Female BALB/c and pre-diabetic NOD mice aged 12-20 weeks or 4-5 weeks (when indicated) were obtained and maintained under specific-pathogen-free conditions at the animal facility of the Oswaldo Cruz Foundation (Rio de Janeiro, Brazil). BALB/c mice was used as a control group of the present study, because gal-3 expression was first standardized in cortical and medullary regions from thymus of this strain (19). In addition, BALB/c mice is considered a suitable control for study of thymic abnormalities of NOD mice (37). Unless stated specifically, at least 3 animals from each group were used in each experiment described below. All protocols for the use and care of animals were approved by the corresponding Ethics Committee for the Use of Experimental Animals (Oswaldo Cruz Institute, Rio de Janeiro), under license numbers L-024/09 and L-004/2014.

Evaluation of onset of diabetes in NOD animals was defined by weekly-based determination of glycemia, using the Prestige Smart System^® blood glucose test strips (Home Diagnostics, Fort Lauderdale, USA) following the manufacturer’s instructions. Mice with glycemia lower than 250 mg/dL were classified as pre-diabetic or non-diabetic (41).

The following primary antibodies were used: rabbit anti-gal-3 (sc-20157, Santa Cruz, Dallas, USA), anti-fibronectin (24951), anti-laminin (24851) from Novotec (Saint Martin-La-Garenne, France), anti-cytokeratin (Z0622, Dako, Carpinteria, USA), rat anti-CD4 (553647), anti-CD8 (553027), anti-CD19 (5502840) purchased from BD Pharmingen™ (San Jose, USA), anti-FoxP3 (14-5773-82, eBioscience, USA), F4/80 monoclonal antibody (ab6640, Abcam, Cambridge, USA), and biotinylated anti-CD11c (117303, Biolegend, San Diego, USA). Alexa Fluor 488-labeled goat anti-rat IgG (A11006), Alexa Fluor 633-labeled goat anti-rabbit IgG (A21070), and Alexa Fluor 488-labeled streptavidin (S11223), purchased from Molecular Probes (Carlsbad, USA), were used as second-step reagents. In triple staining with rabbit-derived primary antibodies, Zenon Alexa Fluor 488 (Z-25302) and Zenon Alexa Fluor 555 (Z-25305) rabbit IgG labeling reagents, both from Molecular Probes (Carlsbad, USA), were applied. Tissue-Tek O.C.T. (4583) was obtained from Sakura, Torrance, USA. Poly-L-lysine (P8920) and the aqueous mounting medium Fluoroshield (F6182) were purchased from Sigma-Aldrich (Saint Louis, USA).

BALB/c and NOD mouse thymuses (n=3/group) were removed, individually embedded in Tissue-Tek O.C.T., and frozen at -70°C. Five μm-thick cryostat sections were settled on poly-L-lysine-covered glass slides, fixed in ice-cold acetone for 10 min, and air-dried. Slides were washed with PBS (P4417, Sigma-Aldrich, Saint Louis, USA), followed by blocking with PBS-BSA 1% for 20 min. For CD4-gal-3-cytokeratin, CD8-gal-3-cytokeratin, CD19-gal-3-cytokeratin, F4/80-gal-3-cytokeratin triple staining, slides were incubated with rat primary antibody for 1h at room temperature (RT), washed with PBS three times for 5 min, incubated with Alexa Fluor 488-labeled goat anti-rat secondary antibody for 45 min at RT, washed with PBS three times, incubated with anti-gal-3 diluted in PBS-BSA 1%-saponin (S7900-25G, Sigma-Aldrich, Saint Louis, USA) 0.1% buffer overnight at 4°C, washed with PBS three times, incubated with Alexa Fluor 633-labeled goat anti-rabbit secondary antibody diluted in PBS-BSA 1%-saponin 0.1% buffer for 45 min at RT, washed with PBS three times, incubated with anti-cytokeratin antibody conjugated with Zenon Alexa Fluor 555 rabbit Ig labeling reagent for 2h and washed with PBS three times.

