Galectin-3 deficiency drives lupus-like disease by promoting spontaneous germinal centers formation via IFN-γ

Germinal centers (GC) are important sites for high-affinity and long-lived antibody induction. Tight regulation of GC responses is critical for maintaining self-tolerance. Here, we show that Galectin-3 (Gal-3) is involved in GC development. Compared with WT mice, Gal-3 KO mice have more GC B cells and T follicular helper cells, increased percentages of antibody-secreting cells and higher concentrations of immunoglobulins and IFN-γ in serum, and develop a lupus-like disease. IFN-γ blockade in Gal-3 KO mice reduces spontaneous GC formation, class-switch recombination, autoantibody production and renal pathology, demonstrating that IFN-γ overproduction sustains autoimmunity. The results from chimeric mice show that intrinsic Gal-3 signaling in B cells controls spontaneous GC formation. Taken together, our data provide evidence that Gal-3 acts directly on B cells to regulate GC responses via IFN-γ and implicate the potential of Gal-3 as a therapeutic target in autoimmunity. Germinal center (GC) is where B cells interact with other immune cells for optimal induction of antibody responses. Here the authors show that galectin-3 regulates GC development by modulating interferon-γ and B cell-intrinsic signaling, such that galectin-3 deficiency mice exhibit lupus-like autoimmune symptoms.

B cell function provides a key line of defence against infections through the production of high affinity antibodies, which are critical for clearing pathogens and preventing reinfections. These antibodies are the main product of long-lived antibody-secreting cells (ASCs) or plasma cells (PCs), which are generated with the help of a distinct subset of CD4 + T cells called T follicular helper (Tfh) cells, within a specialised microenvironment known as the germinal center (GC) [1][2][3] . The GC contains highly proliferative B cells that undergo Ig somatic hypermutation and isotype switching, in which B cells are also selected based on their antigen affinity. Indeed, GCs are an important B cell-tolerance checkpoint in the periphery. IL-4 and IL-21 are the most important cytokines for assisting B cells during the differentiation process to GC B cells, and mice that are deficient for both IL-4 and IL-21 receptors (Il4 −/− Il21r −/− mice) were found to have severe defects in antibody production and GC formation 4 .
Although GC activity is fundamental for health by inducing protective immunity, overreactivity of the GC pathway causes autoimmunity. Dysregulated GC B cell and Tfh responses make decisive contributions to the generation of class-switched autoantibodies and to the development of lupus in several mouse models [5][6][7][8] . Associations between autoimmunity and overactive GCs have also been suggested in human systemic lupus erythematosus (SLE), Sjogren's syndrome, rheumatoid arthritis, and autoimmune thyroid disease 9 . Vinuesa and collaborators 10 have demonstrated that the reduction of GC cell numbers caused by the specific deletion of one allele of Bcl6 ameliorates the autoimmune pathology observed in Roquin san/san (sanroque) mice, which develop GCs in the absence of foreign antigen. A reduction of spontaneous GC formation via the blockade of T cell signaling in different autoimmune experimental models also reduces disease 6,7,11 . Thus, a fine equilibrium between GC formation and number is required for tolerance.
GC formation and B cell differentiation to PCs is strictly regulated by several transcription factors 12 . Bcl-6 and Blimp-1 are transcriptional repressors with the ability to block each other's expression. Bcl-6 is expressed by GC B cells and is required for this phase of B cell development. Bcl-6-deficient mice lack GC B cells and affinity maturation. Constitutive expression of Bcl-6 in B cells in vivo results in a large GC. In addition, three studies have shown that Bcl-6 is a master regulator of Tfh differentiation [13][14][15] . PC differentiation, in contrast, critically depends on Blimp-1 expression and the absence of Bcl-6. Blimp-1 is highly expressed in PCs and controls many genes that are important for PC differentiation. Furthermore, this transcription factor also inhibits genes involved in cellular proliferation, such as Myc and Bcl6, thereby allowing the terminal differentiation of PC 16 .
Although much progress has been made in understanding the factors and rules that govern the generation of GCs and the formation of long-lived ASCs during infections and autoimmunity, the identification of new molecules is crucial to achieve a better understanding of this complex process. This is a pivotal issue not only for the design of vaccines against pathogens and malignant cells but also for therapeutic intervention in autoimmunity.
We and others have shown that the downregulation or deficit of Galectin-3 (Gal-3) leads to an increase in the number of ASCs during the course of Trypanosoma cruzi and Schistosoma mansoni infection 17,18 and promotes peritoneal B1 cell differentiation into PCs 19 . Moreover, B cells with reduced expression of endogenous Gal-3 fail to down-regulate Blimp-1 after IL-4 stimulation 17 . Hoyer and collaborators have demonstrated that Gal-3 is present in naive and memory B cells but almost absent in differentiated B cells, such as CD10 − /IgD − GC B cells, and PCs 20 .
Taken together, these findings support the hypothesis that Gal-3 may influence GC and ASC generation.
Gal-3 belongs to the large growing family of highly conserved β-galactoside binding lectins known as galectins. Gal-3 is the only chimera-type galectin composed of an unusual non-lectin proline and glycine-rich domain coupled to a carbohydrate-recognition domain 21 . Gal-3 has a broad distribution among different types of cells and tissues. It can be localised intracellularly, in the nucleus and cytoplasm, or extracellularly, secreted via the non-classical pathway 22 . Relying on its expression level, the type of cell and the specific immune condition, Gal-3 can be either a positive or a negative regulator of the immune response [23][24][25] . Considering that Gal-3 is expressed in B cells 17,20 and T cells and that the downregulation of Gal-3 improves antigen-specific antibody production 17 , we explored whether Gal-3 could influence the GC reaction and B cell differentiation into ASCs, processes that involve a delicate crosstalk between B and T cell populations.
