Dextran Sulfate Protects Pancreatic β-Cells, Reduces Autoimmunity, and Ameliorates Type 1 Diabetes

In autoimmune type 1 diabetes (T1D), self-tolerance is lost, leading to β-cell destruction (1,2). In T1D in humans and in the NOD mouse, a spontaneous mouse model of T1D, autoreactive T cells target islet-associated antigens and acquire an effector inflammatory phenotype due to costimulatory signals (1–3). In the islet, effector T cells express high levels of interferon-γ (IFN-γ), perforin, and granzyme, leading to tissue invasion and β-cell destruction (1–3). In T1D, effector responses are prevalent over tolerogenic responses, and therapies targeting both blockade of early T-cell activation and enhancement of regulatory T cells (Tregs) are being pursued (4,5). Therapies that at the same time target both the modulation of T-cell activation and the protection of β-cells against known culprits of β-cell destruction in T1D, such as proinflammatory cytokines and endoplasmic reticulum (ER) stress (1–8), could provide an alternative approach for treating the disease.

Low-molecular-weight dextran sulfate (DS) (6,500–10,000 Da) is a sulfated semisynthetic polysaccharide with cytoprotective actions as well as immunomodulatory properties (9–14). DS modulates complement pathways and the coagulation cascade and inhibits the functional maturation of human dendritic cells (DCs) in vitro (9,11,13,14). In addition, DS protects endothelial cells against complement- and NK cell–mediated cytotoxicity in vitro (9,10). DS also protects the vasculature from ischemia/reperfusion injury and reduces the instant blood-mediated inflammatory reaction and the early islet graft loss after intraportal xenotransplantation (10–12). Taken together, these studies suggest that DS might attenuate the proinflammatory effects of immune mediators and exert cytoprotective actions in islets against adverse conditions as in T1D.

Cell surface molecules such as cytotoxic T cell–associated antigen-4 and programmed death-1 (PD-1) are involved in the control of immune tolerance (15,16). PD-1 engagement on activated T cells decreases their proliferation and IFN-γ production (17). PD-1 interacts with programmed death-ligand 1 (PD-L1) that is widely expressed in leukocytes and islets and disruption of PD-1/PD-L1 interaction results in lower T-cell mobility, enhanced T-cell–DC contacts, and accelerated autoimmune diabetes in NOD mice (18–21). Thus, this interaction is critical for limiting T-cell action in the islet and for maintaining peripheral tolerance in a setting predisposed to autoimmunity.

In this study, we analyzed the therapeutic potential of DS for T1D by examining six different parameters: β-cell death, β-cell function, islet transcriptome, islet heparan sulfate (HS)/HS proteoglycan (HSPG) expression, immune regulation, and prevention and reversal of T1D in NOD mice. We found that DS 1) effectively protected β-cells against ER stress and proinflammatory cytokines; 2) improved islet bioenergetics and insulin secretion in proinflammatory conditions; 3) reduced islet chemokine expression; 4) increased islet HS/HSPG expression; 5) increased the number of PD-1⁺CD4⁺, PD-1⁺CD8⁺, and Tregs and diminished the number of IFN-γ⁺CD4⁺ and IFN-γ⁺CD8⁺ cells in vivo; and 6) halted diabetes progression and reversed autoimmune diabetes in NOD mice. These results suggest that DS has therapeutic potential for the treatment of T1D.

Low-molecular-weight DS sodium salt (6.5–10 kDa), HS, and dextran (Sigma-Aldrich), dissolved in saline and sterilized through 0.22-µm filters, were used in these studies. Polysaccharides were added simultaneously to the rest of treatments in vitro. Antibodies are described in Supplementary Table 1.

Human islets from 10 donors (aged 46 ± 4 years; islet purity 90%; and 80% men) (Supplementary Table 2) were obtained through Prodo Laboratories, Inc. Islets from 8-week-old male C57BL/6 mice were isolated as described elsewhere (22). All animal studies were performed in compliance with, and with the approval of, the Icahn School of Medicine at Mount Sinai Institutional Animal Care and Use Committee.

Analysis of β-cell death in vitro was performed in dispersed islet cells as previously described (22). Following treatment with cytokines (100 units/mL interleukin-1β [IL-1β], 1,000 units/mL tumor necrosis factor-α [TNF-α], and 1,000 units/mL IFN-γ) (R&D Systems), 500 nmol/L thapsigargin (Sigma-Aldrich), 40 µmol/L S-nitroso-N-acetylpenicillamine (SNAP) (Sigma-Aldrich), and 100–500 µmol/L DS or dextran, cells were fixed in 4% paraformaldehyde. TUNEL and insulin staining were performed using the DeadEnd Fluorometric TUNEL System (Promega) and a guinea pig anti-insulin (Dako) antibody (22). A minimum of 1,500 β-cells were examined per coverslip.

