Endogenous retroviruses promote homeostatic and inflammatory responses to the microbiota

1Metazoans exist as meta-organisms composed of both the host and its symbiotic microbiota. These complex communities of microbes broadly control host physiology, including numerous aspects of the immune system. The immune system recognizes these resident microbes at all barrier surfaces; this response includes the induction of classical and non-classical T cell responses directed at the microbiota itself. What differentiates these responses from those resulting from pathogenic microbes is that both the initiation of immune responses and accumulation of commensal-specific lymphocytes within tissues occur in the absence of inflammation, a process referred to as homeostatic immunity. Immunity to the microbiota controls numerous aspects of host physiology ranging from antimicrobial defense to tissue repair. Notably, exposure to defined members of the skin microbiota promotes the accumulation of various lymphocyte subsets within the skin that can promote tissue repair. Further, one of the dominant actions of these cells is mediated via the release of interleukin-17A (IL-17A), which can promote antimicrobial responses by keratinocytes.

2Our current understanding of immunity postulates that the induction of immune responses requires inflammation or tissue damage. However, within this framework, it is unclear how the immune system actively recognizes and responds to the microbiota in the absence of classical inflammatory processes. More specifically, the factors responsible for recognizing commensal microbes discretely within tissues remain poorly understood. Based on the evolutionary alliance between the microbiota and the immune system, we hypothesized that such mechanisms may be highly ubiquitous and conserved.

3Of particular interest, the microbiota promotes type I interferon (IFN-I) and antiviral states in tissues; from these findings, the microbiota is thought to protect the host against various viral infections. Recent evidence also supports the idea that, within the gut, the microbiota promotes IFN-I production through viral sensors, including stimulator of interferon genes protein (STING). Together, these observations raise the intriguing possibility that responses to the microbiota may be entwined with responses to endogenous viral elements.

4Retroviruses can integrate into the germline, thus becoming part of the host genome. As such, the long-standing symbiotic relationship between mammals and microbes encompasses not only those with exogenous microbes, such as bacteria, fungi, and viruses, but also endogenous retroviruses (ERVs). Most ERVs are inactive; however, several ERVs express intact open reading frames (ORFs). Numerous physiological and pathological processes rely on the transcriptional activity of ERVs, and thus these retroelements are under constant immune pressure. In several contexts, microbial exposure can control ERV expression; germ-free mice lose basal intestinal expression of several ERV groups, and microbe-derived products, such as Toll-like receptor (TLR) ligands, drive transcription of defined ERVs. However, whether ERVs directly contribute to the induction of responses to the microbiota remains unknown. Further, to what extent heightened expression of ERVs could, in defined settings, contribute to the inflammatory impact of the microbiota has not been addressed.

5Here, we propose that the immune system may have co-opted its endogenous virome as a way to communicate with its exogenous microbiome. Notably, our results support the idea that the intensity and quality of immune responses to the microbiota under both steady-state and inflammatory settings are controlled by the expression of ERVs.

6Skin colonization with commensals promotes the accumulation of various T cell subsets within the skin in a manner uncoupled from inflammatory processes. Notably, skin colonization of specific pathogen-free (SPF) mice with Staphylococcus epidermidis induced a significant increase in both the numbers and frequencies of Tc17 (IL-17A producing CD8+ T cells), Tc1 (IFN-γ producing CD8+ T cells), Th17 (IL-17A producing CD4+ T cells), Th1 (IFN-γ producing CD4+ T cells), mucosal-associated invariant T (MAIT), and IL-17A producing γδ T cells detectable by day 5, peaking at 14 days post-association, and maintained for several months (Figures 1A and S1A;;; ). Keratinocytes represent the primary interface between the host and its microbiota. To uncover upstream mechanisms responsible for the induction of homeostatic immunity to the microbiota, we characterized keratinocyte responses to S. epidermidis colonization prior to the peak of T cell responses (7 days post-association). The gene expression of interfollicular keratinocytes (Figure S1B) exposed to S. epidermidis was distinct and clustered separately from keratinocytes isolated from controls (Figure S1C). Gene Ontology (GO) pathways enriched in the transcriptome of purified keratinocytes from S. epidermidis colonized mice revealed upregulation of numerous genes linked to immune activation compared to keratinocytes from controls (Figures 1B and 1C). In agreement with enhanced T cell accumulation, keratinocytes from S. epidermidis colonized mice expressed elevated levels of chemokines (Figure 1C) associated with lymphocyte homing. Further, and as we have previously shown, colonization increased expression of genes implicated in antigen processing and presentation (; Figures 1C and 1D). Of particular interest, and despite the lack of tissue inflammation, gene expression associated with IFN-I and antiviral responses were among the most upregulated within keratinocytes (Figure 1D). Notably, keratinocytes from colonized mice expressed higher transcript levels of numerous interferon-stimulated genes (ISGs) compared to keratinocytes from control mice (Figure 1D).

7To assess whether IFN-I contributed to immune responses to S. epidermidis, mice were treated with a blocking antibody for the IFN-I receptor, interferon alpha receptor 1 (IFNAR1), or isotype control during colonization. Treatment with anti-IFNAR1 antibody significantly reduced the number of all T cell subsets induced by S. epidermidis as well as their ability to express IL-17A and IFN-γ, with the exception of IL-17A production by CD4+ T cells (Figures 1E and S1D). Thus, S. epidermidis association triggers an antiviral program in keratinocytes and IFN-I promotes T cell responses to this commensal.

8The antiviral signature induced by S. epidermidis colonization supported the idea that response to skin commensals may be associated with the sensing of endogenous viral elements. The mammalian genome contains numerous and diverse endogenous retroelements, the activation of which is potentially sensed by the immune system during microbial colonization. Relevant to our observation, the induction of ERVs has been previously associated with the induction of IFN-I responses. Among other groups, these retroelements include two distinct groups in terms of phylogeny and replication cycle: ERVs, which are derived from ancestral integrations of exogenous retroviruses into the germline and long interspersed nuclear elements (LINEs) that are conserved in all vertebrates but highly expanded in mammals, including humans.

