Immune Checkpoint-Bioengineered Beta Cell Vaccine Reverses Early-Onset Type 1 Diabetes

Abstract

Type 1 diabetes mellitus (T1DM) is a chronic autoimmune disease that results from autoreactive T cells destroying the insulin-producing pancreatic beta (β) cells. The development of T1DM is associated with the deficiency of coinhibitory immune checkpoint ligands (e.g., PD-L1, CD86, and Gal-9) in β cells. Here, we report a new translational approach based on metabolic glycoengineering and bioorthogonal click chemistry, which bioengineers β cells with coinhibitory immune checkpoint molecules that induce antigen-specific immunotolerance and reverse early-onset hyperglycemia. To achieve this goal, we devised a subcutaneous injectable acellular pancreatic extracellular matrix platform for localizing the bioengineered β cells while creating a pancreas-like immunogenic microenvironment, in which the autoreactive T cells can interface with the β cells.

Graphical Abstract

Insulin-dependent diabetes mellitus, also referred to as type 1 diabetes mellitus (T1DM), is a chronic autoimmune disease that results when autoreactive T cells destroy the insulin-producing pancreatic beta (β) cells.^{[1, 2, 3]} The condition leads to hyperglycemia and insulin deficiency.^{[1, 2, 3]} Once it progresses to the symptomatic dysglycemia stage, the islets often rapidly progress to total β cell loss within a year.^{[1, 4, 5]} Every year, more than 1 million new T1DM cases are diagnosed worldwide, and approximately half of these cases are diagnosed in adulthood.^[6] Most T1DM patients maintain their blood glucose levels through multiple daily insulin injections or insulin-pump therapy.^{[1, 2]} However, less than one-third of T1DM patients consistently achieve their target blood glucose levels. Despite the major advances in disease management and care, T1DM remains associated with a considerably higher probability of patients developing acute diseases (e.g., neuropathy, nephropathy, retinopathy, and cardiovascular disease) along with a higher premature death rate compared to the general population.^{[1, 2, 6]} Considerable interest has been expressed toward the development of new treatment strategies for delaying and reversing early-onset T1DM by regaining metabolic control because a substantial β cell mass is present at the early symptomatic stage. Several clinical trials performed in the recent years investigated the use of pro-insulin peptide-based vaccines to reverse early-onset hyperglycemia. The results have, however, been disappointing.^[7]

Immune checkpoints play key roles in regulating self-tolerance in the immune system.^[8] Cancer cells evade immune surveillance by stimulating coinhibitory checkpoint receptors in activated T cells,^[8] such as programmed cell death protein 1 (PD-1), cytotoxic T-lymphocyte-associated protein 4 (CTLA-4), and T cell immunoglobulin mucin 3 (TIM-3). The mutation or deficiency of inhibitory checkpoint molecules is associated with the development of autoimmune diseases.^{[4, 9]} Several recent studies have revealed that the β cell deficiency of coinhibitory immune checkpoint ligands (e.g., PD-L1 (a co-inhibitory ligand for PD-1),^{[4, 10, 11]} CD86 (a coinhibitory ligand for CTLA-4 in activated T cells),^[12] and galectin-9 (Gal-9, a coinhibitory ligand for TIM-3)^[13]) is associated with the T1DM development. In addition, the deficiency of these coinhibitory checkpoint ligands hampers the recruitment, maintenance, and functions of immunosuppressive regulatory T cells (T_reg) in the islets that are essential for maintaining immunotolerance.^{[4, 10]} Recent studies have demonstrated that the intravenous (i.v.) administration of β cells that genetically overexpress PD-L1 can reverse early-onset hyperglycemia in nonobese diabetic (NOD) mice by reducing islet-infiltrating cytotoxic T cells and upregulating the recruitment of islet-infiltrating T_reg.^{[4, 10]} However, the use of genetically engineered β cells requires substantial genetic manipulation, which is not only expensive, but also subject to considerable regulatory oversight. In addition, introducing more than one immune checkpoint molecule through a genetic engineering approach has not yet been fully established. Recent advances in metabolic glycoengineering^[14] and bioorthogonal click chemistry^[15] have enabled the facile chemical decoration of biomacromolecules on targeted cells without needing complex genetic engineering techniques. Here, we hypothesize that β cells engineered with high densities of coinhibitory immune checkpoint ligands can be used as a live-cell vaccines to induce an islet antigen-specific immune tolerance against autoreactive T cells and reverse early-onset hyperglycemia. Accordingly, we devised a subcutaneous (s.c.)-injectable acellular pancreatic extracellular matrix (PAN-ECM) scaffold to localize the bioengineered β cells while creating a pancreas-like immunogenic microenvironment, in which the autoreactive T cells could interface with the β cells (Figure 1a).