Islet protein extracts were analyzed by Western blotting with antibodies against CHOP, phosphorylated (p)Stat1, Jak2, p-p38, pJNK, p-p65 (Cell Signaling Technology), inducible nitric oxide synthase (iNOS) (Santa Cruz Biotechnology), and α-tubulin (Sigma-Aldrich) as previously described (22). Secreted CXCL10 was measured in islet media after incubation of human islets with cytokines and DS by using the human CXCL10 ELISA MAX Deluxe (BioLegend).

Human islets were incubated with DS and cytokines as above for 18 h. Islets were loaded with 100 µmol/L 4-amino-5-methylamino-2′,7′-difluorofluorescein diacetate (Thermo Fisher Scientific), dissociated into single-cell suspensions with phenol red–free TrypLE Express Enzyme (Thermo Fisher Scientific), fixed with 2% paraformaldehyde, and analyzed by flow cytometry.

To detect oxidative stress, we incubated 20 human islet equivalents (IEQs; 1 IEQ is one 125-mm diameter islet) with 20 µmol/L 2′,7′-dichlorodihydrofluorescein diacetate (Abcam), cytokines, and DS for 24 h. Excitation and emission wavelengths were 480 nm and 520 nm, respectively. INS-1 cells were treated for 30–240 min as described above, and reactive oxygen species (ROS) levels were measured as in islet cells. Another set of cells was harvested after incubation and glutathione (GSH) and oxidized GSH (GSSG) contents determined using a GSH/GSSG Ratio Detection Assay Kit (Abcam). Excitation and emission wavelengths were as noted above.

We measured mitochondrial oxygen consumption rates (OCRs) with a Seahorse XFe24 Analyzer (Agilent Technologies). Human islets were treated with cytokines and DS as described above for 24 h. After washing the islets, 150 human IEQs were seeded in each well of the XFe24 islet capture microplate and incubated at 37°C without CO₂. OCR was measured according to the manufacturer’s protocol as five measurements in the following conditions: 2 and 20 mmol/L glucose, 5 µmol/L oligomycin, 1 µmol/L carbonyl cyanide-4-(trifluoromethoxy) phenylhydrazone, and 5 µmol/L rotenone. Raw data were analyzed with WAVE 2.6, and OCR was normalized to the islet insulin content. INS-1 cells (10⁵/well) were seeded in the XFe24 cell culture microplate for 1 day, washed, and then treated with cytokines and DS for 2 h. Analysis was done as in human islets.

Two hundred human IEQs were treated with cytokines and DS as above for 24 h. Islets were washed with Krebs-Ringer buffer containing 6 mmol/L HEPES and 0.2% BSA, transferred into perifusion chambers (Biorep Technologies), and run on dynamic glucose-stimulated insulin secretion (GSIS) as 2 mmol/L glucose (10 min), 20 mmol/L glucose (30 min), 2 mmol/L glucose (10 min), 30 mmol/L KCl (10 min), and 2 mmol/L glucose (10 min). The perifusate was collected at 1-min intervals, and DNA content of perifused islets was measured with the Quant-iT PicoGreen dsDNA assay kit (Thermo Fisher Scientific). Insulin secretion was measured with the Insulin ELISA kit (Mercodia) and normalized by DNA content.

cDNA libraries were made using the Illumina TruSeq mRNA Library Prep kit and sequenced with pair-end 150 bpss on an Illumina HiSeq 4000 instrument. Sequenced reads were trimmed for low-quality bases and adapter sequences using Cutadapt (23). STAR was used to align raw sequencing reads to the human GRCh37 reference genome (24). Raw read counts were calculated using HTSeq-count (25). Differential expression analysis was performed using DEseq2 package (26). A heat map plot was generated using R heatmap package. Expression data were normalized per gene across samples by subtracting their mean and then dividing by their SD. This normalization was performed separately for the three donors to remove human variations. Gene ontology and gene enrichment analysis were performed using DAVID Functional Annotation and gene set enrichment analysis (GSEA) (27), respectively.

Human islets were dissociated and stained with anti–C-peptide Alexa Fluor 647 antibody (BD Pharmingen) (28) and anti-pSTAT1 Alexa Fluor 488 antibody (BioLegend) to determine STAT1 activation in human β-cells. Cells were loaded in a BD LSRFortessa, and results were analyzed with FlowJo X software (Tree Star).

INS-1 cells (10⁴ cells/well) were treated with cytokines as above and DS, HS, dextran, fibroblast growth factor 1 (FGF1) (R&D Systems), or E7090 (selective inhibitor of the tyrosine kinase activities of FGF receptors [FGFRs] 1, 2, and 3) (Chemgood) for 24 h. Each experiment was performed with five replicates per condition. Cell number was measured using CountBright Absolute Counting Beads (Invitrogen). Annexin V staining was performed with FITC–Annexin V and 7-aminoactinomycin D using the FITC–Annexin V Apoptosis Detection Kit (BD Pharmingen). Cells were incubated for 15 min at room temperature in the dark, binding buffer added, and flow cytometry analysis performed (see below). Details on the analysis of heparanase activity and FGF signaling are provided in the Supplementary Material.

