Islet-Expressed CXCL10 Promotes Autoimmune Destruction of Islet Isografts in Mice With Type 1 Diabetes

Treatment of type 1 diabetes (T1D) with exogenous insulin is life saving but cannot completely prevent secondary complications. At the beginning of this millennium, the transplantation of islets of Langerhans had become a huge hope for patients with diabetes, and several clinical trials have been conducted since. It provides T1D patients with functional islets, resulting in a more stable glycemic control than that achieved with exogenous insulin therapy and, thereby, avoids severe hypoglycemia and finally reduces long-term complications (1). Most clinical trials followed the “Edmonton Protocol,” according to which purified islets are injected into hepatic portal vein, settle in the liver, and start insulin production (2). A multicenter trial in North-America and Europe (3) that followed the Edmonton Protocol has confirmed that islet transplantation was indeed successful in establishing insulin independence. After the initial drawback, during which most subjects lost insulin independence within a year (3), and long-term function of the islet grafts could be neither guaranteed nor predicted (4), in subsequent clinical trials (5) the rate of insulin independence 3 years after undergoing transplantation has gradually increased from 27% (years 1999–2002) up to 44% (years 2007–2010). Further, a recent long-term follow-up study (6) in the U.S. demonstrated continued islet function in all seven patients for more than 10 years. An aggressive corticosteroid regimen has been replaced with a combination of milder immunosuppressive drugs, such as sirolimus, tacrolimus, and daclizumab, but many side effects of immunosuppression, such as neutropenia, pneumonia, mouth ulcerations, fever, and worsening genital herpes, still persist (3). The decreased rejection rate observed in the last few years can be attributed partly to changes in the immunosuppressive regimen (7). Thereby, maintenance immunosuppressive therapy was supplemented by anti-inflammatory biologic agents, such as anakinra and etanercept (8).

Graft attrition and rejection are likely due to recurrent alloreactivity in which maintenance immunosuppression fails to abolish the immune response but also to the persistent autoimmunity that was initially mediating the destruction of islets in the pancreas. A hallmark of graft rejection is the extensive infiltration of leukocytes, and it has been suggested that islet function might be also impaired by islet stress. Thus, many biologic agents are being considered as alternative immunomodulatory interventions not only to prevent alloreactive rejection and/or autoimmune destruction of the islet grafts but also to relief islet stress originating from the presence of large cellular infiltrates (9). Data from various groups indicate that CXC-motif receptor 3 (CXCR3) chemokines play an important role in allograft rejection (10–13). Therapeutically, the neutralization of CXCR3 chemokines delays cardiac and skin allograft rejection in MHC-mismatched mice (14–16), and blockade of CXCR3 by anti-CXCR3 antibodies prolongs cardiac and islet allograft survival, in particular when coadministered with the mammalian target of rapamycin (mTOR) inhibitor rapamycin (17). Importantly, the blockade of CXC-motif ligand 10 (CXCL10) with neutralizing antibodies prolonged heart allograft survival in fully MHC-mismatched mice (18). One important difference between many of the above-mentioned transplantation experiments and the situation occurring in T1D patients is the lack of a defined autoimmune setting. In most islet graft models, allogeneic transplantations are performed by transferring islets into MHC-mismatched recipient mice rendered diabetic by the administration of the β-cell toxin streptozotocin (STZ). T1D patients also receive islets from non–HLA-matched donors. However, in addition the patients exhibit an activated immune system with autoreactivity to β-cell autoantigens that are also present in the islet graft. Thus, besides the allogeneic differences, the rejection of islet grafts in T1D patients might be heavily influenced by an “autoimmune” type of destruction. We therefore investigated the role of CXCL10 in an autoimmune destruction setting using the RIP-LCMV mouse model for T1D. Transgenic RIP-LCMV-GP mice express the glycoprotein (GP) of the lymphocytic choriomeningitis virus (LCMV) under control of the rat insulin promoter (RIP) in the β-cells and develop T1D within 10–14 days after LCMV infection (19,20). Previously, we have demonstrated that CXCL10 plays an important role in the attraction of aggressive T cells to the islets and the subsequent destruction of β-cells in the pancreas (21). Importantly, neutralization of CXCL10 in combination with the administration of an anti-CD3 antibody fragment induced a persistent T1D remission in RIP-LCMV mice as well as diabetic NOD mice (22). Here, we demonstrate in the RIP-LCMV model that islet isografts indeed express CXCL10 in situ and that an absence of CXCL10 in the islet isograft results in reduced autoimmune destruction and an accumulation of regulatory T cells (Tregs) in the surrounding cellular infiltrates. In addition, the administration of a neutralizing anti-CXCL10 antibody also reduced the autoimmune destruction rate. Our data indicate that, similar to the autoimmune destruction of β-cells in the pancreas, CXCL10 plays an important role in the autoimmune destruction of islet isografts. Neutralization of CXCL10 might therefore constitute a possible therapeutic intervention to prolong islet graft survival in T1D patients.

