Modulating gut microbiota in a mouse model of Graves' orbitopathy and its impact on induced disease

Abstract

Background: Graves' disease (GD) is an autoimmune condition in which autoantibodies to the thyrotropin receptor (TSHR) cause hyperthyroidism. About 50% of GD patients also have Graves' orbitopathy (GO), an intractable disease in which expansion of the orbital contents causes diplopia, proptosis and even blindness. Murine models of GD/GO, developed in different centres, demonstrated significant variation in gut microbiota composition which correlated with TSHR-induced disease heterogeneity. To investigate whether correlation indicates causation, we modified the gut microbiota to determine whether it has a role in thyroid autoimmunity. Female BALB/c mice were treated with either vancomycin, probiotic bacteria, human fecal material transfer (hFMT) from patients with severe GO or ddH2O from birth to immunization with TSHR-A subunit or beta-galactosidase (βgal; age ~ 6 weeks). Incidence and severity of GD (TSHR autoantibodies, thyroid histology, thyroxine level) and GO (orbital fat and muscle histology), lymphocyte phenotype, cytokine profile and gut microbiota were analysed at sacrifice (~ 22 weeks).

Results: In ddH2O-TSHR mice, 84% had pathological autoantibodies, 67% elevated thyroxine, 77% hyperplastic thyroids and 70% orbital pathology. Firmicutes were increased, and Bacteroidetes reduced relative to ddH2O-βgal; CCL5 was increased. The random forest algorithm at the genus level predicted vancomycin treatment with 100% accuracy but 74% and 70% for hFMT and probiotic, respectively. Vancomycin significantly reduced gut microbiota richness and diversity compared with all other groups; the incidence and severity of both GD and GO also decreased; reduced orbital pathology correlated positively with Akkermansia spp. whilst IL-4 levels increased. Mice receiving hFMT initially inherited their GO donors' microbiota, and the severity of induced GD increased, as did the orbital brown adipose tissue volume in TSHR mice. Furthermore, genus Bacteroides, which is reduced in GD patients, was significantly increased by vancomycin but reduced in hFMT-treated mice. Probiotic treatment significantly increased CD25⁺ Treg cells in orbital draining lymph nodes but exacerbated induced autoimmune hyperthyroidism and GO.

Conclusions: These results strongly support a role for the gut microbiota in TSHR-induced disease. Whilst changes to the gut microbiota have a profound effect on quantifiable GD endocrine and immune factors, the impact on GO cellular changes is more nuanced. The findings have translational potential for novel, improved treatments. Video abstract.

Keywords: Graves’ disease; Graves’ orbitopathy; Gut microbiota; Human fecal microbiota transplant; Microbiome modulation; Murine model; Probiotics; Vancomycin.

Fig. 1

Early-life manipulation treatments modified the lower gastrointestinal microbiota with long-term effects in TSHR-immunized mice. a Rationale of early-life manipulation treatments combined with the hTSHR-immunization procedure, see STAR methods for complete description. b Alpha-diversity indices amongst treatments. Wilcoxon-Mann test BH corrected: ***P < 0.001; **P = 0.019; *P = 0.045. c Significantly differentially abundant genera between immunizations in each treatment. Welch’s t test, BH corrected: ***P < 0.001; **P = 0.011; *P = 0.04. d Random forest confusion matrix of the classification for treatments (all immunizations together). Diagonal boxes represent the number of samples correctly predicted. The RF model accounted for 73.08% accuracy (n mice/group endpoint: control = 20, hFMT = 19, Lab4 = 20, vancomycin = 19). e Top-10 important variables contributing to treatment classification according to the Mean Decrease Gini, related to e. f RandomForest confusion matrix of the classification for immunizations. The RF model accounted for 70.51% accuracy (n mice/immunization endpoint: TSHR = 43, βgal = 35). g Top-10 important variables contributing to immunization classification according to the Mean Decrease Gini, related to g

