Loss of Peripheral Protection in Pancreatic Islets by Proteolysis-Driven Impairment of VTCN1 (B7-H4) Presentation Is Associated with the Development of Autoimmune Diabetes Free

Type 1 diabetes (T1D) is a life-threatening disease of an autoimmune nature, for which the only currently available treatment is continuous insulin administration. Clinical T1D arises as a consequence of the cytotoxic destruction of insulin-producing β cells by abnormally activated autoreactive T cells specific for multiple islet cells’ Ags (1, 2). Accumulating evidence, however, suggests that islet cells do not merely play a role of plain targets of autoimmune destruction, but, on the contrary, possess several protective mechanisms capable of downregulation of autoimmune attack (3, 4). One of such mechanisms is at the center of our investigation.

V-set domain–containing T cell activation inhibitor 1 (VTCN1), also known as B7-H4, B7S1, B7X, is a negative costimulatory molecule, representing one of the newly discovered members of the B7 family (5–7). VTCN1 acts through a not yet identified receptor on T cells, inhibiting T cell activation, proliferation, and cytokine production (5, 6, 8, 9). The persistence of autoreactive T cell responses during T1D prompted several experimental attempts to alleviate diabetogenic autoimmunity via artificial enrichment of VTCN1-mediated coinhibition. Accordingly, matrix surface–bound VTCN1-Ig fusion protein suppressed the proliferation of islet-specific T cell clones derived from T1D patients. Furthermore, the treatment of diabetes-susceptible NOD mice with VTCN1-Ig protein significantly attenuated T1D (10).

Unlike classical costimulatory molecules (B7-1 and B7-2), whose natural expression and action are strictly limited to APCs (11, 12), VTCN1 is also expressed in several nonlymphoid organs and, most importantly, in pancreatic islets (6, 7, 9, 13–15). Consequently, VTCN1 has been hypothesized to not only inhibit classical T cell activation by APCs in the lymphoid compartment, but also to induce T cell tolerance within peripheral target tissues. Supporting this suggestion, upregulated VTCN1 expression was detected in multiple neoplasms (7, 13, 16–18), where it was associated with tumor-protective downregulation of antitumor T cell responses (19).

In the T1D setting, transfection of the VTCN1-Ig construct into human primary islet cells protected them from diabetogenic T cell clones isolated from T1D patients (14). Additionally, ex vivo VTCN1 overexpression in mouse islets shielded them from T cell–induced damage in transplantation experiments (20), whereas in vivo β cell–specific VTCN1 overexpression protected against diabetes induced by both CD4⁺ and CD8⁺ islet-specific clonal T cells (9, 21). Therefore, the distinctive combination of T cell coinhibitory function with expression on islet cells uniquely positions VTCN1 at the interface of pancreatic islets and the immune system.

Despite the growing number of functional studies utilizing genetically manipulated VTCN1 (overexpression and/or deletion), the state of natural VTCN1 on either APCs or islet cells in connection with T1D development is largely unknown. This is why we asked the question of whether a compromised function of endogenous VTCN1 can trigger enhanced vulnerability of islet tissue to diabetogenic autoimmunity. Recently, we unveiled an endogenous pathway of functional VTCN1 inactivation in APCs (particularly in macrophages [Mϕs] and dendritic cells) of NOD mice and T1D patients. Specifically, a gradual loss of membrane-tethered VTCN1 due to a proteolytic cleavage mediated by metalloproteinase nardilysin (N-arginine dibasic convertase 1 [NRD1]) progressed alongside natural T1D development and triggered hyperproliferation of diabetogenic T cells (22). In the present study, we extend our previous findings and dissect a pattern of VTCN1 expression and presentation on islet cells in connection with diabetogenesis. Subsequently, we define a general mechanism of a progressive loss of VTCN1-mediated negative costimulation, which occurs in multiple tissues/cells (islet endocrine cells and APCs) due to the NRD1-dependent diminishment of membrane VTCN1. This mechanism is linked to T1D susceptibility and depends on two separate but synergistic processes. The first process is a result of an increased intracellular NRD1 expression, ultimately leading to enhanced intracellular VTCN1 shedding. The second process includes a systemic upregulation of NRD1 (an enzyme with both intra- and extracellular activities) (23, 24) in multiple tissues, which additionally potentiates VTCN1 proteolysis by extracellular NRD1. In summary, our findings point toward VTCN1 stabilization along with systemic NRD1 inhibition as future strategies for T1D treatment.

All chemicals were from Fisher Scientific (Suwanee, GA) unless stated otherwise. 1,10-Phenanthroline, bestatin, collagenase P, and Histopaque were from Sigma-Aldrich (St. Louis, MO).

Postmortem pancreatic tissue samples from diabetic and control donors were obtained from South Dakota Lions Eye and Tissue Bank.

