Gut Microbiome in BALB/c and C57BL/6J Mice Undergoing Experimental Thyroid Autoimmunity Associate with Differences in Immunological Responses and Thyroid Function

Abstract Experimental models of hyperthyroid Graves’ disease (GD) and Graves’ orbitopathy (GO) are efficiently developed by genetic immunisation by electroporation with human thyrotropin hormone receptor (hTSHR) A-subunit plasmid in female BALB/c (H-2d) mice. We investigated susceptibility in C57BL/6 J (H-2b) animals to allow studies on disease mechanisms in transgenic and immune response gene knock-out mice. Higher numbers of female C57BL/6 J were positive for pathogenic thyroid stimulating antibodies, but induced hyperthyroidism remained at a low frequency compared to BALB/c animals. Assessment of hTSHR specific T cells showed reduced proliferation in C57BL/6 J animals accompanied with anti-inflammatory IL-10, with less pro-inflammatory IFN-γ compared to BALB/c. Whilst the orbital tissue from immune BALB/c mice showed inflammation and adipogenesis, in contrast C57BL/6 J animals showed normal pathology. We characterised the gut microbiota using 16 S ribosomal RNA gene sequencing to explore its possible pathogenic role in the model. Despite being housed under identical conditions, we observed significantly different organisation of the microbiota (beta-diversity) in the two strains. Taxonomic differences were also noted, with C57BL/6 J showing an enrichment of Operational Taxonomic Units (OTUs) belonging to the Paludibacter and Allobaculum, followed by Limibacter, Anaerophaga and Ureaplasma genera. A higher number of genera significantly correlating with clinical features was observed in C57BL/6 J compared to BALB/c; for example, Limibacter OTUs correlated negatively with thyroid-stimulating antibodies in C57BL/6 J mice. Thus, our data suggest gut microbiota may play a pivotal immunomodulatory role that differentiates the thyroid function and orbital pathology outcome in these two inbred strains undergoing experimental GO.


Experimental models of hyperthyroid Graves' disease (GD) and
Graves' orbitopathy (GO) are efficiently developed by genetic immunisation by electroporation with human thyrotropin hormone receptor (hTSHR) A-subunit plasmid in female BALB/c (H-2d) mice. We investigated susceptibility in C57BL/6 J (H-2b) animals to allow studies on disease mechanisms in transgenic and immune response gene knock-out mice. Higher numbers of female C57BL/6 J were positive for pathogenic thyroid stimulating antibodies, but induced hyperthyroidism remained at a low frequency compared to BALB/c animals. Assessment of hTSHR specific T cells showed reduced proliferation in C57BL/6 J animals accompanied with anti-inflammatory IL-10, with less pro-inflammatory IFN-γ compared to BALB/c. Whilst the orbital tissue from immune BALB/c mice showed inflammation and adipogenesis, in contrast C57BL/6 J animals showed normal pathology. We characterised the gut microbiota using 16 S ribosomal RNA gene sequencing to explore its possible pathogenic role in the model. Despite being housed under identical conditions, we observed significantly different organisation of Introduction Autoimmune thyroid diseases (AITD) represent the most common human autoimmune conditions with main clinical presentations of Hashimoto's disease and Graves' disease (GD) [1,2]. Both are characterised by the production of autoantibodies and activated T cells specific for thyroid antigens. The thyroid stimulating hormone receptor (TSHR) present on thyroid follicular cells is the key regulator of thyroid function. The TSHR is also the primary autoantigen in GD, where pathogenic thyroid stimulating antibodies (TSAbs) act as agonists by stimulating the thyroid gland leading to overproduction of thyroid hormone with resultant autoimmune hyperthyroidism [3]. Extrathyroidal expression of TSHR can be detected in a wide variety of cells including adipocytes, bone, fibroblasts, and immune cells such as thymocytes and lymphocytes [4]. It is likely that activation of TSHR in orbital fibroblasts by anti-TSHR antibodies and autoimmune T cells results in inflammatory eye disease known as Graves' orbitopathy (GO), characterised by expansion of the orbital fat tissue and deposition of hyaluronan and fibrosis [5]. The events that link orbital inflammation in GO with the breakdown of immune tolerance to TSHR in the autoimmune process in the thyroid gland remain entirely unknown, although both genetic and environmental factors contribute to the aetiology of the condition [5]. A number of studies have clearly established genetic associations with increased risk for GD, including major histocompatibility complex (MHC) HLA class II haplotypes such as HLA-DR3, genes coding for cytotoxic T lymphocyte-associated antigen (CTLA-4), protein tyrosine phosphatase non-receptor 22 (PTPN22) and thyroid proteins TSHR and thyroglobulin [1,6]. Amongst environmental triggers, bacterial infections such as Yersinia enterocolitica have been implicated by homology of their porins to TSHR proteins [7,8]. Over the last several decades, there has been a dramatic rise in the incidence of autoimmune conditions, including thyroid autoimmune diseases that are believed to arise from substantial changes in the environment, including lifestyle and antibiotic use [1]. These factors may impact the microbiome of intestinal bacteria, fungi and viruses, which are considered to play essential roles in regulating immune tolerance and autoimmunity [8,9].