Eight-week-old NOD/LtJ (NOD) female mice (The Jackson Laboratory) were housed in specific pathogen-free conditions. NOD mice were i.p. injected daily with saline, 2 or 10 mg/kg DS (29), and 10 mg/kg dextran. In the studies with nondiabetic NOD mice (prevention studies), nonfasting blood glucose was measured once a week until the end of the treatments by a portable glucometer (AlphaTRAK 2; Abbott Laboratories), and mice were considered diabetic when blood glucose was >250 mg/dL. In the studies with early-onset diabetic NOD mice (reversal studies), mice with blood glucose >250 mg/dL were treated with saline, DS, and dextran as described above. Once the treatment was initiated, mice were considered diabetic when blood glucose was >250 mg/dL. Plasma or serum insulin was determined by ELISA for mouse insulin (Mercodia). An i.p. glucose tolerance test was performed as described (22). β-Cell proliferation was measured by i.p. injection of BrdU (Amersham Biosciences) 6 h before sacrifice.

Pancreases were fixed for 4 h at room temperature in Bouin fixative solution (Sigma-Aldrich) and then washed with water and 70% ethanol with 1:1,000 Scott’s Tap Water for 3 days. Pancreases were paraffin embedded and sectioned, and the β-cell mass was measured in four nonconsecutive insulin- and hematoxylin-stained sections per mouse using ImageJ (National Institutes of Health) (22). Sections were also stained for BrdU or Ki67 (Thermo Fisher Scientific) and insulin (Dako) to detect β-cell proliferation and for CD3 (Abcam), CD45R/B220 (eBioscience), and insulin to detect immune infiltration in islets. For β-cell death analysis in vivo, TUNEL and insulin immunostaining were performed as described (22). Sections were also stained with hematoxylin and eosin for pathologic evaluation of islet insulitis that was calculated as percent of islets per mouse in each stage of insulitis (22). Details of the studies on sialitis, septic shock, and colitis mouse models are provided in the Supplementary Material.

Eight-week-old NOD female mice were treated daily with 10 mg/kg DS, 10 mg/kg dextran, and saline i.p. for 4 weeks. Spleen and pancreatic lymph nodes (PLNs) were collected, grinded, and made into a cell suspension after lysis of the red blood cells and filtration. Cells (10⁶ cells/mL) were treated for 16 h with 2 μg/mL soluble anti-CD3 and 2 μg/mL soluble anti-CD28 (BioLegend). Surface and intracellular staining of T cells for flow cytometry was achieved with anti–CD8-FITC (eBioscience), anti–CD4-Pacific Blue (BioLegend), anti–IFN-γ–phycoerythrin (eBioscience), anti–PD-1–allophycocyanin (BioLegend), anti–CD25-PerCP-Cy5.5 (BioLegend), and anti–FoxP3-phycoerythrin (eBioscience). Live/dead cells were identified by the Zombie NIR Fixable Viability Kit (BioLegend). Cells were analyzed in the flow cytometer as indicated above.

mRNA expression analysis of immune markers in splenocytes and pancreas, and β-cell markers in human islets, was performed by real-time PCR using specific primers (22) (Supplementary Table 3).

Nine-week-old prediabetic NOD female mice were treated daily with saline or 10 mg/kg DS. One week after initiation of DS or saline treatment, mice were i.p. injected with 500 μg on day 0 and 250 μg on days 2 and 4 of the anti-mouse PD-L1 monoclonal antibody (clone 10F.9G2, rat IgG2b; Bio X Cell). Blood glucose was measured daily.

Statistical analysis was performed with GraphPad Prism 7 software (GraphPad Software, Inc.). The different tests used to perform statistical analysis are described in the figure legends. In all statistical analyses, P < 0.05 was considered significant.

The data sets generated and/or analyzed during the current study are available from the corresponding author on reasonable request. The resources used during the current study are also available from the corresponding author on reasonable request.