For CD11c-gal-3-cytokeratin staining, slides were first incubated with the biotinylated anti-CD11c antibody for 1h, washed with PBS three times, incubated with Alexa Fluor 488-labeled streptavidin for 45 min, washed with PBS three times and stained for gal-3 and cytokeratin. For FoxP3-gal-3-cytokeratin staining, slides were first incubated with anti-FoxP3 antibody diluted in PBS-BSA 1%-saponin 0.1% buffer for 1h, washed with PBS three times, incubated with Alexa Fluor 488-labeled labeled goat anti-rat secondary antibody diluted in PBS-BSA 1%-saponin 0.1% buffer for 45 min at RT, washed with PBS three times and stained for gal-3 and cytokeratin, as above described. For gal-3-fibronectin-cytokeratin and gal-3-laminin-cytokeratin staining, slides were stained for gal-3, incubated with anti-fibronectin or anti-laminin antibody conjugated with Zenon Alexa Fluor 488 rabbit Ig, washed, and incubated with anti-cytokeratin antibody conjugated with Zenon Alexa Fluor 555 rabbit Ig labeling reagent. All slides were ultimately mounted with Fluoroshield using coverslips. Negative controls were performed with unrelated IgG isotypes or incubating the cells with the secondary antibody alone. None controls yielded significant fluorescence signals.

The murine endothelioma cell line tEnd.1, kindly provided by Dr. Thereza Christina Barja-Fidalgo (University of the State of Rio de Janeiro, Brazil), was cultured in RPMI 1640 medium (R6504, Sigma-Aldrich, Saint Louis, USA) supplemented with 10% fetal calf serum (FCS 12657029, Invitrogen Life Technology, Waltham, USA), 2mM glutamine (G8540), 100U/mL streptomycin (S9137) (both from Sigma-Aldrich, Saint Louis, USA) in 8 well Permanox Lab-Tek^® chamber slides (177445, Thermo Scientific Nunc., Rochester, NY, USA) at 37°C in a CO₂ humidified atmosphere. For gal-3 staining, cultures were fixed in methanol for 7 min, followed by blocking with PBS-BSA 1%, incubated with anti-gal-3 primary antibody for 1 h at RT, washed with PBS three times, incubated with Alexa Fluor 488-labeled goat anti-rabbit secondary antibody for 45 min at RT, washed with PBS three times. The slides were mounted with Fluoroshield using coverslips.

Gal-3 expression was evaluated by immunohistochemistry on paraffin sections of thymus from BALB/c and pre-diabetic NOD mouse thymuses aged 4-5 weeks and 16 weeks. Initially, the sections were deparaffinized in three xylol baths for 3 min, hydrated in decreasing concentrations of alcohol for 3 min, washed in running water for 5 min and rinsed in distilled water. For antigen retrieval, sections were incubated with 3M urea (U5378-100G, Sigma-Aldrich, Saint Louis, USA) for 15 min. Next, the sections were washed with three baths of PBS-0.1% Tween 20 (P1379-25ML, Sigma-Aldrich, Saint Louis, USA) for 5 min. To block endogenous peroxidase, the sections were incubated with 3% H2O2 (386790-100ML, Sigma-Aldrich, Saint Louis, USA) diluted in 70% methanol for 5 min. After blocking, the sections were washed again with three baths of PBS-0.1% Tween 20 for 5 min. To block nonspecific binding, sections were incubated with 10% horse serum diluted in 2% PBS-BSA. After this incubation, the sections were incubated with anti-gal-3 for 30 min or overnight at 4°C in a humid chamber. After this incubation, the sections were washed in three baths of PBS-0.1% Tween 20 for 5 min. Next, the sections were incubated with anti-mouse secondary antibody conjugated to peroxidase for 1 h. After three consecutive washes with PBS-0.1% Tween 20, the reaction was revealed using the peroxidase substrate (H2O2) and 3,3-diaminobenzidine (DAB) (K3467, Dako, Glostrup, Denmark). After washing in running distilled water, the sections were counterstained with Harris Hematoxylin for 30 seconds, dehydrated in increasing concentrations of alcohol and clarified in xylene. The slides were mounted in Entellan ^® new (107961, Merck, Darmstadt, Germany). For quantitative analysis of gal-3 positive cells per area, ten random images of each animal’s thymus containing cortical or medullary regions were acquired under a 20x objective, using the Nikon Eclipse E600 microscope (Nikon Instruments Inc., Melville, NY, USA) with the DXM 1200F digital camera system (Nikon Instruments Inc., Melville, NY, USA). The occupied area was determined by a grid of dots and cells positive for gal-3 were counted using the Image J program (NIH). The result was represented by the number of gal-3-positive cells divided by the area occupied.