Generation and screening by PCR of RIP-LCMV-GP mice (H-2^b) were conducted as previously described (19,20). CXCL10-deficient mice (H-2^b) were generated as previously described (23) and have been backcrossed to C57BL/6 mice for >10 years. C57BL/6 and BALB/c mice were purchased from Envigo. CXCL10-deficient RIP-LCMV-GP (RIP-LCMV-GPxCXCL10^−/−) mice have been generated by crossing CXCL10-deficient mice with RIP-LCMV-GP mice. To generate H-2^b/H-2^d mixed-background mice, RIP-LCMV-GP (H-2^b) and RIP-LCMV-GPxCXCL10^−/− (H-2^b) mice were crossed with BALB/c (H-2^d) mice. The resulting F1 generation (RIP-LCMV-GP^+/−; CXCL10^+/−) was intercrossed to re-establish CXCL10 deficiency. Animal experiments have been approved by the local Ethics Animal Review Board, Darmstadt, Germany (V54–19c20/15-F143/66).

Female RIP-LCMV-GP mice were either treated intraperitoneally with a single dose of STZ (Sigma-Aldrich) at 250 mg/kg in 10 mmol/L citrate buffer, pH 4.5, or infected with 10⁴ plaque-forming units LCMV. LCMV-Armstrong was produced as described previously (20). Blood glucose (BG) measurements were performed using a dynaValeo glucometer (dynamiCARE). Mice with nonfasting BG levels >400 mg/dL for two consecutive daily measurements were considered to be suitable for islet engraftment.

Pancreata were harvested from wild-type C57BL/6, CXCL10^−/−, RIP-LCMV-GP, RIP-LCMV-GPxCXCL10^−/− (all H-2^b), RIP-LCMV-GP (mixed H-2^b/H-2^d), or RIP-LCMV-GPxCXCL10^−/− (mixed H-2^b/H-2^d) female donors. Islets were isolated by collagenase P (Roche) injection and Ficoll gradient (Ficoll-Paque Plus; GE Healthcare) separation (24). Islets were then hand picked under a stereo microscope. Prior to transplantation, islets were cultured (RPMI plus GlutaMAX; ThermoFisher Scientific) in 10% FBS (Biochrom) and 1% pen/strep (ThermoFisher Scientific) at 37°C and 5% CO₂ for 24–48 h.

Approximately 500 islets were implanted into the left renal subcapsular space of diabetic mice under isoflurane (AbbVie) anesthesia, as previously described (25). Briefly, the left kidney was exposed through a lateral abdominal incision. A small cut was made in the kidney capsule, and islets were injected using a Vasofix Safety 22-gauge cannula (B. Braun) attached to a Hamilton syringe (Sigma-Aldrich). The cut was cauterized, and the abdominal incision was closed with a 5–0 nonabsorbable suture (Prolene; Ethicon). BG level was determined every day postengraftment for 5 days, and later twice a week. Successful islet engraftment was defined as BG levels <300 mg/dL with a difference in the level before transplantation of at least 200 mg/dL. Thereafter, rejection after successful engraftment was defined by two consecutive weekly BG measurements exceeding 300 mg/dL. For some recipients, a nephrectomy was performed to ensure that normoglycemia was graft dependent. Thereby, the graft-bearing kidney of an anesthetized mouse was exposed, and a clip was used to occlude the renal vessels and ureter before dissection.