Graves' disease occurs spontaneously only in humans, although a spontaneous transgenic mouse model with limitations for pathogenesis in the host has been recently described [10]. Mouse models of induced Graves' disease, including some with orbital manifestations, have been described by immunisation with human (h) TSHR coding plasmids, but have proved difficult to reproduce [11,12]. However, immunisation with hTSHR holoreceptor or a secretory extracellular region of the receptor, termed hTSHR A-subunit by genetic delivery with adenovirus or plasmid electroporation in female BALB/c (H-2d) mice represents the most robust induced model for GD [13][14][15][16]. In other studies, different inbred strains of mice have been evaluated to induce experimental GD by these protocols, which have given varying responses [12,17,18]. In particular, immunisation with hTSHR A-subunit delivery by adenovirus in C57BL/6 (H-2b) mice induced higher levels of TSAbs, but lower rates of hyperthyroidism compared to BALB/c [17,18]. Further analysis showed C57BL/6 animals produced higher titre and more potent blocking TSBAbs, resulting in lower levels of hyperthyroidism [17]. Other studies have shown the MHC H-2 region in different inbred strains of mice to play a minor role in induced hyperthyroidism, with the dominance of non-MHC genes in susceptibility to experimental GD in the model [19,20]. We reported on the induction of orbital manifestations in female BALB/c mice that recapitulate a number of clinical and pathological features present in patients with GO, and showed reproducibility of the model housed in different animal units under variable environments [21,22]. In contrast, genetic predisposition to GO in the mouse model is not known, as only one inbred strain of BALB/c animals with susceptibility to experimental GO has been evaluated [21][22][23], although our earlier preliminary investigations by adenovirus immunisation showed no inflammatory infiltrate in the orbital tissue of C57BL/6 animals [18].
From the studies on the GO model in female BALB/c animals housed under variable environments [21], our studies demonstrate gut microbes as a significant environmental factor influencing the onset of experimental hyperthyroidism. We identified disease-associated taxonomies, which could explain the variations observed in the onset of hyperthyroidism in mice of the same BALB/c strain housed in different animal units [24]. In the present study, we aimed to investigate whether the gut microbiota would also have a role in modulating thyroid function and the induced immune response to influence TSHR-induced disease in different female inbred strains of BALB/c and C57/BL6J mice housed in the same animal unit and fed the same diet [8]. Furthermore, since C57BL/6 J animals have not been fully evaluated for induction of experimental GO [18], the study will also provide information on induced orbital manifestations in this inbred strain to allow studies on transgenic and immune gene knock out animals to be pursued [25].

Animals and immunisations
All animal experiments were performed under UK Home Office regulations of Project License and approved by the Ethical Committee of King's College London, and conducted in a specific pathogen free (SPF) environment in animal care facilities as previously described [21,22]. Female BALB/c (subline BALB/cOlaHsd) and nia beads (Benchmark Scientific, USA) for 3 × 60 s at 5 m/s with 5 min rest in-between (Beadbug microcentrifuge homogeniser, Benchmark Scientific USA). DNA concentration was assessed using Qubit Fluorimetric Quantitation (ThermoFisher Scientific Ltd, UK), following manufacturer's protocol.