Fig. 2

Effect of treatments, immunizations and time on the GO model gut microbiota. a Differences between immunizations at each timepoint and per treatment. PERMANOVA with 999 permutations, ***P = 0.008 and **P = 0.016 (n mice/treatment/immunization at baseline: control TSHR = 10, βgal = 6; hFMT-TSHR = 9, βgal = 6; Lab4 TSHR = 9, βgal = 5; vancomycin TSHR = 12, βgal = 8) (n mice/treatment/immunization at mid-timepoint: control TSHR = 11, βgal = 9; hFMT-TSHR = 5, βgal = 9; Lab4 TSHR = 11, βgal = 11; vancomycin TSHR = 14, βgal = 4). b Firmicutes to Bacteroidetes ratio comparing immunizations in each timepoint and within each treatment group. Welch’s t test with BH correction: ***P = 0.0006. c Rationale of the SourceTracker analysis (see STAR methods). Human GO donors (n = 6), murine controls (n control/timepoint: baseline = 16, mid = 20, final =20) and hFMT-receiving mice (n hFMT/timepoint: baseline = 15, mid = 24, final = 19). d Pie charts: engraftment expressed as average % probability using the control and the test analysis in each timepoint. Unknown: observations not assigned to a specific source at the significant threshold (a = 0.001)

Fig. 3

Mouse TSHR antibody evaluation and impact of gut microbiome modification on thyroid function and morphology. a Total TSHR antibodies (TRAbs) were measured by TSH binding inhibitory immunoglobulin activity to the human TSHR given as % inhibition of bTSH. b Thyroid-stimulating autoantibodies (TSAbs). Stimulating activity is given in cAMP (pmol/mL). Data are presented as mean ± standard error of the mean. Multiplicity adjusted P values are marked as follows ****P < 0.0001, ***P < 0.001, *P < 0.05 (two-way ANOVA). c Total T4 values (μg/dl) in βgal- and TSHR-immunized mice from each treatment group *P < 0.05. d H&E of thyroid slices for thyroid morphology evaluation (in normal, heterogeneous (hetero) or hyperthyroid (hyper)). e Representative thyroid images at × 20 magnification. Heterogeneous thyroids contained normal and hyperthyroid regions (indicated by arrows). f Weights of animals at the endpoint of the experiment. Data are presented as mean ± standard error of the mean. Multiplicity adjusted P value is marked as **P < 0.01 (two-way ANOVA)

Fig. 4

Gut microbiome modification modulates orbital tissue abnormalities in GO mouse model. The orbits of mice were fixed, paraffin embedded, and consecutive slices of the middle orbital area were stained with H&E and evaluated for a brown adipose tissue (BAT); representative images are shown, magnification × 10. b BAT as the percentage of the total fat area. c Atrophic fibers in extra orbital muscles; representative images are shown, magnification × 20, arrows indicate areas with atrophic fibers. d atrophic fibers/total muscle fiber area (a.u.). Data are presented as mean ± standard error of the mean. Multiplicity adjusted P values are marked as ***P < 0.001, **P < 0.01, *P < 0.05 (two-way ANOVA)

Fig. 5

Impact on autoimmune hyperthyroidism and orbitopathy and total disease outcomes of gut microbiota modifications. a Percentages of positive untreated TSHR-immunized mice set to 100% and the changes of positivity in the treated groups Lab4, hFMT or vancomycin were calculated. Changes in positive mice are given in % relative to untreated TSHR-immunized mice. Parameters were normalized using the Z-Score (see the “Materials and methods” section). b Z-Score of autoimmune hyperthyroidism includes values of TSAbs and T4. c Z-Score of orbitopathy evaluated by BAT and atrophic fibers. Data are presented as mean ± standard error of the mean. Statistical analysis was performed by two-way-ANOVA; multiplicity adjusted P values are given as follows: *P ≤ 0.05; **P ≤ 0.01; ***P ≤ 0.001; ****P ≤ 0.0001. d, e Disease classification was based on Z-score values. The number of mice is given in %, see STAR methods for classification parameters. Additionally, the number of mice in each category is shown in Table S2

Fig. 6

Effects of gut microbiome modification on lymph node T cell subsets, cytokine and chemokine levels. a CD4⁺ and b CD25⁺ T cell population from draining lymph nodes of TSHR- and βgal-immunized mice in each treatment group. Data are presented as mean ± standard error of the mean. Multiplicity adjusted P values are marked as ****P < 0.0001, *P < 0.05 (two-way ANOVA). c CCL5 and d IL-4 circulating chemokines/cytokines from sera. Cytokine values are log10 transformed. Multiplicity adjusted P values are marked as *P < 0.05 (two-way ANOVA).