ CCR6⁺ T cells from the skin of naive unassociated or C. accolens–associated mice were cultured overnight at 37°C in 5% CO₂ in 200 µl complete culture medium in a 96-well U-bottom plate in the presence or absence of various concentrations of recombinant IL-1β or IL-23 (Peprotech). Supernatants were collected, and levels of IL-17 were assessed using a bead-based cytokine detection assay (FlowCytomix, eBioscience).

Mice were euthanized 14 d after topical application of C. accolens strain ATCC 49725. Unassociated mice were used as controls. The ears from each mouse were removed and fixed in PBS containing 10% formalin. Paraffin-embedded sections were cut at 0.5 mm, stained with hematoxylin and eosin, and examined histologically.

Data are presented as mean ± SEM or mean ± SD. Group sizes were determined based on the results of preliminary experiments. Mice were assigned at random to groups. Mouse studies were not performed in a blinded fashion. Generally, each mouse of the different experimental groups is reported. Statistical significance was determined using one of the following statistical tests: one-way or two-way ANOVA or two-tailed unpaired Student’s t test under the untested assumption of normality. Two-way ANOVA was used to determine if a specific immune phenotype was affected by two factors, in an independent manner or because of an interaction of factors studied. We used a Holm-Šídák test for correction for multiple hypotheses. Within each group there was an estimate of variation, and the variance between groups was similar. All statistical analysis was calculated using Prism software (GraphPad). Differences were considered to be statistically significant when P < 0.05.

Fig. S1 shows the impact of diverse skin commensal microbes on various populations of immune cells in the skin, the absence of inflammation after cutaneous association with C. accolens, as well as the number of CFUs present on the ear pinnae and the impact of C. accolens after topical application with various doses of bacteria. Fig. S2 shows the absolute numbers of Vγ4⁺ IL-17A⁺ cells in the skin of WT, B2m^−/−, Cd1d^−/−, Clec4e^−/−, Card9^−/−, and Myd88^−/− mice after association with C. accolens, as well as IL-17 secretion by skin γδ TCR⁺ Vγ4⁺ CCR6⁺ cells from naive mice after treatment with IL-1 or IL-23. Fig. S3 shows the change in skin thickness and absolute numbers of Vγ4⁺ IL-17A⁺ cells, neutrophils, inflammatory monocytes, and TCRβ⁺ cells in the skin after IMQ topical application in mice previously associated 2 wk earlier with C. accolens or C. amycolatum compared with unassociated control mice. In addition, Fig. S3 shows the change in skin thickness in unassociated mice receiving a control or high-fat diet, as well as absolute numbers of neutrophils, eosinophils, and inflammatory monocytes in unassociated and C. accolens–associated mice given a high-fat or control diet regimen. Table S1 shows the gene list of the custom NanoString CodeSet used to analyze activation-associated lymphocyte gene expression.

We thank the National Institute of Allergy and Infectious Diseases (NIAID) animal facility staff, in particular, D. Trageser-Cesler and C. Acevedo (NIAID Microbiome Program Gnotobiotic Animal Facility); K. Beacht and J. LeGrand for technical assistance; J. Davis (NIAID Microbiome Program Sequencing); Dr. A.G. Elkahloun (National Human Genome Research Institute Microarray Core); Dr. G. Trinchieri (Myd88^−/− mice); Dr. A. Sher (Clec4e^−/− mice); and Dr. M. Lionakis (Card9^−/− mice). We also thank the Belkaid laboratory for critical reading of the manuscript.

This work was supported by the Division of Intramural Research of the NIAID, a Howard Hughes Medical Institute–Simons Foundation Faculty Scholar Award (to M.A. Fischbach), a Burroughs Wellcome Fund Investigators in the Pathogenesis of Infectious Disease Award (to M.A. Fischbach), and National Institutes of Health grants R01 AI101018, R01 DK110174, and DP1 DK113598 (to M.A. Fischbach). V.K. Ridaura and M.G. Constantinides are Cancer Research Institute Irvington Fellows supported by the Cancer Research Institute. S. Tamoutounour is supported by a European Molecular Biology Organization Long-Term Fellowship.

The authors declare no competing financial interests.

Author contributions: V.K. Ridaura, N. Bouladoux, Y. Belkaid, and M.A. Fischbach designed the studies. V.K. Ridaura, N. Bouladoux, J. Claesen, and Y. Erin Chen performed experiments and analyzed the data. M.G. Constantinides and E.D. Merrill assisted with in vitro co-culture studies and in vivo treatment with IMQ, respectively, and provided valuable intellectual input. S. Tamoutounour performed analysis of keratinocytes and dendritic cell/monocyte/macrophage populations. A.L. Byrd performed microarray data analysis. V.K. Ridaura, N. Bouladoux, and Y. Belkaid wrote the manuscript.