Figure 1.

Immune checkpoint ligand-engineered β cells induce antigen-specific immunotolerance and reverse T1DM. a) The scheme illustrates the bioengineering of a immune checkpoint ligand-functionalized β cell (top) and the mechanism of action of the s.c.-injected β cells (bottom). Functionalized β cells are prepared in two steps: first, native β cells are in vitro metabolically labeled with Ac₄ManNAz to create azide-modified β cells; and second, the DBCO-functionalized co-inhibitory ligands are conjugated to the azide-modified β cells through the strain-promoted azide-alkyne cycloaddition (SPAAC). The tri-functionalized β cells are combined with the ball-milled PAN-ECM to form the s.c.-injectable β cells. The PAN-ECM provides a pancreas-like tissue microenvironment for β cells. The tri-functionalized β cells induce antigen-specific immunotolerance by presenting antigens and binding to the autoreactive T cells via MHC and transmitting inhibitory signals to the T cell through the conjugated immune checkpoint ligands. MHC, major histocompatibility complex; AG, islet-specific antigen; TCR, T cell receptor. b) Representative FACS plots quantifying the level of surface PD-L1 of PD-L1-functionalized and PD-L1-dendrimer functionalized NIT-1 cells. c) CLSM images of unmodified NIT-1 cells and different mono- and tri-functionalized NIT-1 cells after staining with A488-labeled anti-PD-L1, PE-labeled anti-CD86, and PerCP-eFluor 710-labeled anti-Gal-9. d) Representative FACS plots quantifying the level of surface PD-L1, CD86 and Gal-9 of different mono- and tri-functionalized NIT-1 cells. e, PD-1, CTLA-4 and TIM3 expressions of 8.3 T cells after cultured with different immune checkpoint ligand(s)-functionalized NIT-1 cells in the presence of IGRP peptide at an E/T of 5:1 for 24 h. f) Impact of the conjugated immune checkpoint ligand (s) on antigen-specific T cell proliferation. Proliferation of 8.3 T cells after cultured with different immune checkpoint ligand(s)-functionalized NIT-1 cells in the absence of the IGRP peptide for 24 h, as assessed by CFSE-assay. g) Quantification of IFN-γ secretion by 8.3 T cells after incubation with unmodified and functionalized NIT-1 cells at an effector : target (E:T) ratio of 5:1 in the presence and absence of the IGRP peptide (15 h), as determined by the Luminex assay. (n = 3) h) Percentage of cytolysis of the unmodified and functionalized NIT-1 cells after incubation with expanded 8.3 T cells at E:T = 5:1 in the absence and presence of the IGRP peptide for 18 h, as determined by the CytoTox non-radioactive cytotoxicity assay (n = 6). The P values were determined using an unpaired Student’s t test. The error bars represent the standard error of the mean (s.e.m.).

