Commensal orthologs of the human autoantigen Ro60 as triggers of autoimmunity in lupus

Identification of human commensal Ro60 orthologs

GenBank was used to identify all known potential Ro60 orthologs in bacterial species. This list was cross-referenced with the Pathosystems Resource Integration Center (PATRIC) (21) and the Human Microbiome Project (22), resulting in several potential cross-reactive species within human microbiomes (table S1). Commensal bacterial genomes were searched for the presence of the most common bacterial Y RNA, YrlA (Y RNA–like A) RNA, as described (14), which was identified in most species (table S1). The full-length human and com-mensal Ro60 proteins were aligned (fig. S1), and a phylogenetic tree was constructed (Fig. 1A) to compare the sequence homology of commensals with each other and with mammalian Ro60. Bacterial Ro60 orthologs were identified from a wide variety of anatomical niches, including the gut, skin, oral, and nasopharyngeal cavities, airways, and urogenital tract (Fig. 1A).

The earliest and most common Ro60 B cell epitope identified in lupus patients is the region between amino acids 169 and 190 (19, 23). A protein BLAST (Basic Local Alignment Search Tool) search of this peptide sequence against all bacterial genomes specifically identified the bacterial Ro60 orthologs as containing the most similar peptide sequences. Alignment of hRo60 peptide 169–190 with the commensal Ro60 orthologs showed very high amino acid sequence similarity at this epitope (Fig. 1B). Concordantly, alignment with the T cell epitope between amino acids 316 and 335, found in mice to initiate epitope spreading (11), also exhibits high sequence homology with commensal Ro60 orthologs (Fig. 1B). The conserved amino acids may act as anchor residues that allow for B and T cell cross-reactivity between human and bacterial Ro60.

Commensal Ro60 orthologs in human subjects

Subjects with SLE/SCLE and Ro60 autoantibodies (n = 8), SLE without Ro60 autoantibodies (n = 5), and healthy controls (n = 7) completed up to three longitudinal study visits, each 1 month apart, in which microbiome samples were collected from the buccal mucosa, skin over the manubrium, and stool. Blood samples were collected for serum, plasma, and peripheral blood mononuclear cells (PBMCs). A detailed health and medication history and 24-hour diet history was completed by all subjects at each visit (Table 1 and tables S2 and S3). The presence or absence of Ro60 autoantibodies was confirmed by enzyme-linked immunosorbent assay (ELISA) and by immunoprecipitation of Ro60–Y RNA complexes (fig. S2), which had concordant results. Human leukocyte antigen (HLA) typing was performed for the lupus risk alleles HLA-DR3 and HLA-DR15 (Table 1) (24).

High-throughput sequencing of the 16S ribosomal DNA (rDNA) V4 region was performed from the fecal, oral, and skin microbiomes of each study subject. Whereas principal coordinates analysis of weighted UniFrac distances showed expected differences in β-diversity between fecal, oral, and skin microbiomes, no clustering of anti- Ro60–positive versus anti-Ro60–negative patients was observed within different anatomic sites (Fig. 2A). The Shannon-Weiner diversity index was used to calculate α-diversity, and mean values were similar between anti-Ro60–negative and anti-Ro60–positive subjects within each site (fig. S3). Anti-Ro60–negative lupus subjects had slightly lower mean α-diversity values as compared to healthy controls in the fecal and oral samples, but no statistically significant differences in mean α-diversity were calculated between healthy and anti-Ro60–positive subjects. Thus, measures of intra- and inter-sample microbial diversity were similar between anti-Ro60–positive and anti-Ro60–negative patients.