9Skin association with S. epidermidis promoted the expression of several families of endogenous retroelements within the skin (Table 1; Figure 2A). Increased expression of retroelement genes from three major families of endogenous retroelements was also detected in purified keratinocytes: the LINE-1 elements; the intracisternal alpha particle (IAP) elements (an ERV family related to betaretroviruses); and one locus, Chr5 23.7M (also known as Xmv45; Table 1), from the murine leukemia virus (MLV) group, an ERV family related to the extant gammaretrovirus murine leukemia virus (Table 1; Figures 2A, S2A, and S2B).

10We next tested the possibility that ERVs could coordinate responses to the microbiota via engagement of nucleic acid innate sensors. Retroelements can produce several molecular patterns, including double-stranded RNA (dsRNA) produced by abortive transcription and complementary DNA (cDNA) produced by reverse transcription. Increased expression of retroelement loci can create enhanced levels of dsRNA molecules; however, to increase cDNA levels, induced retroelement loci must encode functional copies of the genes required for reverse transcription. For example, LINE-1 elements require their encoded RNA chaperone (ORF1) and reverse transcriptase (ORF2) to be intact to reverse transcribe their mRNA molecules in cis and ERVs require the viral protease/polymerase polyprotein (pro-pol) gene to be functional in order to reverse transcribe. We next assessed whether any retroelement loci upregulated by S. epidermidis colonization carried functional copies of these genes. We examined the expression of the distinct retroelement ORFs separately as a way to broadly assess the genomic integrity of each upregulated retroelement.

11Although both ORF1 and ORF2 are required for retrotransposition, the LINE-1 elements upregulated by S. epidermidis either contained a truncated ORF2 or lacked ORF1 (Figure S2C). All three ERV loci induced by S. epidermidis encoded and transcribed the pro-pol ORF from their genomes; however, only the Chr5 23.7M locus contained a functional gag gene that was in frame with the pro-pol gene as well as a pro-pol gene that was not truncated (Table 1). In line with these observations, keratinocytes from mice associated with S. epidermidis also expressed higher levels of MLV surface glycoprotein (MLV SU) (Figure 2B). This supported the idea that MLV-derived ERVs may contribute to cDNA-induced innate responses in the context of skin microbial colonization.

12We previously showed that the ability of S. epidermidis to induce CD8+ T cells was partially dependent on bacterial cell envelope teichoic acid and lipoproteins and host TLR2 expression. Microbial products and, in particular, TLR ligands can promote the expression of retroviral elements by various cells, including macrophages and B cells. To assess whether such a phenomenon was also true for keratinocytes, primary keratinocytes were exposed in vitro to various TLR agonists and MLV surface expression of the glycoprotein (MLV SU) on keratinocytes was measured. All TLR ligands tested, including ligands for TLR4 (lipopolysaccharide [LPS]), TLR2 (lipotheicoic acid [LTA]), and TLR7/8 (resiquimod [R848]), significantly enhanced the level of MLV SU expression by keratinocytes (Figure S2D). Further, increased MLV SU expression by keratinocytes following S. epidermidis association was reduced in TLR2-deficient mice, supporting the idea that TLR2 signaling contributes, at least in part, to ERV induction in the skin (Figure 2C).

13S. epidermidis teichoic acid (ΔtagO) mutant was unable to induce any assessed ERV family (Figure 2D) following association, and the lipoprotein (Δlgt) mutant still induced MMTV and ETnII ERVs but was unable to induce any other ERV families assessed (Figure 2D). In agreement, the level of expression of MLV SU glycoprotein by keratinocytes following association with each mutant was significantly reduced compared to wild-type (WT) bacteria (Figure 2E). Taken together, these data support the notion that S. epidermidis wall teichoic acids and lipoproteins promote the induction of ERVs in the skin.

14We next assessed whether cDNA synthesis of endogenous retroelements controlled T cell responses to S. epidermidis. Mice were treated with a combination of antiretrovirals, tenofovir and emtricitabine, which are nucleotide and nucleoside reverse transcriptase inhibitors, respectively, or vehicle control prior and during association (Figure 2F). These compounds prematurely terminate nascent cDNA synthesis during reverse transcription. Imaging of skin revealed that antiretroviral treatment profoundly inhibited T cell accumulation within the epidermis following S. epidermidis association (Figure 2G). Antiretroviral treatment significantly reduced the absolute number of S. epidermidis specific CD8+ T cells (f-MIIINA:H2-M3+; ) induced by S. epidermidis within the skin (Figure 2I), but not in the spleen (Figure S2G). Antiretroviral treatment also significantly decreased the absolute number of all T cell subsets induced by S. epidermidis association and significantly reduced IL-17 expression by CD8+, MAIT, and γδ T cells within the skin but did not impact T cells in secondary lymphoid organs (except γδ T cells; Figures 2H, S2E, and S2F). Antiretroviral treatment had no impact on ERV expression, supporting the idea that the impact of this treatment was associated with reduced responses to cDNA rather than changes in ERV mRNA levels (Figure S2H).

15Of note, the genome of S. epidermidis did not contain homologs of the conventional bacterial reverse transcriptases ltrA and RT-Cas1. Antiretrovirals did not impact the growth rate or overall growth yield of S. epidermidis in vitro and did not impair the ability of S. epidermidis to colonize the skin (Figures S2I and S2J). Thus, host reverse transcriptase activity is required to promote cognate responses to S. epidermidis in the skin. Further, other common skin commensals, such as Staphylococcus aureus and Staphylococcus xylosus, also promoted ERV expression within the skin (Figure 2J), supporting the idea that expression of retroelements could represent a general response to microbiota colonization.