To demonstrate how metabolic glycoengineering and bioorthogonal click chemistry allow the facile decoration of β cells with coinhibitory immune checkpoint ligands, we employed PD-L1 as a model ligand and tested two strain-promoted alkyne-azide cycloaddition (SPAAC) functionalization strategies on azide-modified NIT-1 cells (insulinoma cells established from islet β cells of NOD mice; Figure 1a; Figure S1, Supporting Information). Although we did not investigate in this pilot study, β cells can be bioengineered from human stem cell-derived β cells^[16] for reversion of T1DM in human. Azide-modified NIT-1 cells were obtained by in vitro culturing NIT-1 cells in the presence of N-azidoacetylmannosamine-tetraacylated (Ac₄ManNAz; Figure S1a, Supporting Information). During the Ac₄ManNAz metabolism, ManNAz was incorporated into mucin-type O-linked glycoproteins on the cell membrane of the NIT-1 cell membrane.^[14] The presence of azide (N₃) group on the surface of the modified NIT-1 cells was confirmed by labeling with dibenzocyclooctyne (DBCO)-functionalized Alexa Fluor 488 (A488) (Figure S1a, Supporting Information). We then conjugated PD-L1 onto the azide-modified NIT-1 cells through either a monovalent DBCO ligand or a multivalent DBCO-functionalized polyamidoamine dendrimer anchor (Dend; Figure 1b; Figures S2-S4, Supporting Information) through SPAAC at physiological conditions. The target degree of functionalization was 10 µg of the functionalized PD-L1 per million cells (Figure S1b, Supporting Information). For quantification, Texas Red (TR)-labeled PD-L1 was utilized in the functionalization study. Consequently, one million functionalized NIT-1 cells were found to contain 1.4 µg of TR-PD-L1-DBCO or 4.4 µg of TR-PD-L1-Dend (Figure S5, Supporting Information). The higher conjugation efficiency recorded for the dendrimer conjugation strategy can be explained by the multivalent effect offered by the DBCO-functionalized dendrimer anchor (14 conjugated DBCO ligand per dendrimer after being functionalized with one PD-L1 ligand; Figure S3, Supporting Information). Flow cytometry (FACS; Figure 1b; Figure S6a, Supporting Information) and confocal fluorescence microscopy (CLSM; Figure S6b, Supporting Information) revealed that the as-prepared PD-L1-Dend-functionalized NIT-1 cells contained approximately 26-fold more surface PD-L1 than those functionalized through the monovalent DBCO ligand. Time-dependent FACS and CLSM studies demonstrated that the PD-L1 expression levels of both types of PD-L1-functionalized NIT-1 cells gradually declined after conjugation (Figure 1b; Figure S6, Supporting Information) due to mitotic division and glycan/membrane recycling. The PD-L1 expression of the PD-L1-DBCO-functionalized NIT-1 cells decreased to the background level within 3 days after functionalization, whereas the PD-L1-Dend-functionalized NIT-1 cells maintained a persistent level of PD-L1 expression for at least 5 days (Figure 1b; Figure S6, Supporting Information). These results indicate that the multivalent dendrimer anchor provided a more effective conjugation and prolonged the retention of the conjugated coinhibitory checkpoint molecule on the NIT-1 cell surface. This phenomenon is likely due to multivalent conjugation leading to the aggregation of surface-conjugated molecules.^[17] In vitro toxicity studies confirmed that neither of the metabolic labeling methods affected the proliferation of the functionalized NIT-1 cells (Figure S7, Supporting Information). Therefore, we used the dendrimer-based conjugation strategy to engineer CD86- and Gal-9-mono-functionalized NIT-1 cells, along with PD-L1/CD86/Gal-9-tri-functionalized NIT-1 cells. It is estimated that all 4 different mono- and tri-functionalized NIT-1 cells contained approximately 4.4 µg of immune checkpoint ligand(s) that were conjugated to the NIT-1 cells through the DBCO-functionalized dendrimer. The CLSM and FACS studies confirmed the successful decoration of NIT-1 cells with coinhibitory immune checkpoint ligands (Figure 1c,d; Figure S8-S9, Supporting Information).