Our study reveals previously unrecognised functions of Gal-3 in the regulation of GC responses. The absence of Gal-3 in mice drives an excess of IFN-γ, which leads to aberrant GC formation and autoantibody production. This study reveals Gal-3 as a key factor in the development of IFN-γ-mediated lupus-like disease.

Results
Gal-3 has a critical function in GC formation. To investigate the Gal-3 expression pattern in differentiated B cells, purified splenic B cells were stimulated with LPS or anti-CD40 plus IL-4 to induce differentiation into ASCs. Figure 1a shows that Gal-3 protein was detectable in splenic B cells and its expression decreased to undetectable and very low levels in LPS-stimulated and anti-CD40 plus IL-4-activated B cells, respectively. To define the kinetics and pattern of Gal-3 downregulation upon B cell differentiation induced by LPS, splenic B cells were cultured alone or with LPS for different times. While the Gal-3 protein level rose in B cells cultured for 48 and 72 h without stimuli, it became completely undetectable after 48 h of LPS stimulation (Fig. 1a).
As LPS is an inducer of PC differentiation 26 , the abovedescribed results suggest that a decrease in Gal-3 expression can be associated with antibody secretion. Thus, we next compared Gal-3 expression between peritoneal B1 and splenic B2 cells with differential abilities to spontaneously secrete Igs. We observed no differences in Gal-3 expression among these two B cell subsets ( Supplementary Fig. 1a). Additionally, Gal-3 was also detectable in T cells, and this expression was substantially lower than in B cells ( Supplementary Fig. 1a). In agreement with previous reports 27 , specific fluorescence staining for Gal-3 was detected in both the nucleus and cytoplasm of non-stimulated resting B cells ( Supplementary Fig. 1b, upper panels), and a remarkable loss of fluorescence intensity was observed when B cells were stimulated with LPS ( Supplementary Fig. 1b, lower panel). These findings indicate that the differentiation of B cells to ASCs is not associated with a sub-cellular redistribution of Gal-3 protein but rather a reduction of protein synthesis. Indeed, stimulation of B cells with LPS induced a down-regulation of the transcript encoding Gal-3 ( Supplementary Fig. 1c). Gal-3 mRNA levels were significantly down-regulated after 24 and 48 h of LPS stimulation with respect to the transcript level at 12 h (p < 0.001 by one-way ANOVA). In agreement with the protein expression level (Fig. 1b), B cells cultured in the absence of stimulation had similar Gal-3 mRNA levels at 12 and 24 h, but a significant upregulation (p < 0.01 by one-way ANOVA) was detected after 48 h of culture with medium ( Supplementary Fig. 1c).
To address the role of Gal-3 in B cell immunobiology in vivo, we took advantage of Gal-3 KO mice. We first compared the concentration of serum antibodies and the frequency of splenic and bone marrow ASCs in Gal-3 KO and WT mice. We found that, without purposeful immunization or infection, Gal-3 KO mice displayed significantly (p < 0.001 by unpaired Student's ttest) higher concentrations of IgM, IgG2c, and IgG3 in serum (Fig. 1c), as well as higher frequencies of IgM and IgG ASCs in spleen and bone marrow (Fig. 1d), in comparison to WT mice. Considering that production of antibodies under homoeostatic conditions relies on different B cells subsets and can be produced by ASCs derived or not from the GC, we examined the frequency of different mature B cells (follicular, FO, and marginal zone, MZ, B cells), GC B cells and PCs in spleen from Gal-3 KO and WT mice. We found no significant differences (p > 0.05 by unpaired Student's t-test) in the absolute number of FO and MZ B cells between the groups of mice (Fig. 1e), but Gal-3 KO mice had a significantly greater number of GC (B220 + FAS + GL7 + CD3 − CD11b − CD11c − ) and PC (B220 int CD138 + CD3 − ) B cells ( Fig. 1f-h) compared with their WT counterparts (p < 0.001 and p < 0.01 by unpaired Student's t-test, respectively). The phenotype of GC B cells was confirmed by low levels of IgD and CD38 ( Supplementary Fig. 2a-c) and high levels of CXCR4 and CD40 expression, and a high frequency of Ki-67 + cells within B220 + GL7 + FAS + B cells, in comparison to the phenotype exhibited by non-GC B cells (B220 + GL7 − FAS − ) (Supplementary Fig. 2d-f, respectively). Immunofluorescence analysis of spleen sections revealed larger spontaneous PNA+ + GCs within the follicles in Gal-3 KO mice, whereas WT mice showed a complete absence of GCs (Fig. 1i). When we analysed the presence of Tfh cells, identified as CD4 + CXCR5 + ICOS + Foxp3 − T cells, in Gal-3 KO mice, we observed that, in clear contrast to the few Tfh cells observed in WT mice, the spleen of Gal-3 KO mice exhibited an exacerbated frequency of Tfh cells (Fig. 1j). We verified the Tfh cell identity based on the high PD-1 and Bcl-6 expression (Supplementary Fig. 3a-d). We also examined Tfr cells, a subset of CD4 + Foxp3 + Treg cells within the GC that coopt a CXCR5 + ICOS + phenotype and limit the magnitude and output of the GC response. Among CD4 + CXCR5 + ICOS + T cells, the frequency of Tfr cells was reduced in Gal-3 KO mice in comparison to WT (Fig. 1k). However, the total Treg (CD4 + Foxp3 + CD25 + ) frequency and number were similar in both groups of mice ( Supplementary Fig. 4a), and there were no differences in expression of the regulatory molecules CTLA-4 or CD39 on Treg cells between WT and Gal-3 KO mice ( Supplementary Fig. 4b, c, respectively). Of note, the Tfh/Tfr ratio, which is an important factor in humoral immunity because it dictates the magnitude of antibody responses 28 , was increased in Gal-3 KO mice. Taken together, these results indicate that the absence of Gal-3 drives spontaneous GC formation and, consequently, Igs production.