NOD/ShiLtJ (NOD), NOD.CB17-Prkdc^scid/J (NOD-scid), NOD.Cg-Prkd^cscid Tg(CAG-EGFP)10sb/KupwJ (NOD-scid-EGFP), NOD.Cg-Rag1^tm1Mom Tg(TcraBDC12-4.1) 10Jos Tg(TcrbBDC12-4.1)82Gse/J (NOD.BDC12-4.1), B6.129S7-Rag1^tm1Mom/J (B6-Rag1^−/−), and B6.NOD-(D17Mit21-D17Mit10)/LtJ (B6^g7) mice were purchased from The Jackson Laboratory. B6.G9C8 mice, transgenic for TCR derived from insulin-specific CD8⁺ T cell clone G9C8, and H-2K^d MHC allele (25) were a gift of Dr. A. Chervonsky (University of Chicago, Chicago, IL). B6.NOD-(D17Mit21-D17Mit10)-Rag1^tm1Mom (B6^g7-Rag1^−/−) mice were generated by crossing B6-Rag1^−/− with B6^g7 mice, then intercrossing F₁ animals to produce the F₂ population, which was typed by FACS for presence of T cells in peripheral blood and by PCR for presence of H-2^g7 using 5′-TGCACTTGCATAAGGAAAAC-3′ and 5′-GACTTTGGGGCCTACTTATG-3′ as forward and reverse primers, respectively. Only both H-2^g7–positive and T cell–negative mice were used. To generate bone marrow (BM) chimeric mice, B6^g7 or EGFP-positive (NOD × NOD-scid-EGFP) F₁ (NOD-EGFP) mice were used as either BM donors or recipients. All mice were maintained in the Sanford Research Animal Facility according to the National Institutes of Health guidelines for animal use.

Female NOD mice at 6 wk of age were injected daily (i.p.) with either vehicle or 10 mg/kg bestatin (see Fig. 3A). Two groups (n = 7/group) were treated for 4 wk, the other two (n = 12–14/group) were treated until diabetes development. To analyze the proliferation of insulitis-forming cells, all animals in the 4 wk treatment groups received i.p. injections of 2.5 mg/kg 5-ethynyl-2′-deoxyuridine (EdU) for the last four consecutive days of treatments. Mice were then euthanized and Mϕs, pancreatic lymph nodes (PLNs), spleens, and pancreata were collected for analysis. From 12 wk of age, all mice were monitored daily for urine glucose using Diastix strips. Mice with urine glucose ≥250 mg/dl on 2 consecutive days were confirmed diabetic by a blood glucose analysis and euthanized. Mϕs were collected from diabetic mice.

Adoptive transfers were performed by i.v. injection of 10⁷ splenocytes from diabetic NOD mice into irradiated (600 rad) B6^g7 or nonirradiated B6^g7-Rag1^−/− or NOD-scid recipients as previously described (26). Sham-transferred mice were not irradiated and received PBS. Starting from day 18 posttransfer, recipient mice were injected i.p. for 4 consecutive days with 2.5 mg/kg EdU per day. Mice were then sacrificed and pancreatic cryosections were prepared.

To generate chimeric mice, lethally irradiated (950 rad) 6- to 7-wk-old female recipients were injected i.v. with 1 × 10⁷ BM cells, isolated from tibias and femurs of donor mice as previously described (27). NOD-EGFP mice were used to distinguish between recipient and donor cells in the chimeric animals. To confirm the chimerism, recipient PBMCs were analyzed 3 and 10 wk after BM transfers for EGFP positivity by FACS. Mice were euthanized 10 wk after BM transplantations. Peritoneal Mϕs and pancreata were collected for further analyses.

Thioglycollate-elicited mouse peritoneal Mϕs (for the coculturing experiments and from B6^g7/NOD chimeras) or peritoneal cavity–residential Mϕs (from bestatin-treated mice) were collected in PBS and plated for immunofluorescence (onto coverslips) or RNA isolation (in a six-well plate). Only the adherent cells were then used for analyses.

Islets were isolated by collagenase digestion as described (28). Handpicked islets (n > 50) were then either lysed in 1 ml TRIzol (Life Technologies) for isolation of RNA, or dispersed for islet cell/T cell cocultures. Undispersed islets were cultured to either conditioned medium or in vitro treatments. The effects of NRD1 inhibitors were evaluated after in-culture treatments of freshly isolated whole islets for 24 h with either bestatin (10 μM), phenanthroline (20 μM), or vehicle solution. The islets were then fixed with ice-cold methanol and VTCN1 levels were analyzed by immunofluorescence. To prepare conditioned medium, islets (>20 per well) were cultured in 48-well plates for 16 h in 200 μl serum-free RPMI 1640.

For in vitro T cell activation assays, isolated islets were dispersed into single-cell suspensions in cell-dissociation buffer (Life Technologies) for 15 min at 37°C. Dispersed islet cells were then washed, depleted from plastic and nylon wool–adherent APCs for 1 h at 37°C, and incubated for 24 h in 96-well plates with RPMI 1640 medium containing 10% FBS in the presence or absence of 10 μM bestatin.

G9C8 or BDC 12-4.1 T cells were isolated from spleens of B6.G9C8 or NOD.BDC12-4.1 mice, respectively, via mechanical disruption of spleens followed by the lysis of RBCs. In some experiments G9C8 T cells were purified from bulk spleen population using the pan T cell purification kit II (Miltenyi Biotec). Isolated T cells were then labeled with CFSE (Life Technologies) and 1 × 10⁵ cells were added to the cultures of dispersed islets containing either bestatin or vehicle solution. In some experiments, 2 × 10⁴ NOD peritoneal Mϕs were incubated with disrupted islets for 24 h in the presence of either bestatin or vehicle solution before addition of BDC12-4.1 T cells. After 5 d of cocultures, the nonadherent cells were collected, washed, and analyzed by flow cytometry for CFSE dilution. On day 3, 100 μl conditioned medium was collected for IL-2 ELISA analysis, and then fresh medium with 5 U/ml rIL-2 (BioLegend) was added to the cocultures. Nonspecific activation of G9C8 T cells with anti-CD3/anti-CD28–coated beads (Thermo Fisher Scientific) in the presence/absence of bestatin was carried out for 5 d in the same experimental conditions as cocultures.