Immunostained tissues were analyzed by confocal microscopy using a Leica TCS SP5 device (Leica Microsystems CMS GmbH, Mannheim, Germany) from the Confocal Microscopy Laboratory Station (Biology Institute, Fluminense Federal University, Niteroi, Rio de Janeiro). With respect to gal-3 quantitative fluorescence and colocalization analyses, at least six photomicrographs comprising cortical and medullary regions of the thymic lobules (n=3/group) were randomly obtained with 63x objective at zoom 2. Thymic cortical and medullary regions were defined by cytokeratin staining. Image J (NIH) and LAS AF Lite (Leica Microsystems) softwares were used to quantify fluorescence and colocalization, respectively. The result of gal-3 quantification was represented by fluorescence mean which is calculated by the sum of the gray values of all pixels divided by total pixel numbers. For gal-3 colocalization with CD4, CD8, CD19, fibronectin, and laminin in giant PVS, at least six photomicrographs comprising blood vessels containing cytokeratin-negative regions of NOD mouse thymus sections were selected. Before beginning the colocalization analysis, each photomicrography was adjusted through the establishment of percentage values for threshold and background parameters. A scatter diagram called a cytofluorogram was displayed for each photomicrography. This diagram shows the distribution of gray-scale values of fluorescent signals in two detection channels. A digital image was also generated with colocalizing fluorescence signals in white to facilitate visualizing the colocalization of two given molecules. The result was represented by the colocalization rate (%) which is calculated by the ratio of the area of colocalizing fluorescence signals by the area of the image foreground; the latter indicating the difference between the total area of the image and the area of the image background.

To analyze the effect of gal-3 on thymocyte migration, we performed migration assays in 6.5 mm diameter transwell chambers bearing 5 µm pores (3421, Corning Inc., New York, USA) (19, 34). The membranes were previously treated with fibronectin from human plasma (F2006) or BSA (A7906) alone (10 μg/mL), both from Sigma-Aldrich (Saint Louis, USA), for 1h at 37°C, followed by blocking with PBS-BSA 0.5% for 45 minutes. For thymocyte isolation, thymuses from BALB/c and NOD mice were removed and mechanically macerated in a glass potter grinder with 1mL RPMI 1640. Next, thymocyte suspension were counted in a Neubauer chamber. In those membranes treated only with BSA we added 2.5x10⁶ thymocytes from both mice treated or not with recombinant mouse gal-3 (1197-GA, R&D Systems, Minneapolis, MN.) to the upper chamber (in 100 μL fugotaxis buffer containing 1% BSA in RPMI 1640 medium) and 600 μL of the same buffer in the lower chamber. After 3 hours of migration at 37°C in the CO₂-humidified atmosphere, we washed the lower compartment of the membranes with 200μL of 1% RPMI 1640-BSA, collected the cells that migrated to the lower chamber, and counted them in a Neubauer chamber (21, 37).