Armenian hamster-anti-mouse CXCL10 antibody (clone 1F11) (26) was purified from hybridoma cell supernatant, as described previously (22). Diabetic RIP-LCMV-GP mice engrafted with RIP-LCMV-GP islets were injected intraperitoneally with 100 μg of anti-CXCL10 antibody in PBS three times a week starting 1 day before engraftment. A total of nine injections was administered.

For CXCL10 ELISA, ∼400 islets were isolated from four C57BL/6 or CXCL10^−/− mice and cultured in 250 μL/well RPMI 1640 plus GlutaMAX (ThermoFisher Scientific) supplemented with 1% BSA (Carl Roth) and 1 mmol/L HEPES (Sigma-Aldrich), at 37°C and with 5% CO₂ for 2 days. Islets and supernatants were separated by centrifugation (1,000 rotations/min, 2 min). Islet pellet was homogenized using a glass potter in 250 μL of culture medium. A commercially available ELISA (Quantikine Colorimetric Sandwich ELISA; R&D Systems) for mouse CXCL10 was used according to manufacturer instructions. Samples were assayed in duplicate.

RIP-LCMV-GP and RIP-LCMV-GPxCXCL10^−/− islets were isolated and cultivated for 2 days at 37°C in serum-free RPMI 1640 plus GlutaMAX (∼400 islets in 2 mL). The lower chamber was filled with the supernatant of the islet culture (conditioned islet medium) only or supplemented with 20 ng/mL recombinant mouse CXCL10 (PeproTech). Splenocytes were isolated from RIP-LCMV-GP mice at day 12 postinfection and seeded in the upper chamber of a 96-Transwell insert (5-µm pore size; Corning Costar) at 2 × 10⁵ cells/well in RPMI 1640 plus GlutaMAX (ThermoFisher Scientific) containing 1% BSA (Carl Roth). After incubation for 2 h at 37°C, migrated cells were stained for surface expression of CD8, CD4, and CXCR3 by flow cytometry (FACSCantoII; BD Biosciences). The following antibodies were used: V450-conjugated rat-anti-CD4 (BD Biosciences), brilliant violet 510-conjugated rat anti-CD8 (BioLegend), and allophycocyanin-conjugated rat anti-mouse CXCR3 (R&D Systems).

Kidneys bearing islet grafts and pancreata were immersed in Tissue-Tek optimum cutting temperature (O.C.T.) (Sakura Finetek) and quick frozen on dry ice. The 6-µm cryosections (CM1850 cryostat; Leica) were fixed in 100% ethanol or 50% ethanol/50% acetone at −20°C. Sections were sequentially incubated with primary and biotinylated secondary antibodies and avidin peroxidase conjugate (Vector Laboratories). The primary antibodies used were as follows: rat anti-mouse CD4, CD8α, and CD19 (all from BD Biosciences); rat anti-mouse CD11b and biotinylated rat anti-mouse CD11c (both from eBioscience); rabbit anti-mouse CXCL10 (PeproTech); rabbit anti-mouse CXCR3 (Zymed); rat anti-mouse F4/80 (Abcam); goat anti-NKp46 (R&D Systems); and guinea pig anti-swine insulin and rabbit anti-glucagon (both from Dako). The secondary antibodies used were as follows: goat anti-guinea pig, goat anti-rabbit, and rabbit anti-rat (all biotinylated from Vector laboratories). Images were acquired with an Axioscope2 Microscope (Zeiss).