Multiplex PCR primers targeting the V1-V2 plus regions of the 16 S ribosomal (r) RNA gene were used (Supplementary Table 1 S), and 10 K paired-end sequencing was performed on Illumina Miseq (Illumina, USA) at the Research and Testing Laboratory LLC (Lubbock, USA). Mothur v1.36 [26] was used to process and cluster quality-filtered sequencing reads into Operational Taxonomic Unit (OTU) at the 97 % identity cut-off. SILVA (http://www.arb-silva.de) was used as the bacterial 16 S gene reference database and taxonomic assignment from phylum to genus was performed using the Ribosomal Database Project (RDP) naïve Bayesian classifier with 80 % cut-off [27]. Each sample library was sub-sampled according to the smallest library size to compare sequencing libraries. Statistical analysis and figures were produced in R (version 3.4.1) and p-values less than 0.05 were considered significant.

Statistical analyses
Microbial counts from the mouse groups were compared through the Analysis of variance (one-way ANOVA) and Tukey's multiple comparison test with adjusted p-values. Statistical analysis was performed using GraphPad Prism version 6.00 (GraphPad Software, La Jolla California USA, www.graphpad.com).
For metataxonomics data, statistical analysis was performed in R (v3.4.1). The weighted Unifrac matrix [28] was calculated as a beta-diversity measurement in Mothur and represented in a non-metric dimensional scaling (NMDS) plot using the R Vegan package. Dissimilarities between strains were assessed with the non-parametric permutational analysis of variance (PERMANOVA) implemented in the Adonis function. The hierarchical clustering of genera was performed using the Ward agglomeration method. Heatmap of the top-20 genera most abundant amongst samples was created using "a heatmap function" of the NMF R package with scaled genus abundances (e. g., centre and standardise each column separately to column Z-scores). Operational Taxonomic Units (OTUs) capable of discriminating between strains were identified with a RandomForest algorithm using OTUs with non-zero counts as predictors and 10 000 trees. Differential abundance of discriminating OTUs between strains was assessed using the non-parametric Wilcoxon test. Correlations between genus abundances and disease features (e. g., anti-TSHR antibodies and thyroid hormone fT4 levels) or T-cell proliferation rates, IL-10 and γ-IFN-responses were calculated using the Spearman coefficient (Rho) and represented in a correlation plot. p-Values less than 0.05 were considered as significant.

Immunological and thyroid function studies
We studied antibody and thyroid function in C57BL/6 J mice immunised with hTSHR A-subunit plasmid by electroporation and compared it to our standard BALB/c model of GD and GO induced by the same procedure. We have shown earlier that the majority of BALB/c C57BL/6 J mice of 6-8 weeks age were purchased from Harlan UK (now called Envigo UK). Animals were allowed access ad libitum to water and No.1 Maintenance Special mouse diet and maintained conventionally in open cages in one room and co-housed at a maximum of 4 animals per cage, segregated according to immunisation group. BALB/c (n = 5) and C57BL/6 J (n = 7) animals were immunised by i.m. injection of 50 μg pTriEX1.1 neo-human TSHR A-subunit plasmid (termed hTSHR A-subunit plasmid) followed by leg muscle injection site electroporation, with a total of four injections every three weeks and animals sacrificed nine weeks after the last immunisation [21,22]. All animals were sacrificed on the same day for collection of serum and tissue samples for this study. As controls, BALB/c (n = 4) and C57BL/6 J (n = 3) were immunised with 50 μg β-Gal plasmid as described above. One C57BL/6 J control animal died during immunisations for no apparent reason. Assessment of thyroid disease by thyroid function tests, antibodies to TSHR, splenic T cell stimulation with purified preparations of recombinant TSHR extracellular region protein (rhTSHR-E, amino acids 20 to 419) for proliferative and cytokine responses and histological evaluation of thyroid and orbital tissues was performed as described [21,22]. Histological analysis was conducted blindly without knowledge of the origin of samples.