16We have previously shown that T cell responses to the microbiota promotes several aspects of tissue physiology, including wound healing, via, in part, their action on keratinocytes. As such, we assessed the impact of antiretrovirals on keratinocyte responses at the peak of T cell responses to S. epidermidis via single-cell RNA sequencing (scRNA-seq) (Figures S3A and S3B). S. epidermidis induced expression of genes associated with wound healing, antigen presentation on major histocompatibility complex (MHC) class I, and antiviral defense (Ifitm3) in clusters from the interfollicular epidermal basal layer that were significantly decreased by antiretroviral treatment (Figure 3A). T cell responses to commensals promotes the development of organized cellular clusters within the epidermis that are surrounded by keratinocytes expressing a discrete program associated with antigen presentation, including high level of MHC class II (MHC-II), and antimicrobial defense. MHC-II+ keratinocytes display enhanced proliferation rates and higher levels of activation than MHC-II− keratinocytes following S. epidermidis association. In agreement, genes linked with antigen presentation via MHC-II that are expressed in defined clusters of interfollicular epidermal and infundibular basal keratinocytes were significantly decreased by antiretroviral treatment (Figures S3A and S3C). More particularly, one cluster (cluster 12), enriched in genes associated with MHC-I and MHC-II antigen presentation, antimicrobial peptides, and wound healing, was only observed in S. epidermidis associated mice and was the most significantly decreased following antiretroviral treatment (Figures 3B and S3C). In line with the concept that MHC-II-expressing keratinocytes may represent “hot spots” for microbiota-induced T cells responses, MHC-II+ keratinocytes expressed higher level of MLV SU than MHC-II− keratinocytes following commensal colonization (Figure S3D). The impact of antiretroviral treatment on keratinocytes was further confirmed by their decreased proliferation and expression of MHC-II following S. epidermidis association compared to controls (Figure 3C). Thus, antiretroviral treatment, via its action on T cell responses and/or keratinocytes, impacted the ability of S. epidermidis to promote a gene expression program consistent with antimicrobial defense and tissue repair.

17Naive and S. epidermidis colonized mice, treated or not with antiretroviral drugs, were subjected to back-skin punch biopsy, and 5 days later, the epidermal tongue of proliferating keratinocytes was measured. In line with reduced T cell responses, antiretrovirals significantly impaired the ability of S. epidermidis to promote wound healing compared to vehicle-treated mice (Figures 3D and 3E). Thus, inhibition of host reverse transcriptase has a broad impact on T cell responses to S. epidermidis and functionally inhibited homeostatic immunity to the microbiota.

18We next sought to identify DNA sensors potentially involved in these responses. Retroviruses have been previously shown to activate cGAS to produce cGAMP, which binds to and activates the adaptor protein STING. Under steady-state conditions, numbers of CD8+, CD4+ T cells, and IFN-γ-expressing CD8+ T cells (Tc1) were significantly decreased in the skin of cGas−/− and Sting−/− mice (Figure S4A). Thus, in SPF mice, the tonic accumulation of type 1 T cells in the skin depends on nucleic-acid-sensing pathways. Following S. epidermidis association, the absolute number of all T cell subsets analyzed as well as their ability to produce cytokines were significantly reduced in both cGas−/− and Sting−/− mice compared to WT controls (Figures 4A and S4B). The antiviral signature and upregulation of ERV expression by keratinocytes following S. epidermidis association supported a primary role for these cells in sensing ERV-derived cDNA molecules (Figures 1D, 2A, and 2B). Under steady-state conditions, deletion of STING in keratinocytes (Krt14Cre+Stingf/f) had no impact on microbial colonization or ERV expression (Figures S4C and S4D). On the other hand, lack of STING only in keratinocytes (but not in CD11c-expressing cells, which include dendritic cells [DCs]) was sufficient to reduce the numbers of Tc1 and Th1 cells within the skin (Figures 4B and S5). Deletion of STING in keratinocytes significantly reduced the number of all T cell subsets as well as cytokine production following association with S. epidermidis in a manner comparable to that seen in Sting−/− mice and without impacting the level of ERV expression (Figures 4A, 4B, S4D, and S4E). Together, our results propose that keratinocyte responses to microbial colonization promotes a discrete upregulation of ERVs resulting in cGAS/STING activation that is required for the induction of homeostatic T cell responses to a skin microbe (Figure S4F).

19Although homeostatic immunity to the microbiota promotes immune fitness, dysregulated responses to these symbionts can also promote tissue inflammation. Heightened expression or sensing of ERVs has also been associated with inflammation and disease states. To test the possibility that ERVs could contribute to the pathogenic impact of the microbiota, we employed a model of dietary intervention, having previously shown that a high-fat diet can promote inflammatory response to the microbiota. The consumption of a diet rich in saturated fatty acids promotes inflammation in a nuclear factor κB (NF-κB)-dependent manner; a pathway previously linked to the induction of ERVs, and obesity increases the prevalence and severity of inflammatory disorders. Mice were placed on a high-fat or control diet for 6 weeks prior to association with S. epidermidis (Figure 5A). At this time point, mice showed modest weight gain and did not develop skin inflammation (Figures S6A and S6B). High-fat diet alone did not impact keratinocyte proliferation or MHC-II expression (Figures S6C and S6D), whereas the number of Tc1, IL-17A-producing MAIT, and γδ T cells within the skin was increased compared to controls (Figure 5D). As previously described, S. epidermidis association of mice fed a regular diet did not result in inflammation (Figures 5B and 5C). On the other hand, when applied to mice fed a high-fat diet, S. epidermidis promoted a significant increase in skin thickening associated with a hyperplastic epidermis and hyperkeratosis (Figures 5B and 5C). Altered skin physiology was detectable as early as 3 days post-association and plateaued at 7 days and was associated with a significant increase in the number of all T cell subsets and cytokine production potential compared to controls (Figures 5B, 5D, and S6E).

20In contrast to high-fat diet alone, S. epidermidis colonization under high-fat conditions increased the expression of genes associated with inflammatory responses in skin, including cell proliferation, positive regulation of apoptosis, chemotaxis, and the IL-36 signaling pathway (Figures 5E and S6F). Genes associated with keratinization and antigen presentation pathways were more significantly induced by S. epidermidis colonization under high-fat conditions than in the context of a regular diet (Figures 1D and 5E). Altered activation and inflammatory state of keratinocytes was further confirmed by a significant increase in MHC-II expression and proliferation in response to S. epidermidis in high-fat-diet-fed mice compared to controls (Figures S6C, S6D, and S6F).

21We next assessed the level of ERV expression under the inflammatory conditions. Multiple families of ERV were strongly induced in the skin of mice fed a high-fat diet and colonized with S. epidermidis (Figure S6G). Further, mice colonized with S. epidermidis under a high-fat regimen highly expressed MLV ERV loci in keratinocytes relative to keratinocytes isolated from mice fed a high-fat diet alone (Figure 5F). The induced retroelements were distinct from those induced in the context of a high-fat diet or S. epidermidis alone (Figure 5F). Numerous ERV loci were induced by S. epidermidis colonization in the skin of mice fed a high-fat diet, all of which encoded a functional reverse transcriptase (Figure 5G). Consistent with these observations, keratinocytes from mice fed a high-fat diet and associated with S. epidermidis also expressed higher levels of MLV SU compared to mice fed a high-fat diet alone or mice fed a control diet and associated with S. epidermidis (Figures 2B and S6H). Thus, high-fat diet promotes heightened ERV expression in response to S. epidermidis association.