To test the ability of different bioconjugated coinhibitory immune checkpoint ligands to inhibit islet antigen-specific T cell activation and killing, we performed islet-specific glucose-6-phosphatase catalytic subunit related protein (IGRP)-specific cytotoxic T cell (8.3 T cell) immunosuppression and cytotoxicity assays with different functionalized NIT-1 cells in the presence of islet-specific IGRP_206–214 peptide. The exogenous IGRP_206–214 peptide directly inserts into the MHC class I complex in the NIT-1 cells^[18] and presents it to the IGRP-specific 8.3 cytotoxic T cells, which have specific T cell receptors that recognize this antigen and kill the IGRP-presenting cells.^{[4, 19]} Similar to previous studies in cancer models,^[20] the presence of active functionalized checkpoint molecules on the NIT-1 cells upregulated the expressions of the corresponding checkpoint molecules on the 8.3 T cells (Figure 1e; Figure S10, Supporting Information). In the presence of the exogenous IGRP_206–214 peptide, 8.3 T cells underwent a rapid proliferation (Figure 1f,g; Figures S11, Supporting Information) and showed strong cytotoxicity (≈ 85%; Figure 1h; Figure S12, Supporting Information) against the targeted unmodified NIT-1 cells at an effector-to-target (E/T) ratio of 5:1. No significant specific lysis was observed in the absence of the exogenous IGRP_206–214 peptide (Figures S11, Supporting Information). The PD-L1-Dend-functionalized NIT-1 cells significantly inhibited the T cell activation (Figure 1f), reduced the IFN-γ in supernatant by approximately 50% after co-cultured for 15 h (Figure 1g), and decreased IGRP-specific lysis by approximately three folds compared with the unmodified NIT-1 cells (Figure 1h; Figure S12, Supporting Information). By contrast, PD-L1-DBCO-functionalized NIT-1 cells only slightly reduced the specific lysis (Figure 1g; Figure S12, Supporting Information). Both CD86-Dend- and Gal-9-Dend-mono-functionalized NIT-1 cells also significantly inhibited the 8.3 T cell proliferation (Figure 1f), reduce IFN-γ in tissue co-culture supernatant (Figure 1g) and IGRP-specific lysis (Figure 1h; Figure S12, Supporting Information), but they were 62 – 85% less effective than the PD-L1-Dend-functionalized NIT-1 cells (Figure 1f,g,h). The lower inhabitation efficiency recorded for the CD86-Dend-functionalized NIT-1 cells can be explained by competing for binding between the co-inhibitory receptor CTLA-4 and co-stimulatory receptor CD28 with the conjugated CD86, whereas PD-L1 only binds to the co-inhibitory receptor PD-1.^[21] Similar to previous studies, membrane-bound Gal-9 is less effective to inhibit T cell activation because the activation of Gal-9/TIM-3 co-inhibitory immune checkpoint pathway involves internalization of Gal-9/TIM-3 complex.^[22] The tri-functionalized NIT-1 cells showed comparable T cell inhibition efficiency as the PD-L1-Dend-mono-functionalized NIT-1 cells because the PD-L1/PD-1, CD86/CTLA-4, and Gal-9/TIM3 immune checkpoint pathways synergistically inhibit T cell activation,^[23] even though it only has one-third of the conjugated PD-L1 compared with the PD-L1-Dend-mono-functionalized NIT-1 cells.

We reasoned that the prolonged retention of the bioconjugated coinhibitory immune checkpoint ligands on the NIT-1 cell membranes could improve their ability to anergize the autoreactive T cells and, consequently increase its ability to reverse recent-onset diabetes. Therefore, we intrapancreatically administrated TR-labeled PD-L1-functionalized NIT-1 cells to healthy non-diabetic NOD mice to investigate the in vivo retention of differently bioconjugated PD-L1 (Figure S13, Supporting Information). The ex vivo fluorescence imaging performed 1 week after the administration indicated that the TR-labeled PD-L1-Dend-functionalized NIT-1 cell grafts contained 4.5-fold greater TR-labeled PD-L1 compared to cells conjugated through the DBCO ligand (Figure 2a(i),(ii); Figure S13a, Supporting Information). An immunohistological (IHC) study confirmed that the PD-L1 expression levels of the TR-labeled PD-L1-Dend-functionalized NIT-1 cell grafts were, on average, three times higher than those of unmodified NIT-1 cell grafts (Figure 2a(iii); Figure S13, Supporting Information).

Figure 2.