IL-2 concentrations were analyzed using a mouse IL-2 Platinum ELISA kit (eBioscience, San Diego, CA) following the manufacturer’s recommendations.

Frozen in OCT compound, pancreata were cryosectioned into 5-μm-thick sections and placed on Superfrost glass slides. Slides were fixed in 100% acetone at −20°C, air dried, and then stored at −80°C before processing. Cryosections were stained with either goat anti-mouse VTCN1 (2 μg/ml; R&D Systems) or mouse anti-human VTCN1 (5 μg/ml; eBioscience) and then visualized with the respective secondary Alexa Fluor 488–conjugated donkey anti-goat or donkey anti-mouse Abs (Jackson ImmunoResearch Laboratories). Pancreatic β and α cells were stained with guinea pig anti-insulin (1:100) and rabbit anti-glucagon (1:100), respectively (both from Abcam), followed by the corresponding secondary donkey Abs conjugated either with Alexa Fluor 594 or Alexa Fluor 647 (Jackson ImmunoResearch Laboratories). For staining of Mϕs, rat anti-mouse F4/80 (5 μg/ml) (eBioscience) was used. Peritoneal Mϕs were fixed in 4% paraformaldehyde and then stained with rat anti-mouse VTCN1 (5 μg/ml; R&D Systems) followed by donkey anti-rat conjugated with Alexa Fluor 594 (Jackson ImmunoResearch Laboratories). Pancreatic sections from animals injected with EdU were stained for insulin and CD3 and assessed for T cell proliferation via detection of EdU incorporation into the DNA of replicating cells, as previously described (29). CD3⁺ cells in the insulitis area were counted and the percentage of EdU⁺ cells was calculated. More than 130 islets with >30,000 infiltrating cells per group were analyzed.

To analyze insulitis, pancreatic sections were stained for insulin, and islet infiltration was graded in a blinded fashion from 0 (intact islets) to 4 (virtually no β cells within the islet infiltrate) and then divided into three subcategories: 1) no insulitis, islets with grade 0; 2) mild insulitis, islets with grades 1 and 2; and 3) severe insulitis, islets with grades 3 and 4. The percentage of islets falling into each subcategory was calculated for each experimental group.

All images were acquired on a Nikon A1 confocal microscope. Imaging conditions were kept constant for all samples within an experiment. Results were expressed as relative fluorescent units of VTCN1 staining calculated from at least 100 individual cells or >30 islets.

Total RNA was purified from Mϕs or pancreatic islets using Direct-zol RNA MiniPrep kit (Zymo Research) from samples collected in TRIzol. Only RNAs with an RNA integrity number >8.0 were used for generation of cDNA and subsequent real-time quantitative PCR (RT-qPCR) analysis. cDNA was generated using a GoScript reverse transcription system (Promega).

RT-qPCR analysis using primers specific for VTCN1, NRD1, proprotein convertase subtilisin/kexin type 1 (PCSK1), or PCSK2 was performed as previously described (22).

All Abs were purchased from eBioscience unless stated otherwise. To analyze regulatory T cells (Tregs), single-cell suspensions from spleens and PLNs were blocked with 10 μg/ml anti-mouse CD16/CD32 and then incubated for 30 min with 2.5 μg/ml anti–CD4-FITC and 2 μg/ml anti–CD25-allophycocyanin. The cells were then washed, incubated in fixation/permeabilization buffer, and stained for 30 min with 5 μg/ml PE-conjugated anti-Foxp3 Ab.

Analysis of diabetogenic NRP-V7 mimotope–reactive CD8⁺ cells was performed in splenocytes and PLN cells that were consequently incubated with 10 μg/ml PE-conjugated H-2K^d-KYNKANVFL pentamer (ProImmune) and 2.5 μg/ml FITC-conjugated anti-CD8 for 10 and 30 min, respectively. Cells were then washed and stained with the TO-PRO-3 (Life Technologies) viability marker. The viability marker 7-aminoactinomycin D (7-AAD) (eBioscience) was used to detect the dead cells during analyses of in vitro–activated G9C8 or BDC12-4.1 T cells and in peripheral blood monocytes collected from chimeric mice. For analysis of cocultured BDC12-4.1 T cell proliferation, the cells were also stained for 1 h with 5 μg/ml allophycocyanin-conjugated anti-CD69 (BioLegend). Samples were then recorded on an Accuri C6 cytometer (BD Biosciences) and analyzed with FlowJo software (Tree Star). All gating strategies are shown as supplemental material (see Supplemental Figs. 3, 4).

Serum-free medium was conditioned by pancreatic islets for 24 h at 37°C. Cell lysates were prepared in RIPA buffer as described before (22). The samples (5 μg total protein) were subjected to electrophoresis through a 4–12% NuPAGE Bis-Tris polyacrylamide gel. The membranes were blocked with 1% casein/0.1% Tween 20 in PBS and probed with goat anti-VTCN1 (LifeSpan BioSciences) or rabbit anti-NRD1 (Proteintech). HRP-conjugated anti-goat or anti-rabbit (1:5000) (Jackson ImmunoResearch Laboratories) were the secondary Abs followed by chemiluminescent substrate (SuperSignal West Dura substrate, Pierce). The membranes were then scanned on a UVP BioSpectrum 500 imaging system.

Statistical analysis was performed using GraphPad Prism 6 (GraphPad Software). Differences in T1D incidence rates were assessed using log-rank survival curve analysis. Differences in pairwise comparisons between groups were assessed using a two-tailed Student t test with a significance threshold of p ≤ 0.05.