Gal-3-deficient B cells express GC-associated molecules.
To investigate whether Gal-3 affects the transcriptional programme of B cells, we performed gene expression profiling by microarray using purified B cells from spleens of WT and Gal-3 KO mice (Fig. 2a). Heat maps highlighted the most differentially expressed transcript of genes that are relevant for GC biology (Fig. 2a). Remarkably, the levels of transcripts encoding Igs and molecules associated with the GC B cell phenotype and function, such as Gcet2 29 and H2B 30 ; the indicator of proliferation Mki67 and the critical component of antibody class-switching Aicda were upregulated in B cells from Gal-3 KO mice. In accordance with spontaneous GC formation, the array analysis showed an upregulation of two genes encoding miRs associated with the GC   Fig. 2 B cells from Gal-3 KO mice had increased expression of genes related to GC formation and costimulatory molecules. a Heat maps of the expression of selected genes related to the GC reaction determined in B cells from WT and Gal-3 KO mice (n = 2 for both strains). b Statistical analysis of the frequency of CD80 + , CD86 + , IL-21R + , CXCR4 + , Ki-67 + , and CD69 + B220 + B cells from the spleen of 8-week-old WT and Gal-3 KO mice (n = 6 for both strains). c Radar plots summarising the B cell phenotype observed in the spleen of Gal-3 KO and WT mice (n = 6 for both strains). Data are the mean ± s.e. m. *p < 0.05, **p < 0.005 (unpaired Student's t-test). a Data are representative of two independent experiments. b, c Data are representative of three independent experiments Fig. 1 Downregulation of Gal-3 during PC differentiation and increased concentration of Igs and spontaneous GC formation in Gal-3 KO mice. a Splenic B cells were cultured with medium, LPS, or anti-CD40 plus IL-4 for 72 h, and Gal-3 expression was evaluated by western blot analysis. The Gal-3 signal was normalised to β-Actin (n = 5 for each condition). b Splenic B cells were cultured medium or LPS from 12 to 72 h. The Gal-3 signal was normalised to calreticulin (n = 3 for each condition). c Quantification by ELISA of IgM, IgG1, IgG2c, IgG2b, and IgG3 in sera from 8-week-old WT and Gal-3 KO (n = 11 for both strains) mice. d Representative images (left) from the ELISPOT assay of IgG-ASCs in the spleen and BM and statistical analysis (right) of IgM and IgG ASCs in the spleen and BM from 8-week-old WT and Gal-3 KO mice (n = 5 for both strains). e Numbers of splenic MZ B cells and FO B cells determined by flow cytometry in WT and Gal-3 KO mice (n = 6 for both strains). f Numbers of splenic GC B cells and PCs determined by flow cytometry in WT and Gal-3 KO mice (n = 6 for both strains). g, h Representative plots and statistical analysis of splenic GC B cells (g) and PCs cells (h) in WT (n = 6) and Gal-3 KO (n = 8) mice. i Immunofluorescence of spleen sections (7 μm) from WT and Gal-3 KO mice, indicated as KO, stained with PNA (green), anti-B220 (red), and anti-CD3 (blue). Magnification: ×100 (top) and ×200 (bottom) (n = 6 for both strains). Scale bar: 200 μm (upper panels), 100 μm (lower panels). j Representative plots and statistical analysis of splenic Tfh cells in WT and Gal-3 KO mice (n = 6 for both strains). k Representative plots and statistical analysis of the ratio between splenic Tfh and Tfr cells in WT (n = 6) and Gal-3 KO (n = 8) mice. Data are the mean ± s.e.m. *p < 0.05, **p < 0.005, ***p < 0.001. ns, not significant (unpaired Student's t-test). a-k Data are representative of three independent experiments. WT mice are indicated in green and Gal-3 KO mice in blue reaction: miR-155 31,32 and miR-217 33 . The absence of Gal-3 in B cells resulted in a downregulation of miR-181b-1, the decrease of which is also associated with the GC response and lupus 34 . The gene and surface molecule expression profile of Gal-3 KO-derived B cells highlighted that the absence of Gal-3 favors an activation phenotype that is prone to differentiate into GC B cells. To gain further insights into the overall impact of the absence of Gal-3 signaling on B cells, we determined the expression of markers involved in the GC reaction. We observed that IL-21R, a key cytokine receptor involved in the regulation of the Bcl-6 and GC response 35 , was expressed at a higher frequency on B cells from Gal-3 KO in comparison to WT (Fig. 2b, c). We also found a higher percentage of CD80-expressing and CD86-expressing B cells in Gal-3 KO mice compared with their WT counterparts Gal-3 acts directly on B cells to regulate GC responses. Next, we wanted to address whether Gal-3-deficient B cells have the capability to spontaneously differentiate and secrete Igs in the absence of Ag stimulation and within a Gal-3-sufficient physiological environment. To achieve this goal, we performed two different adoptive transfer experiments. First, total purified splenic B cells from WT or Gal-3 KO mice were transferred to B celldeficient μMT hosts, and the levels of antibodies were quantified in sera two weeks after reconstitution (Fig. 3a). Sera from mice transferred with Gal-3-deficient B cells presented a higher concentration of IgG2c and IgG3 than sera from recipients of B cells from WT (Fig. 3a). As expected, sera from μMT that did not receive cells had undetectable Igs. In a second approach, we transferred sorted GL7 − Fas − B cells (non-GC B cells) from WT or Gal-3 KO mice into WT recipients (Fig. 3b). Sera from WT mice that received non-GC Gal-3deficient B cells had a higher concentration of IgG2c and IgG3 than unmanipulated WT mice or WT mice transferred with Gal-3-sufficient non-GC B cells (Fig. 3b).