Traditional microbial cultures of mouse gut contents
After sacrifice, murine intestines were snap-frozen and stored in sterile containers at -80 °C. For subsequent analysis, whole intestines were thawed and directly afterwards placed on a sterile padding. The entire intestines were dissected into several parts to obtain the feces by scraping with a sterile inoculation loop. The contents of these parts from each individual mouse were collected in a sterile container, mixed and stored as one fecal sample at -80 °C, and was used for 16 s rRNA extraction and traditional culture. The gut contents (1 g) were diluted in 9 ml pre-reduced maximum recovery diluent (CM0733, Oxoid, Basingstoke, UK) with 20 % v/v glycerol and mixed by vortexing. After that, 10-fold serial dilutions were prepared and 100 μl of each dilution plated onto different culture media under aerobic or anaerobic conditions (Anaerobic Workstation, AW400SG, Elektrotek, Keighley, West Yorkshire, UK). Specific media, culture conditions and dilutions used to grow different microorganisms were as described in the Supplementary section. Briefly, bacterial counts were made from total aerobe and total anaerobe bacteria as well as Enterobacteriaceae, enterococci, coliforms, staphylococci, yeast, Bacteroides, lactobacilli, bifidobacteria and clostridia.
Identification of bacteria was performed by Gram staining, colony morphology, the presence of spores, catalase reaction and partially by the API system (BioMerieux, Marcy-l'Étoile, France). Viable bacterial cell counts were enumerated and all counts were recorded as the numbers of log 10 colony forming units per gram of sample. Counts data were Box-Cox transformed before statistical analysis.
DNA extraction, 16 S ribosomal RNA gene sequencing and microbiome analysis Total genomic DNA was extracted from the gut contents previously isolated with the QiAmp Fast DNA Stool Mini Kit (Qiagen Ltd, UK), following manufacturer's procedure. A bead-beating step was included to lyse bacterial cells in the presence of 1 ml InhibitEX buffer (Qiagen Ltd, UK) and a mixture of sterile 0.1 mm silica and zirco-animals undergo experimental GD and GO at nine weeks after the end of immunisation [21,22]. All C57BL/6 J animals challenged with hTSHR A-subunit plasmid showed high levels of anti-TSHR antibodies when measuring the TSH binding inhibitory immunoglobulins (TBII) activity, similar to the immune BALB/c animals (▶Fig. 1a). Control animals immunised with β-Gal plasmid scored negative for TBII activity (▶Fig. 1a). Assessment of biologically functional anti-TSHR antibodies as TSAbs showed five out 7 (70 %) positive immune C57BL/6 J animals, compared to 2 out of 5 immune BALB/c mice (40 %). Both strains also developed TSH stimulating blocking antibodies (TSBAbs) (▶Fig. 1b). We evaluated thyroid function by measurement of serum total T4 (TT4) concentrations, which showed 1 out of 7 (15 %) immune C57BL/6 J animals to be hyperthyroid, compared to 3 out of 5 (60 %) immune BALB/c mice (▶Fig. 1c). Control animals in both strains immunised with β-Gal plasmid remained euthyroid with no signs of illness (▶Fig. 1c).

Thyroid gland and orbital histology
Histological examination of thyroid glands from hTSHR A-subunit immunised C57BL/6 J animals showed one hyperplastic gland with a microfollicular pattern with cuboidal cells lining the follicles and with increased cellularity (▶ Fig. 1d, arrow). Interestingly, the serum TT4 concentrations from the animals with hyperplastic thyroid gland indicated euthyroid status. Examination of thyroid glands from the remaining immune C57BL/6 J animals showed standard architecture with no signs of inflammation, similar to control β-Gal plasmid immunised animals (▶ Fig. 1e). In contrast, all thyroid glands from hTSHR A-subunit immunised BALB/c animals showed activated status (not shown), similar to our earlier studies [21,22]. Examination of orbital tissue from C57BL/6 J animals immunised with TSHR A-subunit plasmid showed a histological pattern that was similar to that obtained with control β-Gal plasmid immunised animals, showing that this inbred strain was resistant to the induction of orbital disease (▶ Fig. 1f). In contrast, all TSHR A-subunit plasmid BALB/c animals showed signs of orbital disease, manifest by either inflammation or expansion of adipose and connective tissue (not shown) [21,22].