For detection of insulin, glucagon, and CD31, 6-μm tissue sections were fixed in 50% ethanol/50% acetone and permeabilized with 0.1% Triton X-100 (Sigma-Aldrich). After avidin-biotin (Vector Laboratories) and 10% FBS blocking steps, sections were incubated for 2 h at room temperature in a humidified chamber. The primary antibodies used were as follows: guinea pig anti-swine insulin antibody and rabbit anti-glucagon antibody (both from Dako); and rat anti-mouse CD31 antibody (BD Biosciences). The secondary antibodies were as follows: Alexa Fluor 488–conjugated donkey anti-rabbit IgG and Alexa Fluor 594–conjugated goat anti-rat IgG (both from Life technologies). Biotinylated goat anti-guinea pig IgG and allophycocyanin-conjugated streptavidin (SouthernBiotech) were applied to detect insulin. For CD3 and FoxP3 costaining, sections were fixed with 4% paraformaldehyde at room temperature and then in 100% ethanol at −20°C, permeabilized with 0.1% Triton X-100 (Sigma-Aldrich), and blocked with 10% FBS. The primary antibodies used were as follows: rat anti-FoxP3 (eBioscience) and Armenian hamster anti-mouse CD3e (BD Biosciences). The secondary antibodies used were as follows: Alexa Fluor 594–conjugated goat anti-rat IgG (Life Technologies) and fluorescein isothiocyanate (FITC)–conjugated goat anti-Armenian hamster IgG (eBioscience). Sections were counterstained with DAPI (Sigma-Aldrich). All images were collected using a confocal microscope (LSM510 META; Zeiss). The number of FoxP3⁺ T cells per area of infiltrated immune cells was then determined with using Keyence Software (BZ-9000) (Supplementary Fig. 1).

The multi-epitope ligand cartography (MELC) technology has been described previously (27). Kidneys were removed at day 12 after transplantation, immersed in Tissue-Tek O.C.T. compound, and quick frozen on dry ice. The 6-µm cryosections were cut using a Leica CM1850 cryostat (Leica) and applied on silane-coated coverslips. Sections were fixed in 4% paraformaldehyde in PBS, permeabilized with 0.1% Triton in PBS, and blocked with 3% BSA in PBS for 1 h at room temperature. The sample was placed on the stage of an inverted, wide-field fluorescence microscope (HC PLAN APO 20×/0.70 PH2, model DM IRE2; Leica). By a robotic process, the slices were first incubated for 15 min with predetermined fluorescence-labeled antibodies (FITC–anti-F4/80 [BioLegend]; FITC–anti-CD4 [SouthernBiotech]; FITC–anti-CD11b [AbD Serotec]; and PE–anti-CD11c, FITC–anti-CD45, and FITC–anti-MHC II [Miltenyi Biotec]) and rinsed with PBS. Afterward, the phase contrast and fluorescence signals were imaged by a cooled charge-coupled device camera (Apogee KX4; Apogee Instruments; 2× binning results in images of 1,024 × 1,024 pixels; final pixel size 286 × 286 nm²). To delete the specific signal of the antibody before addition of the next, a bleaching step was performed. A postbleaching image was recorded and subtracted from the following fluorescence tag image during the data analysis. With the use of the corresponding phase-contrast images, fluorescence images produced by each antibody were aligned pixel by pixel. Images were corrected for illumination faults using flat-field correction.

