Female MRL/Mp-Fas^lpr (MRL/lpr, stock number 000485) mice were purchased from The Jackson Laboratory (Bar Harbor, ME) and bred and maintained in a specific pathogen-free facility according to the requirements of the Institutional Animal Care and Use Committee (IACUC) at Virginia Tech (Animal Welfare Assurance Number: A3208-01). CO₂ was used for euthanasia according to the IACUC protocol. All experiments were performed in accordance with the National Institutes of Health guide for the care and use of Laboratory animals (NIH Publications No. 8023, revised 1978). Vancomycin (2 g/L) was given in the drinking water during the indicated periods of time. The drinking water containing vancomycin was replenished once a week. Endotoxin-free E. coli bacterial DNA was purchased from InvivoGen (San Diego, CA). Eighty micrograms per mouse of bacterial DNA was prepared in phosphate buffered saline (PBS) then orally gavaged to vancomycin-treated mice once a week for 4 consecutive weeks at 4, 5, 6, and 7 weeks of age. All experiments were performed at least twice.

Urine was collected weekly starting from disease onset at 8 weeks of age, and all samples were stored at −20°C and analyzed at the same time with a Pierce Coomassie Protein Assay Kit (Thermo Scientific). When mice were euthanized at 15 weeks of age, kidneys were fixed in formalin for 24 h, embedded in paraffin, sectioned, and stained with periodic acid-Schiff (PAS) stain by Virginia Tech Animal Laboratory Services (ViTALs) at Virginia-Maryland College of Veterinary Medicine. Slides were read with an Olympus BX43 microscope. All the slides were scored in a blinded fashion by a board-certified veterinary pathologist. Glomerular lesions were graded on a scale of 0 to 3 for each of the following five categories: increased cellularity, increased mesangial matrix, necrosis, the percentage of sclerotic glomeruli, and the presence of crescents. Tubulointerstitial (TI) lesions were graded on a scale of 0 to 3 for each of the following four categories: presence of peritubular mononuclear infiltrates, tubular damage, interstitial fibrosis, and vasculitis.

Spleens were collected and mashed in 70-µm cell strainers with complete media (RPMI 1640, 10% fetal bovine serum, 1 mM sodium pyruvate, 1% 100× MEM non-essential amino acids, 10 mM HEPES, 55 μM 2-mercaptoethanol, 2 mM L-glutamine, 100 U/ml penicillin–streptomycin, all from Life Technologies, Grand Island, NY). Red blood cells were lysed with RBC lysis buffer (eBioscience). For surface staining, cells were blocked with anti-mouse CD16/32 (eBioscience), stained with fluorochrome-conjugated antibodies, and analyzed with BD FACS Aria II flow cytometer (BD Biosciences, San Jose, CA). For intracellular staining, Foxp3 Fixation/Permeabilization kit (eBioscience) was used. Zombie Aqua fixable viability kit (Biolegend) was used to exclude dead cells. Anti-mouse antibodies used in this study include the following: CD3-APC, CD4-PE-Cy7, CD8-PerCP/Cy5.5, IL-10-PerCP/Cy5.5, IFNγ-APC/Cy7, IL-17A-FITC, CD44-FITC, CD62L-APC/Cy7, Foxp3-PE, B220-PE, CD19-APC (Biolegend); RORγt-PE (eBioscience). Flow cytometry data were analyzed with FlowJo.

Spleen and kidney were homogenized with Bullet Blender homogenizer (Next Advance), and total RNA was extracted with RNeasy Plus Universal Kit (Qiagen) according to the manufacturer’s instructions. Genomic DNA was removed by digestion with RNase free DNase I (Qiagen). Reverse transcription was performed by using iScript cDNA Synthesis Kit (Bio-Rad). Quantitative PCR was performed with iTaq Universal SYBR Green Supermix (Bio-Rad) and ABI 7500 Fast Real-Time PCR System (Applied Biosystems). Relative gene expression was calculated using L32 as the housekeeping gene. Reactions were run in triplicate. Primer sequences for mouse L32, IL-17A, Has1, Has2, Has3, IL-1β, TNFα and IL-6 are listed in Supplemental Table I .