22To assess a potential role for ERVs in other skin inflammatory settings, we evaluated ERV expression in an experimental model of psoriasis. Imiquimod treatment of mice was associated with a strong upregulation of the locus also upregulated in response to S. epidermidis (Chr5 23.7M), with imiquimod inducing expression of a defective locus (Chr2 68.6M) and heightened levels of MLV SU expression by keratinocytes compared to controls (Figures S6I and S6J). To assess the potential relevance of our findings to settings of human skin inflammation, we analyzed skin samples from psoriatic patients. Our results revealed a strong induction of retroelement loci in psoriatic lesions relative to normal skin within the same patient and, more particularly, specific retroelement families, including two LINE-1 families, L1P1 and L1HS, and the ERV family HERVK9 (Figure 5H). Thus, skin inflammation in mice and humans can be associated with enhanced ERV expression.

23We next assessed whether, in the context of high-fat diet, IFN-I and cytosolic nucleic acid sensors contributed to the pathology induced by S. epidermidis. The ability of S. epidermidis to cause skin inflammation and promote T cell responses in the context of a high-fat diet was significantly reduced in cGas−/−, Sting−/−, and Ifnar1−/− mice (Figures 6A, 6B, and S7A). Defects in DNA sensing more significantly impacted T cell responses to S. epidermidis than defects in IFN-I signaling, supporting the idea that other inflammatory mediators contributed to skin inflammation under these conditions (Figures 6B and S7A).

24We next assessed the contribution of keratinocyte-specific nucleic acid sensing. Skin pathology induced by S. epidermidis in mice fed a high-fat diet was significantly reduced in Krt14Cre+Stingfl/fl as compared to control Krt14Cre-Stingfl/fl mice (Figure 6C). In mice fed a high-fat diet, the absolute number of IL-17A-producing γδ T cells and MAIT cells, but not other cell subsets, was significantly reduced in Krt14Cre+Stingfl/fl compared to controls (Figure 6D). Following S. epidermidis association, the absolute number of all T cell subsets induced by S. epidermidis was significantly reduced in the skin of high-fat fed Krt14Cre+Stingfl/fl mice compared to controls (Figure 6D). The ability of CD8+ T cells to produce IL-17A and IFN-γ and MAIT and γδ T cells to produce IL-17A was also significantly reduced in Krt14Cre+Stingfl/fl mice compared to control mice (Figures 6D and S7B). Thus, DNA sensing by keratinocytes contributes to the ability of the microbiota to promote inflammation in the context of metabolic alteration.

25To evaluate a potential role for ERVs and reverse transcriptase in driving the inflammatory responses caused by microbes under high-fat-diet conditions, mice were treated with vehicle or antiretrovirals prior to S. epidermidis association and thereafter (Figure 6E). Under these settings, skin inflammation caused by S. epidermidis was significantly reduced compared to controls (Figure 6F). This level of reduction was comparable to that observed in STING-deficient mice or that with a keratinocyte-specific STING deletion (Figures 6A and 6C). Antiretroviral treatment was also associated with significant reduction in the number of all T cell subsets as well as a decrease in the ability of CD8+ T cells to produce cytokines (Figures 6G and S7C). Taken together, these data support the idea that the ERV/cGAS/STING/IFN-I axis plays a fundamental role in context-specific responses to the skin microbiota.

26Here, we propose that the host endogenous virome dominantly controls the alliance between the microbiota and the immune system. Our results also propose that the level of ERV expression fundamentally controls the threshold of tissue activation and dictates the ability of the host to respond in homeostatic or inflammatory manner to the microbiota.

27The mechanisms underlying the physiological initiation of homeostatic immunity to the microbiota are unclear. Our present work supports the idea that the host endogenous virome may represent a fundamental adjuvant of these responses. Thus far, our model proposes that discrete sensing of the microbiota by keratinocytes promotes the expression of defined ERVs and that these ERVs reverse transcribe, resulting in cytosolic accumulation of ERV-derived cDNA. The cGAS/STING signaling pathway then senses these cDNAs. These discrete keratinocyte hot spots could promote a favorable environment for DCs to capture microbiota-derived antigens and migrate to lymph nodes. Following priming, commensal-specific T cells migrate back to the skin, where their accumulation and ERV-activated keratinocytes may also promote T cell function. In the context of non-classical T cells that expand in tissues, such as MAIT cells, local responses to ERVs may be sufficient to control their accumulation in the skin, post-microbiota association. The result of this multi-kingdom dialog is the induction and accumulation of a network of commensal-specific T cells able to broadly promote tissue physiology, including tissue repair (Figure S4F).

28Although ERVs represent a sizable proportion of the mammalian genome, most ERVs are inactive. However, many ERVs have retained retroviral characteristics, including the ability to reverse transcribe and, in some cases, form virus-like particles. As such, these elements are under constant immune pressure and in a state of equilibrium with the host, a dialog proposed to contribute to the tonic activation of the immune system. TLR signaling induces these elements, a phenomenon that we confirmed also occurs in keratinocytes. In line with our observations, S. epidermidis derived lipoproteins and teichoic acids, together with host expression of TLR2/Dectin-1, are required for the optimal induction of commensal-specific T cells. In agreement with a role for these pathways in promoting ERV expression, host expression of TLR2 and S. epidermidis expression of teichoic acids and lipoproteins are required for optimal induction of ERVs in the skin. Of particular interest, S. epidermidis colonization promoted the selective induction of ERVs. We speculate that these defined sequences may have lost a repressor sequence or gained a transcription factor binding site that is active in keratinocytes. In support of this hypothesis, the strong responsiveness of Chr5, 23.7M to LPS stimulation in lymphocytes is, at least in part, due to the LPS responsiveness of the lncRNA AI506816 promoter near the Chr5, 23.7M integration site. Of interest, the induction of ERV families was variable among skin commensals, a phenomenon that could contribute to the ability of defined microbes to impose unique immune signatures. These observations also raise an intriguing possibility that microbe- and cell-type-specific expression of ERVs may control local immune microenvironments. In support of this, expression of retroelements is tissue specific and expression of defined ERV families is lost in the absence of the microbiota. Whether a role for ERVs is restricted to sites with low microbial biomass, such as the skin or the lung remains unclear, but based on the prevalence of retroelements in the mammalian genomes, all responses to the microbiota may be, at least in part, ERV dependent.