To mimic thymocyte export under the influence of gal-3 to the periphery, we performed a reverse transendothelial migration assay (38), in which we used the murine thymic endothelium cell line t.End.1. These cells were plated on the lower compartment of Transwell^® system membranes with 8 µm pores (3422, Corning Inc., New York, USA) for 2 hours to allow adhesion. Then the cells were cultured in 500μL of RPMI 1640 and added to the upper and lower chambers for 48 hours at 37°C in a CO₂-humidified atmosphere. After this incubation period, 2.5x10⁶ thymocytes from BALB/c or NOD mice, treated or not with recombinant gal-3, were placed in the upper chamber (in 100 μL of migration buffer) and 600 μL of the same buffer were added to the lower chamber. After overnight migration, thymocytes were collected in the lower chamber and counted in a Neubauer chamber.

Data distribution was analyzed using the D’Agostine-Pearson normality test. Statistical differences were analyzed by unpaired Student’s t-test or one-way ANOVA, followed by post hoc Tukey’s test using GraphPad Prism 5 software (GraphPad^® Software, San Diego, USA). The results were expressed as mean ± standard error (SE). The differences between the analyzed groups were considered statistically significant when p<0.05.

We initially investigated the expression of gal-3 in the thymus of pre-diabetic NOD mice. To study the in situ expression of gal-3 in TEC, we double-stained thymus sections of BALB/c ( Figures 1A–C ) and NOD ( Figures 1D–F ) for gal-3 and cytokeratin, followed by confocal microscopy analysis. Similar to what we had previously shown for BALB/c mice (19), we observed a stronger gal-3 immunostaining in medullary ( Figure 1F ) than in cortical regions ( Figure 1E ) of NOD mouse thymuses, including both epithelial and non-epithelial (cytokeratin-negative) cells. Quantitative analysis of the colocalization between gal-3 and cytokeratin in NOD mouse thymuses showed a significant enhancement of colocalization in the medulla, as compared to the cortex ( Figure 1G ). Importantly, we noticed a significant increase of colocalization in the cortex as well as in the medulla of NOD mouse thymuses in comparison to BALB/c, suggesting that NOD cortical and medullary TEC express more gal-3 than the corresponding BALB/c TEC ( Figure 1G ).

Next, we double-stained thymus sections with anti-F4/80 and anti-gal-3 antibodies ( Figure 2 ). We noticed that NOD mouse thymuses had higher numbers of F4/80⁺ macrophages ( Figure 2C ) than the corresponding BALB/c controls ( Figure 2A ). Interestingly, many BALB/c and NOD cortically located macrophages were also gal-3-positive ( Figures 2A, C ), although most medullary macrophages were negative for the lectin in the thymuses of both mouse strains ( Figures 2A, C ). Quantitative colocalization analysis of gal-3 and F4/80 revealed a statistically significant increase in the colocalization rate in the thymic cortex of BALB/c and NOD, as compared with the respective medullary regions. Yet, NOD cortical (but not medullary) macrophages presented higher gal-3 colocalization than BALB/c controls ( Figure 2E ).

CD11c-positive DCs also expressed gal-3 in both BALB/c and NOD mouse thymuses, with an increased colocalization of these two molecules in the medulla as compared to the cortex in both BALB/c ( Figure 2C ) and NOD mouse strains ( Figure 2D ). Quantitative colocalization analysis revealed that both cortical and medullary DCs in the NOD thymus, express more gal-3, when compared to respective BALB/c controls ( Figure 2F ).

Since gal-3 is secreted by thymic microenvironmental cells and interacts with ECM glycoproteins such as fibronectin and laminin, we also evaluated the colocalization of gal-3 with these molecules ( Figure 3 ). No differences in gal-3/fibronectin colocalization rates were observed between the cortex and medullary regions of BALB/c or NOD mouse thymuses ( Figure 3E ). However, a significant decrease in colocalization was observed in the NOD mouse thymus, when compared to BALB/c controls ( Figure 3E ).

