Host genetics but not commensal microbiota determines the initial development of systemic autoimmune disease in BXD2 mice

Abstract

Objective.

To determine the extent of gut microbiome in influencing systemic autoimmunity, we generated germ-free (GF) BXD2 lupus mice, which otherwise develop spontaneous germinal centers (GCs) and high titers of serum autoantibodies.

Methods.

The GF status was confirmed by gut bacterial culture. The autoimmune phenotypes in 6- and 12-mo-old gnotobiotic GF BXD2 mice and specific pathogen-free (SPF) BXD2 mice were compared. Serum levels of autoantibody were measured using ELISA. Histologic sections of kidney and joints were evaluated. Flow cytometry was used to analyze GC and age-associated B cells (ABCs). CD4⁺ T cells were analyzed for PD-1⁺ICOS⁺ activated T cells, follicular T-regulatory cells (Tfr, Foxp3⁺CD25⁺ PD-1⁺CXCR5⁺), and PMA/ionomycin stimulated IL-17A⁺ or interferon-gamma (IFN-ɣ)⁺ PD-1⁺ICOS⁺ T cells.

Results.

At 6-mo of age, the GF status did not affect splenomegaly, GC B cells, ABCs or serum autoantibodies except for IgG anti-histone. GF BXD2 mice exhibited a significantly higher percent of Tfr cells, compared to the SPF counterpart. At 12-mo-old, however, there were significantly diminished IgG autoantibodies and a lower percent of GC B cells and ABCs in GF BXD2 mice. Following stimulation, PD-1⁺ICOS⁺ CD4 T cells expressed significantly lower IL-17A but not IFN-ɣ in GF BXD2 mice, compared to SPF mice. Both SPF and GF BXD2 mice developed equivalent renal and joint disease with no significant differences in severity.

Conclusion.

Our results suggest a model in which genetics play a dominant role in determining the initial development of autoimmunity. In contrast, gut microbiomes may regulate the persistence of certain aspects of systemic autoimmunity.

Introduction

Commensal microbiota consist of a large community of microbes that populate different niches of the human body and play an important role in health and disease (1). The effect of the germ-free (GF) status on T and B cell subpopulations has been less well studied in autoimmune mouse models. Although earlier studies showed that GF MRL-Fas^lpr/lpr mice displayed similar disease severity as their specific-pathogen-free (SPF) counterpart (2), the role of intestinal microbiota in SLE development remains elusive. The development of systemic autoimmunity in MRL-Fas^lpr/lpr mice is heavily influenced by a homozygous mutation in the Fas^lpr gene that results in defective activated lymphocyte apoptosis. The effect of microbiota might be minimized in lupus mouse models with single dominant mutations and studies using a complex genetic mouse model of lupus might be more relevant to SLE in humans.

We have extensively characterized the genetically complex BXD2 mouse model of systemic autoimmunity (3-5). A distinct feature of autoimmunity in the BXD2 mouse is the development of age-dependent splenomegaly and large well-developed GCs as a result of abnormal activated follicular T-helper (Tfh) cell stimulation. Adult BXD2 mice develop autoantibodies including anti-DNA, anti-histone, rheumatoid factor (RF) as well as autoantibodies related to marginal zone macrophage (MZM) apoptotic debris clearance defects such as anti-MARCO (6). To determine if microbiota plays an accelerating or inhibitory role in development of autoimmunity of BXD2 mice, we generated GF BXD2 mice. The results suggest that in this complex mouse model of lupus, genetic factors play a dominant role in determining the initial susceptibility of autoimmunity while gut microbiome influences the progression of the disease.

Materials and Methods

Mice

C57BL/6 (B6) and BXD2 recombinant inbred mice were purchased from Jackson Laboratory and housed in the University of Alabama at Birmingham (UAB) Mouse Facility in specific pathogen-free (SPF) conditions. Gnotobiotic GF BXD2 mice were generated and housed in GF condition at the UAB Gnotobiotic and Genetically-Engineered Mouse Core (Supplemental Methods). All procedures on the mice were approved by the UAB Institutional Animal Care and Use Committee. All SPF and GF BXD2 mice were sacrificed at 6- or 12-mo of age. Control was 6-mo-old SPF B6 mice. Unless specified, all mice were female mice.