Cytotoxicity induced by cytokines is one of the culprits of β-cell demise in T1D (1,2,8). However, whether DS can protect β-cells against cytokines is unknown. DS completely protected mouse (Fig. 1A and B) and human (Fig. 1C and D) β-cells against death induced by the combination of IL-1β, TNF-α, and IFN-γ. The effect on β-cell survival was intrinsic to β-cells because DS also protected INS-1 cells against cytokines (Supplementary Fig. 1A–C). In contrast, the nonsulfated version of DS, dextran, did not protect human β-cells and INS-1 cells against cytokines (Supplementary Fig. 1D–F). Interestingly, DS was slightly but significantly more potent in protecting INS-1 cells against cytokines (Supplementary Fig. 1E and F) than the same dose of the sulfated polysaccharide HS, known to play an important role in islet basement membrane, β-cell biology, and the pathology of T1D (30). Sulfated polysaccharides protect FGFs from denaturation and facilitate their binding to FGFRs (31,32). HS has been suggested to protect β-cells from cytokines by potentiating FGF signaling (32). In this study, we report that DS enhanced FGF1 signaling (ERK1/2 phosphorylation), and this effect was blocked by E7090, the selective inhibitor of FGFR1, -2, and -3 (Supplementary Fig. 1G). Furthermore, FGF1 together with DS significantly increased INS-1 cell survival against cytokines compared with DS alone (Supplementary Fig. 1F, bottom panel). However, FGFR signaling blockade did not eliminate DS prosurvival effects in INS-1 cells, suggesting alternative prosurvival mechanisms induced by DS.

Based on the protective actions of DS in human β-cells, we next determined whether DS treatment modified the transcriptional profile induced by cytokines in human islets. As shown in Supplementary Fig. 2A–C, DS significantly modified the expression of 693 genes, cytokines significantly modified 7,718 genes, and cytokines plus DS significantly altered 6,246 genes in human islets compared with vehicle treatment. Of these genes, DS significantly upregulated 217 genes more than twofold and significantly reduced 353 genes >50% (NOS2, CXCL10, CXCL9, and FAS among them) in human islets treated with cytokines (Fig. 1E). Gene set enrichment analysis identified oxidative phosphorylation, pancreatic β-cells, pyruvate metabolism, ATP synthesis, mitochondrial protein complex, oxidoreductase activity, and insulin secretion as the top biological processes upregulated by DS in cytokine-treated human islets (Fig. 1F). In contrast, apoptosis, ROS, inflammatory response, and TNF-α and IFN-γ signaling pathways were the main biological processes downregulated by DS in cytokine-treated human islets (Fig. 1F). Analysis of pathways that include genes critical for inflammatory response, apoptosis, and TNF-α and IFN-γ response revealed these were highly decreased in human islets treated with cytokines and DS compared with cytokines alone (Fig. 1G). A closer examination of the genes downregulated by DS in the presence of cytokines also revealed a large number of chemokines and chemokine-receptor genes (Fig. 1H). To correlate these observations with chemokine production, we focused on CXCL10 and found that the increase in CXCL10 secretion induced by cytokines in human islets was significantly attenuated by DS (Fig. 1I). Taken together, these results indicate that DS protects β-cells from cytokine-mediated cytotoxicity and reduces the expression on proinflammatory genes in human islets.

DS treatment significantly decreased the activation of cytokine-induced intracellular targets mediating β-cell death such as JAK2/STAT1, p65–nuclear factor-κB, and iNOS, but not JNK and p38 MAPK, suggesting specific modulation of proapoptotic signals by DS (Fig. 2A and B). These results confirmed the GSEA shown in Fig. 1F, indicating that DS decreases IFN-γ, TNF-α, and IL-1β signaling pathways in human islets. Integrated pathway analysis of genes modulated by cytokines and DS in the IFN-γ response pathway using Ingenuity Pathway Analysis showed STAT1 as the central node of these pathways and chemokines CXCL1, -9, -10, and -11 and the receptor CXCR3 as prominent edges (Supplementary Fig. 2D). Indeed, STAT1 activation (pSTAT1) specifically in human β-cells is clearly decreased in islets incubated with DS and cytokines compared with cytokines alone (Fig. 2C).

DS significantly reduced NO formation induced by cytokines in human islets (Fig. 2D) and human β-cell death induced by SNAP (NO donor) (Fig. 2E), suggesting that impairment of NO formation and action in an inflammation context could contribute to the protective effects of DS in human β-cells. In addition, cytokine-induced ROS formation was also reduced by DS in human islets and INS-1 cells (Fig. 2F and Supplementary Fig. 3A), and conversely, the antioxidant pathway GSH/GSSG was increased by DS (Supplementary Fig. 3B). Collectively, these results suggest that DS protects β-cells against cytokines by downregulating cell death and inflammation signals and decreasing reactive nitrogen species and ROS.

DS treatment significantly reduced human β-cell death and CHOP expression induced by thapsigargin (ER stress inducer) in human islets (33) (Fig. 2G and H). Taken together, these results clearly indicate that DS is a novel protector of β-cells in cytotoxic environments that cause β-cell destruction in T1D.