Anti-dsDNA IgG was measured according to a previously described method (31). Relative anti-dsDNA IgG levels were calculated as the ratio between anti-dsDNA IgG (U/ml) and total IgG (mg/ml), and fold changes are shown with the ratio of the control group set as 1. Total IgG and cytokine concentrations were determined with mouse IgG (Bethyl Laboratories), IL-10, IL-6, IFNγ (Biolegend) and IL-35 (LifeSpan BioSciences) ELISA kits, respectively, according to the manufacturers’ instructions. Samples were run in duplicate.

IL-10 producing B cells were isolated from the spleens and mesenteric lymph nodes (MLNs) of sex- and age-matched (6 and 7 weeks of age, or 11 and 12 weeks of age) female MRL/lpr donor mice with the Mouse Regulatory B Cell Isolation Kit, purchased from Miltenyi Biotec (Gladbach, Germany) following the manufacturer’s protocol. Single cell suspensions were prepared from spleens and MLNs and B cells were enriched by using the Regulatory B cell Biotin-Antibody cocktail, followed by addition of Anti-Biotin MicroBeads and magnetic separation on LD Columns (Miltenyi Biotec). Regulatory B cells were enriched following in vitro stimulation of purified splenic B cells. Specifically, enriched B cells were stimulated for 16 h with 10 µg/ml lipopolysaccharide (LPS), followed by 5 h of 50 ng/ml phorbol myristate acetate (PMA) and 500 ng/ml ionomycin. Stimulated cells were labeled with Regulatory B Cell Catch Reagent (Miltenyi Biotec) and incubated for 45 min. The supernatant was removed, and cells were resuspended in buffer and labeled with Regulatory B Cell Detection Antibody. Anti-PE microbeads were mixed into the cell suspension and cells were magnetically separated on LS columns (Miltenyi Biotec) to capture IL-10 secreting Breg cells by positive selection. For adoptive transfer, each recipient mouse was injected with one million of these isolated Breg cells per injection through the tail vein. Two injections either at 6 and 7 weeks of age (pre-disease stage) or at 11 and 12 weeks of age (active-disease stage) were carried out. Mice in the control cell group were injected with cells depleted of Breg cells at 6 and 7 weeks of age. Specifically, the non-B cell fraction from the initial B-cell enrichment was combined with the negative fraction of stimulated B cells following IL-10 capture and magnetic isolation to constitute the control cell population used for adoptive transfer.

Fecal microbiota samples were obtained by taking a mouse out of the cage and collecting a fecal pellet. To avoid cross-contamination, each microbiota sample was collected by using a new pair of sterile tweezers. Samples were stored at −80°C. All samples were processed at the same time. Sample homogenization, cell lysis, and DNA extraction were performed as previously described (13). PCR was performed, and purified amplicons were sequenced bidirectionally on an Illumina MiSeq at Argonne National Laboratory. 16S rRNA gene sequencing data were analyzed as described previously (13). The datasets generated and analyzed during the current study are available in the NCBI SRA (BioProject #PRJNA529260). For gas chromatography measurement of short chain fatty acids (SCFAs), fecal samples were weighed then acidified using phosphoric acid immediately before analysis at Virginich Tech Environmental and Water Resources Research Facility. Quantification was performed with a Hewlett-Packard 5890 gas chromatograph fitted with a flame ionization detector coupled with a Nukol GC column and a Hewlett-Packard 7673 GC/SFC injector. The injector settings were: temperature -200°C; carrier-hydrogen; injection mode-split (ratio 2:1). The temperature program was: initial temperature 80°C held for 3 min, then increase temperature at a rate of 6°C per minute to 140°C and hold for 1 min. The Flame Ionization Detector Settings were: temperature-250°C; hydrogen flow-35 ml/min; air flow-350 ml/min; makeup flow (nitrogen)-15 ml/min; total makeup (makeup + column flow)-30 ml/min. The levels of individual SCFAs are shown as µmol/g feces.