To investigate whether presentation of endogenous VTCN1 by islet cells can be tied to the development of diabetogenic autoimmunity, we compared the dynamics of VTCN1 levels in Langerhans islets in diabetes-susceptible NOD and diabetes-resistant B6^g7 mice. Immunofluorescent staining of pancreatic sections revealed an age-dependent progressive loss of VTCN1 protein from NOD islets (Fig. 1A, left panel). Such loss became noticeable at ∼9–10 wk of age and led to widespread decline of VTCN1 by 15 wk of age, which is the average time of T1D onset in our NOD colony. Conversely, B6^g7 animals had stable intraislet VTCN1 levels throughout the observation time. The analysis of mean fluorescence intensities confirmed a statistically significant decrease of VTCN1 signal in NOD versus B6^g7 islets (Fig. 1A, right panel).

VTCN1 mRNA, however, increased with age in islets of diabetes-prone NOD animals, whereas an opposite nonsignificant trend toward age-dependent downregulation of VTCN1 expression characterized the B6^g7 islets (Fig. 1B). The augmentation of VTCN1 gene expression was even more evident and, importantly, was paralleled by the significant increase in metalloproteinase NRD1 gene expression when islets from prediabetic NOD mice were directly compared with the ones from age-matched B6^g7 animals (Fig. 1C). These findings echoed our recent report describing a highly similar phenotype of progressive VTCN1 protein loss (coupled with compensatory VTCN1 mRNA upregulation) in Mϕs and DCs from T1D-succeptible subjects, which occurred due to proteolysis of cell-associated VTCN1 by the metalloproteinase NRD1 (22). Hence, we compared dynamics of islet NRD1 expression and observed a significant age-dependent elevation of NRD1 mRNA in NOD, but not B6^g7, islets (Fig. 1D). Accordingly, levels of NRD1 proteinase were also augmented in prediabetic NOD islets in comparison with age-matched B6^g7 (Fig. 1E). Moreover, the amounts of intracellular NRD1 enzyme increased in NOD islets with age, paralleling T1D progression. To confirm that proteolytic shedding is indeed the mechanism liable for the observed VTCN1 decline, we cultured freshly isolated NOD and B6^g7 islets and detected elevated levels of soluble VTCN1 (sVTCN1) immunogenic fragments in medium conditioned by islets of prediabetic NOD mice (Fig. 1F).

To characterize VTCN1 expression in different types of pancreatic endocrine cells and to examine whether VTCN1 loss observed in islets of T1D-succeptible mice extends to human T1D patients, we analyzed pancreatic sections immunostained for VTCN1, insulin, and glucagon. In human subjects, both glucagon-expressing α cells and insulin-expressing β cells were VTCN1⁺ (Supplemental Fig. 1A), although the VTCN1 signal was generally higher in α cells. Likewise, in murine islets both α and β cells expressed VTCN1 (Supplemental Fig. 1B); however, contrary to humans, VTCN1 immunoreactivity in β cells was considerably stronger than in α cells. Evaluation of VTCN1 immunofluorescence in islets from a small cohort of T1D patients revealed that, similarly to NOD mice, the islet remnants from a representative T1D donor displayed substantially reduced VTCN1 levels relative to islets from a representative healthy subject (Supplemental Fig. 1C). Interestingly, a small number of undersized islets in the T1D patient were strongly positive for both insulin and VTCN1 (Supplemental Fig. 1D).

Taken together, these results suggest that proteolytic loss of VTCN1 in multiple islet cells accompanies the course of T1D development in both humans and mice.

Besides NRD1, pancreatic endocrine cells express two functional proprotein convertases, PCSK1 and PCSK2, which by in silico analysis using the Eukaryotic Linear Motif database (http://elm.eu.org/) are predicted to target VTCN1 cleavage. Hence, we compared mRNA expression of these enzymes as well as of PCSK1 inhibitor (PCSK-1N) in islets isolated from NOD and B6^g7 mice. A nonsignificant trend toward upregulation of all these molecules was found in NOD islets (Supplemental Fig. 2). Therefore, to ratify that the proteolytic clearance of VTCN1 from pancreatic islets is mediated by NRD1, we used two inhibitors known to inhibit NRD1, but not PCSK1 or PCSK2: 1,10-phenanthroline and bestatin (30–32) for ex vivo in-culture treatments of nondispersed VTCN1-low islets freshly isolated from 13- to 15-wk-old NOD mice. Both inhibitors stabilized islet-associated VTCN1, as confirmed by the increase of VTCN1 immunofluorescence (Fig. 2A), thus implying that NRD1 is indeed the mediator of VTCN1 shedding from islet cells. Similar ex vivo bestatin treatment of VTCN1-high islets from 6-wk-old either B6^g7 or NOD mice did not change islet VTCN1 levels in all animals analyzed (not shown).