Multivalent dendrimer anchor enhances the retention of the bioconjugated PD-L1 on NIT-1 cells and increases the efficacy of engineered NIT-1 cells in reversing T1DM. a) (i) Quantification of TR-labeled PD-L1 in the intrapancreatic parenchymal-injected (IPPI) NIT-1 cells 1 week after inoculation, as determined by ex vivo fluorescence imaging: each mouse was inoculated with 2×10⁶ cells. Rt, percentage of the inoculated TR-labeled PD-L1 retained on the NIT-1 cell graft. (ii, iii) Representative (i) CLSM and (ii) IHC images of the NIT-1 cell grafts 1 week after inoculation. E_PDL1, represents the relative active PD-L1 expression quantified from the representative IHC images (n = 4 mice per group). The P values were determined using an unpaired Student’s t test. The error bars represent the s.e.m. b) (i-ii) Blood glucose of the NOD mice before and after the intrapancreatic administration of the unmodified or PD-L1-functionalized NIT-1 cells (2×10⁶ cells per mouse). (iii) Survival curves of the NOD mice after treatment with unmodified or PD-L1-functionalized NIT-1 cells. NOD mice received treatment within 4 days of diabetes onset (n = 6 mice for the non-treatment group, and n = 9 mice for all three treatment groups). The error bars denote the mean ± s.e.m.

We next tested whether the slower PD-L1 detachment rate affected the ability of the PD-L1-functionalized NIT-1 cells to reverse early-onset hyperglycemia (blood glucose > 250 mg/dl; Figure 2b). We intrapancreatically administered different types of PD-L1-functionalized NIT-1 cells to recent-onset hyperglycemic NOD mice (Figure 2b(i); Figure S14, Supporting Information) to maximize the interface and between the functionalized NIT-1 cells the autoreactive T cells. Approximately two-thirds of the mice treated with the PD-L1-Dend-functionalized NIT-1 cells showed an initial response to the treatment (i.e., became normoglycemic for at least 3 weeks after treatment; Figure 2b(ii); Figure S14a,b, Supporting Information). In addition, the treatment significantly prolonged the overall survival (Figure 2b(iii); Figure S14c, Supporting Information). However, only one-third of the mice treated with the PD-L1-DBCO-functionalized NIT-1 cells showed an initial response to the treatment (Figure 2b(i)-(ii); Figure S14a,b Supporting Information). The weaker therapeutic efficiency of the PD-L1-DBCO-functionalized NIT-1 cells is due to less PD-L1 conjugated to the NIT-1 cells and a faster detachment rate of the conjugated PD-L1-DBCO ligand. Moreover, the treatment only slightly prolonged survival compared to those treated with the non-functionalized NIT-1 cells and unconjugated PD-L1 (Figure 2b(iii); Figure S14c, Supporting Information).

Guided by these findings, we performed a correlative study to compare the efficiencies of different coinhibitory immune checkpoint ligand-functionalized NIT-1 cells in reversing new-onset hyperglycemia. In a protocol similar to the earlier in vivo therapeutic study, the functionalized β cells were intrapancreatically administered to allow a direct interface with the autoreactive T cells (Figure 3a(i)). As compared with the earlier efficacy study, three-fourth of the early-onset diabetic mice treated with the PD-L1-Dend-functionalized NIT-1 cells partially responded to the treatment. More than half of the treated mice reversed to a normoglycemic state for at least 50 days after the treatment (Figure 3a(i)–(iii); Figure S15, Supporting Information). Although CD86^[12] and Gal-9^[13] play critical roles in regulating immunotolerance, most early-onset hyperglycemic mice did not respond well to treatments with the CD86- and Gal-9-functionalized NIT-1 cells (Figure 3a(i)–(ii); Figure S15, Supporting Information). The treatment with the CD86-functionalized NIT-1 cells only slowed the progression of hyperglycemia and prolonged the overall survival by 18 days (Figure 3a(ii)–(iii); Figure S15, Supporting Information).

Figure 3.