Next, we assessed the impact of endogenous Gal-3 deficiency on the ability of B cells to produce antibodies in vitro. To achieve this goal, we cultured sorted GL7 − Fas − B cells from WT or Gal-3 KO mice with sorted CD4 + T cells from WT or Gal-3 KO mice in the presence of anti-IgM and anti-CD3 (Fig. 3c). Gal-3 KO B cells produced larger amounts of IgG2c and IgG3 than WT B cells, independently of the presence of CD4 + T cells from Gal-3 KO or WT mice. All the results described above support the concept that B cell-intrinsic Gal-3 deficiency promotes spontaneous antibody secretion and, probably, spontaneous GC formation.
Increased IFN-γ-producing CD4 + T cells in Gal-3 KO mice. Because Gal-3 KO mice had a particular serum Ig isotype profile, we investigated whether this pattern correlated with a specific type of T cell response defined by cytokine production. Cytokines secreted by Tfh cells and other cells within the GC niche influence both the growth, survival, and differentiation of ASCs as well as the differentiation of Tfh cells in an autocrine manner 35,36 . We observed that, following anti-CD3/CD28 stimulation, purified CD4 + T cells from Gal-3 KO mice produced significantly higher (p < 0.01 by unpaired Student's t-test) amounts of IFN-γ, IL-21, and IL-4 and similar levels of IL-6 and IL-2 compared with CD4 + T cells from WT (Fig. 4a). Accordingly, the frequency of splenic IFN-γ-producing CD4 + T cells in Gal-3 KO mice was 2.5-fold higher than that observed in WT controls (Fig. 4b), while no difference in the frequency of IFN-γ-producing CD8 + T cells, DC cells, NK cells or NKT cells ( Supplementary Fig. 5a-d, respectively) was observed between the experimental groups.
Interestingly, Gal-3 KO mice also exhibited a higher frequency of IFN-γ-producing B220 + cells than WT controls (Fig. 4c). Additionally, Tfh contained the highest frequency of IFN-γproducing cells and displayed the highest IFN-γ expression compared with the non-Tfh and Tfr T cell subsets (Fig. 4d). To investigate whether the B cell-intrinsic deficiency in Gal-3 expression affected IFN-γ production, we used the chimeric mice described in Fig. 3d, in which only B cells lacked Gal-3 signaling.  Fig. 3d (n = 7 for each group). g Frequency of splenic Ki-67 + B220 + B cells (left) and Ki-67 + CD4 + T cells (right) from mixed BM chimeric mice shown in Fig. 3d (n = 7 for each group). h Schematic of mixed BM chimeric mice containing a 70:30 ratio of CD45.1 Gal-3 +/+ WT:CD45.2 Gal-3 KO BM cells. i, j Representative plots and statistical analysis of the frequency of splenic GC B cells (i) and Tfh cells (j) from mixed BM chimeric mice shown in Fig. 3h (n = 9). k Frequency of splenic Ki-67 + B220 + B cells (left) and CD4 + T cells (right) from mixed BM chimeric mice shown in Fig. 3h (n = 9). Data are the mean ± s.e.m. *p < 0.05, **p < 0.005, ***p < 0.001 (unpaired (a-h) and paired (i-k) Student's t-test). Data are representative of three (a-c) or two (d-k) independent experiments Selective Gal-3 deficiency in B cells increased the frequency of IFN-γ-producing CD4 + T cells (Fig. 4e), thus indicating that Gal-3 signaling in B cells is necessary to regulate IFN-γ production by CD4 + T cells. Next, we decided to evaluate IFN-γ production in those chimeric mice described in Fig. 3h sharing Gal-3-sufficient cells and Gal-3-deficient cells. We observed an increased frequency of IFN-γ-producing CD4 + T cells among Gal-3 KO CD45.2 + cells in comparison to Gal-3-sufficient WT CD45.1 + cells (Fig. 4f). To analyse how IFN-γ influence B cell responses in Gal-3 KO mice, we first compared the expression of IFN-γR on B and T cells from WT and Gal-3 KO mice. We observed that B cells exhibited higher expression of INF-γR in comparison to T cells and that INF-γR expression was higher on the cells from Gal-3 KO mice with respect to WT. Interestingly, GC B cells (B220 + CD4 − FAS + GL7 + ) had higher expression of INF-γR than non-GC B cells (B220 + CD4 − FAS − GL7 − ; Fig. 4g). Considering that IFN-γ plays a role in T-bet induction in B cells and is crucial for IgG2c class switching [37][38][39] , we also examined Tbet expression in lymphocytes from WT and Gal-3 KO mice. Regarding INF-γR, lymphocytes from Gal-3 KO mice expressed higher levels of T-bet than their counterparts from WT mice (Fig. 4h, i). We found that GC B cells expressed the highest levels 25 1.  of T-bet with respect to other populations indicated (Fig. 4h) and that T-bet expression was significantly higher (p < 0.001 by unpaired Student's t-test) in Tfh than in non-Tfh cells (Fig. 4i).
