Dynamics of the gut microbiota in NZB/W F1 mice.
Like in MRL/lpr mice, the disease phenotype in NZB/W F1 mice resembles human SLE and is characterized by high levels of antinuclear antibodies, hemolytic anemia, proteinuria, and progressive immune complex glomerulonephritis. These mice have been used as a model for human SLE since the early 1960s, as these mice and humans with this multifactorial disease have similarly complex genomic landscapes. Similar to human SLE, which has a strong female bias, the disease is most pronounced in female NZB/W F1 mice. The average life span for females is 8 months, with disease onset at approximately 5 months (20 weeks) of age. Thus, disease progression is much slower in NZB/W F1 mice than in MRL/lpr mice.
To determine the dynamics of the gut microbiota during lupus progression in female NZB/W F1 mice, we analyzed fecal pellets collected at three predisease time points (10, 14, and 18 weeks of age) and three post-disease-onset, or diseased, time points (23, 28, and 33 weeks of age). The structure and diversity of the lupus-associated microbiotas changed continuously over time (
Fig. 1). The unweighted UniFrac distance-based principal-coordinate analysis (PCoA) showed that the gut microbiotas were distinct at three predisease time points but clustered together at the diseased time points (
Fig. 1A; also see Fig. S1 in the supplemental material). Importantly, the UniFrac distance between predisease and diseased time points was greater than the distance observed among the three predisease time points. In addition, the gut microbiotas split into two groups along the PC1 axis (
P < 0.01, permutational multivariate analysis of variance [PerMANOVA]), and the split happened at ∼20 weeks of age, suggesting a dramatic change in the gut microbiotas upon the onset of SLE-like symptoms. It is worth noting that we cannot rule out the possibility that cage effects are driving the differences in
Fig. 1A, and we do not have a wild-type control for the effects of aging on the microbiota.
We next determined the change in microbiota diversity during lupus progression in NZB/W F1 mice. The number of operational taxonomic units (OTUs) in these mice increased significantly from the predisease stage to the diseased stage (
P < 0.001) (
Fig. 1B), suggesting increased bacterial diversity as the disease progressed. This is consistent with other lupus-prone mouse models, in which lupus-associated increases in microbiota diversity were also observed (
9,
11). Individual bacterial species fluctuated over the time period tested, and the species with significant changes are shown in
Fig. 1C. Specifically, significant increases from the predisease stage to the diseased stage were observed for several bacterial species in the genera
Clostridium,
Dehalobacterium,
Lactobacillus,
Oscillospira,
Dorea (family
Lachnospiraceae),
Bilophila (family
Desulfovibrionaceae), and AF12 (family
Rikenellaceae) and an unnamed genus within the family
Ruminococcaceae (
P < 0.01 in all cases).
Akkermansia muciniphila and a species within the genus
Anaerostipes (family
Lachnospiraceae), however, significantly decreased from the predisease stage to the diseased stage (
P < 0.01). These results suggest that the composition of the gut microbiota changed markedly from before to after the onset of lupus disease in NZB/W F1 mice, with greater diversity and increased representation of several bacterial species as lupus progressed from the predisease stage to the diseased stage. It is worth noting that these changes may also be caused by the maturation of bacterial communities as the mice aged.
To determine whether a common treatment for SLE could reverse the changes in the gut microbiota, we treated NZB/W F1 mice with 2 mg/kg body weight dexamethasone (Dex) from 20 to 34 weeks of age (14 weeks of treatment). Prior studies showed that Dex suppresses the development of disease in NZB/W F1 lupus-prone mice (
21). Treatment with Dex altered the microbiota as the animals aged, compared to the vehicle-treated controls (
Fig. 2). The overall microbiota structure with Dex treatment was distinct from that with vehicle treatment (control), as shown in the unweighted UniFrac-based PCoA plot (
P < 0.01) (
Fig. 2A). Interestingly, instead of decreasing the bacterial diversity, which was already high in diseased mice (
Fig. 1B), Dex appeared to further increase the diversity, with a significantly higher Shannon index and more observed OTUs (
P < 0.001 for both cases) (
Fig. 2B; also see Fig. S2). Notably, the community variability decreased (
Fig. 2A) while diversity increased (
Fig. 2B) in Dex-treated mice, compared to control mice. We suggest that high diversity in the Dex-treated group may lead to a more stable community and thus lower observed variability than for the control group.