Intrapancreatic administration of the PD-L1-, CD86- and Gal-9-tri-functionalized β cells reverses T1DM and limits insulitis in NOD mice. a)(i,ii) Blood glucose levels of NOD mice after intrapancreatic administration of the unmodified or mono-/tri-functionalized NIT-1 cells (2×10⁶ cells per mouse). (iii) Survival curves of NOD mice after IPPI with different functionalized NIT-1 cells; all groups received treatment within 4 days of diabetes onset (n = 7 mice for the non-treatment group and the control group administered with unmodified NIT-1 cells, n = 8 for all other treatment groups). The error bars denote the mean ± s.e.m. b) Representative immunohistochemical staining of pancreatic sections from control and treatment groups 3 weeks after diabetes onset: (i) hematoxylin and eosin (H&E), (ii) CD3/insulin, and (iii) CD4/FoxP3. c)(i) Distribution of insulitis scores of a given set of non-treatment and treatment group mice. The islets were scored on the following scale: score 0: no infiltration; score 1: peri-insulitis, score 2; infiltrative insulitis of < 50% of the islet; score 3: infiltrative insulitis in > 50% of the islet. Islets were analyzed blinded and was given a score of between 0 and 3. (ii) Percentage of FoxP3⁺ CD4⁺ T cells. (n = 5 mice for all non-treatment and treatment groups, except n = 4 mice for the IPPI of the unmodified NIT-1 cells). Five 5-µm thick representative sections, which were 350 µm apart were analyzed for each mouse.

We then investigated whether the combination of PD-L1, CD86, and Gal-9 could reverse newly onset hyperglycemia. The same amount of NIT-1 cells co-functionalized with PD-L1, CD86, and Gal-9, or a 1:1:1 combination of three different mono-functionalized NIT-1 cells was intrapancreatically administered. Unexpectedly, only a quarter of the hyperglycemic mice treated with the combination of the mono-functionalized NIT-1 cells showed initial treatment responses (i.e., diabetic-free two weeks post-treatment) to the treatment and achieved a long-term survival (Figure 3a(i)–(iii); Figure S15, Supporting Information). Seven out of eight of the hyperglycemic mice treated with the tri-functionalized NIT-1 cells showed initial treatment response (i.e., diabetic-free two weeks post-treatment) (Figure 3a(i); Figure S15, Supporting Information), whereas only five out of eight of the mice treated with PD-L1-Dend-mono-functionalized NIT-1 cells demonstrated an initial treatment response. Half of the treated mice were diabetic-free for at least 50 days (Fig. 3a(i)) and achieved a long-term survival (Figure 3a(iii); Figure S15, Supporting Information). Although the survival benefits of the tri-functionalized NIT-1 cells were comparable to those of the PD-L1-functionalized NIT-1 cells (p = 0.9648), the tri-functionalized cells only contained one-third of the conjugated PD-L1. In addition, the mice treated with the tri-functionalized NIT-1 cells showed higher initial response rate than those treated with the PD-L1-functionalized NIT-1 cells (Figure 3(a)(i)–(iii); Figure S15, Supporting Information). in in vivo In contrast to the in vitro cell immunosuppression and cytotoxicity assays, the in vivo correlative study demonstrated that decorating co-inhibitory immune checkpoint ligands in close proximity to the tri-functionalized NIT-1 cells (7 out of 8 showed an initial treatment response, MS = 81 days) allowed a more effective and simultaneous coinhibition of multiple immune checkpoint pathways and induced a more robust antigen-specific T cell exhaustion than the combination of three mono-functionalized NIT-1 cells (2 out of 8 demonstrated an initial treatment response, MS = 39 days; Figure 3a (i),(iii)).

Histological examination of the endogenous pancreas showed an increased number of intact islets in the NOD recipients of the PD-L1/CD86/Gal-9-trifunctionalized NIT-1 cells, and a lower insulitis^[24] than NOD recipients of the unmodified NIT-1 cells (Figure 3b,c; Figure S16, Supporting Information). Although the NOD recipients of PD-L1-Dend-functionalized versus tri-functionalized NIT-1 cells retained comparable numbers of islets (47 ± 6 versus 54 ± 5), the number of intact islets was reduced and the insulitis score increased in the NOD recipients of the latter (Figure 3b,c; Figure S16, Supporting Information). Interestingly, after the transfer of tri-functionalized NIT-1 cells, the islets contained 31% more FoxP3-expressing CD4⁺ T cells than those of NOD mice receiving unmodified NIT-1 cells (Figure 3b,c; Figure S17, Supporting Information). The absence of insulin staining by transferred NIT-1 cells (at 3 weeks after injection) suggested that remission was retained by insulin production by endogenous β cells, which was itself correlated with the insulitis reduction (Figure S18, Supporting Information).