Absence of Gal-3 favors autoimmunity. Spontaneous generation of GC is linked to autoimmunity 5 , and IFN-γ has also been associated with spontaneous GC formation and lupus development 40 . Based on our results, it is conceivable that Gal-3 KO mice may develop autoimmune manifestations. Considering that progression of autoimmune diseases is age-related and that secretion of autoantibodies precedes the development of clinical manifestations of lupus 41 , we evaluated the presence of antinuclear antibodies (ANA) and tissue damage in 8-month-old Gal-3 KO mice. We observed that Gal-3 mice exhibited higher antibody titres (median 1/640) than WT mice (median 1/40) (Fig. 5a). The most frequent patterns of positivity were as follows: nucleolar, nuclear homogeneous and nuclear fine speckled (Fig. 5c), which are frequently associated with SLE. To confirm the findings of the ANA positivity measured with the HEp-2 assay and to measure isotype-and subclass-specific serum ANA titres, we performed ELISA and immunofluorescence with Chritidia luciliae slides for specific nuclear antigens. Gal-3 KO mice had significantly elevated (p < 0.05, p < 0.01, and p < 0.001 by unpaired Student's ttest) levels of serum IgM, IgG2c, and IgG3 specific for dsDNA, Sm/RNP and histone, hallmarks of SLE, in comparison to WT mice ( Fig. 5b-d). No significant differences (p > 0.05 by unpaired Student's t-test) were observed in specific IgG1 autoantibodies (Fig. 5d) in serum samples. We also performed a double blinded score study to evaluate kidney damage in Gal-3 KO compared with WT mice. Kidneys from 8-month-old Gal-3 KO mice exhibited an expanded double layer around the glomeruli with thicker tubules, glomerular hypercellularity and hypercellularity in the tubulointerstitium compared with age-matched control mice (Fig. 5e). To characterise the type of infiltrate in the tubulointerstitium, we performed immunofluorescence of kidney slides. We found well-organised lymphoid-like structures wherein B and T cells were intimately interacting with each other (Fig. 5f). Finally, we measured urea and creatinine and 24-h urine proteins, which are important clinical features of renal function. We observed higher concentrations of creatinine and urine proteins in Gal-3 KO compared with WT mice, suggesting a loss of renal function (Fig. 5g, h).
Considering that the emergence of GCs could be agedependent, we examined the time-course of these structures in both groups of mice. GC B cells were barely detectable in spleen of WT and Gal-3 KO WT mice at weaning (3 weeks old) ( Supplementary Fig. 6a), but this cell subset expanded with age and was significantly increased (p < 0.05 and p < 0.01 by unpaired Student's t-test, respectively) in spleen of young adult (8-week-old) and aged (8-month-old) Gal-3 KO mice. Remarkably, and accompanying the pronounced proliferation of the GC B cell population, Tfh cells were also increased in 8-month-old Gal-3 KO mice in comparison to their counterparts in WT mice ( Supplementary Fig. 6a).
Compatible with the marked increase in GC B cells, a significantly higher (p < 0.05 by unpaired Student's t-test) percentage of B cells from 8-month-old Gal-3 KO mice expressed Bcl-6 ( Supplementary Fig. 6b) and exhibited a centroblast phenotype characterised by high expression of CXCR4 42 (Supplementary Fig. 6d). Aged Gal-3 KO mice also had an increased frequency of B cells with the plasmablast/PC phenotype (B220 + CD4 − Blimp-1 + ) in comparison to WT mice (Supplementary Fig. 6c). Remarkably, B cells from 8-month-old Gal-3 KO presented a higher percentage of CD80 + ( Supplementary  Fig. 6e) and CD86 + (Supplementary Fig. 6f) cells than their WT counterparts.
In agreement with our previous results, aged Gal-3 KO mice presented abnormally high levels of serum IFN-γ (Fig. 5i). Considering that Gal-3 negatively regulates Th17 polarisation 43 and that Th17 cells may be responsible for aberrant selection of self-reactive GC B cells and autoantibody formation in BXD2 mice 44 , we also evaluated IL-17 production in Gal-3 KO mice. No increased levels of IL-17A were detected in sera from aged Gal-3 KO mice (Fig. 5i).