29Antiretroviral treatment not only reduced T cell accumulation within the skin but also impaired subsequent responses of keratinocytes, including decreased expression of genes involved in antiviral responses and wound healing. In agreement, antiretroviral treatment had a significant impact on S. epidermidis accelerated wound repair. Epithelial cells not only dictate the induction of responses directed toward the microbiota but also provide specific checkpoints allowing functional licensing and retention of T cells within tissues. For instance, we previously showed that MHC-II expression by keratinocytes was required for homeostatic Th1 responses to the microbiota. Indeed, the microbiota promoted the expression of genes associated with T cell homing, as well as numerous genes associated with MHC class I and MHC-II antigen presentation and processing pathways, with a substantial fraction of these impacted by antiretroviral treatment. Of note, MHC-II+ keratinocytes express elevated levels of MLV SU, indicating that these cells may represent hot spots for the initiation (e.g., via DC activation) and/or accumulation (e.g., via chemokine production) of T cell responses to the microbiota. Further, ERV-activated keratinocytes may play an important role in the local presentation of microbiota-derived antigens to commensal-specific T cells.

30In agreement with a role for STING, IFN-I also promoted commensal-specific T cell responses. Within the skin, IFN-I could act on numerous cell types, including keratinocytes, DCs, or T cells. The microbiota has been shown to promote a broad antiviral program within tissues, which was associated with their ability to enhance protective immunity against viral infections. The skin microbiota also induces IFN-I to promote tissue repair. Within the gut, microbiota-induced IFN-I has also been proposed to rely on viral sensors, such as STING and mitochondrial antiviral signaling protein (MAVS). Thus, the ability of the microbiota to promote IFN-I-dependent responses at all barrier sites, including protective immunity and tissue repair, may be controlled by ERV activity and innate responses against these endogenous retroelements.

31Pathologies that increasingly affect humans, such as allergies, autoimmunity, and inflammatory disorders, have all been linked to altered composition or responses to the microbiota. Although a pathogenic link between ERVs and the microbiota has not been previously reported, retroelements are well recognized for their pathogenic potential. Our results propose that enhanced ERV expression may contribute to inflammatory responses to the microbiota. Notably, we showed that antiretroviral treatment or specific deletion of STING in keratinocytes alleviated inflammation caused by S. epidermidis in mice fed a high-fat diet. Our data from murine model of psoriasis and human psoriatic lesions also support the idea that skin inflammation may be generally associated with enhanced expression of ERVs. These results are in line with previous work demonstrating that ERV-derived nucleic acids and protein expression are frequently increased in numerous settings, including infection, autoimmunity, and cancer. Furthermore, increased retroelement expression coupled with mutations in cytosolic-DNA-sensing genes have been implicated in inflammation and autoimmunity. Aberrant accumulation of self-DNA promotes inflammatory processes in a cGAS-STING-dependent manner, and a link between ERVs and cGAS has been demonstrated in the context of defined inflammatory mouse models and proposed in defined human settings.

32Whether increased levels of ERV expression above a defined set point could be sufficient to promote inflammation remains to be addressed, but tissue context is a likely critical determinant. Indeed, altered activation status of keratinocytes imposed by nutritional stress or altered regulatory landscape is expected to also impact responses to ERVs. How a high-fat diet primes the skin for enhanced ERVs expression in the context of microbial exposure remains unclear. Numerous factors associated with metabolic alterations, including enhanced cytokine levels (e.g., IL-1), enhanced microbial translocation, or responses to saturated fatty acids, could contribute to an altered tissue state of activation. Although inflammation caused by S. epidermidis in mice fed a high-fat diet was significantly reduced following antiretroviral treatment or in mice lacking STING specifically in keratinocytes, this reduction was not absolute, supporting the idea that additional cellular partners and innate pathways could also contribute to these inflammatory processes. Of relevance to our findings, obesity has been linked to both altered microbiota composition and increased severity and development of numerous inflammatory skin disorders, including atopic dermatitis and psoriasis. How microbiota-induced ERV expression contributes to the etiology of these disorders remains to be addressed.

33Together, our work proposes that tissue physiology may be controlled by a fundamental alliance between the exogenous microbiota and the endogenous virome. Our work also supports the idea that ERV responses to the microbiota should be integrated into our understanding of diseases and inflammatory states.

34In our study, we showed that endogenous retroviral transcripts were induced by various skin commensals. Whether these specific ERV loci are the primary molecules being sensed by cGAS/STING needs to be further confirmed. It is possible that numerous retroelements are tonically sensed by cGAS/STING and that antiretroviral treatment reduces the sensing of all elements. As ERV RNA is also sensed by immune cells (via melanoma differentiation-associated protein 5 [MDA5]), we cannot exclude a possible role for sensing of ERV RNA in commensal induced T cell immunity. Additionally, we present data showing that retroelements are induced in lesional keratinocytes from patients with the autoinflammatory disorder psoriasis. However, at this time, it is unclear whether retroelement sensing plays a role in the pathogenesis of this disease.

35Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Yasmine Belkaid (ybelkaid@niaid.nih.gov).

36The S. epidermidis NIHLM087 ΔtagO and NIHLM087 Δlgt mutants used in this study are available upon request.

37RNA Sequencing datasets generated during this study are available at the NCBI GEO: GSE160688. The published data are available at NCBI GEO under the accession GSE54456.