Proliferative and secretory cytokine responses in antigen specific T cells
To gain insight into the serological and pathological differences in these two inbred strains immunised by electroporation with hTSHR A-subunit plasmid, we studied unfractionated splenic T cells for their proliferative responses to purified rhTSHR-E antigen. Strong proliferative T cell responses to rhTSHR-E at different doses of 30, 10 and 3 μg/ml antigen were obtained in BALB/c animals undergoing experimental GD and GO, whilst C57BL/6 J cells showed poor antigen-specific proliferative responses (▶ Fig. 2a). To further investigate differences in T cell recall responses between the two inbred strains of mice after immunisation with hTSHR A-subunit, we ▶Fig. 1 Evaluation of thyroid function, antibody and histology in BALB/c and C57BL/6 J mice immunised with hTSHR-A subunit plasmid. a: Measurement of total anti-TSHR antibodies in serum using TSH binding inhibitory immunoglobulin (TBII) activity in competition with bovine TSH. Data show significant induction of TBII activity in all hTSHR A-subunit plasmid immunised mice in both strains. Borderlines indicate mean + 3 SD of control mice immunised with β-Gal for BALB/c and C57BL/6 J in left vertical and right vertical axis respectively. b: Measurements of blocking TSBAbs and stimulating TSAbs in individual animals by bioassay. The ordinate shows percent blocking of TSH-mediated stimulation for TSBAbs (left axis) or cAMP generated for TSAbs (right axis). The dotted line indicates the mean + 3 SD for TSAbs of the control β-Gal immunised mice, the joined points in TSABs and TSBAbs indicate the respective measurements in the same animal. c: Total T4 values in the serum of hTSHR A-subunit immunised BALB/c mice plotted on the left vertical axis and C57BL/6 J projected on the right axis. Borderline shows mean + 3 SD of control mice immunised with β-Gal for each strain. d: A representative histology of thyroid gland retrieved from C57BL/6 J mouse immunised with hTSHR-A subunit plasmid showing hyperplastic gland with increased thyrocyte cellularity (arrowed). e: A representative histology of normal thyroid gland retrieved from control C57BL/6 J mouse. f: Orbital histology of C57BL/6 J mouse without any apparent inflammation or extensive adipogenesis. The accumulation of pigmented dendritic cells is marked by an arrow. determined their cytokine responses to rhTSHR-E antigen. Splenocytes co-cultured with hTSHR-E antigen show significantly higher anti-inflammatory IL-10 responses with concomitant lower pro-inflammatory γ-IFN-responses in C57BL/6 J animals, compared to BALB/c mice which show strong pro-inflammatory and proliferative responses to the TSHR antigen (▶Fig. 2b,c).