Next, we examined the functional impact of NRD1 inhibitor–induced VTCN1 conservation for downregulation of anti-islet immune responses. For this study, dispersed into single-cell suspension VTCN1-low islets from 13-wk-old NOD mice were pretreated with either bestatin or vehicle for 24 h and then cocultured for 5 d (with inhibitor still present) with CFSE-labeled bulk spleen cells or purified T cells from B6.G9C8 mice. B6.G9C8 animals express rearranged α- and β-chains of TCR from insulin B chain (aa 15–23)–specific CD8⁺ T cell clone G9C8, and also its MHC H-2K^d restriction allele (25). Therefore, these mice are an excellent source of diabetogenic CD8⁺ T cells of a single G9C8 clonotype, which are capable of the direct β cell recognition (33). Whereas purification decreased the absolute numbers of dividing G9C8 T cells, the overall proliferative behavior of purified T cells in inhibitor-treated cocultures with islet cells (not shown) was very similar to that of the bulk G9C8 cells, which are illustrated in Fig. 2B and 2C. Specifically, bestatin-treated islet cells prompted a statistically significant reduction of proliferative capacity of cocultured G9C8 T cells when compared with vehicle-treated islets (Fig. 2B). Moreover, the production of the proproliferative cytokine IL-2 by G9C8 cells cocultured with bestatin-treated islets was also decreased (Fig. 2C). To control for the possibility that the decline in G9C8 T cell propagation is due to a direct anti-proliferative effects of bestatin, rather than the consequence of bestatin-induced VTCN1 preservation, we used Ag-independent activation of G9C8 cells by anti-CD3/CD28 beads. The presence of bestatin in these control reactions did not affect the propagation of G9C8 cells (Fig. 2B), indicating that stabilization of endogenous VTCN1 in islet cells and, in particular, in β cells, via NRD1 inhibition directly reduces proliferation of diabetogenic CD8⁺ T cells.

Previously we reported that a proteolysis-dependent decrease of VTCN1 levels on APCs from diabetes-prone NOD mice potentiates in vitro hyperactivation and proliferation of autoreactive T cells (22). To reveal whether such a mechanism contributes to the peripheral tolerance mediated by islet-resident APCs, presenting β cell Ags to diabetogenic CD4⁺ T cells, we used CFSE-labeled splenocytes isolated from NOD.BDC12-4.1 mice in our coculture experiments. NOD.BDC12-4.1 mice bear Rag1 recombinase deficiency coupled with the transgenic expression of both α- and β-chains of TCRs from highly pathogenic I-A^g7–restricted insulin B chain (aa 9–23)–reactive CD4⁺ T cell clones (34). Analysis of BDC12-4.1 cells cocultured, as described above, with bestatin- or vehicle-treated suspensions of islet cells from 13-wk-old NOD mice revealed a significant bestatin-induced reduction in proliferative rates of BDC12-4.1 T cells (Fig. 2D). The increase of APC load (via addition of exogenous VTCN1-low peritoneal Mϕs from 17-wk-old NOD mice) in these conditions markedly augmented proliferative responses of BDC12-4.1 T cells, whereas the significant bestatin-mediated reduction of BDC12-4.1 cell proliferation was kept intact (Fig. 2D).

Collectively, these results suggest that bestatin-induced NRD1 inhibition stabilizes VTCN1 on multiple types of islet cells (in particular, islet-resident APCs and β cells), which cocontribute to the peripheral tolerance via downregulation of anti-islet T cell responses.

To test whether the systemic NRD1 inhibition in vivo will alleviate VTCN1 loss on the islet cells and peripheral APCs and, consequently, alter the course of diabetes, 6-wk-old NOD female mice were injected (i.p.) daily with 10 mg/kg bestatin (Fig. 3A). Administration of bestatin moderately but significantly delayed (by 3 wk, on average) the onset of disease (Fig. 3B, left) and produced significant stabilization of VTCN1 levels on peritoneal Mϕs. Interestingly, this effect was much more prominent early in the treatment course (4 wk of administration, 10 wk of age) than at the latest time points (diabetic animals) (Fig. 3B, right). Similarly to the peritoneal Mϕs, VTCN1 levels on the islets of mice treated for 4 wk with bestatin showed a trend toward stabilization, even though this effect did not reach the threshold of statistical significance (Fig 3C, left panel, p = 0.1). Such an incomplete islet phenotype is most likely the consequence of the i.p. route of bestatin delivery, which produced higher bestatin concentrations affecting peritoneal Mϕs and lower intraislet concentrations. Alternatively, even partial islet VTCN1 stabilization in bestatin-treated mice coincided with the significant decrease in numbers of proliferating cells within islet infiltrates, as well as with the significant delay in insulitis formation (Fig. 3C, middle and right panels), pointing toward partial activation of peripheral tolerance within the islets.

Finally, the analysis of T cell populations in spleens and PLNs of bestatin-treated (4 wk of treatment) animals revealed that whereas inhibitor did not affect the relative sizes of both CD4⁺ and CD8⁺ T cell populations, it produced a significant elevation of splenic Tregs, as well as a statistically significant decrease in numbers of highly diabetogenic NRP-V7 mimotope–specific CD8⁺ T cells (35) (Fig. 3D). These later observations mimic the earlier reported effects of VTCN1-Ig treatments of autoimmune mice (10, 36) and thus validate VTCN1-stabilizing, self-tolerance–promoting effects of systemic bestatin administration.

As infiltration of multiple immune cells into peri-islet vicinity (insulitis) is the major T1D trait in both mice and humans (37–39), we assessed whether the loss of VTCN1 from islet cells is induced by insulitis formation or, on the contrary, precedes and even potentiates it. For this, we adoptively transferred splenocytes from acutely diabetic NOD mice into 15- to 20-wk-old B6^g7 recipients, allowed 20 d for islet infiltrates to progress, and then evaluated VTCN1 levels in recipients’ islets. Despite the formation of massive peri-islet infiltrates, recipient B6^g7 mice maintained stable islet VTCN1 levels, which did not differ from the ones observed in the islets of sham-transferred mice (Fig. 4A), implying that insulitis does not induce VTCN1 decline.

Approaching the paradigm of connection between VTCN1 loss and insulitic inflammation from an opposite angle, we analyzed VTCN1 immunofluorescence in islets of NOD-scid mice, which have a diabetes-susceptible NOD genetic background but lack autoimmunity due to the Prkdc^scid mutation–caused absence of functional T and B lymphocytes (40). Similarly to NOD mice, islets from NOD-scid animals displayed progressive age-dependent loss of VTCN1 (Fig. 4B), suggesting that the diabetes-prone NOD background predisposes for such loss independently of lymphocyte signaling.