MRL/lpr mice were treated with 2 g/L of vancomycin in the drinking water starting at 3 weeks of age. To avoid the possibility of chemical reactions between vancomycin and SCFAs in the drinking water, SFCAs were administered by oral gavage from 3 weeks of age to 8 weeks of age. Sodium acetate, sodium propionate and sodium butyrate (Sigma) were administered by oral gavage, at 1 g/kg of body weight, diluted in water, five times per week. Mice were then monitored during disease progression, and urine samples were collected weekly. Mice were sacrificed for sample collection at 15 weeks of age.

Analysis of non-sequencing data was performed with non-parametric Mann–Whitney test for the comparison of two groups, and non-parametric Kruskal–Wallis test for the comparison of three or more groups. Unless specified, the PBS-treated groups were used as the control. The results were considered statistically significant when p < 0.05. All analyses were performed with GraphPad Prism (San Diego, CA).

To explore the role of the microbiota in modulating the induction of SLE, we treated MRL/lpr mice with antibiotics before disease initiation. We chose vancomycin because it is not absorbed in the gastrointestinal tract (32), so any effect on the immune system should be mediated by local effects on the gut microbiota. As previously noted, both female (15) and male MRL/lpr mice ( Figures S1A–C ) benefited from oral vancomycin treatment during the active-disease stage, i.e., 9 to 15 weeks of age. However, when we extended the time course of vancomycin treatment starting at 3 weeks and ending at 15 weeks of age, which covered both pre-disease and active-disease stages, the beneficial effects of vancomycin disappeared in both female ( Figures 1A–E ) and male MRL/lpr mice ( Figures S1A–C ). This is consistent with the lack of phenotype change between germ-free and specific pathogen-free MRL/lpr mice (14). We, therefore, hypothesized that the effects of gut microbiota on lupus was time-dependent, and questioned whether oral vancomycin treatment during the pre-disease stage, i.e., 3 to 8 weeks of age, could promote lupus disease development. We decided to focus on female mice from then on as lupus has a strong female bias. Strikingly, splenomegaly in female MRL/lpr mice was significantly aggravated as a result of vancomycin treatment before disease onset ( Figure 1A ). In addition, the ratio of anti-double stranded (ds)DNA autoantibodies to total IgG in the serum was significantly elevated ( Figure 1B ). As kidney inflammation (or lupus nephritis) affects more than half of SLE patients (33), we assessed renal function by measuring proteinuria, renal lymph node (RLN) weight, and kidney histopathology. Both proteinuria and the RLN weight were significantly increased when mice were treated with oral vancomycin before disease onset ( Figures 1C, D ). Early vancomycin treatment also resulted in significantly higher pathologic scores in glomerular and tubulointerstitial (TI) evaluations ( Figure 1E ). These results indicate that gut microbiota has a time-dependent influence on the development of lupus, and that vancomycin-mediated modulation of gut microbiota before disease initiation could exacerbate lupus in female MRL/lpr mice.