Similar to previously reported microbial compositions in healthy individuals (22), microbiome samples in study subjects from the feces, oral mucosa, and skin were primarily composed of four microbial phyla: Firmicutes, Bacteroidetes, Proteobacteria, and Actinobacteria (Fig. 2B). These four phyla represented on average 96.4% (±5.7%) of fecal samples, 98.4% (±1.6%) of oral samples, and 97.2% (±2.9%) of skin samples. The relative abundance of operational taxonomic units (OTUs) from anti-Ro60–positive and anti-Ro60–negative subjects was compared by linear discriminant analysis effect size using LEfSe (25), and P values were adjusted for the false discovery rate using the Benjamini-Hochberg procedure. This analysis revealed no significantly different bacterial OTUs in the fecal, oral, or skin microbiomes between anti-Ro60–positive and anti-Ro60–negative subjects. The same analysis was performed comparing healthy and lupus subjects, and no significantly different OTUs were found between groups at any of the three anatomic sites. Previous work showed a lower Firmicutes/Bacteroidetes ratio in the fecal microbiomes of subjects with lupus (26). This was also true in our study subjects. The fecal Firmicutes/Bacteroidetes ratio was 1.3 in healthy subjects and 0.75 in lupus subjects (P = 0.0039). No difference in this ratio was observed between anti-Ro60–positive and anti-Ro60–negative lupus subjects (0.70 versus 0.80, P = 0.47) or between lupus subjects with or without the HLA-DR3 or HLA-DR15 alleles. In summary, major dysbiosis of the oral, skin, and fecal microbiomes was not observed in anti-Ro60–positive lupus patients.

Given that 16S rDNA V4 region sequencing does not allow for discrimination at the species level, genetic strategies were developed to specifically detect commensal Ro60 orthologs. Because the prevalence, abundance, and role of most of the identified commensal species with Ro60 orthologs are poorly understood, three commensal species with the highest similarity to hRo60 and one common and abundant gut com-mensal were chosen for study and assessed by quantitative real-time polymerase chain reaction (qPCR) of genomic DNA extracted from human microbiota samples. Abundance of Propionibacterium propionicum (P. prop) Ro60 (on the buccal mucosa and skin), Corynebacterium amycolatum (C. amyc) Ro60 (on the skin), Actinomyces massiliensis (A. mass) Ro60 (on the buccal mucosa), and Bacteroides thetaiotaomicron (B. theta) Ro60 (in the feces) was compared with the amount of total 16S rDNA to calculate bacterial load of each species. These commensal bacteria with Ro60 orthologs were commonly identified in subjects with and without anti-Ro60 antibodies, including lupus subjects and healthy controls (Fig. 2C). The presence of the bacteria alone would not be expected to trigger autoimmunity in healthy individuals but may be a source of antigen in susceptible patients with genetic predispositions such as HLA polymorphisms that present the appropriate peptides. There was no difference in the mean bacterial load between anti-Ro60–positive and anti-Ro60–negative patients except in P. prop from chest swabs because of three Ro60-positive patients (SLE01, SLE02, and SCLE01) that were densely colonized with P. prop. Notably, the skin swab from patient SCLE01 was within an SCLE eruption on the chest (fig. S4) and contained relatively high levels of P. prop. Although these species represented a small component of the oral and skin micro-biota, all subjects tested had at least one of the four Ro60 ortholog–containing species and many subjects had all four species, indicating exposure to orthologous Ro60. Given that the absolute number of bacteria present in the human microbiota is similar to that of human cells, even species of low relative abundance could represent an ample source of orthologous Ro60. The skin surface swab method may also underrepresent anaerobes such as Propionibacteria that proliferate in the anaerobic environment found deep within the hair follicles (27).

To investigate whether P. prop is not only present in superficial skin swabs from lupus patients but also localized in situ within skin lesions, we performed bacterial fluorescence in situ hybridization (FISH). Skin biopsies from seven anti-Ro60–positive SCLE patients were stained with a eubacterial-specific (EUB338) (28) and a P. prop–specific 16S rDNA FISH probe (Fig. 2D) (29). There was high overlap between the signals for EUB338 and P. prop. Biopsies from five psoriasis patients were used as controls for inflammatory skin disease. Five of seven cutaneous SCLE biopsies stained positive for P. prop, whereas P. prop was not detected within psoriasis lesions, which may reflect the decreased cutaneous microbial diversity observed in psoriasis (30). We could detect these Ro60 commensals deeper than the surface of the epidermis, as was previously shown for the cutaneous microbiome (31). Overall, these data support the significance of Ro60 ortholog colonization within lesions of SCLE patients.