38C57BL/6 J (wild-type [WT] mice), cGas–/– (B6(C)-Cgastm1d(EUCOMM)Hmgu/J), Sting–/– (B6(Cg)-Sting1tm1.2Camb/J), Krt14Cre+ (B6N.Cg-Tg(KRT14-cre)1Amc/J), Cd11cCre+ [B6.Cg-Tg(Itgax-cre)1-1Reiz/J], Tlr2–/– (B6.129-Tlr2tm1Kir/J) and Stingflox/flox (B6;SJL-Sting1tm1.1Camb/J) mice were purchased from the Jackson Laboratory. Ifnar1–/– (B6.129S2-Ifnar1tm1Agt/Mmjax) were obtained either through the NIAID-Taconic exchange program or from Jackson Laboratory. For experiments with knockout mice, littermate controls were used as wild-type controls, and/or wild-type (C57BL/6J) and knockout mice were co-housed at 3-5 weeks of age for 2-3 weeks in the same cage, prior to the start of experimental manipulations. All mice were bred and maintained under SPF conditions at an American Association for the Accreditation of Laboratory Animal Care (AAALAC)–accredited animal facility at 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 NIAID under an Animal Study Proposal (LHIM3E) approved by the NIAID Animal Care and Use Committee. Sex- and age-matched mice between 4 and 8 weeks of age were used for each experiment. Female and male mice were used for experiments involving conditional knockout mice, while only female mice were used for experiments involving only conventional mice.

39NIHLM087 (wt), S. epidermidis NIHLM087 ΔtagO, S. epidermidis NIHLM087 Δlgt, Staphylococcus aureus AD04.E17 and Staphylococcus xylosus 42C08 were cultured for 18 h in Tryptic Soy Broth at 37°C without shaking. For topical association of bacteria, each mouse was associated with a bacterial suspension (∼109 colony-forming unit [CFU]/mL) across the surface of the ear pinnae or across the entire back skin using a sterile cotton swab. Application of bacterial suspension was repeated 4 times every other day for a total of four times or performed daily for 7 days as indicated before analysis.

40Mice were injected intraperitoneally with 1 mg of either anti-mouse IFNAR1 antibody (clone MAR1-5A3; BioXCell) or mouse IgG1 isotype control (clone MOPC-21; BioXCell) 1 day before the initial application of S. epidermidis NIHLM087; each mouse then received 0.5 mg of either antibody at days 3, 6, 9 and 12 post association.

41Mice were provided a combination of Tenofovir disproxil fumarate and Emtricitabine (both from ACROS Organics) based on the effective doses previously reported in the literature or human-to-animal dose translation studies (100 mg/kg for Tenofovir and 60 mg/kg Emtricitabine;;; ). Both antiretroviral drugs were dissolved in water and a cocktail containing 12.5 mg/mL of Tenofovir and 7.5 mg/mL of Emtricitabine was administrated daily by gavage to mice in a total volume 200 μL starting 7 days prior to S. epidermidis colonization or 2 weeks after the start of the high-fat diet regimen.

42Tenofovir disproxil fumarate and Emtricitabine (both from ACROS Organics) were dissolved in Tryptic Soy Broth, and then mixed at concentrations ranging from 10 mg/ml to 0.64 μg/ml with a culture of S. epidermidis NIHLM087 at an OD600 of 0.1. The bacterial culture was then incubated for 12 hours and the OD600 was measured every 10 minutes using a Biotek Synergy HTX plate reader.

43Wounding and quantitation of wound healing were performed as previously described. Briefly, male mice in the telogen phase of the hair cycle were anesthetized with ketamine/xylazine, shaved with clippers and punch biopsies were performed on the back skin. A 6-mm biopsy punch was used to partially perforate the skin and iris scissors were then used to cut epidermal and dermal tissue to create a full thickness wound in a circular shape. Back-skin tissue was excised 5 days after wounding, fixed in 4% paraformaldehyde in PBS, incubated overnight in 30% sucrose in PBS, embedded in OCT compound (Tissue-Tek), frozen on dry ice, and cryo-sectioned (20-mm section thickness). Sections were fixed in 4% paraformaldehyde in PBS, rinsed with PBS, permeabilized with 0.1% Triton X-100 in PBS (Sigma-Aldrich), and blocked for 1 hour in blocking buffer (2.5% Normal Goat Serum, 1% BSA, 0.3% Triton X-100 in PBS). Sections were then incubated overnight with the primary antibody chicken anti-mouse Keratin 14 antibody (Poly9060, BioLegend) diluted 1:400 in blocking buffer and in presence of rat gamma globulin and anti-CD16/32. After washing with PBS, a secondary antibody conjugated with Alexa647 (goat anti-chicken, Jackson ImmunoResearch) was added for 1 hour at room temperature. Slides were then washed with PBS, counterstained with DAPI and mounted in Prolong Gold. Wound images were captured with a Leica DMI 6000 widefield epifluorescence microscope equipped with a Leica DFC360X monochrome camera. Tiled and stitched images of wounds were collected using a 20 × /0.4NA dry objective. Images were analyzed using Imaris software (Bitplane).

44The ear skin of topically associated or unassociated control mice were swabbed with a sterile cotton swab previously soaked in tryptic soy broth. Swabs were streaked on Columbia blood agar plates. Plates were then placed at 37°C under aerobic conditions for 18 hours. Colony-forming units (CFU) on each plate were enumerated and the number of CFU was reported per cm2 of skin.

45The 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 weeks old) for 6 weeks before topical association with S. epidermidis. 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 at 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.

468-12-week-old C57BL/6J mice were treated daily for 5 consecutive days on each ear pinnae with 10 mg of 5% imiquimod cream (Aldara Cream 5%).

47Cells from the ear pinnae were isolated as previously described Briefly, ears were excised and separated into ventral and dorsal sheets. Ear pinnae were digested in digestion media (RPMI 1640 media supplemented with 2 mM L-glutamine, 1 mM sodium pyruvate and nonessential amino acids, 55 mM β-mercaptoethanol, 20 mM HEPES, 100 U/ml penicillin, 100 μg/ml streptomycin and 0.25 mg/ml Liberase TL purified enzyme blend, Roche), and incubated for 2 hours at 37°C and 5% CO2. Digested skin sheets were homogenized using the Medicon/Medimachine tissue homogenizer system (Becton Dickinson). To separate epidermis from dermis, ear pinnae were first digested for 45 min with 500 CU Dispase (Becton Dickinson) in HBSS without calcium and magnesium. Epidermis was then peeled from dermis with curved forceps, washed in PBS and successively cut with scissors, digested with Liberase TL digestion media for 1 h at 37°C and 5% CO2, homogenized by pipetting up and down, and finally filtered through 70-μm cell strainer.