Cytokines attenuate GSIS by restricting mitochondrial respiration in β-cells (34,35). Therefore, we next determined whether DS could improve mitochondrial respiration in human islets treated with cytokines. Detailed real-time respiratory analysis of human islets (Fig. 3A and B) revealed that DS treatment reversed cytokine-induced impairments in basal respiration, ATP synthesis, maximal respiration, proton leak, nonmitochondrial respiration, and coupling efficiency (Fig. 3B and C). DS also recovered the production of ATP in INS-1 cells treated with cytokines for 2 h, enough time to induce a pronounced decrease in ATP levels without altering cell numbers (Supplementary Fig. 4). This improvement in ATP production induced by DS correlated with a significant improvement in GSIS in perifused human islets (Fig. 3D and E). In addition, gene expression of β-cell markers such as PDX1, MAFA, and PAX6, corrected by insulin content, negatively correlated with cytokine treatment, and the expression was upregulated by DS cotreatment (Fig. 3F). Therefore, DS can protect the bioenergetics and function of the β-cell in inflammation conditions.

Because DS protects β-cells against cytokines and ER stress and reduces inflammation and chemokine production in islets, we next analyzed whether short-term administration of DS could potentially protect prediabetic NOD mice from developing T1D. We treated 8-week-old prediabetic NOD female mice daily with 10 mg/kg DS i.p. (28) for 4 weeks. No significant changes in the percentage of diabetic mice, body weight, blood glucose, plasma insulin, β-cell mass, or β-cell proliferation were observed following DS treatment (Supplementary Fig. 5A–H). However, glucose tolerance was significantly improved and β-cell death and insulitis were significantly diminished in DS-treated mice (Supplementary Fig. 5E–J).

Because DS treatment decreased insulitis in NOD mice, we next analyzed whether DS administration could widely affect the immune response in mice. First, the number and size of foci of infiltrated mononuclear cells in submandibular salivary glands (sialitis) was similar in saline- and DS-treated NOD mice (36) (Supplementary Fig. 5K). Second, 4-week treatment with DS did not alter the response of mice to a septic shock induced by Escherichia coli lipopolysaccharide injection (37) (Supplementary Fig. 5L and M). These studies suggest a potential beneficial effect of DS treatment on ameliorating β-cell death and insulitis in prediabetic NOD mice without alteration of other immune responses. Unfortunately, daily treatment for only 4 weeks with DS was not enough to prevent the development of diabetes in NOD mice later in life (Supplementary Fig. 5N), suggesting the need for continuous treatment with DS to potentially ameliorate diabetes.

Administration of large doses of 35–55 kDa DS in the drinking water rapidly induces colitis in mice (38). To determine whether daily i.p. administration of 6.5–10 kDa DS in NOD mice could cause intestinal alterations, we analyzed multiple intestinal morphology parameters in DS-treated NOD mice. Four-week daily i.p. injections of 10 mg/kg DS (6.5–10 kDa) did not significantly alter body weight (Supplementary Fig. 5B), stool consistency (not shown), or intestinal morphology and immune infiltration in these mice compared with mice in which 3% DS (35–55 kDa) was supplemented in the drinking water for 1 week (Supplementary Fig. 5O and P). Therefore, i.p. administration of 10 mg/kg of 6.5–10 kDa DS does not induce colitis in mice.

Four-week DS treatment also increased HS and HSPG staining in islets of NOD mice (Supplementary Fig. 6A). Furthermore, in vitro treatment of human islets with DS also reduced the decrease in HSPG expression induced by cytokines (Supplementary Fig. 6B and C). The increase in HS/HSPG presence in islets treated with DS could be the result of inhibition of heparanase activity. Indeed, HS and DS similarly inhibited heparanase activity (Supplementary Fig. 6D). Collectively, these results suggest that DS may contribute to maintain basement membrane integrity and HS/HSPG expression in islets and reduce heparanase activity, an enzyme involved in T1D development (39).

Interestingly, administration of 10 mg/kg of nonsulfated dextran in NOD mice did not induce any improvement in glucose homeostasis, β-cell mass, β-cell proliferation, β-cell survival, or insulitis (Supplementary Fig. 7A–I), indicating that only the sulfated version of dextran, DS, induces beneficial effects in prediabetic mice and has the potential to ameliorate diabetes in NOD mice.

We next tested whether continuous chronic treatment with DS would decrease diabetes development in prediabetic NOD mice. For this purpose, we administered daily 2 or 10 mg/kg DS i.p. to 10-week-old prediabetic NOD female mice for 11 weeks. Interestingly, 10 mg/kg DS significantly reduced the percentage of diabetic mice (20%) compared with saline (70%) or the lower dose of DS (60%) (Fig. 4A). Mice treated with 10 mg/kg DS displayed lower blood glucose compared with saline-treated mice (Fig. 4B). At the end of the treatment, 10 mg/kg DS-treated NOD mice showed significantly increased plasma insulin and β-cell mass compared with saline-treated mice, reaching similar values to those in 10-week-old nondiabetic NOD mice (Fig. 4C and D). To determine whether the increase in β-cell mass was due to enhanced β-cell proliferation or survival, we performed costaining for insulin and BrdU or TUNEL. As shown in Fig. 4E–G, DS treatment decreased TUNEL-positive β-cells, while the percentage of BrdU-positive β-cells was similar in saline- and DS-treated mice. Taken together, these results indicate that continuous DS treatment decreases diabetes development, β-cell death, and islet immune cell infiltration and enhances β-cell mass and insulin availability in NOD mice.