T cell Ro60 cross-reactivity in lupus subjects

To test the hypothesis of ortholog cross-reactivity in autoreactive human lupus CD4⁺ T cells, a T cell library assay developed by Sallusto and colleagues was used (32). T cells were isolated from blood of anti- Ro60–positive patients, sorted into T cell subsets, and expanded in vitro with phytohemagglutinin (PHA) and human recombinant interleukin-2 (IL-2) for 10 to 14 days. CD4⁺ CD25⁻ T cell subsets were sorted into memory CCR6⁻ and CCR6⁺ subsets; CCR6 was used as a marker to enrich for T helper 17 (T_H17) cells frequently implicated in autoimmunity and inflammatory diseases (33). Memory T cells were expanded with IL-2/PHA, rested, and then stimulated with autologous, irradiated monocytes pulsed with whole, heat-killed P. prop. A stimulation index (SI) of 5 was used to define significant reactivity (Fig. 3, A and B) (32). T cell clones with SI ≥ 5 were collected, expanded, and restimulated with either recombinant hRo60 or an hRo60 T cell peptide that initiated epitope spreading in mice, amino acids 316 to 335 (11). Several P. prop–reactive T cell clones proliferated in response to hRo60 and the pathogenic Ro60 T cell peptide p316-335 (Fig. 3C). Similarly, expanded memory T cells reactive to recombinantly expressed Ro60 from the Ro60 ortholog–containing gut commensal B. theta (Fig. 3, D and E) also showed cross-reactivity to hRo60 or the p316-335 Ro60 peptide (Fig. 3F). The fact that several commensal- reactive or commensal Ro60–reactive clones proliferated in response to the hRo60 protein or peptide (Fig. 3, C and F) supports the idea that microbiota-reactive T cell clones cross-react with the major Ro60 autoantigen and could initiate Ro60 autoreactivity in vivo.

Next, we decided to test the reverse situation: the activation of hRo60-autoreactive T cells with Ro60 commensals. Freshly isolated anti-Ro60–positive patient memory T cells were first stimulated with recombinant mouse Ro60 or hRo60 protein (Fig. 4, A and B) and wells with SI ≥ 5 were subcloned as above to obtain Ro60 autoreactive T cell clones. After expansion, Ro60 clones were restimulated with heat-killed Ro60 orthologous commensal bacteria, identifying a CCR6⁺ T cell clone that proliferated in response to P. prop but not B. theta (Fig. 4C), supporting the concept of Ro60 ortholog epitope-specific cross-reactivity. TCR sequencing confirmed the presence of a single clonotype (TRBV9, CDR3 sequence CCAGCAGCCAAAGCCGCGGGAGATACGCAGTAT). Because CCR6 can be either a skin or mucosal marker, it is possible that the identified cross-reactive clone was derived from a skin-homing T cell as it reacted to a Ro60 epitope from the skin commensal P. prop but not the gut commensal B. theta. Differential reactivity may be due to subtle differences in sequence of the commensal Ro60 T cell epitope recognized by this clone or the Ro60 expression level in each commensal. Cytokine secretion studies to analyze the functional po-larization of the cross-reactive T cell clones revealed a diverse phenotype after restimulation with commensals (Fig. 4D), similar to previous studies of pathogen-specific memory T cell clones (34). The cross- reactive P. prop clone secreted proinflammatory cytokines interferon-γ (IFN-γ), granulocyte-macrophage colony-stimulating factor (GM-CSF), tumor necrosis factor–α (TNFα), and IL-6, as well as IL-13, IL-9, and the T follicular helper cytokine IL-21.

As a negative control, CD4⁺ T cells from a healthy tetanus toxoid (TT)–vaccinated donor were stimulated with TT to obtain toxin-specific clones. After expansion, restimulation with hRo60 or whole, heat-killed Ro60 commensal bacteria (B. theta or P. prop) did not result in proliferation (fig. S5). Similarly, CD4⁺ T cells from an anti-Ro60–negative lupus patient did not proliferate when stimulated with hRo60 (fig. S6).