To investigate the immunological mechanism(s) by which early vancomycin treatment worsened lupus, we first examined IL-6 and IL-17 as we had previously shown that diminished IL-6 and IL-17 levels contributed to the attenuated disease phenotype upon vancomycin treatment during the active-disease stage (15). We found that whereas long-term vancomycin treatment (3 to 15 weeks of age) did not change the serum level of IL-6, it was significantly upregulated in mice treated from 3 to 8 weeks only ( Figure 1F ). We next quantified splenic T-helper 17 (Th17) cells and CD4⁻CD8⁻ double-negative (DN) T cells, two major cellular sources of IL‑17 in both human and mouse lupus (34). A significant imbalance towards Th17 cells over regulatory T (Treg) cells was noted ( Figure 1G ). Moreover, the percentage and absolute number of DN T cells were significantly increased in mice treated with vancomycin before disease onset ( Figure 1H ). These data suggest a higher level of pro-inflammatory IL-17 cells as a result of treating mice with vancomycin before disease onset. Indeed, the transcript level of Il17a was significantly increased in both spleen and kidney ( Figure 1I ). The proportion of CD4⁺ and CD8⁺ T cells did not change ( Figures S1D, E ); however, an elevated level of IFNγ, which is known to promote lupus in both humans and MRL/lpr mice (35), was found in the serum of mice treated with vancomycin before disease onset ( Figure 1J ). Further analysis showed that oral vancomycin treatment before disease onset resulted in significantly more IFNγ-producing CD4⁺ T cells (or Th1 cells) in the spleen ( Figure 1K and Figure S1F ).

We next analyzed the phenotype of immune cells in 8-week-old mice right after vancomycin treatment before disease onset, and found a significantly reduced percentage of IL-10 producing Breg cells ( Figure 1L ), but not of Treg cells ( Figure S1G ). Breg cells produce large amounts of IL-10 and IL-35, two anti-inflammatory cytokines known as key mediators of the regulatory function of Breg cells (19). We found that treatment with vancomycin before disease onset significantly decreased the serum levels of both IL-10 and IL-35 at 8 weeks of age ( Figures 1M, N ), suggesting a functional loss of Breg cells after vancomycin treatment during the pre-disease stage. However, the percentage of Breg cells in 15-week-old mice did not differ ( Figure S1H ). This is likely related to the increase in Breg inducers during active disease, including self-DNA complexes and multiple elevated pro-inflammatory cytokines (26, 36). In addition, the bias towards DN T cells was already present at 8 weeks of age in mice treated with vancomycin ( Figure S1G ), suggesting early initiation of IL-17 production. Interestingly, administration of neomycin from 3 to 8 weeks of age led to minimal changes in RLN weight, proteinuria and circulating IL-6 ( Figure S2 ), suggesting a vancomycin-specific effect on exacerbation of lupus nephritis. Neomycin, a systemically absorbed antibiotic, might have additional effects apart from its action on the gut microbiota. Notably, anti-dsDNA autoantibodies were increased with neomycin ( Figure S2B ) but did not seem to play a pathogenic role. The implications of this observation are unclear.

Taken together, these results suggest that vancomycin-mediated modulation of gut microbiota before disease onset significantly reduced Breg cells, resulting in systemic inflammation that may contribute to lupus exacerbation in female MRL/lpr mice.

To directly address the hypothesis that Breg cells served to protect mice from developing SLE, we adoptively transferred Breg cells into vancomycin-treated mice at different disease stages. Compared to untreated mice and mice treated with vancomycin only, early transfer of enriched IL-10 producing Breg cells (90% purity) at 6 and 7 weeks of age significantly reduced weights of secondary lymphoid organs ( Figure 2A ) and the level of anti-dsDNA autoantibodies at 15 weeks of age ( Figure 2B ). In addition, adoptive transfer of Breg cells before disease onset significantly ameliorated lupus nephritis with reduced proteinuria, smaller RLN and less severe glomerular pathology ( Figures 2C, D ). Importantly, transfer of cells depleted of IL-10 producing Breg cells, which included 29% activated IL-10⁻ B cells and 70% non-B cells ( Figure 2E ) but no introduction of Breg cells ( Figures 2E, F ), during the same time frame did not have the protective effects ( Figures 2G–J ). Furthermore, although Breg cells were increased in number regardless of the time of transfer ( Figure 2F ), Breg cells injected during the active-disease stage were not beneficial ( Figures 2G–J ). The lack of beneficial response to late Breg injection suggests that the immunosuppressive function of Breg cells may be more pronounced during the pre-disease stage of lupus development.