To address whether DS treatment could revert diabetes progression in early-onset diabetes, we treated diabetic NOD female mice (blood glucose levels >250 mg/dL) with DS or saline for 10 weeks and analyzed blood glucose weekly and β-cell homeostasis at the end of the treatment. Only 40% of mice treated with DS displayed blood glucose levels >250 mg/dL at the end of the 10-week treatment versus 100% of the saline-treated mice (Fig. 5A). DS-treated mice displayed lower blood glucose (Fig. 5B), an aspect not observed with nonsulfated dextran treatment (Supplementary Fig. 7J), and significantly increased plasma insulin compared with saline-injected mice (Fig. 5C). Analysis of β-cell mass clearly showed a significant difference between DS-treated and saline-treated diabetic NOD mice (Fig. 5D). β-Cells were almost undetectable in diabetic NOD mice treated with saline but easily detectable in the pancreas of NOD mice treated with DS, albeit surrounded by both T and B cells (Fig. 5E and F). Collectively, these results suggest that DS protects β-cells and potentially modulates the immune system, ameliorating the development and progression of early-onset T1D in NOD mice.

Because DS treatment is capable of reversing T1D in a large percentage of early-onset diabetic NOD mice, we next analyzed whether DS treatment could reduce the expression of genes involved in activation or repression of T cells in splenocytes from NOD mice treated daily with 10 mg/kg DS for 4 weeks. We chose 4 weeks of treatment for these and following studies because glucose homeostasis, body weight, and general body condition were similar in both groups of treated mice (Supplementary Fig. 5). Splenocytes from DS-treated NOD mice displayed significant downregulation of expression of cytokine genes such as Ifn-γ and Il-17 and activation markers such as T-bet, Rorγt, and Pu.1 (Fig. 6A). DS treatment also increased the expression of the anti-inflammatory cytokine Il-10 (Fig. 6A). In contrast, treatment with 10 mg/kg nonsulfated dextran did not alter the expression of any of these genes in splenocytes (Supplementary Fig. 7K). In addition, DS treatment reduced the expression of genes such as Il-1β, Tnf-α, Cxcl9, and Cxcl10 in pancreas (Fig. 6B). These results clearly portray a decreased inflammatory signature induced by DS in vivo in NOD mice.

PD-1 expression by effector T cells regulates islet-reactive effector cells by inhibiting infiltration of the islet and limiting diabetes (21). To address whether DS could be modulating PD-1 expression and T-cell activation, we performed flow cytometry analysis of PD-1 and IFN-γ in CD4⁺ and CD8⁺ cells in spleen and PLN from NOD mice treated daily with 10 mg/kg DS or dextran for 4 weeks. Treatment with DS, but not with dextran, significantly increased the number of PD-1⁺CD4⁺ cells in spleen and PLN and the number of PD-1⁺CD8⁺ cells in PLN (Fig. 6C, Supplementary Fig. 7L, and Supplementary Fig. 8A and B). Furthermore, DS-treated mice display a significantly reduced number of IFN-γ⁺CD4⁺ and CD8⁺ in spleen and PLN, an effect not seen in dextran-treated mice (Fig. 6D, Supplementary Fig. 7L, and Supplementary Fig. 8C and D). Importantly, DS treatment did not alter the numbers of total lymphocytes and CD4⁺ and CD8⁺ cell populations (Supplementary Fig. 9). These data indicate that in vivo treatment with DS, but not with dextran, induced an increase in the immunoregulatory marker PD-1 in T cells and a decrease of activated T cells that could potentially participate in the amelioration of diabetes in NOD mice. Indeed, cotreatment of 9-week-old prediabetic NOD female mice with a blocking monoclonal antibody against PD-L1 abrogated the beneficial effects of DS of preventing diabetes development in NOD mice (Fig. 6E).

Tregs are essential regulators of peripheral immune tolerance (40,41), and a decrease in their number or suppressive function can lead to autoimmune disease development, including T1D (42–44). Analysis of FoxP3 mRNA expression showed that DS significantly increased the expression of this transcription factor in splenocytes (Fig. 6A). This result encouraged us to examine whether DS could modify the numbers of Tregs in vivo. The number of FoxP3⁺CD25⁺CD4⁺ cells in both spleen and PLN was significantly increased by ∼50% in DS-treated compared with saline-treated NOD mice (Fig. 6F). These effects were not observed in mice treated with nonsulfated dextran (Supplementary Fig. 7K and L). Collectively, these data suggest that DS modulates Treg numbers, and this could contribute to the amelioration of diabetes in NOD mice.