Microbiome 16S rRNA gene sequencing and data analysis

Mouse fecal samples were collected from different cages of SPF B6 and BXD2 mice. Bacterial genomic DNA from mouse fecal samples was extracted using the Fecal DNA Isolation Kit from Zymo Research (cat. no. D6010) following the manufacturer’s protocol (7). The PCR primers specific for amplification of the V4 region of the 16S rRNA gene were as previously described (7). All samples were run on an Illumina MiSeq DNA chip (read depth 30,000-40,000 reads per sample) at the Microbiome/Gnotobiotics Shared Facility at the University of Alabama at Birmingham (UAB)(7). The analysis of the 16S rRNA gene data steps were as previously described (7). The sequencing data are available at the NCBI Sequence Read Archive (SRA) website (BioProject ID: PRJNA746948).

Gut bacterial culture

All work was carried out in an anaerobe chamber (Coy Laboratories, Ann Arbor, MI) using a gas mix of 90% N2, 5% CO2, and 5% H2 as previously described (8). Mouse ceca were removed from the mice while in the anaerobe chamber. The colony forming unit at Day 2 and Day 7 was quantitated using ImageJ software (9).

Flow cytometry analysis and sorting

For analysis of the surface marker and intracellular expression of certain cytokines and transcriptional factors, single-cell suspensions of spleen were prepared for standard cell-surface staining or intracellular staining according to the manufacturers’ instructions. Anti-mouse antibodies were BioLegend pacific blue anti-CD19 (clone 6D5), pacific blue anti-CD4 (clone GK1.5), allophycocyanin anti-IFN-ɣ (clone XMG1.2), FITC anti-CD21 (clone 7E9), Brilliant violet anti-IgD (clone 11-26c.2a), BD Biosciences PE anti-CD95 (clone Jo2), PE-Cy7 anti-CXCR5 (clone 2G8), PE anti-Foxp3 (clone FJK-16s ), and ThermoFisher eFluoro 660 anti-GL7 (clone GL7), FITC anti-PD-1 (clone J43) and Alexa Fluor 647 anti-IL-17A (clone eBio17B7) and APC anti-CD25 (clone PC61.5), FITC anti-ICOS (clone 7E.17G9). APC-eFluor® 780 Organic Viability Dye (eBioscience) was used to exclude dead cells from analysis.

To determine the cytokine-producing T-cell, whole spleen cells were cultured and stimulated for 5 h with phorbol myristate acetate (PMA; 50 ng/ml; Sigma-Aldrich) and ionomycin (750 ng/ml; Sigma-Aldrich) in the presence of GolgiPlug (BD Biosciences) for 4 hours. Cells were stained with surface markers and then subjected to fix and permeabilized with Cytofix/Cytoperm solution (BD Biosciences) following intracellular staining (3). Standard flow cytometry (LSRII, BD Biosciences) was applied for data acquisition. Data were analyzed with FlowJo_v10 software.

Histology

Kidney and joint histology analysis was carried out in paraffin sections as previously described (10). Kidney and joint disease severity scoring was evaluated by a certified histopathologist in a blinded manner as described in the Supplementary Methods section (11, 12).

ELISA

The levels of IgG, IgM, and circulating autoantibodies were measure by ELISA method as previously described (5, 6). MARCO was ordered from R&D system, Inc. The other autoantigens were purchased from Sigma-Aldrich. All antibodies used for ELISA assay were purchased from Southern Biotech (Birmingham, AL).

Statistical analysis

Results are mean ± SD or mean ± standard error of the mean, shown in the figure legends. The p values of less than 0.05 were considered significant. Multiple comparisons were performed using Tukey’s multiple comparisons test. All analyses were performed using GraphPad Prism software (La Jolla, CA) unless otherwise indicated.

Results

Altered microbiome composition in the gut of BXD2 mice

16S rRNA gene analysis reveals that the fecal microbiota of lupus-prone BXD2 mice was characterized by a significant reduction in the Verrucomicrobia phylum (p=0.0010), compared to normal B6 mice that were housed in the same facility (Figure S1). Compared to B6 mouse, the enriched family and genus in the distal gut in BXD2 mice included uncultured Bacteroidales bacterium, uncultured Rikenellaceae_Alistipes_bacterium, unclassified Lachnospiraceae, uncultured Lachnospiraceae, unclassified Ruminococcaceae, and two Bacteroides species (acidifaciens and massiliensis dnLKV3). Significantly decreased species in BXD2 mice were Akkermansiaceae_Akkermansia_uncluture bacterium and Bacteroides ovatus V975 (Figure S2, S3).