48For detection of basal cytokine production potential, single- cell suspensions from ear pinnae skin were cultured ex vivo for 2.5 hours at 37°C in 5% CO2 in complete medium (RPMI 1640 supplemented with 10% fetal bovine serum, 2 mM L-glutamine, 1 mM sodium pyruvate, 1 mM nonessential amino acids, 20 mM HEPES, 100 U/ml penicillin, 100 μg/ ml streptomycin, and 50 mM β-mercaptoethanol) containing 50 ng/ml of phorbol myristate acetate (PMA) (Sigma-Aldrich) and 5 μg/ml of ionomycin (Sigma-Aldrich), and a 1:1000 dilution of brefeldin A (GolgiPlug, BD Biosciences).

49Primary neonatal mouse keratinocytes were isolated from newborn pups as described previously. In brief, 2-3 days old pups were sterilized thoroughly with betadine followed by washing with 70% ethanol. Their skin was stripped off and floated on 0.25% trypsin (GIBCO, Grand Island, NY, USA) overnight (16–18 h) at 4 °C. Epidermal layer was separated from dermis and chopped in medium containing 8% FBS and 1.3 mM calcium (high calcium (HiCa) medium). The cells were filtered through 100 μm cell strainer (Corning Falcon) and seeded in 24-wells plates in medium containing 0.05 mM calcium and 8% Chelex-treated FBS (low calcium (LoCa) medium) at 37 °C in a 7% CO2 atmosphere. After 18 h, loosely attached or floating rounded cells and cell aggregates, representing suprabasal and cornifying cells, were removed by washing with PBS. Culture media was changed every 2 days. By day 4 after plating, cells were collected and stimulated with S. epidermidis (MOI of 100), LPS (1 μg/ml), R848 (5 μg/ml), LTA (10 μg/ml) for 24 hours.

50f-MIIINA:H2-M3 specific CD8+ T cells from spleen were subjected to magnetic bead-based enrichment. In brief, spleen cells were prepared to a single cell suspension and stained for 1 hour at room temperature in the dark with f-MIIINA:H2-M3-streptavidin-phycoerythrin (PE) tetramer in complete RPMI and in the presence of anti-mouse CD16/32 and rat IgG. Samples were then incubated with StemCell anti-PE antibody cocktail (Stem Cell) for 15 minutes at room temperature in the dark and then incubated with StemCell Easysep Magnetic Particles for 10 minutes at room temperature in the dark. Samples were then enriched using a StemCell EasySep Magnet Multistand (StemCell).

51Single-cell suspensions were incubated with combinations of fluorophore-conjugated antibodies against the following surface markers: CD3e (145-2C11), CD4 (RM4-5), CD8b (53- 6.7), CD45 (30-F11), TCRβ (H57-597), MHCII (M5/114.15.2), CD49f (eBioGoH3), CD90.2 (53–2.1), CD31 (MEC13.3), CD34 (RAM34), Sca-1 (D7) and γδTCR (eBioGL3) in Hank’s buffered salt solution (HBSS) for 30 min at 4°C and then washed. LIVE/DEAD Fixable Blue Dead Cell Stain Kit (Invitrogen Life Technologies) was used to exclude dead cells. Cells were then fixed and permeabilized with the Foxp3/Transcription Factor Staining Buffer Set (eBioscience) and stained with fluorophore-conjugated antibodies for at least 60 min at 4°C. For transcription factor (Foxp3), Ki-67 and intracellular cytokine staining, cells were stained with fluorophore-conjugated antibodies against Foxp3 (FJK-16 s), Ki-67 (16A8), IFN-γ (XMG-1.2), and IL-17A (eBio17B7) for 60 min at 4°C. For MAIT cells or f-MIIINA:H2-M3 staining, cells were stained in RPMI complete media for 1 hour at room temperature with 5-OP-RU mMR1 tetramer or f-MIIINA:H2-M3 tetramer (provided by the NIH Tetramer Core Facility). Each staining was performed in the presence of purified anti-mouse CD16/32 (clone 93) and purified rat gamma globulin (Jackson Immuno- research). All antibodies were purchased from eBioscience, Biolegend, BD Biosciences, or Miltenyi Biotec. Ecotropic and non-ecotropic ERV envelope protein (MLV SU) was detected using the 83A25 monoclonal antibody. Cells were incubated with hybridoma 83A25 supernatant or rat IgG2a isotype control, followed by staining with a biotinylated anti-rat IgG2a antibody (clone RG7/1.30, BD Biosciences), and streptavidin-PE-Cy7 (BioLegend). For MLV SU expression analysis, the iMFI, which reflects the total functional response and is calculated by multiplying the frequency by the MFI, was used. Cell acquisition was performed on a BD Fortessa X-20 flow cytometer using FACSDiVa software (BD Biosciences) and data were analyzed using FlowJo software (TreeStar).

52Interfollicular epidermal keratinocytes were purified by cell sorting from the epidermis of ear pinnae from unassociated or S. epidermidis-associated mice treated or not with antiretroviral drugs (when indicated). Mice were daily associated for 7 days (for bulk RNA-sequencing) or every other day a total of four times (for single-cell RNA sequencing), and the experiments were performed at day 7 or 14 post association, respectively. Cell suspensions obtained from the epidermis of ear pinnae were incubated with the following antibodies: anti-CD16/32 (93), anti-CD31 (MEC13.3), anti-CD34 (RAM34), anti-CD45 (30-F11), anti-CD49f (eBioGoH3), anti-Sca-1 (D7) in the presence of DAPI. Interfollicular epidermal keratinocytes were sorted by flow cytometry on a FACSAria (BD Biosciences) as DAPI–CD45–CD31–CD34–CD49f+Sca-1+cells.

53Keratinocyte RNA was extracted using the RNEasy Micro Kit (QIAGEN). RNA-seq libraries were prepared using Illumina TruSeq Stranded Total RNA Library Prep for paired-end sequencing and sequenced on a HiSeq4000. Sequencing reads were mapped to the C57BL/6 mouse genome (GRCm38: mm10) with STAR using default parameters (Dobin et al., 2013). Gene expression was assessed using HOMER 4.11, specifically the analyzeRepeats function with parameters mm10, -count exons -condenseGenes -noadj, and differential expression was calculated with DESeq2. Principal component analysis (PCA) was done with R. Differentially expressed genes were defined as genes that had a False Discovery Rate (FDR) < 0.05 and Fold Change (FC) > 2. Gene ontology analysis was done with metascape. Sequencing data is deposited in the NCBI SRA under PRJNA664825.