To determine whether a major T cell epitope was targeted by the autoreactive and P. prop cross-reactive T cell clone, we tested the previously identified KRFLAAVDVSASMNQ peptide 369–383 within mammalian Ro60 (20) and the corresponding Ro60 mimic region in P. prop (KRTMLALDVSGSMCS) and C. amyc (KRTLLSLDVSASMHW) (Fig. 5A). As above, human CD4⁺ lupus T cells were sorted into CCR6⁻ and CCR6⁺ populations and hRo60-reactive T cell clones were identified using a T cell library assay (Fig. 5, B and C). hRo60-reactive clones were restimulated with the hRo60 peptide 369–383, bacterial mimic peptides, and whole, heat-killed bacterial lysates (Fig. 5D). The clones proliferated strongly in response to the commensal mimic peptides as well as to whole, heat-killed commensal bacteria including B. theta, likely due to high conservation of the amino acid sequence of this hRo60 T cell peptide. Cytokine secretion arrays showed a diverse cytokine profile after stimulation with commensal Ro60 mimic peptides (fig. S7), supporting a heterogeneous T cell phenotype, as described for human pathogen-specific memory T cell clones [see Fig. 4 and (34)].

Because the frequency of autoantigen-specific T cells in the peripheral blood is low, and to confirm cross-reactivity of T cell clones with another method, major histocompatibility complex (MHC) class II tetramers were synthesized for isolation of hRo60 peptide–reactive memory CD4⁺ T cells ex vivo from a Ro60-positive patient. The tetramer was loaded with the hRo60 peptide KRFLAAVDVSASMNQ (peptide 369–383, as used in Fig. 5) (20). CD45RA⁻CD45RO⁺CD25⁻tetramer⁺ single T cells were expanded with IL-2 and PHA and then restimulated with the hRo60 tetramer peptide and with Ro60 commensal bacteria (Fig. 5E). The tetramer-positive clone proliferated significantly (P < 0.05) in response to the Ro60 ortholog commensal C. amyc, and to a lesser degree also to B. theta and P. prop, which contain a slightly different Ro60 similarity to the hRo60 peptide from C. amyc (Fig. 5A). Overall, these data support commensal Ro60 T cell cross-reactivity using both T cell library assays and tetramer-sorted memory CD4⁺ T cells.

Commensal Ro60 bound by human lupus serum antibodies

In addition to T helper cell cross-reactivity, antibody cross-reactivity with commensal Ro60 orthologs could also drive human anti-Ro60 responses in lupus patients. To assess for the presence of antibodies reactive with bacterial Ro60 RNPs in patients’ sera, immunoprecipitations were performed with human sera and lysates from cultured P. prop and B. theta, because a putative ortholog of the most conserved bacterial Y RNA, YrlA, had been identified in each of these species (14). RNA was extracted from immunoprecipitated complexes and subjected to Northern blotting to detect YrlA. Human sera from anti-Ro60–positive subjects with SLE and SCLE immunoprecipitated RNPs containing the P. prop YrlA RNA (n = 4 of 6), whereas anti- Ro60–negative SLE subjects and healthy controls did not (n = 3 of 3) (Fig. 6A), suggesting the presence of cross-reactive Ro60 autoanti-bodies in most lupus patients. Oligonucleotide probes to the YrlA RNA 5′ stem region, 3′ stem region, and loop region all bound P. prop YrlA RNA by Northern blot, confirming the predicted bacterial YrlA RNA sequence (Fig. 6B) (14). Some lupus subjects with Ro60 autoantibodies did not immunoprecipitate P. prop Ro60 RNPs containing YrlA RNA (n = 2 of 6), highlighting the heterogeneity among SLE patients and the possibility that other bacterial Ro60 orthologs with different conservation of hRo60 B cell epitopes than P. prop could be cross-reactive with these sera. Because bacterial YrlA RNAs contain a domain that is structurally similar to transfer RNA (tRNA) (14), the Northern blots were reprobed for the far more abundant P. prop tRNA-proline- GGG as a negative control. None of the human sera immunoprecipitates contained this tRNA, supporting specificity of the assay (Fig. 6C). To confirm the presence of cross-reactive antibodies in anti-Ro60– positive patients, sera from a larger cohort of 63 well-characterized SLE patients from Brigham and Women’s Hospital (called Harvard sera thereafter) were obtained as a validation cohort. Anti-Ro60 status was first determined by ELISA (fig. S8A) and confirmed by immunoprecipitation of hRo60–Y RNA immune complexes (fig. S8B). Twenty anti-Ro60–positive patient sera selected based on Ro60 ortholog reactivity by ELISA were subjected to immunoprecipitation of P. prop Ro60–YrlA RNA, which again demonstrated cross-reactivity in most patients (Fig. 6D).