GF status did not affect the initial development of autoimmune B cells

To directly determine if gut microbiome influenced autoimmunity in the BXD2 mice, Genobiototic GF BXD2 mice were generated. The development of megacolon was an overt feature in GF BXD2 mice (data not shown). Culture of cecal contents for 2 and 7 days confirmed the absence of bacteria in GF BXD2 mice (Figure S4). At 6 months of age, the spleen of SPF and GF BXD2 mice were equivalent in size and weight and both were larger than control SPF B6 mice (Figure S5A, S5B). Spleen weight but not cell count was significantly higher in 12-mo-old SPF BXD2 mice, compared to all other mice (Figure S5B, S5C). There was a significantly increased in the percent and number of Fas⁺GL-7⁺ GC B cells in 12-mo-old, but not 6-mo-old, SPF BXD2 mice, compared to age-matched GF BXD2 mice (Figure 1A, 1B).

Figure 1. Age-dependent effects of germ-free status on B-cell dysregulation in BXD2 mice.

A, Representative flow cytometry plots showing Fas⁺GL-7⁺ GC within CD19⁺ B cells. B, The percentage (left) and absolute count (right) of Fas⁺GL-7⁺ GC B cells in the spleen of each strain. C, Left: Representative flow cytometry plots of CD19⁺ CD21⁻IgD⁻ B cells in the indicated mouse spleen. Right: Quantification of the number of CD19⁺ CD21⁻IgD⁻ B cells. D, Left: Representative plots of CD11c⁺T-bet⁺ cells within the IgD⁻CD21⁻ (top) and IgD⁺ (bottom) B cells. The mean percent (±SD) of CD11c⁺T-bet⁺ cells in each strain is shown in each plot (data with different alphabets are significantly different from each other). Right: The number of CD11c⁺T-bet⁺IgD⁻CD21⁻ CD19⁺ cells in the spleen. All bar graph plots represent the mean±SD from 3 independent assays. Each data point represents the results from an individual mouse (B6, N=7; 6 mo-old SPF BXD2, N=6; 6 mo-old GF BXD2, N=6; 12-mo-old SPF BXD2, N=13; 12 mo-old GF BXD2, N=9; *P < 0.05 **P < 0.01 ***P < 0.005, **** P < 0.0001; ANOVA test and Tukey’s multiple comparisons test for significance between groups).

CD11c⁺T-bet⁺ IgD⁻CD21⁻ age-associated B cells (ABC) have been associated with autoantibody production and aberrant T-cell activation in mouse models of lupus (13). At 6 months of age, there was no significant difference in the percent and number of IgD⁻CD21⁻ and CD11c⁺T-bet⁺ IgD⁻CD21⁻ ABCs in SPF BXD2 mice and GF BXD2 mice compared to B6 mice (Figure 1C, 1D). However, at 12 months of age, ABCs accumulated significantly in SPF but not the GF BXD2 mice. This has resulted in a significantly higher number of IgD⁻CD21⁻ B cells (Figure 1C, p=0.0009) and ABCs (Figure 1D upper, p=0.0023) in the BXD2 SPF mice, compared to the GF BXD2 mice. There was also a significantly higher percent of CD11c⁺T-bet⁺ in IgD⁺ B cells in 12-mo-old SPF BXD2 mice, but not GF BXD2 mice, compared to 6-mo-old mice (Figure 1D lower), suggesting these B cells were activated at the pre-switched stage in 12-mo-old SPF BXD2 mice.

GF status did not affect the initial development of autoantibodies and age-related autoimmune disease

BXD2 mice exhibited a characteristic increase in sera levels of multiple autoantibodies that have undergone class-switch recombination (4, 5). There were equivalent levels of total IgG, IgG anti-DNA, rheumatoid factor (RF), and anti-MARCO in 6-mo-old SPF compared to GF BXD2 mice, with total IgG, anti-DNA, anti-histone, and anti-MARCO as being significantly higher than B6 control mice (Figure 2A, 2B). In 12-mo-old BXD2 mice, the total IgG (p=0.0014), as well as the anti-DNA (p=0.0372), rheumatoid factor (RF)(p=0.0313), and anti-MARCO (p<0.0001) autoantibodies, were significantly higher in SPF mice compared GF mice (Figure 2C, 2D). Serum IgM levels in the SPF and GF BXD2 mice at both 6 and 12 months of age were nearly indistinguishable (Figure S6). At 12-mo-old, SPF and GF BXD2 mice developed equivalent histopathological features of renal and joint diseases (Figure S7, S8).

Figure 2. Age-dependent effects of GF status on IgG autoantibody in BXD2 mice.