These results raised the important functional question: What are the islet-shielding effects of endogenous VTCN1 presented by islet cells in vivo? Because by 15 wk of age B6^g7 and NOD-scid mice display high and low VTCN1 levels on their islets, respectively (Fig. 4C), we used 17- to 20-wk-old animals of these strains as recipients of adoptive transfers of NOD splenocytes. The VTCN1 levels on recipients’ islets were examined 3 wk after the transfers. Formation of insulitic lesions did not influence VTCN1 levels, which remained stably high in B6^g7 and stably low in NOD-scid recipient mice (Fig. 4C). We then evaluated the T cell proliferation within the insulitis area of adoptively transferred mice. For these experiments, 15-wk-old NOD-scid recipients were compared with B6^g7-Rag1^−/− mice, which, similarly to B6^g7 animals, maintain high VTCN1 levels on islets (not shown) and are lymphopenic equally to NOD-scid mice. The latter characteristic is important to eliminate the influence of the lymphopenic environment on homeostatic proliferation of transferred cells. The percentage of dividing CD3⁺ T cells within peri-islet infiltrates was significantly higher in NOD-scid recipient mice, correlating with their lower islet-associated VTCN1 levels (Fig. 4D). Moreover, the percentage of insulitis-free islets in B6^g7-Rag1^−/− mice was significantly higher when compared with NOD-scid mice (Fig. 4D, right panel). Therefore, VTCN1 on islet cells may function as an endogenous protective molecule, which attenuates proliferation of autoreactive islet-infiltrating T cells, shielding the pancreatic islets from diabetogenic autoimmunity.

Because mice of NOD background succumb to the progressive NRD1-mediated loss of islet-protective VTCN1, we explored a possibility to interfere with such loss. Accordingly, we asked whether stable introduction of B6^g7-originated myeloid cells, which display low internal NRD1 and steady VTCN1 levels (22), will be able to prevent and/or delay loss of VTCN1 from NOD islets. For this, we constructed chimeric mice by transplanting BM from B6^g7 donors into lethally irradiated NOD-EGFP recipients (B6^g7→NOD-EGFP). Reciprocal BM transfers from NOD-EGFP mice into B6^g7 recipients (NOD-EGFP→B6^g7) as well as control B6^g7→B6^g7 and NOD-EGFP→NOD BM transplantations were also performed. At the time of the transfers, all recipient mice were 6–7 wk of age, which is before the start of noticeable VTCN1 loss. Although it is impossible to assess the effects of lethal irradiation on islet VTCN1 levels, sublethal irradiation doses used in our adoptive transfer experiments did not affect either progressive loss of islet VTCN1 in NOD mice (not shown) or the stable islet-tethered VTCN1 levels in B6^g7 animals (Fig. 4C). Ten weeks after the BM transplantations, >90% of PBMCs in chimeric mice were of the donor’s origin (Supplemental Fig. 3). Stable engraftment of B6^g7 BM did not alleviate VTCN1 shedding in NOD islets, as VTCN1 immunofluorescence in B6^g7→NOD-EGFP chimeras was not significantly different from NOD-EGFP→NOD animals (Fig. 5A). This indicated that VTCN1 loss is predominantly an islet cell–autonomous process intrinsic for NOD islets. NOD-EGFP→B6^g7 chimeric mice, however, exhibited islet VTCN1 levels significantly lower than for B6^g7→B6^g7 chimeras, but nevertheless, significantly higher levels than for NOD-EGFP→NOD chimeric animals (Fig. 5A). This suggests that a long-term introduction of NOD-originated hematopoietic cells can induce partial destabilization of islet VTCN1 in B6^g7 animals through an extrinsic (islet nonautonomous) mechanism.

To investigate in more detail the nature of the systemic VTCN1-ablating stimuli produced by NOD BM-derived cells, we analyzed the phenotype of peritoneal Mϕs isolated from the constructed chimeric animals. In agreement with our previous findings demonstrating massive VTCN1 shedding in APCs of diabetes-prone NOD mice (22), NOD BM-originated Mϕs exhibited a significant VTCN1 decrease independently from the host’s environment that they developed and resided in (NOD-EGFP→B6^g7 and NOD-EGFP→NOD chimeras versus B6^g7→B6^g7 mice, Fig. 5B), confirming the inherent ability of NOD Mϕs to autonomously shed VTCN1. Furthermore, the similar decline in VTCN1 levels was observed in B6^g7-originated Mϕs from B6^g7→NOD-EGFP chimeras, indicating that nonhematopoietic cells of NOD mice induce strong acceleration of VTCN1 loss from B6^g7-derived cells. Interestingly, mRNA analysis showed increased VTCN1 expression only in Mϕs maturing in cellular environment of NOD, but not of B6^g7 BM recipients (Fig. 5C, left panel), pointing toward upregulation of VTCN1 transcription as a likely compensatory feedback mechanism activated in response to strong systemic induction of VTCN1 loss. Finally, analysis of NRD1 expression, which was found to be elevated in Mϕs from chimeric animals harboring any NOD component (either hematopoietic or nonhematopoietic; Fig. 5C, right panel), allowed us to conclude that: 1) augmentation of NRD1 expression is the characteristic trait of T1D-succeptible NOD background; 2) NRD1 is the most likely candidate for the sought-after systemic VTCN1-destabilizing factor; and 3) the increased NRD1-mediated shedding is the major cause of VTCN1 decline observed in both Mϕs and islet cells in conjunction with T1D development (Fig. 6).