Blockade of IFN-γ signaling prevents GC reactions and lupus. We next investigated whether disease can be ameliorated by blocking IFN-γ. To achieve this goal, 4-week-old female Gal-3 KO mice, which have not yet developed GCs, were treated with 500 µg monoclonal antibody (mAb) anti-IFN-γ every 3 days for 3 weeks. At the end of this treatment, anti-IFN-γ-treated mice had reduced numbers of GC B cells and Tfh cells than mice treated with isotype control (Fig. 6a, b, respectively). IFN-γblocked mice exhibited a total absence of PNA + cells inside the follicle (Fig. 6c); as expected, untreated Gal-3 KO mice presented PNA + GC B cells (Fig. 6c). Interestingly, the follicular B cells of IFN-γ-blocked Gal-3 KO mice expressed IgD, indicating that, in the absence of IFN-γ activity, B cells from Gal-3 KO mice did not experience class-switch recombination (CSR) (Fig. 6d). In addition, IFN-γ blockade in Gal-3 KO mice significantly decreased the frequency of IFN-γ-producing B and T cells (Fig. 6e, f, respectively) (p < 0.05 and p < 0.05 by unpaired Student's t-test, respectively) and the frequency of IFN-γR-expressing non-GC B220 + B cells (Fig. 6g) compared to mice treated with an isotype control. Finally, we determined that anti-IFN-γ-treated Gal-3 mice had undetectable ANA (Fig. 6h) and presented diminished mononuclear cell infiltrates in the kidney (Fig. 6i) in comparison to untreated Gal-3 KO mice. Together, the results demonstrate Fig. 4 Gal-3 KO mice exhibited augmented IFN-γ production. a Quantification of IFN-γ, IL-21, IL-4, IL-6, and IL-2 by ELISA in the supernatants of purified CD4 + T cells from the spleen of WT and Gal-3 KO mice cultured with anti-CD3 and anti-CD28 for 72 h (n = 4 for each condition). b Representative plots and statistical analysis of the frequency of splenic IFN-γ-producing CD4 + T cells of WT and Gal-3 KO mice (n = 5 for both strains). c Representative plots and statistical analysis of the frequency of splenic IFN-γ-producing B220 + B cells of WT and Gal-3 KO mice (n = 5 for both strains). d Representative plots and statistical analysis of the frequency of splenic IFN-γ-producing non-Tfh, Tfh, and Tfr cells of Gal-3 KO mice (n = 5). e Representative plots and statistical analysis of the percentage of splenic IFN-γ-producing CD4 + T cells in the mixed BM chimeras shown in Fig. 3d containing an 80:20 ratio of μMT: Gal-3 +/+ WT or μMT:Gal-3 KO BM at 10 weeks after transplantation (μMT+Gal-3 +/+ n = 7; μMT+Gal-3 −/− B cells n = 7). f Representative plots and statistical analysis of the percentage of splenic IFN-γ-producing CD4 + T cells in the mixed BM chimeras shown in Fig. 3h  . c Immunofluorescence of frozen spleen sections (7 μm) from Ctrl and αIFN-γ mice stained with anti-CD4 (green), anti-B220 (red) and PNA (blue). Magnification: ×100 (top) and ×400 (bottom) (n = 5 for each group). Scale bar, 200 μm (upper panels) and 50 μm (lower panels). d Immunofluorescence of frozen spleen sections (7 μm) from Ctrl and αIFN-γ mice stained with PNA (green), anti-IgD (red) and anti-B220 (blue). Magnification: 4×00 (n = 5 for each group). Scale bar, 50 μm. e Representative plots and statistical analysis of the frequency of IFN-γ-producing B220 + cells in the spleen of Ctrl and αIFN-γ mice (n = 5 for each group). f Representative plots and statistical analysis of the frequency of IFN-γ-producing CD4 + T cells in the spleen of Ctrl and αIFN-γ mice (n = 5 for each group). g Representative plots and statistical analysis of the frequency of B220 + GL7 − FAS − non-GC IFN-γR + B cells in the spleen of Ctrl and αIFN-γ mice (n = 5 for each group). h ANA IgG determined with Hep-2 slides in sera from Ctrl and αIFN-γ mice. Magnification: ×100 (top) and x400 (bottom) (n = 5 for each group). Scale bar, 200 μm (upper panels) and 50 μm (lower panels). i Kidney sections from Ctrl and αIFN-γ mice stained with H&E. Magnification ×100 (top) and ×400 (bottom) (n = 5 for each group). Scale bar, 100 μm (upper panels) and 20 μm (lower panels). Ig-Control Gal-3 KO mice (Ctrl) are indicated in blue and αIFN-γ-treated Gal-3 KO mice (αIFN-γ) in violet. Data are the mean ± s.e.m. *p < 0.05 (unpaired Student's t-test). a-i Data are representative of two independent experiments that in Gal-3 KO mice, IFN-γ overproduction leads to (induces and is required to sustain) a pathogenic and lupus-inducing GC response.

Discussion
Our findings provide the first evidence for the involvement of Gal-3 as a key molecule in the GC reaction. Previously, we and others have found that Gal-3 regulates B cell niches in lymph node tissues and inhibits PC differentiation in vitro 17 and in vivo 45 . Our studies revealed that Gal-3 is expressed by resting B cells, as previously reported 17 , and is downregulated after activation with a T-independent (LPS) or T-dependent (CD40L+IL-4) stimulus that favor B cell differentiation into PCs, suggesting that the downregulation of Gal-3 may be necessary to favor the ASC programme. Accordingly, we demonstrated that mice lacking Gal-3 have hypergammaglobulinemia and an increased frequency of ASCs in the spleen and BM. We found increased levels of IgG2c and IgG3 in Gal-3 KO mice. These Ig subclasses efficiently activate stimulatory FcγRs and complement, and they display the highest pathogenic activity of the Igs 46 . It is well known that the productions of these Ig subclasses is promoted by IFN-γ 47 and that they are closely involved in the development and progression of autoantibody-mediated autoimmune diseases, such as lupus. Here we show that, without purposeful immunization or infection, Gal-3 KO mice exhibit spontaneous GC generation in the spleen. The increase in number of GC B cells was time-dependent since GC structures appeared after 4 weeks of birth and increased with age. The increase in B cells with GC phenotype was accompanied by an expansion of Tfh cells. Additionally, the Tfh:Tfr ratio was significantly higher in these mice compared with their WT counterparts. It has been described that the balance between Treg and effector T cell is critical for proper control of the quality and magnitude of adaptive immune responses and for establishing or breaching tolerance to self-and non-self-antigens 48,49 . Therefore, the increase in the ratio of Tfh to Tfr cells in Gal-3 KO mice could have an impact on promoting the development of GCs in Gal-3 KO mice, resulting in uncontrolled GC reactions and subsequent development of an autoimmune disease 50,51 .