Recognizing that intrapancreatic administration of engineered cells is difficult for clinical translation, we examined an engineered s.c. injectable platform using decellularized pancreatic extracellular matrix (PAN-ECM) scaffold. The PAN-ECM scaffold provides the necessary organ-specific microenvironment for NIT-1 cells to persist^[25]. The PAN-ECM scaffold was isolated from healthy murine pancreata through a spin-decellularization method.^[26] The decellularized and delipidized PAN-ECM was lyophilized and ball-milled before further use (Figure S19, Supporting Information). Proteomic analysis revealed that the spin-decellularization protocol preserved the physiological levels of pancreatic matrisomes and non-matrisome proteins (Figure 4a; Table S1, Supporting Information), which are important for guiding the cell migration, stimulating the cell proliferation, and modulating the cellular response.^[27] An in vitro study revealed that the tri-functionalized NIT-1 cells proliferated and spontaneously formed three-dimensional spheroid colonies with the PAN-ECM in a serum-free culture medium (Figure 4b; Figure S20, Supporting Information). In contrast, the NIT-1 cells did not survive in the serum-free culture medium under the same in vitro culture conditions (Figure S21, Supporting Information). An ex vivo study on early-onset hyperglycemia NOD mice confirmed that autoreactive T cells spontaneously engaged the s.c.-inoculated cell-free PAN-ECM (Figure 4c; Figure S22, Supporting Information). To demonstrate that the PAN-ECM can improve the retention of s.c.-administered β cells, we s.c. inoculated the carrier-free CFSE-labeled NIT-1 cells and CFSE-labeled NIT-1 cell-embedded PAN-ECM into a site close to the pancreatic lymph nodes in healthy NOD mice (Figure 4d; Figure S22, Supporting Information). The ex vivo fluorescence imaging performed 1 week after the inoculation confirmed that the CFSE-labeled NIT-1 cell embedded PAN-ECM group retained more CFSE-labeled NIT-1 cells at the injection site (Figure 4d; Figure S23, Supporting Information). In contrast, no carrier-free CFSE-labeled NIT-1 cells were located at the injection site (Figure 4d; Figure S23, Supporting Information).

Figure 4.

Subcutaneous injection of functionalized β cells reverses T1DM. a) Volcano plot (left) showing a quantitative comparison between native and de-celled murine pancreata. The green rectangle encompasses the proteins considered to be retained in the de-celled samples (fold change > 1). The table (right) summarizes the matrisome proteins retained in the PAN-ECM (n = 3 biological replicates). b) Representative field emission scanning electron microscopy images of the (i) ball-milled PAN-ECM, (ii) NIT-1 cells cultured with low concentration of the ball-milled PAN-ECM (20 µg/well), and (iii) NIT-1 cells cultured with a high concentration of the ball-milled PAN-ECM (40 µg/well) in serum-free cell culture media. NIT-1 cells were seeded at a density of 4×10⁴ cells per well. The collagen from the ball-milled PAN-ECM is easily visualized (in the absence of cells). The cells grew around the exposed collagen fibers as the PAN-ECM concentration increased. c) Representative H&E and CD3/insulin staining of sections of cell-free PAN-ECM 7 days after inoculation of pre-diabetic and diabetic NOD mice. No insulin staining was observed in either cell-free PAN-ECM specimens. d) Quantification of different formulations of the CSFE-labeled NIT-1 cell graft 1 week after injection. Cells were s.c. injected near the pancreatic lymph nodes, as determined by ex vivo fluorescence imaging. The retention study was performed in non-diabetic NOD mice (n = 5 mice per group). e) (i-ii) Blood glucose of diabetic NOD mice after s.c. injection of different tri-functionalized NIT-1 cell formulations: the tri-functionalized NIT-1 cells were s.c. injected near the pancreatic lymph nodes, and a booster was administered 14 days later. (iii) Survival curves of diabetic NOD mice after s.c. treatment with the tri-functionalized NIT-1 cells (n = 5 mice for the non-treatment group, n = 8 for all treatment groups). f) Representative immunohistochemical staining: (i) H&E, (ii) CD3/insulin, and (iii) CD4/FoxP3 staining in the serial pancreatic sections from non-treatment and treatment group mice 3 weeks after the onset of diabetes. g) Insulitis scores (i), and percentage of FoxP3⁺ CD4⁺ T cells (ii) for non-treatment (n=3) and treatment groups (n=5).