VTCN1 was identified as a negative costimulatory molecule with primary function of downregulating immune responses by reducing T cell proliferation and cytokine production (5–7). VTCN1 gene expression is not limited to APCs, but it also occurs in several peripheral nonlymphoid tissues. Therefore, in addition to backing the anergic state of naive T cells populating lymphatic organs (41), VTCN1-induced signaling can also negatively modulate already stimulated T cells, contributing to induction of peripheral tolerance (21, 42). Such a variety of expression patterns and functional activities suggests that VTCN1 plays a far more important role in the regulation of immune response than was previously appreciated.

Defective VTCN1 expression and/or presentation was coupled with exacerbation of several autoimmune diseases harboring hyperactivation of autoreactive T cells, namely T1D, rheumatoid arthritis (RA), and multiple sclerosis (8, 9, 21, 43). Moreover, the delineation of defective VTCN1 presentation in RA-associated autoimmunity was directly related to the destruction of cell-associated VTCN1 and release of sVTCN1 fragments into the periphery (8). Extending these studies into the context of experimental and natural T1D, we recently demonstrated that high blood sVTCN1 concentrations in NOD mice and T1D patients are accompanied by almost complete loss of VTCN1 from the APC membranes (22). Based on these results, we suggested that sVTCN1 may serve as an early marker of clinical T1D in pediatric patients.

In contrast to the autoimmunity, VTCN1 levels on tumor cells and tumor-infiltrating APCs were found to be elevated in multiple neoplasms, providing for hypoactive antitumor T cell responses (44–47). Interestingly, such augmentation of VTCN1 in the context of cancer development was also combined with elevated peripheral blood sVTCN1, which was proposed as a potential prognostic marker for metastatic cancer spread and consequent poor outcome in cancer patients (19, 48). It seems, however, that such a similar high sVTCN1 phenotype is a result of different mechanisms governing VTCN1 metabolism with opposite outcomes, that is, inhibited T cell responses in cancer or hyperactivated ones in autoimmunity.

Because VTCN1 was found to be expressed in the endocrine pancreas (9, 15, 21), we started this study with an intent to examine the functionality of endogenous VTCN1 in islet cells during T1D development and to assess the natural VTCN1 capability to modulate autoimmune processes. We show that defective VTCN1 presentation from pancreatic islets, due to an increased shedding, precedes T1D development (Fig. 1). Surprisingly, despite the observed difference in the expression patterns between mice and humans, both α and β cells were found to be VTCN1⁺ (Supplemental Fig. 1). The drastic decrease of cell-associated VTCN1 levels detected in islets of human T1D patients was not limited to β cells, but, surprisingly, it affected several endocrine cell types (Supplemental Fig. 1C), suggesting the presence of a general defect in VTCN1 presentation. Moreover, the fact that VTCN1 loss in islets occurred despite elevated VTCN1 mRNA expression (Fig. 1B, 1C) indicates the involvement of posttranslational control mechanisms, such as protein degradation and/or shedding, in regulation of this process. Importantly, similarly to the islets, elevated VTCN1 gene expression was also unable to compensate VTCN1 protein loss in Mϕs from T1D-prone animals and human T1D patients (Fig. 5B, 5C) (22). Taken together, these observations suggest that a general proteolytic mechanism responsible for defective VTCN1 presentation on multiple cell types is tied with the initiation and/or progression of diabetogenic autoimmunity (Fig. 6).

Accordingly, we suggested the metalloproteinase NRD1 as the common VTCN1-shedding enzyme responsible for impaired VTCN1 presentation in multiple tissues (Fig. 6). Stabilization of VTCN1 by NRD1 inhibitors in isolated islets (Fig. 2A), supplemented by the involvement of NRD1 proteinase in VTCN1 proteolysis in Mϕs during T1D progression (22), clearly validates this suggestion. Moreover, application of the NRD1 inhibitor bestatin ex vivo restored the functionality of endogenous VTCN1 on NOD islets, as evidenced from the decreased proliferation and cytokine production of G9C8 CD8⁺ T cells and BDC12-4.1 CD4⁺ cells (Fig. 2D) in coculturing experiments (Fig. 2B–D). These results evidently stressed the importance of endogenous VTCN1 presentation on both intraislet APCs and β cells for the local control of autoimmunity. In agreement, several studies have also shown that ex vivo VTCN1 overexpression protects islet transplants/β cells from allograft rejection (20, 42, 49, 50). However, our results indicate that VTCN1 overexpression by itself will not address the shedding-dependent loss of functional VTCN1 protein. Therefore, development of better and more specific inhibitors of NRD1-directed VTCN1 proteolysis, as well as design of VTCN1-stabilizing agents of a different nature, should be an important direction for future studies. This conclusion is in complete agreement with the observed partially protective antidiabetic effects of continuous bestatin administration in NOD mice, which alleviated, to a certain extent, autoimmune responses via systemic decrease of NRD1 activity (Fig. 3).