Intestinal microbiota of BALB/c and C57BL/6 J animals undergoing experimental disease
We analysed the bacterial contents derived from the whole intestines of BALB/c (n = 5) and C57BL/6 J (n = 7) mice immunised with the hTSHR A-subunit plasmid. Viable counts were obtained from standard microbiology cultures for both the aerobes (including coliforms and enterococci), anaerobe bacteria (such as lactobacilli, bifidobacteria, and Bacteroides) and also yeasts (▶table 1); however, bacteri-▶Fig. 2 Cell proliferation and cytokine expression in antigen-specific splenic T cells. The culture of total splenocytes from BALB/c and C57BL/6 J immunised with hTSHR-A subunit plasmid were incubated with 30, 10 and 3 μg/ml rhTSHR-E antigen for five days. Tracking of fluorescent intensity of proliferation dye (CFSE) was used to detect antigen-specific T-cell proliferation. a: T cell proliferation assay on TSHR-A subunit immunised BALB/c mice at three different doses of antigen projected on the left vertical axis and similarly for C57BL/6 J mice on the right axis. The dotted line indicates the mean + 3 SD for T cell proliferation of the appropriate control mice immunised with the β-Gal plasmid. b: Secretion of γ-IFN into the culture medium of splenocytes incubated with 30 μg/ml rhTSHR-E antigen. Note the different units on the y-axis. This data indicates that splenocytes from BAL-B/c mice (left axis) immunised with TSHR A-subunit produce significantly higher level of γ-IFN than TSHR A-subunit plasmid immunised C57BL/6 J mice (right axis). Control mice of both strains showed similar values and were used to generate the borderlines. c: Secretion of IL-10 into the culture medium of splenocytes incubated with 30 μg/ml rhTSHR-E antigen. This data indicates that splenocytes from C57BL/6 J mice tend to produce higher IL-10 level than of BALB/c mice. Although higher IL-10 secretion in C57BL/6 J mice even from control suggests this as a systemic difference between two strains, one-way ANOVA (Kruskal-Wallis test) and post hoc test, as well as non-parametrical tests for comparison between TSHR A-subunit immunised mice in C57BL/6 J and BALB/c mice show a significant difference that presented in the figure. Mann-Whitney test and Kolmogorov-Smirnov test were used for non-parametrical tests.  al contents scored by this method were not significantly different between immunised BALB/c and C57BL/6 J mice (▶table 1). We next performed the 16 S rRNA gene sequencing (metataxonomics) in order to characterise the overall microbiota composition. A total of 1214 Operational Taxonomic Units (OTUs) with more than 10 counts across samples were obtained from good-quality subsampled sequences, accounting a mean 98 % Good's coverage per OTU definition (median value 0.97 ± 0.003), and used for subsequent analysis.
Significant spatial separation between BALB/c and C57BL/6 J bacterial communities (beta-diversity) was observed based on the weighted UniFrac metric (p = 0.025; ▶ Fig. 3a), largely confirmed by the Ward Hierarchical Clustering based on the twenty most abundant genera (▶ Fig. 3b). However, there was no variation in the richness and diversity (alpha diversity) of the bacterial communities between the two immune strains (Supplementary ▶Fig. 1S).
We used the RandomForest algorithm to discriminate between the two inbred strains. The gut microbiota composition was shared mainly by the two strains, accounting an overall accuracy of the classification algorithm of near 42 %. However, we were able to observe the enrichment of two Operational Taxonomic Units (OTUs) from genus Paludibacter (p = 0.005), and Allobaculum (p = 0.01) in C57BL/6 J immune mice (▶Fig. 3c). Other distinguishing OTUs be-tween the two strains belonged to the genera Limibacter (p = 0.0184), Anaerophaga (p = 0.0184) and Ureaplasma (p = 0.0182) (▶ Fig. 3c).

Correlation of gut microbiota in BALB/c and C57BL/6 J animals undergoing experimental disease
When exploring associations between disease features and the gut bacterial composition, a different pattern of correlations was identified within the two strains (▶Fig. 4). A higher number of genera correlated significantly with disease features in C57BL/6 mice, both positively (Rho > 0.5) and negatively (Rho < 0.5). The genus Limibacter OTUs were higher in C57BL/6 J mice, and correlated negatively with TSAb levels (Rho -0.85, p = 0.013). Whereas in the same mouse strain, the genus Marvinbryantia correlated negatively with TSAb levels (Rho -0.85) and positively with TSBAb (Rho 0.788). In BALB/C mice, TBII levels and TT4 correlated positively to each other (Rho > 0.8). In both strains, levels of IFN-γ and IL-10 were strongly positively correlated with each other and with the genus Erysipelotrichaceae incertae sedis (Rho > 0.8). Correlations with genera and T-cell proliferation provided similar features among stimulations in BALB/c animals, but not in the C57BL/6 J, apart for genus Bilophila, which were negatively correlated with all the T-cell proliferation doses (Rho -0.79) (▶Fig. 4).

Discussion
In our recent study, we demonstrated that the gut microbiota composition is a significant environmental factor that may influence the onset of clinical hyperthyroidism in our BALB/c model of GD and GO induced in different institution environments [24]. The variability of the gut microbiota could also depend on the genetic background of mouse strains [19,20,29,30], which in turn might contribute to differences in the disease model. Therefore we combined immunological investigations and detailed study of the gut microbiota composition in BALB/c and C57BL/6 J following immunisation with hTSHR A-subunit plasmid.