T1D is an autoimmune disease in which central and peripheral mechanisms of immune tolerance fail, leading to the destruction of insulin-producing cells by autoreactive lymphocytes. Therapeutic approaches to stop this destructive process, gain immune tolerance, preserve residual insulin secretion, and enhance β-cell regeneration are being pursued (4,45–47). However, therapies with single drugs have shown modest efficacy thus far (45–47). DS is a sulfated semisynthetic polysaccharide with islet-protective functions in an islet transplant setting, a modulator of the maturation of human DCs in vitro, and a protector of endothelium from complement and NK-cell cytotoxicity (8–12). Therefore, DS could be of therapeutic value in a setting in which immune regulation needs to be reestablished and β-cell mass needs to be preserved, such as in T1D. In this study, we show that DS protects human β-cells in vitro against ER stress and cytokines, reduces both cytokine-mediated mitochondrial dysfunction and impaired insulin secretion, increases islet HS/HSPG expression, and reduces chemokine production in islets. At the same time, DS increases the number of PD-1⁺ T cells, reduces the number of IFN-γ⁺ T cells, and enhances the number of Tregs in vivo in NOD mice, aspects that could suggest the improvement of a tolerogenic response in T1D, leading to amelioration of diabetes. Indeed, DS treatment prevents the development of the disease, an effect that is lost by blocking PD-1/PD-L1 interaction with an anti–PD-L1 antibody. Most importantly, we find that DS is capable of stopping β-cell death, preserving β-cell mass and circulating insulin levels, and decreasing diabetes incidence in early-onset diabetic NOD mice. The beneficial effects of DS on β-cell survival, immune modulation, and reversal of diabetes were not observed with the nonsulfated version of DS, dextran, suggesting that the sulfate groups in DS are essential for the observed effects. Collectively, these studies highlight the potential therapeutic value of DS for the treatment of T1D.

The NOD mouse model, which spontaneously develops T1D, has greatly helped in understanding the etiopathogenetic molecular mechanisms of this disease (48–50). Although therapies regulating early T-cell activation have been successful in preventing the disease in NOD mice, very few have provided some success in reversing the progression of the ongoing disease (45–49). In this study, we report that treatment with DS not only prevented the development of T1D but also reversed diabetes in 60% of early-onset diabetic NOD mice. Importantly, continuous daily treatment with DS was required to ameliorate T1D in these mice. Furthermore, DS treatment neither affected sialitis nor changed the response to a septic shock in mice, suggesting that general immune function is not altered during DS treatment.

DS treatment clearly decreased β-cell death in NOD mice. Recently, it has been reported that ER stress in β-cells is one of the potential triggers of β-cell dysfunction and death in the initiation of T1D (6,7). Furthermore, proinflammatory cytokines are involved in the destruction of β-cells in T1D (8). In the current studies, we found a pronounced protection of human β-cells against thapsigargin or the combination of IL-1β, TNF-α, and IFN-γ. DS decreased CHOP expression and diminished the activation of JAK2/STAT1 and nuclear factor-κB/iNOS signals without altering other intracellular signals such as JNK and p38. This suggests that there is differential modulation of intracellular signals by DS that leads to protection against cell death and that DS might be inducing prosurvival signals to counteract the potential β-cell death induced by JNK and p38 pathway activation. The effect of DS on β-cell survival was β-cell intrinsic and similar to the one induced by HS, a known component of the islet basement membrane and inducer of β-cell survival in proinflammatory conditions (30). HS has been shown to further enhance FGF binding to FGFR (31,32). Interestingly, DS further enhanced FGF1 signaling but DS-mediated prosurvival effect in INS-1 cells was not dependent on FGFR signaling activation, suggesting alternative mechanisms.

As previously shown in endothelial cells, DS decreased the activation of JAK2/STAT1 by IFN-γ and diminished the production of NO in human islets (51–53). The decrease in cytokine signaling induced by DS led to decreased chemokine production, diminished proinflammatory gene expression, improved mitochondrial function, and enhanced insulin secretion in human islets. Although some of the functional, metabolic, and gene-related beneficial effects of DS can be associated to the decrease in β-cell death, DS provided beneficial effects to the β-cell (2 h) even before cytokines induced β-cell death. Furthermore, similarly to HS, DS also inhibited heparanase activity, an enzyme that has been shown to play key roles in mouse β-cell survival and autoimmune diabetes (39). DS also increased the presence of HS and HSPG in islets in a proinflammatory environment, perhaps due to the decrease in heparanase activity. Nevertheless, the detailed analysis of the multiple mechanisms involved in DS-protective effects warrants further studies. In contrast, DS did not show any proliferative action on β-cells in NOD mice. This result highlights the opportunity for a combination therapy of DS with β-cell mitogens to improve the treatment of T1D.