Further extending our findings of endogenous VTCN1 downregulating anti-islet diabetogenic responses into an in vivo setting, we demonstrated that the proliferation of islet Ag-reactive T cells within insulitic lesions was reversely correlated with the levels of VTCN1 presented by the islet cells (Fig. 4C, 4D). This result is consistent with the previous observations that VTCN1 overexpression is protective against immune responses in two different systems: allograft survival of pancreatic islet transplants (20, 42) and in the AI4αβ/B6.H2^g7 severe diabetes mouse model (21). Moreover, the inflammatory influence of massive multicell insulitis, which is widely accepted to be the source of cytokines, growth factors, and other stimuli driving proliferation of diabetogenic T cells within islet infiltrates, was trumped by VTCN1-mediated inhibitory signaling, as intrainfiltrate T cells proliferated significantly better in low-VTCN1 presenting NOD islets in comparison with high-VTCN1 B6^g7 islets (Fig. 4D). Therefore, the intraislet coinhibition, delivered by proteolysis-uncompromised endogenous VTCN1, serves as a powerful local mechanism for self-protection against diabetogenicity.

Contrary to common expectations, our adoptive transfer experiments demonstrated that stability of VTCN1 protein on islet cells was independent from accumulation of peri-islet lymphoid infiltrates. VTCN1 levels were steadily high in heavily infiltrated islets of B6^g7 animals, but they displayed a gradual age-dependent decline in the infiltrate-free islets of NOD-scid mice (Fig. 4A–C). Hence, VTCN1 degradation originates at least in part within the islet cells of diabetes-prone animals, being therefore a cell-autonomous process characteristic for T1D-prone NOD background.

A more detailed investigation into the balance of VTCN1 loss, however, revealed that long-term engraftment of NOD BM cells into B6^g7 hosts jeopardized islet VTCN1 levels through an extrinsic, systemic mechanism (Fig. 5A). Additionally, the nonhematopoietic NOD cellular environment also induced accelerated VTCN1 loss in B6^g7 BM-originated Mϕs (Fig. 5B). Taken together, these data demonstrate that the persistent phenomenon of proteolytic impairment of functional VTCN1 presentation, detected in multiple cell types in conjunction with T1D susceptibility, is regulated in two synergistically acting manners: cell-autonomous and systemic. The metalloproteinase NRD1, which has been found in several cell types, including pancreatic islets (Fig. 1C, 1D) and immune cells (Fig. 5C), where its activity is reported to be required for Ag processing and generation of cytotoxic T lymphocyte epitopes (51), appears to fit both of these modes. NRD1 protein is mainly localized in the cytosol, but a significant proportion of the active enzyme is secreted through an unconventional secretory pathway and distributed on the cell surface (23). In agreement, our experimental data show upregulation of NRD1 gene expression in islets of diabetes-prone NOD mice (Fig. 1C), as well as in Mϕs from BM-chimeric animals harboring any NOD component (Fig. 5C). Moreover, elevation of NRD1 mRNA was accompanied by a gradual increase of NRD1 protein secretion from NOD islets, which paralleled the progression of T1D (Fig. 1D). Therefore, NRD1 is the most likely constituent to deliver both cell-intrinsic VTCN1 ablation (through its intracellular activity) and systemic extrinsic VTCN1 loss (by its proteolytic actions in an extracellular, secreted mode). In summary, our results open the door for new studies addressing an important quest to identify the genetic and environmental factors responsible for upregulation of NRD1 expression and enzymatic activity with respect to autoimmunity in general and T1D in particular. Our present and previous findings strongly indicate that such factors are shared between autoimmunity-prone NOD mice and a substantial cluster of T1D patients. The elevated serum sVTCN1 levels (8), combined with an increase in PBMC NRD1 expression in RA patients (52), provide an initial insight into this quest and hint that NRD1-dependent impairment of VTCN1-mediated negative costimulation is a general autoimmunity-associated pathway. In agreement, systemic NRD1 deficiency was recently reported to significantly reduce liver inflammation and protect mice from diet-induced nonalcoholic steatohepatitis (53). Our results imply a possibility that stabilization of VTCN1 in liver cells of these mice is a conceivable explanation for such a phenotype.

Interestingly, VTCN1 seems to mimic another endogenous mechanism of β cell protection, namely the programmed cell death-1 protein (PD-1)/programmed cell death–ligand 1 (PD-L1) pathway. It has previously been shown that NOD mice double-deficient for PD-L1 and PD-L2 as well as PD-1–deficient animals developed accelerated diabetes with 100% penetrance in both male and female mice (4, 54). In light of these studies, the generation of VTCN1-knockout mice on the NOD background would be an interesting future direction that would provide us with valuable information about the VTCN1 importance in diabetes development in comparison with the PD-1/PD-L1 pathway. Nevertheless, the existence of such redundancy in the negative costimulation signals on the periphery might be crucial for providing multiple levels of defense of pancreatic islets against autoimmune destruction.

Our present data have multiple implications for a variety of disease conditions. For example, contrary to T1D, characterized by strong VTCN1 diminishment from the surface of multiple cell types, in several cancers VTCN1 shedding is not associated with loss, but rather with an increase of membrane-tethered VTCN1. Possible explanations could include an intratumor loss/downregulation of NRD1 expression/activity or an overcompensatory excessive VTCN1 expression. Subsequently, the control of NRD1 functionality appears to be a very attractive target for future investigations aimed at the development of novel therapeutics against either certain cancers or autoimmune conditions. Finally, our data show that disrupted control of costimulation, evidenced by VTCN1 loss in both APCs and pancreatic islets, ultimately results in altered balance of control of immune responses.

We thank Drs. A. Chervonsky and V. Varnasi (Department of Pathology, University of Chicago) for the transgenic B6.G9C8 mice. We are also grateful to Brady Roby for critical editing of the manuscript and to Drs. Satoshi Nagata and Tomoko Ise (Sanford Research) for helpful discussions. We are deeply thankful to the staff of the South Dakota Lions Eye and Tissue Bank for help with human tissue specimen collection.

The authors have no financial conflicts of interest.