We next investigated whether the PD-L1/CD86/Gal-9-tri-functionalized NIT-1 cell-embedded PAN-ECM formulation could be used as a therapeutic vaccine to reverse early-onset hyperglycemia. We s.c. administered the β cell vaccine in a site near the pancreatic lymph nodes in recent-onset hyperglycemic NOD mice (Figure 4e). A booster was administered 2 weeks after the initial treatment (Figure 4e). All hyperglycemic mice treated with the tri-functionalized NIT-1 cell-embedded PAN-ECM vaccine demonstrated a complete initial response, with approximately 60% reversing to normoglycemia for more than 50 days after the initial treatment (Figure 4e(i)–(ii); Figure S24, Supporting Information). In addition, more than 60% of the treated mice achieved a long-term survival (Figure 4e(iii); Figure S24, Supporting Information), whereas the median survival of the non-treated mice was only 39 days (Figure 4(iii); Figure S24, Supporting Information). The control studies confirmed that the s.c. administration of the carrier-free tri-functionalized NIT-1 cells and non-functionalized NIT-1 cell-embedded PAN-ECM did not trigger significant immune responses in the course of the hyperglycemia reversion (Figure 4(i)–(iii); Figure S24, Supporting Information). Most of the early-onset hyperglycemic NOD mice also showed an initial response to a single s.c. injection of the tri-functionalized NIT-1 cell-embedded PAN-ECM formulation, but the response rate reduced by half, and the median progression-free survival time significantly decreased from 68 days to 18 days (Figure S25, Supporting Information).

In the histological examination, the NOD mice receiving tri-functionalized NIT-1 cell-embedded PAN-ECM exhibited 2.7-fold more intact islets than the untreated group, and a lower insulitis score (Figure 4f,g; Figure S26, Supporting Information). Furthermore, the NOD mice receiving tri-functionalized NIT-1 cells embedded in PAN-ECM exhibited a 60% more in FoxP3⁺ CD4⁺ T cells in the endogenous islets than NOD recipients of the unmodified NIT-1 cells (Figure 4f,g). Histological analysis one week post-inoculation confirmed that CD3⁺ T cells infiltrated the graft of tri-functionalized NIT-1 cells in the PAN-ECM scaffold (Figure S27, Supporting Information). Furthermore, a temporal analysis indicated no proliferation of the NIT-1 cells after s.c. inoculation (Figure S28, Supporting Information).

In this study, we demonstrated that β cells containing high levels of immune checkpoint ligands (e.g., PD-L1, CD86, and Gal-9) can be produced through metabolic glycoengineering and bioorthogonal click reaction. The co-inhibitory immune checkpoint ligand-functionalized β cells effectively induced islet antigen-specific T cell exhaustion in vitro. To the best of our knowledge, this is the first study showing that the intrapancreatic administration of PD-L1/CD86/Gal-9-tri-functionalized NIT-1 cells can reverse early-onset hyperglycemia in NOD mice. A new s.c.-injectable vaccine based on PD-L1/CD86/Gal-9-tri-functionalized NIT-1 cell-embedded PAN-ECM was successfully engineered and used to reverse early-onset hyperglycemia. We showed that the acellular PAN-ECM functioned as a scaffold for the localization of the functionalized β cells and it also regenerated an immunogenic pancreas microenvironment in which β cells could interface with autoreactive T cells to evoke strong antigen-specific effector T cell inhibition. In contrast to the use of PD-L1-overexpressed platelets to reverse early-onset T1DM through non-specific binding to inflamed pancreas tissue,^[28] our immune checkpoint ligands-functionalized β cell vaccine platform can directly induce islet-antigen specific T cell exhaustion without affecting other healthy organs. The key advantage of this whole cell vaccine platform is to allow the functionalized β cells to present a broad range islet-specific antigens to the autoreactive T cells without prior known of a specific antigen that causing T1DM. Thus, the live-cell vaccine platform reported herein may be applicable to other autoimmune diseases. In a more general sense, this platform may be useful for generating a broad range of antigen-specific cytotoxic T cell responses, from immunity to tolerance.