54Sequencing reads were mapped to the appropriate host genome (C57BL/6 mouse genome (GRCm38: mm10) or Homo sapiens (hg38)) with STAR using stringent mapping parameters such that reads aligned to multiple regions of the genome were considered unmapped (–outFilterMultimapNmax 2). Retroelement expression was assessed using HOMER 4.11 by defining retroelements using the annotation provided by the Genome-based Endogenous Viral Element Database (gEVE) for GRCm38 or GRCh38 where appropriate. The gEVE database identifies specific chromosomal regions as expressing a specific ORF; these chromosomal coordinates were directly used in HOMER to specify regions of the genome that encoded a retroelement. Reads were analyzed using the analyzeRepeats function with parameters mm10, -count genes -noadj, and differential expression was calculated with DESeq2. Differentially expressed retroelements were defined as loci that had a False Discovery Rate (FDR) < 0.05.

55A 4 mm biopsy punch was collected from the center of ear pinnae and submerged in RNAlater (Sigma) and stored at −80°C. Total tissue RNA was isolated from skin tissue using the RNeasy Fibrous Tissue Mini Kit (QIAGEN), as per manufacturer’s instructions. DNase-treated total RNA (1 mg) was reverse transcribed into cDNA with Omniscript RT kit (QIAGEN) following manufacturer’s instructions. qPCR was performed using the iQ SYBR Green Supermix (Life Technologies) on a QuantStudio 6 Flex Real-Time PCR (Life Technologies) using the previously published primer sequences listed (Table S1;;;;;;; ). The housekeeping gene Gapdh was used to normalize the Critical Threshold (CT) values. Analysis was conducted with the ΔCT method, using the algorithm Value = 2(CT value of Gapdh – CT value of target) x 104. The data are represented as fold increase of S. epidermidis-associated relative to unassociated mice.

56Amino acid sequences for ltrA related to Group II Introns in other Staphylococcus members (NCBI accession numbers: WP_115926208.1, WP_015728874.1) as well as a representative RT-Cas1 gene (WP_013659858.1) were aligned to the Staphylococcus epidermidis NIHLM087 genome using tBLASTn with an e-value cutoff of 0.1.

57Murine ERV loci encoding potential active RT were identified using a prediction model based on conservation of active site amino acid residues. The following motifs in the RVT_1 domain of RTs (http://pfam.xfam.org/family/PF00078) were used: YxDD and DLKDAF, involved in Mg2+ ion coordination and polymerization site; [WY]xxxLPQG and QDLREVNK involved in template binding and polymerization site. All motifs were found to be conserved in the active RTs of HIV-1, GLN2, MLV, MMTV and HERVK and motif YxDD was additionally conserved in LINE-1 ORF2p. Template binding motifs were further validated based on the reported structure of HIV-1 RT (PDB: 1RTD).(Murine LINE-1 loci encoding potential active RT were identified based on existing annotation for the presence of an intact ORF2p in L1base (http://l1base.charite.de; ).

58WT mice (n = 5 per group) were unassociated or S. epidermidis-associated and treated with vehicle or antiretroviral drugs by oral gavage, beginning at 1 week before S. epidermidis association for a total of 3 weeks. At day 14 post association mice ears was collected and digested as described above to separate epidermis from dermis. Isolated cells were stained with antibodies against surface markers and TotalSeq-A hashtag oligonucleotide (HTO) antibodies (BioLegend) (Table S2). Hash tags with not enough reads (due to insufficient staining) were excluded from the downstream analysis. Interfollicular epidermis keratinocytes were sorted as live (DAPI−) CD31−CD45−CD34−Sca-1+CD49f+. The Chromium Single Cell Controller (10X Genomics) was overloaded with 45,000 cells (3,000 cells per sample) per lane and the library was prepared following the 10X Genomics guidelines. The hashtag library was processed according to guidelines from. A total of 4 lanes were prepared and libraries were sequenced on an Illumina Nextseq500 (Next Seq 500/550 High Output Kit v2, Illumina). For processing, cellranger 4.0.0 (10X Genomics) was used to map the raw sequencing data to the mm10 reference genome and HTO and mRNA libraries were processed together with cellranger count. Only singlet cells with less than 10% mitochondrial contamination were used for downstream analysis, leaving 26,646 cells. Downstream analysis was performed using Seurat version 3.1.4. Data was normalized, scaled, principal components analysis was performed, and neighbors were found using 11 dimensions. Uniform manifold approximation and projection (UMAP) reduction was performed on this dataset with 25 dimensions. Clusters were assigned using marker genes defined in earlier studies on epidermal keratinocytes. Differentially expressed genes between treatment groups were identified running FindMarkers using the non-parameteric Wilcoxon rank sum test between each treatment. To identify differentially expressed genes between clusters FindAllMarkers were run. P values were adjusted using a Bonferroni correction.

59Ear pinnae were split with forceps, fixed in 1% paraformaldehyde solution (Electron Microscopy Sciences) overnight at 4°C and blocked in 1% BSA, 0.25% Triton X blocking buffer for 2 hours at room temperature. Tissues were first stained with anti-CD4 (RM4-5, eBioscience), anti-CD8α (clone 53- 6.7, eBioscience), anti-CD49f (GoH3, eBioscience) antibodies overnight at 4°C, washed three times with PBS and then mounted with ProLong Gold (Molecular Probes) antifade reagent. Ear pinnae images were captured on a Leica TCS SP8 confocal microscope with a 40X oil objective (HC PL APO 40X/1.3 oil). Images were analyzed using Imaris Bitplane software.

60Mice were euthanized 14 days after topical application of S. epidermidis. 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 μm, stained with hematoxylin and eosin, and examined histologically.

61Groups were compared with Prism V7.0 software (GraphPad) using the two-tailed unpaired Student’s t test, one-way analysis of variance (ANOVA) with Holm-Šidák multiple-comparison test, or two-way ANOVA with Holm-Šidák multiple-comparison test where appropriate. Differences were considered to be statistically significant when p ≤ 0.05.

62This study did not generate any additional resources.