In contrast, none of the patient sera from the Yale cohort coimmuno-precipitated the B. theta YrlA RNA (fig. S9A), although YrlA RNA was detected in total RNA extracted from B. theta at the late log phase of bacterial growth (fig. S9B). To explore whether human antibodies could bind recombinant B. theta Ro60 protein (BtRo60) in the absence of YrlA RNA, Western blots were performed. hRo60, BtRo60, and B. theta lysates were run on a protein gel and blotted with serum from human anti-Ro60–positive lupus patients (Fig. 7, A and B). Lysates from two other commensal bacteria, Propionibacterium acnes and Roseburia intestinalis, which do not contain Ro60 orthologs, were used as negative controls. Anti-Ro60–positive patients had serum antibody binding to both hRo60 (60 kDa) and recombinant BtRo60 (~56.5 kDa), as well as a positive ~60-kDa band in the lane containing the total B. theta bacterial lysate. Translation of BtRo60 can initiate at one of three methionines, giving rise to potential proteins of 57.7, 57.6, and 56.5 kDa. The recombinant protein, which migrates faster than the bands detected in extracts, corresponds to the shortest version, suggesting that in vivo, the larger species may predominate. Protein bands of varying sizes were seen in the P. acnes lysate as might be expected, given the role of P. acnes in acne pathogenesis. None of the sera bound to proteins of the gut symbiont R. intestinalis. These results indicate that despite the lack of antibodies to B. theta RNPs, anti-Ro60–positive lupus patients from two independent patient cohorts had serum antibodies that bound the BtRo60 ortholog protein.

Gnotobiotic mouse monocolonization with Ro60 commensals

Although the Yale lupus sera did not immunoprecipitate BtRo60, the Western blots showing patient sera binding to BtRo60 may be biologically relevant, because binding of hRo60 antibodies to the human protein can occur in the absence of Y RNAs (35, 36). Thus, Ro60 antibodies may exist that cross-react with only the YrlA RNA–free form of the BtRo60. Because autoantibodies, including against Ro60, are found in a low frequency and titer in the general population (37, 38), cross-reactivity with some Ro60 orthologs may contribute to benign autoreactivity that may only spread to pathogenic epitopes depending on the HLA haplotype of the individual and the Ro60-directed T helper cell responses that we also demonstrated to be cross-reactive with B. theta. To study the general concept of commensal ortholog cross-reactivity in vivo, we next monocolonized germ-free (GF) wild- type C57Bl/6 mice with the common gut commensal, B. theta (n = 15). After 3 or 5 months of monocolonization, sera from all mice were tested and found to be positive for anti-hRo60 IgG by ELISA as compared to age- and sex-matched GF C57Bl/6 mice (Fig. 8A). Subsets of these mice were also treated orally with 0.1% imiquimod (IMQ) or 1 to 2% dextran sulfate sodium salt or the combination of both to promote inflammation and barrier irritation, but the anti-hRo60 IgG titers of these animals were not different from untreated animals (table S5), which is why they were treated as one cohort. Although the mouse anti-Ro60 antibodies did not immunoprecipitate hRo60 from cell extracts, none of the sera were positive for anti-insulin antibodies by ELISA, which were used as a negative control to test for nonspecific polyreactivity (table S5). C57Bl/6 mice monocolonized with a different gut commensal, Enterococcus gallinarum, were also negative for anti-Ro60 antibodies by ELISA (table S5). These data support selective B cell cross-reactivity between a commensal Ro60 ortholog and hRo60 in vivo.

To determine activation of T cells by Ro60, proliferation of cells collected from mesenteric lymph nodes (MLNs) and spleens was measured 72 hours after stimulation with bacterial lysate or recombinant protein. BtRo60 induced cellular proliferation from splenocytes in a dose-dependent manner (Fig. 8B), strongly supporting that mono-colonization led to a systemic T cell–dependent anti-Ro60 immune response. Whole bacterial lysates from cultured B. theta also led to highly significant (P < 0.0001) proliferation, albeit at lower levels than to recombinant protein. MLN cells also proliferated in response to higher concentrations of BtRo60 as well as whole bacterial lysate (Fig. 8C).