We next examined immune cell populations and cytokine production with or without adoptive transfer of Breg cells. Breg cell transfer during the pre-disease stage significantly decreased the percentage of DN T cells and altered the balance of Treg/Th17 towards Treg cells ( Figure 2K and Figure S2E ). Additionally, the percentage of IFNγ-producing CD4⁺ T (or Th1) cells significantly decreased in mice receiving Breg cells before disease onset ( Figure 2L and Figure S2F ). The level of IFNγ in the circulation was not significantly affected, although there was a trend for decline. These results suggest that increasing the abundance of activated Breg cells in the vancomycin-treated mice—where Breg function was impaired—reduced IL-17 and IFNγ producing cells leading to attenuation of systemic manifestations and prevention of the progression of nephritis in MRL/lpr mice. Whether activated Breg cells mitigated lupus in a IL-10 dependent manner, or through the production of other cytokines such as IL-35, will be investigated in the future.

The components of the gut microbiota removed by oral vancomycin may be responsible for inducing Breg cells at the pre-disease stage before disease onset in MRL/lpr mice. Analysis of gut microbiota revealed that the diversity and richness of commensal bacteria were quickly—within 2–3 weeks after oral vancomycin was started—and significantly reduced to similar levels regardless of the treatment time frame ( Figure 3A ). However, the composition of the gut microbiota at 15 weeks of age was quite different between the two vancomycin-treated groups ( Figure 3B ), indicating that the same antibiotic could result in different changes of gut microbiota when given at different times. Focusing on the pre-disease period where vancomycin exacerbated autoimmunity, we found significant changes in the gut microbiota at several taxonomical levels. At the phylum level, in vancomycin-treated young mice, Bacteroidetes were completely removed and replaced by Proteobacteria whereas Firmicutes remained unchanged ( Figure 3C ). Within the phylum Firmicutes, changes in multiple classes were observed, specifically, the decline of Clostridia and Erysipelotrichia but increase of Bacilli ( Figure 3D ). Within the class Clostridia, every major family decreased, including Clostridales and Lachnospiraceae ( Figure S3A ). Interestingly, within the class Bacilli, whereas the order Lactobacillales was more abundant with vancomycin treatment ( Figure 3E ), not every Lactobacillus species increased its abundance: L. animalis flourished, L. intestinalis decreased, whereas L. frumenti remained unchanged ( Figure S3B ). At the order level, the major alterations were expansions of Lactobacillales and Enterobacteriales along with complete removal of Bacteroidales and Clostridiales ( Figure 3E ).

As oral vancomycin can remove many producers of short chain fatty acids (SCFAs) (37, 38), we quantified the levels of fecal SCFAs with gas chromatography. In female MRL/lpr mice, the level of total SCFAs in the feces increased from 3 to 5 weeks of age then plateaued ( Figure S4A ). Vancomycin significantly lowered the level of fecal SCFAs, consistent with vancomycin-mediated removal of Clostridia ( Figure S4B ). Further analysis showed that the three most abundant SCFAs—acetate, propionate and butyrate—were all significantly reduced in the feces with vancomycin treatment ( Figure S4C ). In contrast, the fecal heptanoate level significantly increased although its role in immune regulation and autoimmunity is unknown. In light of the known effect of SCFAs on Treg cells (39), we examined whether restoration of SCFAs could promote Breg cells and attenuate lupus in vancomycin-treated mice. Our result showed that supplementation with SCFAs did not increase the percentage of Breg cells ( Figure S4D ). Consequently, SCFA supplement did not attenuate either systemic autoimmunity ( Figure S4E ) or lupus nephritis ( Figure S4F ). These results suggest that the reduction of SCFAs upon vancomycin administration was not the reason for Breg impairment and disease exacerbation.