Although a higher number of C57BL/6 J immunised animals were positive for TSAbs (71 %) compared to BALB/c mice (40 %), only 1 of 7 (15 %) were hyperthyroid, compared to 60 % in BALB/c mice. Analysis of histology showed only one C57BL/6 J thyroid gland with hyperplastic features of activated follicular cells. Of more importance, in C57BL/6 J animals immunised with hTSHR A-subunit plasmid orbital morphology was comparable to that in control β-Gal plasmid immunised animals, whilst similarly immunised BALB/c mice displayed orbital pathology.
Our serological results confirm studies of other groups [17] who used adenovirus delivery of hTSHR A-subunit to immunise female C57BL/6 mice and found that they had lower susceptibility to hyperthyroidism than BALB/c mice. The authors attributed the difference to high titres of blocking TSBAbs in the C57BL/6 mice [17]. However we are confident that our immunisation method, that is, plasmid immunisation with hTSHR A-subunit plasmid by electroporation, is superior to other modes of genetic delivery since in female BALB/c mice we were able to induce a strong immunological response and onset of both thyroid and orbital pathology, even resulting in breakdown of immune tolerance when using syngeneic antigen to induce disease [31]. The striking difference between the two strains of inbred mice immunised with hTSHR A-subunit plasmid by electroporation led us to examine the immunological profile in more detail.
Ex vivo analysis of splenic cells from immune BALB/c animals showed robust proliferative responses to the receptor protein compared to those in C57BL/6 J animals. In addition to variation in T cell proliferative responses, a clear difference between anti-and pro-inflammatory secretory cytokines after antigen-specific stimulation and T-cell proliferation rates ( %) in BALB/c and C57BL/6 J immunised with hTSHR-A subunit. Tcell_30, Tcell_10 and Tcell_3 represent the T-cell proliferation ( %) at 30, 10 and 3 µg/ml rhTSHR-E antigen stimulation respectively. TRAK refers to TBII activity measured as % inhibition of luminescence-labelled bTSH binding; IL_10 and INFg refer to IL-10 and IFN-γ respectively. Rho, Spearman correlation coefficient is shown in tiles. The strength of the correlation coefficient is represented by the colour change. Only correlations with p < 0.05 are shown. of splenic T cells suggested a systemic distinction between the two inbred strains in response to our immunisation protocol. Earlier studies have shown the MHC haplotype to play minimal roles in induction of experimental GD, suggesting that non-MHC genes make major contribution to the aetiology of the disease in the mouse model [20,21]. These non-MHC genes may lead to a shift in the immune response towards humoral (higher levels TSAbs) and prevent cellular reactions (lower T cell proliferative and pro-inflammatory cytokines) in C57BL/6 J mice.
Our metataxonomic data, comparing GO models in the same BALB/c strain but from different vendors and housed in separate animal units, demonstrated more marked variation (e. g., significant differences in the alpha as well as beta diversity indices and from standard microbiology data) than those reported here between distinct H-2 strains in the same laboratory [24]. However significant taxonomic differences between the two strains were observed. TSHR-immunised C57BL/6 J mice showed an increased abundance of OTUs belonging to the genera Limibacter, Paludibacter and Allobaculum compared to BALB/c mice. The same genera were found to be enriched in an ACE2 -/-colitis model in C57BL/6 mice, associated with an increased susceptibility to develop DSS colitis. Dietary tryptophan and nicotinamide treatments were able to realign the microbiota composition and restore the phenotype to wild-type [32]. We observed a down-regulation of the metagenomic tryptophan metabolism pathway, imputed from the 16 S rRNA gene sequencing, in C57BL/6 J mice but not BALB/c, with both strains in the same animal unit and room and fed the same diet ( Supplementary ▶ Fig. 2S).