GF autoimmune-prone nonobese diabetic (NOD) mice were then monocolonized with B. theta to determine whether the Ro60 ortholog responses were different and possibly pathogenic with a genetic background that can lead not only to type 1 diabetes but also to lupus and Sjögren’s syndrome under certain circumstances (39, 40). After only 2 weeks of monocolonization and before the onset of diabetes, sera were positive for anti-hRo60 IgG antibodies by ELISA and spleen and MLN cells from NOD mice proliferated vigorously in response to BtRo60 in a dose-dependent manner (Fig. 8, D to F). The purified cells also proliferated in response to recombinant hRo60 in a dose- dependent manner, suggesting ortholog cross-reactivity because the mice had no previous exposure to the human antigen. Cells from MLN were approximately equally responsive to hRo60 and BtRo60, whereas splenic cells had a similar response to hRo60 with a more robust proliferation response to BtRo60, possibly because of the proximity of the MLN to the source of the antigen. Intriguingly, MLN incubation with high-dose B. theta lysates led to inhibition, reminiscent of the tolerogenic response to B. theta in the human gut (41). This suggests that an inflammatory response to BtRo60 may be dampened by the natural high-density colonization with this common human gut commensal and may explain the heterogeneity seen in responses to different Ro60 orthologous commensals as above. Notably, MLN and spleen cells did not proliferate in response to the control protein bovine serum albumin (BSA) or the unrelated human autoantigen β₂-glycoprotein-I (B2GP1) (42). Together, these data support com-mensal Ro60 ortholog cross-reactivity as an instigator of hRo60 auto-reactivity in vivo. During the study period, however, none of the mice developed overt autoimmunity.

To this end, we decided to use an inducible model for lupus in the GF setting and then test whether B. theta monocolonization leads to signs of autoimmunity. We therefore repeated the monocolonization in GF C57Bl/6 mice followed by 8 weeks of topical treatment with IMQ, a TLR7 agonist that has been shown to induce a lupus-like syndrome in mice under specific pathogen–free (SPF) conditions (43, 44). After 8 weeks of monocolonization with or without IMQ treatment (until 16 weeks of age), sera and organs were collected from all mice for autoimmune studies (Fig. 8, G to L). The sera were tested for anti- hRo60, anti–double-stranded DNA (dsDNA), and anti-RNA IgG levels by ELISA and compared to IMQ-treated age- and sex-matched GF C57Bl/6 mice. B. theta monocolonization plus IMQ led to increased levels of anti- hRo60 compared to all other conditions but did not enhance other lupus autoantibody titers, likely due to the short study period of 8 weeks that does not yet allow for epitope spreading, which takes years in humans (1, 44). Despite this, lupus nephritis–like immune complex deposition composed of C1q, C3, IgG, IgA, and IgM (45–47) was detectable in 88% of all glomeruli of the IMQ-treated B. theta–monocolonized mice (37 of 42 glomeruli) compared to only 4 of 49 glomeruli involved in IMQ-treated GF mice (Fig. 8L shows representative images of immunofluorescence), indicating a systemic autoimmune effect. Last, we performed flow cytometric analysis of immune cell subsets from gut and peripheral tissues to determine the effects of B. theta monocolonization on these tissues. A significant increase of CD11b⁺MHCII⁺ monocytes (P < 0.05) and CD11b⁺CD11c⁺MHCII⁺ dendritic cells (DCs; P < 0.01) was noted that was enhanced further by IMQ (fig. S10, A to C), suggesting that antigen-presenting cells are recruited to the gut and may interact with B. theta in vivo to present commensal Ro60 antigens. Small intestinal plasmacytoid DCs (pDCs)—a major source of type I IFN in SLE (48)—were increased as well by B. theta monocolonization alone (fig. S10, D and E). B. theta can induce both anti-inflammatory regulatory T cells (T_regs) in the colon and proin-flammatory T_H17 in the ileum (41, 49), but its effects on T_regs in secondary lymphoid organs are not clear. We unexpectedly found that B. theta reduced T_reg numbers in the spleen independently of IMQ treatment (fig. S10, F and G), creating an immunostimulatory environment also in the periphery that may allow for antigen-specific immune responses to Ro60, as shown in Fig. 8 (A to F). Together, these in vivo data in wild-type and autoimmune-prone mice demonstrate that B and T cell responses to Ro60 occur after monocolonization with a human gut commensal that produces Ro60 orthologs and that monocoloni-zation with a bacterial Ro60 ortholog enhances systemic lupus–like disease in a lupus mouse model.