We next identified the bacterial species with significant changes from the pre-disease-onset stage to the post-disease-onset stage, and we investigated whether Dex treatment significantly reversed such changes. Only one bacterial species, i.e., “
Lactobacillaceae,
Lactobacillus, other,” fulfilled these criteria (
Fig. 2C). The relative abundance of this species increased significantly, from 0.01% at the pre-disease-onset stage to 10% at the post-disease-onset stage, and Dex was able to significantly decrease the abundance to 1% (
P < 0.01 in both cases). BLAST analysis showed that the OTU sequences within
Lactobacillaceae,
Lactobacillus, other, were at least 98% identical to those of
Lactobacillus murinus,
Lactobacillus kimchicus, and
Lactobacillus senmaizukei; among the three, only
Lactobacillus murinus was isolated from mice (the other two were isolated from kimchi and Japanese pickle, respectively). However, there could be unsequenced isolates that would be represented by the sequence obtained from the 16S analysis; therefore, the identity of the isolate(s) is unknown, but the isolate(s) could be one or more of the three detected species. We further studied Spearman association coefficients for this bacterial species and two measurements of SLE disease state, namely, renal function and systemic autoimmunity (
Fig. 2D). The renal function was calculated as a composite score reflecting the level of proteinuria and the histopathological score, whereas systemic autoimmunity was calculated as a composite score reflecting the level of autoantibodies against double-stranded DNA (dsDNA) and the weight of spleen (see Materials and Methods for details). Correlation analysis showed that
Lactobacillaceae,
Lactobacillus, other, was positively associated with renal function (correlation efficient of 0.38;
P = 0.094) and systemic autoimmunity (correlation efficient of 0.42;
P = 0.067), although the associations were not statistically significant. These results suggest that a greater abundance of a group of lactobacilli (for which a species assignment could not be made) in the gut microbiota may be associated with more severe clinical signs in female NZB/W F1 mice. This is distinct from findings observed for female MRL/lpr mice, in which a greater abundance of
Lactobacillus reuteri was found to be associated with disease attenuation (
9,
10).
Gut microbiota dysbiosis in human SLE patients.
To study the differences in the gut microbiota of patients with SLE, compared to patients without immune-mediated diseases, we enrolled 14 patients with active SLE and 17 non-SLE controls. Although the overall community structures could not be separated by unweighted UniFrac-based PCoA analysis (
Fig. 3A), SLE subjects had significantly lower diversity, measured with the Shannon index (
P < 0.05, Mann-Whitney test) (
Fig. 3B). Unlike the published comparison between healthy controls and SLE patients, in which the
Firmicutes/
Bacteroidetes ratio was significantly lower in SLE patients in remission (
12), the ratios were not significantly different for our cohort of SLE patients versus non-SLE controls (
P > 0.05) (
Fig. 3C). In addition, the abundance of the bacterial phylum
Proteobacteria (a representation of facultative anaerobic Gram-negative bacteria) was significantly greater in the gut microbiota of SLE patients (
P < 0.05) (
Fig. 3D). This is consistent with the increased serum levels of lipopolysaccharide (LPS) endotoxin in SLE patients that were observed by others (
22). Three bacterial species appeared to be differentially represented in SLE patients, compared to non-SLE controls (
P < 0.05, based on both nonparametric Mann-Whitney tests and DESeq2 analysis) (
Fig. 3E); the species were within the genera
Odoribacter and
Blautia (family
Lachnospiraceae) and an unnamed genus (family
Rikenellaceae). BLAST analyses of the OTU sequences were performed, and the identified species are presented in
Table 1. These results suggest that, compared to controls without immune-mediated diseases, SLE patients with active lupus disease possessed an altered gut microbiota that differed in several bacterial species and was less diverse, with increased representation of Gram-negative bacteria. Medication may influence the composition and diversity of the gut microbiota (
Table 2) but, due to the small sample size, we did not attempt to analyze the effects of medication.