In BALB/c animals undergoing experimental GD and GO counts from Clostridium-IV and Anaerotruncus genera were positively correlated with TSAbs levels, whilst Robinsoniella, Pseudobutyrivibrio and Acetitomaculum were positively correlated with TSBAbs. In C57BL/6 J mice OTUs from Robinsoniella positively correlated with TSAbs whilst Marvinbryantia and Dorea were positively correlated with TSBAbs. Although Robinsoniella counts were similar in both strains, they were positively correlated with TSBAb levels in BALB/c but with TSAb levels in C57BL/6 J immune mice.
How the different bacterial genera in the gut microbiome of hTSHR A-subunit plasmid immunised BALB/c and C57BL/6 J mice might trigger production of autoantibodies remains to be investigated. Recent studies have indicated gut microbiome to produce a range of different short chain fatty acids (SCFAs) as fermentation products that can modulate immune responses, including autoimmune responses in the model type 1 diabetes in NOD mice [33,34]. Moreover, SCFAs have also been shown to modulate survival of antibody secreting plasma cells [35,36], which in our mouse model of GD and GO may potentially influence the autoreactive long lasting plasma B cells secreting TSAbs or TSBAbs. This may provide an explanation for the association of the gut microbiome on the induced GD and GO model described in this study.
The prevalence of Gram-positive (Robinsoniella) or Gram-negative (Bilophila) bacteria in the gut microflora has a differential influence in the immunologic response [37][38][39]. Despite the phylogenetic levels, it has been shown that Gram-positive bacteria induced much more IL-12 and IFNγ, more TNF-α, and slightly more IL-1β than did Gram-negatives, which instead induced more IL-6, IL-8, and IL-10 than did Gram-positives [40]. Also, monocytes produced higher levels of interleukin 12 and tumour necrosis factor in response to Gram-positive than in response to Gram-negative bacteria. In contrast, dendritic cells secrete large amounts of IL-12, TNF-α, IL-6, and IL-10 in response to Gram-negative but have been practically unresponsive to Gram-positive bacteria [38]. The lack of a response to the Gram-positive strains has been shown to correlate with lower surface expression of Toll-like receptor 2 (TLR2) on dendritic cells than on monocytes, and gamma interferon increases the expression of TLR4 on dendritic cells, potentiating the cytokine response to the Gram-negative strains [38]. The fundamental difference in innate immune responses to Gram-positive and Gram-negative bacteria, which crosses taxonomic borders, probably reflects differences in cell wall structure. Thus, the composition of the microbial community and gut bacterial abundance in BALB/c and C57BL/6 J animals immunised with hTSHR A-subunit plasmid might show potential associations in modulating immunological profile which subsequently altered thyroid function and orbital pathology in the mouse model.
Recent studies in Hashimoto's disease and other autoimmune conditions such as multiple sclerosis and type 1 diabetes are beginning to provide compelling links with gut microbiota in disease pathogenesis [41][42][43][44][45][46]. However, studies on Hashimoto's disease in the human and experimental mouse model provide contrasting results on the role of gut microbiota in disease development [41,47]. Treatment of mice undergoing experimental Hashimoto's thyroiditis with probiotic strains such as Lactobacillus rhamnosus and Bifidobacterium lactis exhibited neither inhibitory nor stimulatory effects on disease development, whereas recent studies on human patients link gut microbial dysbiosis with development of destructive (Hashimoto's) disease [41,47], and reviewed in [48]. On the contrary, for experimental GD, our studies provide strong evidence for cross reactivity between TSHR and the cell surface porins of infectious agents such as Yersinia enterocolitica [7,8]. In other autoimmune conditions such as mouse models of retinal autoimmunity, cross reactivity of gut microbial antigens with pathogenic T cell determinants have been shown to trigger autoimmunity, highlighting immune privileged sites with sequestered antigens such as the eye to be susceptible to pathogenic T cells activated in the gut [49]. However, the nature of the gut microbial antigen that cross reacts with eye antigens to activate the pathogenic T cells remain to be elucidated [50]. An understanding of how the composition of the gut bacterial microbial community, and abundance of specific bacterial strains, influences autoimmune responses in GD and GO may further advance our knowledge on disease pathogenesis of these common conditions (Fig. 3S).