Study design

This was a pilot proof-of-concept study using observational human subject data, ex vivo human sample experiments, and in vivo mouse experiments to demonstrate the ability of Ro60 ortholog bacteria to induce autoimmunity. Primary data are located in table S5, and details are described in Supplementary Materials and Methods. All human subject protocols were approved by the Yale Human Investigations Committee and in accordance with the Declaration of Helsinki. A signed document of informed consent was obtained from all study subjects. Exclusion criteria were as follows: ongoing chronic infection, antibiotic or probiotic use in the last 90 days, topical antibiotic or antimicrobial use in the last 7 days, bathing or tooth brushing in the last 8 hours, major gastrointestinal surgery in the last 5 years, gastrointestinal bleeding history, inflammatory bowel disease, bulimia or anorexia nervosa, morbid obesity, uncontrolled diabetes mellitus, malignancy in the past year, and known excessive alcohol use. Power calculations were performed using a known microbiome reference set. Assuming an SD of 8% for a commensal genus in each group, the power to detect a difference between groups with a 0.05 significance level would be 80% with n = 10 in each group. Lupus patients (n = 18; 16 with SLE and 2 with SCLE) and age-matched (±5 years) and sex-matched healthy controls (n = 11) were enrolled over a 2-year period. All study subjects completed up to three study visits for the collection of detailed health and diet histories, whole blood, and oral, skin, and fecal microbi-ota sampling following a larger microbiome study protocol at Yale (ClinicalTrials.gov identifier: NCT02394964).

The V4 regions of rDNA extracted from human skin, oral, and fecal swabs were sequenced to determine the composition of the micro-biota and subjected to qPCR to validate the presence of target species. Human sera were used to immunoprecipitate bacterial Ro60. T cell libraries were generated using one out of three visits from each human subject studied for each library. Researchers were not blinded.

Three cohorts of GF mice were monocolonized with B. theta. Serum, lymph node, spleen, gut, and kidney samples were collected after termination of the experiment. Serum was used for Ro60 ELISA, lymph nodes and spleens were used for proliferation assays in response to bacteria and proteins, spleens and gut tissues were used to isolate cells for flow cytometry, and kidneys were used for immunofluores-cence. One monocolonized C57Bl/6 animal was found dead before the end of the study and excluded from analysis. No samples were otherwise excluded.

Bacteria and culture conditions

C. amyc strain SK46 was obtained from the Biodefense and Emerging Infections Research Resources Repository (BEI), and single colonies were grown in Brain Heart Infusion Medium broth (BD Difco) aerobically at 37°C. P. prop strain F0230 was obtained from BEI and grown in Modified Reinforced Clostridial Medium (BD Difco) anaerobically at 37°C. A primary human isolate of B. theta (59) was grown in Gut Micro-biota Medium anaerobically at 37°C to an OD₆₀₀ (optical density at 600 nm) value of 1.6 to 1.8 for all experiments (late log phase, see fig. S5). A. mass strain F0489 was obtained from BEI and grown anaerobically in Actinomyces Broth (BD Difco) at 37°C. Additional details are available in Supplementary Materials and Methods.