Through removal of gut microbiota, vancomycin would also decrease the bacterial load and thus bacteria-derived DNA, the latter of which had been shown to suppress lupus in mice (40). Interestingly, in our mice, vancomycin led to a significant decrease in the total bacterial load in the gut and a significant reduction of bacterial DNA in the circulation ( Figure 4A ). We therefore asked whether bacterial DNA could be the Breg inducer that would protect vancomycin-treated mice from disease exacerbation. Notably, recent studies have shown a role of oral antigens in the induction of tolerance involving Breg cells (41–46), but previous studies have not shown a role of orally administered nucleic acids in tolerance. We, therefore, sought to determine whether orally gavaged endotoxin-free E. coli DNA, administered once a week from 4 to 7 weeks of age at 80 µg/dose, would induce Breg-mediated oral tolerance in vancomycin-treated female MRL/lpr mice. Strikingly, oral administration of bacterial DNA led to a significant increase of Breg cells in the spleen and MLN ( Figures 4B and S5A, B ) without affecting Foxp3⁺ Treg cells, Foxp3⁺RORγt⁺ Treg cells, or IL-10 producing CD4⁺ T cells ( Figures S5C, D ). As a result, in vancomycin-treated mice given oral bacterial DNA, the spleen and MLN weights ( Figure 4C ), serum level of anti-dsDNA autoantibodies ( Figure 4D ), proteinuria ( Figure 4E ), and kidney glomerular score ( Figure 4F ) all significantly decreased. These results suggest that orally administered bacterial DNA can induce a form of oral tolerance and be protective against lupus progression, and that the reduction in circulating bacterial DNA by vancomycin treatment during the pre-disease stage may be the cause of lupus exacerbation.

We next investigated the effects of bacterial DNA on inflammatory mediators in the circulation and their cellular sources in the spleen. Serum IL-6 was significantly decreased by oral DNA treatment ( Figure 4G ). Correspondingly, the ratio of Treg cells to Th17 cells was significantly increased in the spleen ( Figures 4H and S5E ). DN T cells, another major IL-17 producer, were significantly reduced, together with a significant increase of the proportion of CD8⁺ T cells ( Figure 4I ). In both spleen and kidney, the expression of Il17a mRNA was significantly decreased following bacterial DNA treatment ( Figure 4J ). In addition, IFNγ production from splenic CD4⁺ T cells was inhibited, resulting in a significant decrease of serum IFNγ ( Figure 4K ). Moreover, the expression of CD44 in CD8⁺ T cells was significantly elevated in the spleen of bacterial DNA-treated mice ( Figure 4L ), resulting in a significantly higher percentage of CD44⁺CD8⁺ T cells ( Figure 4M ). CD44 is important for T cell migration into inflammatory sites such as the nephritic kidney (47). Because of their potential protective effect against murine lupus (48–50), more CD8⁺ T cells infiltrating the kidney may be beneficial. Furthermore, treatment with bacterial DNA significantly decreased the transcript levels of type 1 and type 2 hyaluronan synthases (Has) in the kidney ( Figure 4N ), enzymes that mediate the secretion of hyaluronate (HA) (51). HA is known as the principal extracellular CD44 ligand (51). CD44–HA interaction mediates the recruitment of diverse immune cells, in particular T cells, and contributes to disease activity in lupus (52–54). In patients and mice with active lupus nephritis, the secretion of HA is enhanced in kidney, which correlates with lymphocyte infiltration and kidney damage (55, 56). Importantly, the inhibition of HA in lupus-prone mice improved disease parameters, at least partially by reducing the expression of pro-inflammatory cytokines, including IL-1β, TNFα, and IL-6, in the kidney (57). We noted that the expression of TNFα mRNA in the kidney was also significantly reduced ( Figure 4O ). This may be another mechanism by which the administration of bacterial DNA attenuated lupus nephritis in vancomycin-treated female MRL/lpr mice. Notably, the administration of bacterial DNA to mice treated with vancomycin did not significantly alter the composition of the small intestine microbiota ( Figure S6 ), suggesting that bacterial DNA may have bypassed the gut microbiota to induce tolerance in these mice.