ELISA data of total IgG (A, C) and IgG anti-DNA, anti-histone, RF, and anti-MARCO (B, D) from the indicated strains are shown as optical density absorbance at 450-650 nm (OD_450-650). Bar graph showing the mean (±SD) of sera antibody levels. Each data point represents the results from an individual mouse (6-mo-old: B6, N=7; SPF BXD2, N=11; GF BXD2, N=9; 12-mo-old: B6, N=5, SPF BXD2, N=8; GF BXD2, N=7). Results are shown as the mean ± SD from 3 independent assays. Statistical differences were measured by ANOVA-test and Tukey’s multiple comparisons test for significance between groups (*P < 0.05, **P < 0.01, ***P < 0.005 and **** P < 0.0001).

Lower activated CD4 T cell in aged GF BXD2 mice

We previously reported that activated CD4 T-helper cell is required for the formation of GCs of BXD2 mice (3). There was a not statistical increase in the CD4⁺ PD1⁺ICOS⁺ T cells in SPF BXD2 and GF BXD2 mice compared to B6 mice at 6 months of age (Figure 3A). At 12 months of age, however, there was a further increase in PD1⁺ICOS⁺ T cells in SPF BXD2 (p<0.0001) and GF BXD2 mice (p=0.0016) compared to B6 mice (Figure 3A). The increase in activated CD4⁺ PD1⁺ICOS⁺ T cells was especially apparent in SPF mice from 6- to 12 months of age (Figure 3A, p=0.0161). There was no difference in the percent of activated CD4⁺ PD1⁺ICOS⁺ T cells between 6- and 12-mo-old GF BXD2 mice (Figure 3A, p=0.4753).

Figure 3. Age-dependent effects of germ-free status on CD4 T cell development in BXD2 mice.

A, Representative flow cytometry plots (upper) and bar graphs showing the percent of ICOS⁺PD-1⁺ cells within CD4⁺ T cells (lower) in the indicated mice. (B, C) Spleen cells from 12-mo-old SPF and GF BXD2 mice were stimulated with PMA and ionomycin. B, Representative flow cytometry plots (left) and quantification of the percent of ICOS⁺PD-1⁺ cells within the CD4⁺ T cells (right). C, Representative flow cytometry (lef) and the percent (right) of IL-17A-expressing and IFN-γ expressing ICOS⁺PD-1⁺ CD4⁺ T cells. D, Representative flow cytometry plots (upper) and quantification of the percent of Foxp3⁺CD25⁺ cells within the CXCR5⁺PD-1⁺ subset of CD4 T cells (lower). For all panels, results are shown as the mean±SD from 3 independent assays. Each data point represents the results from an individual mouse. Statistical differences were measured by ANOVA-test and Tukey’s multiple comparisons test was performed for significance between groups (A, D, B6, N=7; 6-mo-old SPF BXD2, N=6; 6-mo-old GF BXD2, N=6; 12-mo-old SPF BXD2, N=13; 12-mo-old GF BXD2, N=9) or two-tailed Student’s t-test (B, C, N=8 in each group)(*P < 0.05, **P < 0.01, ***P < 0.005,and **** P < 0.0001).

PD1⁺ICOS⁺ T cells in BXD2 mice can produce IL-17, IL-21, or IFN-ɣ (3). The percent of IL-17A- and IFN-ɣ expressing PD1⁺ICOS⁺ CD4 T cells in 12-mo-old mice following PMA/Ionomycin stimulation was analyzed. There was a significant increase in the percent of PD-1⁺ICOS⁺ CD4 T cells in SPF BXD2 mice compared to GF BXD2 mice (Figure 3B, p=0.0064). Further, within PD-1⁺ICOS⁺ CD4 T cells, there was a significantly increased IL-17A producing CD4 T cells in SPF BXD2 mice, compared to GF BXD2 mice (Figure 3C, p=0.0061), whereas IFN-ɣ expressing cells were not significantly different (Figure 3C, p=0.1348).

It has been shown that a prominent human commensal, Bacteroides fragilis, induces the development of Foxp3⁺ Treg cells in the gut (14) and that an altered Schaedler flora (ASF) colonization is required for colonic Treg generation (15). These results suggest an essential role of microbiota in the formation of Treg cells in the gut. In contrast to the evidence of the essential role microbiota in Treg generation in the gut, there was a significantly higher frequency of Foxp3⁺CD25⁺ CXCR5⁺PD-1⁺CD4⁺ follicular Treg cells in the spleen of 6-mo-old GF BXD2 compared to SPF BXD2 mice (Figure 3D, p=0.0183). Despite the diminished autoimmune profiles in the spleen of 12-mo-old GF BXD2 mice, Tfr frequency also significantly declined at this age (Figure 3D, p=0.0003).