To better understand the relationship between Gal-3 and the B cell differentiation programme, we compared gene expression levels between B cells from WT or Gal-3 KO mice. Gal-3 KO mice showed increased expression of genes related to B cell activation and GC-formation, such as Mki67, Aicda, Lilrb4, Itga1, Ig heavy chain-encoding genes and others. Interestingly, we found increased transcript levels of miRs, which coordinate GC formation, CSR/SHM, PC differentiation, and memory B cell differentiation, such as miR-155 31 and miR-217 33 . These short non-coding RNAs would individually target multiple factors in addition to regulating AID, Bcl-6 and Blimp-1. Likewise, we found quite a few miRs with as yet unidentified functions, which were differentially expressed in B cells from Gal-3 KO and WT mice and could influence GC reaction development. Given the ability of Gal-3 to bind to different molecules, such as RNA 52 ; adaptor proteins involved in protein transport, such as Alix 53 ; transcription factors, such as Bcl-2 54 ; and components of the nuclear ribonucleoprotein complex 55 , it is likely that Gal-3 could act a gene regulator, either directly or indirectly.
Analysis of molecules that have a critical role in regulating the GC reaction revealed that Gal-3 KO mice had higher percentages of CD80 + and CD86 + B cells, and after a TLR-stimulated B cells from Gal-3 KO mice had increased expression of CD80 in comparison to B cells from WT mice. It has been reported that B cell expression of CD80 plays a critical role in both early and late GC responses 56 . Moreover, high expression of CD80 and CD86 molecules is evident on freshly isolated B cells in patients with SLE 57 . In our model, the increased percentage of CD80 and CD86 B cells in Gal-3-deficient mice possibly facilitated Tfh induction, which in turn could allow the GC response. This hypothesis favors the induction of B cell activation and autoantibody production. It is possible that the increase in expression of CD80 and CD86 resulted from the increased concentrations of IFN-γ detected in Gal-3 KO mice rather than a direct effect of Gal-3 in regulating the expression of these proteins, as it has been reported that IFN-γ increases the expression of these molecules on B cells 58 .
A question that arises is whether spontaneous GCs development was induced by the lack of Gal-3 expression specifically in B cells or if B and T cells are co-protagonists of the exacerbated GC reaction. Adoptive transfer of naive B cells into B cell-deficient μMT mice or WT mice showed that, in any case, serum levels of IgG2c and IgG3 were increased only in mice transferred with Gal-3-deficient B cells. The data suggest that B cells deficient in Gal-3 can differentiate and secrete Igs independently of a Gal-3sufficient environment. In addition, co-cultures of naive B and T cell from WT and Gal-3 KO mice showed that stimulated Gal-3deficient B cells produced higher levels of Igs in comparison to Gal-3-sufficient B cells and that the addition of Gal-3-deficient T cells did not increase that production. Using chimeric mice, we were able to show that the lack of Gal-3 in B cells is, by itself, sufficient to trigger disruption of immunological tolerance and to favor the generation of spontaneous GCs. By contrast, evidence provided by other groups has shown that the absence of Gal-3 would also regulate certain functions of CD4 + T cells. Huan-Yuan Chen et al. found that Gal-3 is an inhibitory regulator of Tcell activation and functions intracellularly by promoting TCR downregulation 53 . In our work, we also observed that CD4 + cells could produce more IFN-γ (probably by signals provided by Gal-3-deficient B cells), and we hypothesise that, although not responsible for triggering spontaneous GC formation, they contribute to maintaining and propagating the autoimmune condition mediated by IFN-γ.
Spontaneous GC B cells and Tfh cells play a crucial role in generating high-affinity pathogenic autoantibodies in many autoimmune diseases 7,10,59-62 . Accordingly, in Gal-3 KO mice, we found ANAs (including anti-dsDNA, anti-Histone and anti-SM/RNP) and kidney disease, both of which are cardinal features of systemic autoimmune diseases, such as lupus. Interestingly, Gal-3 KO mice also presented well-organised lymphoid-like structures wherein B and T cells intimately interacted within each other. This observation suggests that these T:B aggregates in the kidneys promote the local secretion of pathogenic autoantibodies that contribute to the lupus-like phenomena. Thus, Gal-3 KO mice exhibited higher concentrations of creatinine in sera and 24h urine proteins compared with WT mice, suggesting a loss of renal function.
Interestingly, according to the serum Ig isotype profile, we observed high levels of serum IFN-γ and increased numbers of IFN-γ-producing cells in Gal-3 KO mice. Increased IFN-γ expression alone can result in the development of systemic autoimmunity, as documented by the finding that epidermal transgenic expression of IFN-γ leads to anti-dsDNA and antihistone autoantibodies and glomerulonephritis 63 . The ability of IFN-γ to promote IgG class switching to more pathogenic autoantibodies (IgG2c and IgG3 in mice) and activation of IgG Fc receptors and complement could also contribute to disease severity 64 . Mutation of ROQUIN (Rc3h1) in Roquin san/san mice, another model of lupus, leads to reduced decay of IFN-γ mRNA, resulting in increased IFN-γ signaling and accumulation of Tfh cells, culminating in an increase in GC B cells and autoantibodies 7 . Notably, IFN-γR plays a crucial role in the development of autoimmune conditions, and it was clearly demonstrated recently in several mouse models 7,65,66 . Interestingly, in Gal-3 KO mice, B cells as well as T cells expressed higher levels of IFN-γR in comparison to WT, suggesting that the absence of Gal-3 conditioned both lymphocyte populations toward a better IFN-γ signaling response.