T cell cloning, tetramers, and proliferation assay

Human PBMCs were isolated from whole blood and immunomagnetically separated by CD19, CD14, and CD4 (STEMCELL Technologies). Viable CD4⁺ T cell were sorted into CCR6⁻ memory (CD45RA⁻CD45RO⁺ CD25⁻CCR6⁻) and CCR6⁺ memory (CD45RA⁻CD45RO⁺CD25⁻CCR6⁺) populations of at least 97% purity for use in T cell library assays as previously described (32). Library screening was carried out by stimulation of 250,000 T cells per well with irradiated (45 Gy) autologous monocytes (~25,000 per well). The monocytes were pulsed for 3 hours with recombinant mouse Ro60 (100 μg/ml) or hRo60 protein or TT (1 μg/ml). Positive control wells were expanded and then restimulated at a ratio of 1:10 with irradiated monocytes that were pulsed with either whole, heat-killed bacteria or commensal mimic peptides. Negative control wells contained monocytes only to assess any background signal. After 64 hours, culture supernatants were removed for cytokine measurement using a bead-based immunoassay (LEGENDplex, BioLegend). Cell proliferation was measured either by measuring [³H]-thymidine incorporation on a scintillation β-counter (PerkinElmer) or alternatively by nonradioactive ATP measurement using the ATPlite Kit (PerkinElmer), as indicated in the figure legends. Additional details are available in Supplementary Materials and Methods.

Immunoprecipitation and Northern blotting

After lysing P. prop bacteria by cryogenic grinding, the lysates were incubated with human sera, followed by incubation with Protein A Sepharose CL-4B (Pharmacia). Beads were washed and RNA was extracted using phenol/chloroform/isoamyl alcohol (50:50:1), fractionated in 6% polyacrylamide/8 M urea gels, transferred to Hybond (Amersham), and RNA cross-linked to the membrane as described (60). Hybridization with [γ-³²P]ATP-labeled oligonucleotides was performed as described (61). Oligonucleotides were as follows: P. prop YrlA RNA, 5′-ATCCCTGATAACCGATCCCCTGCGG-3′ and 5′-CAACCTCCT-GATCCCTGATAAC-3′; P. prop tRNA^Pro, 5′-TTGTCGGGCTGACAG-GATTTG-3′. Additional details are available in Supplementary Materials and Methods.

Mice

Animal care and handling was approved by the Yale Institutional Animal Care and Use Facility and in accordance with the National Institutes of Health (NIH) Guide for the Care and Use of Laboratory Animals. GF mice were housed in gnotobiotic isolators, and monocolonization was achieved by oral gavage with 0.2 ml of thawed B. theta at ~1.3 × 10⁹ cells/ml in 20% glycerol, followed by confirmation after 2 weeks by Sanger sequencing of the 16S region of DNA extracted from the fecal pellet. Four male and 12 female 6-week-old C57Bl/6 mice were monocolonized for 3 months (n = 7) or 5 months (n = 8) followed by terminal blood collection and organ collection. A second group of C57Bl/6 GF mice was treated with topical IMQ three times weekly from age 8 to 16 weeks to mirror an inducible lupus model, as previously described under SPF conditions (44). Fourteen C57Bl/6 mice remained GF, and 6 mice were monocolonized with B. theta. Half of each group was treated with topical IMQ and housed in gnotobiotic isolators. Three 11-week-old male and three 6-week-old female GF NOD mice were monocolonized and sacrificed after 2 weeks. Serum anti-Ro60 IgG was measured by ELISA. Cells from MLNs and spleen were plated at 100,000 cells per well and stimulated with B. theta, Ro60, or control proteins. Proliferation was measured after 72 hours using an ATP assay. For flow cytometric analysis, cells from spleen and small intestinal lamina propria were stained on the surface with fluorescently conjugated antibodies specific to CD3ε, CD4, CD8, CD45, CD45R (B220), PDCA-1, CD11c, CD11b, and MHCII and intracellularly stained with fluorescently conjugated antibody specific for FoxP3 (BD Biosciences). Kidneys were immunofluorescently stained with anti-IgG/IgA/IgM, anti-C1q, and anti-C3 (Abcam). Additional details are available in Supplementary Materials and Methods.

Statistical analysis

Plotting of data and statistical analysis were performed using GraphPad Prism software. Unless otherwise stated, statistical significance was determined by the unpaired two-tailed Student’s t test, and differences were considered statistically significant if P < 0.05. P values are represented using * for P < 0.05, ** for P < 0.01, *** for P < 0.001, and **** for P < 0.0001. Shannon-Weiner diversity index and the false discovery rate were calculated using R software (62).