DS modulates the maturation of human DCs in vitro, leading to decreased T-cell proliferation and diminished cytokine secretion (13). In our studies, we found that DS treatment increased the number of PD-1⁺CD4⁺ and PD-1⁺CD8⁺ cells in spleen and PLN in vivo in NOD mice. T-cell activation is induced by two signals: 1) the interaction with the specific antigen presented by MHC and 2) the interaction with the costimulatory antigen nonspecific signal, both through receptors present on the T-cell surface and the antigen-presenting cells (54). A state of anergy or apoptosis occurs when costimulation is not engaged (54,55). Besides the costimulatory signals, inhibitory costimulatory receptors provide negative stimuli influencing T-cell activation and the maintenance of peripheral tolerance (17,55). If the balance between positive and negative costimulatory signals is broken in favor of the first one, autoimmunity may arise (15,17). PD-1 constitutes one of the negative costimulator receptors in T cells (18). During the presentation of tissue antigens to CD8⁺ T cells, PD-1/PD-L1 interaction crucially controls the effector differentiation of autoreactive T cells to maintain self-tolerance. One of the first pieces of evidence for the protective effect of PD-L1 in T1D was observed when blocking antibodies to PD-1/PD-L1 were administered to NOD mice, strongly accelerating diabetes development (19–21). In this study, we report that DS increased the number of PD-1⁺CD4⁺ and PD-1⁺CD8⁺ T cells, suggesting that this upregulation could modulate the effector differentiation of autoreactive T cells enhancing self-tolerance. Importantly, treatment with anti–PD-L1 antibody eliminated the beneficial effect of DS on the prevention of diabetes in NOD mice. This suggests that upregulation of inhibitory costimulatory signals such as PD-1/PD-L1 by DS at the effector site could be of therapeutic importance for the treatment of autoimmune diabetes. How DS stimulates PD1 expression in T cells at the cellular and molecular level warrants further studies.

In humans, no alterations in Treg numbers but loss of Treg phenotype and defective suppressive capacity have been observed in peripheral blood cells isolated from subjects with T1D (56,57). Treg depletion and FoxP3 deficiency in genetically modified mice rapidly lead to development of autoimmune diabetes (40–42). Interestingly, Tregs infiltrating in the islet in mice have high levels of FoxP3 but decreased expression of the high-affinity IL-2 receptor CD25 and the prosurvival protein Bcl-2 (58). Therefore, if alterations in Treg number, phenotype, or function are contributors to T1D, increasing these parameters can be beneficial for the treatment of the disease (40–42,59). Importantly, our results indicate that DS treatment in vivo leads to an increase in the number of CD25⁺FoxP3⁺ cells, potentially increasing the suppressive capacity of T cells. Taken together, these results suggest that DS is beneficial for Tregs in diabetes, highlighting a potential therapeutic future for DS in this context.

In summary, we have found that one single molecule, DS, is a β-cell prosurvival factor, a suppressor of proinflammatory signals, and a Treg- and PD-1–mediated immune modulator with potential therapeutic effects for the treatment of early-onset T1D. Because DS is commercially available and clinically used, this drug could be a potential therapeutic agent for T1D in humans.

Acknowledgments. The authors thank Drs. Andrew F. Stewart, Donald K. Scott, Rupangi Vasavada, and Juan Carlos Alvarez-Perez (Icahn School of Medicine at Mount Sinai, New York, NY) for helpful comments during the development of these studies and Drs. Jayalakshmi Laksmipathi and Jordi Ochando and Xinyi Yang (Icahn School of Medicine at Mount Sinai) for technical help. The authors also thank Prodo Laboratories, Inc. (Aliso Viejo, CA), for providing human islets for these experiments and the Human Islet and Adenoviral Core of the Einstein-Mount Sinai Diabetes Research Center (New York, NY) for providing expertise on the analysis of islet perifusion and mitochondrial bioenergetics.

Funding. This work was supported in part by grants from the National Institutes of Health (DK-113079 and DK-020541-38) and the Department of Defense (W81XWH-17-1-0363 and W81XWH-17-1-0364) to D.H. and A.G.-O. and the American Diabetes Association/F.M. Kirby Foundation (1-14-BS-069) to A.G.-O.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. G.L., F.R.-P., J.Z., Z.Z., T.Z., S.V., C.R., C.B., P.C., M.P.S., and J.G.G. researched data, contributed to discussion, and reviewed and edited the manuscript. D.H. and A.G.-O. contributed to discussion and wrote and edited the manuscript. A.G.-O. is the guarantor of this work and, as such, had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.