Discussion

In the present study, we have identified that there was only a minor difference in the development of certain IgG autoantibodies and immune dysregulation in 6-mo-old GF BXD2 compared to SPF BXD2 mice, although the persistence of IgG autoantibody levels was diminished in 12-mo-old GF BXD2 mice. As IgM autoantibody levels in 12-mo-old mice were comparable between GF and SPF BXD2 mice, the results suggest that immune abnormalities that lead to autoantibody class-switch recombination was compromised in 12-mo-old GF mice. Indeed, there was a significantly lower GC response, lower development of ABCs, and a lower percent of activated CD4 T cells that express IL-17 but not IFN-ɣ in GF BXD2 mice at an older age.

Chervonsky (1) has proposed that mouse models of RA (such as the K/BxN mouse model), as well as mouse models of SLE (MRL-Fas^lpr/lpr and NZB) can develop autoantibodies independent of the contribution of microbiota species, but pathobionts can amplify the disease (16, 17). Previous studies in the NZB mouse suggest that in the absence of commensal bacteria, autoantibodies including anti-DNA and ANA develop, but disease severity may be delayed or attenuated (17). Consistent with these previous results, the present findings suggest that although the composition of the BXD2 commensal bacteria does not influence the initial development of autoimmune phenotypes, they enable the continued progression of immune dysregulation in older mice. Although we did not find differences in renal and joint disease in older GF BXD2 mice, this is not meant to imply that certain taxa that are different from the commensal SPF bacteria in BXD2 mice could not affect the onset of autoimmunity as microbiota has been shown to influence inflammatory cytokines that can act on both T and B cells in SLE (18).

An interaction between the microbiome and T cells, especially promotion of the Th17 lineage (19), as well as modulation of Tregs is well established (14, 15). These previous studies were carried out in mice in which the commensal microbiome composition was modulated rather than eliminated and resulted in altered immune phenotypes and autoimmune disease. The present studies however focused on immune phenotype analysis in a mouse model of lupus that were derived and maintained under gnobiotic germ-free conditions. Under such conditions, we did not observe alterations in activated CD4 T cells, GC B cells, and ABCs in 6-mo-old GF BXD2 mice compared to SPF BXD2 mice. The only significant difference between 6-mo-old GF and SPF BXD2 mice was a higher percent of Tfr and lower IgG anti-histone in GF mice. At 12 months of age, there was diminished PD-1⁺ICOS⁺ IL-17 producing CD4 T cells and a lower percentage of GC B cells and ABCs in GF BXD2 mice compaired to SPF BXD2 mice. The present studies suggest a model in which the initial development of autoimmunity is predisposed by the host genetic composition in the BXD2 mice. In this model, CD4 T cell and B cell tolerance loss to self-antigen stimulation plays a dominant role in leading to T and B cell activation and autoantibody formation. It is on this established autoimmune phenotype that commensal gut microbiome can interact with the host immune system and modify the progression of autoimmune disease.

A question remains to be addressed is if host autoimmune responses also influence gut microbiome composition. Fecal microbiome composition analysis revealed BXD2 enriched families included Bacteroidaceae, Lachnospiraceae, and Ruminococcaceae. Interestingly, the last two families were also enriched MRL-Fas^lpr/lpr mice (16). The similar dysbiosis in these lupus strains suggests a possibility that host autoimmune phenotypes may also regulate gut microbiome distribution.

In summary, these results have implications for the role of microbiota in SLE. First, lupus can be predisposed due to genetic factors that create the permissive inate and adaptive immune system that enables autoantigen stimulation leading to autoimmunity (3-5). Likewise, in some highly predisposed individuals, autoimmunity and SLE may not require the gut microbiome for initial development. Second, chronic immune modulators such as microbiota that stimulate toll-like receptors or different subsets of T cells can act as an autoimmune modulator to alter the subsequent course of disease.

Abbreviations used in this article:

ABC

age-associated B cells

B6

C57BL/6

BXD2

The second C57BL/6 x DBA/2 recombinant inbred strain

GF

germ-free

GC

germinal center

NZB

New Zealand Black mouse strain

SPF

specific pathogen-free

Tfh

follicular T-helper cells

Tfr

follicular T-regulatory cells

RF

rheumatoid factor

MARCO

macrophage receptor with collagenous structure