A relationship has been observed between the lack of Gal-3 and increased levels of IFN-γ. Higher serum levels of IFN-γ have been shown in melanoma-bearing Gal-3 KO mice 67 , and CD4 + T cells incubated with Gal-3 knockdown DCs produced more IFN-γ than control cells 68 . Recently, Tseng and collaborators showed that Gal-3 may decrease IFN-γ signaling by facilitating AKT/ GSK-3b/SHP2 signaling 69 .
Additionally, a possible link has been proposed between commensal microbiota and autoimmunity 70 . We observed that B cells from Gal-3 KO mice were more sensitive to LPS (Supplementary Fig. 6g) than B cells from WT mice. These results, together with the observation that GCs appear in Gal-3 KO mice after the suckling period, lead us to hypothesize that the microbial flora could be a stimulus that induces B cells to differentiate into GC B cells. However, when Gal-3 KO mice were treated with a mixture of antibiotics to deplete the murine intestinal microbiota, no changes were observed in the frequency of either splenic GC B cells or Tfh in comparison to untreated mice ( Supplementary  Fig. 7a, b). Likewise, co-housed experiments between mutant Gal-3 KO and WT mice revealed no induction of GCs in WT mice nor prevention of spontaneous GCs in Gal-3 KO mice (Supplementary Fig. 7b), suggesting that ligands from the gut microbiota are not involved in the induction of the self-reactive GCs in Gal-3 KO mice.
Herein, we report complete dependence of the presence of selfreactive GCs on IFN-γ in Gal-3 KO mice. IFN-γ blockade limited the induction of spontaneous GC responses, reducing GC B cell and Tfh cell frequencies and Ig class switching as well as preventing the appearance of autoantibodies and kidney infiltrate; these findings demonstrated that IFN-γ overproduction was required to sustain lupus-like disease. In conclusion, we revealed a new undefined role of Gal-3 expression in controlling GC formation and, consequently, immune tolerance. Our findings demonstrate that the absence of Gal-3 in B cells intrinsically favors GC formation and highlight the potential for therapeutic targeting of this pathway in autoimmunity. μMT mice (B6.129S2-Ighmtm1Cgn/J; JAX:002288) were obtained from The Jackson Laboratories (USA). All animals were bred and housed under barrier, temperature and humidity-controlled conditions at the Animal Facility of CIBICI-CONICET, Facultad de Ciencias Químicas, Universidad Nacional de Córdoba. Animals were euthanized with carbon dioxide (CO 2 ) followed by cervical dislocation.

Methods
Cell culture. Spleens were obtained and homogenised through a tissue strainer. Bone marrow (BM) cells were isolated by flushing femurs and tibias of mice with RPMI 1640. Cells were resuspended in a Tris-ammonium chloride buffer for 5 min to lyse red blood cells.
The specific cell population and gating strategy in each case are described in Supplementary Fig. 8.
Urea and creatinine determination. Urea and creatinine concentrations were determined on a Dimension RxL Max system (enzymatic/colorimetric method and Jaffé reaction, respectively; Siemens Healthcare Diagnostics), following the manufacturer's instructions.
Proteinuria. Urine samples were collected 24 h by placing mice in custom-made mouse metabolic cages. Urine total protein was determined on a Dimension RxL Max system (colorimetric method; Siemens Healthcare Diagnostics), following manufacturer's instructions.
DNA microarrays. For gene-expression analysis, purified B cells from 8-week-old WT and Gal-3 KO mice were lysed with TRIzol reagent. Total RNA was extracted with the RNAeasy Mini Kit (Qiagen). RNA quality was verified in an Agilent Bioanalyser and measured with a Nanodrop 1000 (Thermo Scientific). Microarray experiments were performed on GeneChip Mouse Gene 2.1 ST arrays (Affymetrix). The working lists were created by filtering probes with detection P values <0.05 for all chips and discarding overlapping probes. Each dataset was derived from two biologically independent replicate samples. Independent samples were compared by computing fold ratios and filtered at a 1.2-fold threshold for Venn diagrams and pathway analysis. For pathway analysis, GenBank accession numbers were mapped to the Ingenuity database (IPA, http://www.ingenuity.com) to retrieve relevant biological processes. Heat maps were generated using the web interface programme Matrix2png.
Statistical analysis. All data are expressed as mean ± s.e.m. Replicates were biological replicates and n represents the number of individual mice analysed per experiment. Mice were assigned to groups at random for all mouse studies and, where possible, mixed among cages. Careful consideration has been given to calculate the number of mice needed to give statistically significant results, while using the minimum number of mice possible. Numbers of mice for each experiment were decided after consultation with a statistician at UNC. Experiments were performed independently in triplicate to control for experimental variation. No exclusion was performed. Statistical comparisons were determined using unpaired or paired twotailed Student's t-test or one-way analysis of variance (ANOVA) followed by Tukey-Kramer multiple comparison test when appropriate. Statistical analyses were performed using GraphPad Prism software (GraphPad). Data followed Gaussian distribution with similar variance between groups that were compared. Differences were considered statistically significant when the P values <0.05.
Data availability. The microarray data have been deposited in the ArrayExpress database under accession number E-MTAB-6053. The authors declare that the data supporting the findings of this study are available within the article and