Laminin was found mainly in the medulla of both BALB/c and NOD thymuses (42, 43). Like fibronectin, no differences between the thymic regions were observed in both mouse strains when the colocalization analysis of gal-3 and laminin was performed ( Figure 3F ). Conversely, no decrease was observed when the cortex and the medulla of NOD mouse thymus were compared to the respective regions of BALB/c controls ( Figure 3F ).

As previously described, one striking thymic alteration in the NOD mouse thymus is the presence of giant PVS, where mature thymocytes are progressively accumulated (35, 36, 38). To define the giant PVS, we performed immunolabeling of thymus sections with anti-cytokeratin, and we analyzed the expression of gal-3 within these structures (negative for cytokeratin). We found a heterogeneous distribution of gal-3 inside the giant PVS ( Figures 4A, B ). We observed an accumulation of fibronectin fibrils within these spaces, with gal-3 being colocalized with some of them ( Figure 4B ). A similar pattern was seen for the colocalization of gal-3 and laminin ( Supplementary Figure 1 ).

Since mature thymocytes are retained within the giant PVS, we investigated the presence of gal-3 labeling, in these cells. For that, NOD mouse thymus sections were labeled for gal-3, cytokeratin, and CD4 ( Figure 5A ), CD8 ( Figure 5B ) or FoxP3 ( Figure 5C ). CD4⁺ thymocytes intermeshed with gal-3 were found inside these structures ( Figure 5A ), as well as a punctuate gal-3 labeling around CD8⁺ and FoxP3⁺ cells within the giant PVS ( Figures 5B, C ).

Differently, we did not observe any F4/80⁺ macrophages in most PVS analyzed, except for the few gal-3-negative-macrophages located near the border of some PVS ( Figure 5D ). Conversely, we did detect gal-3 bearing DCs in giant PVS ( Figures 5E, G ).

Clusters of B cells were described in the giant PVS of NOD mouse thymus (35). Interestingly, we observed a clear colocalization of gal-3 with CD19⁺ cells in intra-PVS clusters of B cells ( Figures 5F, H ). Colocalization rates of gal-3 with CD4, CD8, CD19 or CD11c in the giant PVS were compared ( Figure 5I ). CD11c and CD19 colocalization with gal-3 was clearly seen ( Figure 5I ). Although no colocalization between FoxP3 (nuclear staining) and gal-3 was observed, FoxP3⁺ cells are positioned intermingled in a gal-3-rich environment ( Supplementary Figure 2 ).

Together, these data suggest a close relationship of leukocytes with gal-3 within the giant PVS observed in pre-diabetic NOD mouse thymuses.

Considering our previous data showing that gal-3 is capable of interfering with the interaction of thymocytes with thymic microenvironmental cells modulating their migration and adhesive properties (19), we evaluated the possible influence of gal-3 on NOD thymocyte migration. Firstly, we confirmed that thymocytes from NOD mice have a significant decrease in the migratory response towards fibronectin when compared to control BALB/c derived thymocytes. Importantly, a decrease in the migratory response of NOD thymocytes was also seen in response to gal-3 itself ( Figure 6A ).

Given the previous data showing that mature thymocytes from NOD mice are accumulated within giant PVS that occupy a large part of the medullary thymic region (35), and considering that NOD thymocytes have a deficient migratory response in response to gal-3, we decided to evaluate the role of this lectin in thymocyte egress using a reverse transendothelial cell migration assay. We initially assessed gal-3 expression in the mouse thymic endothelial line tEnd.1, demonstrating that these cells produce gal-3 ( Supplementary Figure 3 ). To carry out the transendothelial migration assay, we cultured the thymic endothelial cells in a transwell membrane in order to form a monolayer which the thymocytes would be able to cross, mimicking their exit through the blood-thymus barrier, as previously described (38). We observed that NOD thymocytes exhibited deficient reverse transendothelial migration compared to BALB/c thymocytes ( Figure 6B ). Although preliminary, our data suggest that gal-3 is involved in the migration disturbs observed in the thymus of pre-diabetic NOD mice.