We used two diabetes mouse models for our engraftment experiments. First, we used STZ-treated diabetic C57BL/6 mice as recipients to demonstrate CXCL10 production by transplanted islets (Fig. 1A), and, second, we used RIP-LCMV-GP mice to evaluate the role of CXCL10 in islet isograft destruction (Fig. 1A). To demonstrate CXCL10 expression, ∼500 isolated islets of C57BL/6 mice were transplanted under the renal capsule of C57BL/6 mice that were chemically rendered hyperglycemic (>300 mg/dL) by a single dose of 250 mg/kg STZ. Such syngeneic grafts were well tolerated, and at day 100 postengraftment the isografts were still fully functional, expressing both insulin and glucagon (Fig. 1B). The pattern of CD31 staining suggests well-vascularized isografts (Fig. 1B). Diabetes remission (<300 mg/dL) was restored rapidly after engraftment and remained stable until the end of the observation period at day 100 (Fig. 1C). CXCL10 protein was already expressed in isolated islets that have been kept in culture for 2 days. As determined by ELISA, ∼400 islets isolated from C57BL/6 mice produced ∼1,603 pg (±143 pg) CXCL10, whereby ∼920 pg (±77 pg) were found in the culture supernatant and ∼683 pg (±176 pg) in the homogenized islet fraction (Fig. 1D). To demonstrate functional CXCL10 expression, we performed a migration assay using the culture supernatant (conditioned islet medium) from 2-day RIP-LCMV-GP or RIP-LCMV-GPxCXCL10^−/− islet cultures and RIP-LCMV-GP splenocytes isolated at day 12 after LCMV infection. This setting was chosen to reflect the subsequent engraftment experiments. Migrated splenocytes were quantified and identified by flow cytometry. In general, we found that more CD8 than CD4 T cells have migrated (ratio ∼7:1). Importantly, the addition of conditioned medium from C57BL/6 islets increased the migration of CD8 and CD4 T cells substantially (Fig. 1E). The migration of CD8 T cells was significantly (P = 0.0034) reduced when conditioned medium from CXCL10^−/− islets was used (Fig. 1E). The islet supernatant–mediated increase in the migration of CD8 T cells was reduced by ∼30% with supernatant from CXCL10^−/− compared with C57BL/6 islets (Fig. 1E). Similarly, we detected a tendency (P = 0.1374) toward a decreased migration of CD4 T cells in the absence of CXCL10 (Fig. 1E). In contrast, the addition of external recombinant CXCL10 further increased the migration (Fig. 1E). These data indicate that conditioned islet medium contains several factors that enhance the migration of CD8 and CD4 T cells. These factors include, but are not restricted to, the chemokine CXCL10. The expression of CXCL10 was also detected in situ at day 2 after transfer of C57BL/6 islets, indicating that the isolation and/or the engraftment procedure induced inflammatory processes (Fig. 1F). CXCL10 expression was still detectable at day 100 post-transfer, however, to a lesser degree (Fig. 1F). Also, in this syngeneic setting only a few CD4 and CD8 T cells were detected in the isografts by day 100 (Fig. 1F). No CXCL10 expression has been found in CXCL10^−/− islet isografts (Supplementary Fig. 2).

To investigate the influence of CXCL10 in an autoimmune setting, RIP-LCMV-GP mice were infected with 10⁵ plaque-forming units of LCMV and engrafted with ∼500 islets isolated from RIP-LCMV-GP or RIP-LCMV-GPxCXCL10^−/− mice on days 12–14 postinfection (Fig. 2). Only LCMV-infected RIP-LCMV-GP mice that displayed BG levels >400 mg/dL by day 12–14 after infection were used as recipients. All mice achieved normal BG levels within 1–3 days after transfer, confirming the establishment of functional islet isografts (Fig. 2A). However, whereas normal RIP-LCMV-GP islet isografts were destroyed within 5–30 days in all mice, RIP-LCMV-GPxCXCL10^−/− islets were tolerated in 58% of recipients (11 of 19 recipients). In such mice, the BG levels remained normoglycemic for at least 100 days (Fig. 2A). To confirm that normoglycemia was achieved by the engrafted islets and not due to islet regeneration or stress relief in the pancreas, we used two recipient mice to perform insulin staining of pancreas sections and nephrectomy. Insulin staining of several sections at different positions of the pancreas confirmed that nearly all β-cells were indeed destroyed (data not shown). Left nephrectomy to remove the islet isografts in two nondiabetic mice (days 40 and 104 postengraftment) led to a prompt increase in BG levels to pre-engraftment levels 1 day after graft retrieval (Supplementary Fig. 3). Sham-operated mice or untreated mice showed no changes in BG levels and had to be sacrificed because of severe clinical symptoms by week 4–6 after LCMV infection (data not shown).

Consecutive sections of islet isografts obtained at days 40 (diabetic mice) and 100 (nondiabetic mice) after engraftment were stained for insulin and cellular infiltrates (Fig. 2B). Destroyed RIP-LCMV-GP islets show a disturbed architecture with only few remaining insulin-producing β-cells. In contrast, intact RIP-LCMV-GPxCXCL10^−/− islets found in nondiabetic recipient mice show strong insulin staining and a largely unscathed appearance. Destroyed RIP-LCMV-GPxCXCL10^−/− islets have a phenotype similar to those of destroyed normal RIP-LCMV-GP grafts (Fig. 2B). Cellular infiltration was restricted to the graft site, as neighboring kidney tissue was relatively devoid of infiltrating cells (Fig. 2B). Particularly striking is the accumulation of CD4 T cells in destroyed grafts of both groups, whereas only a few CD4 T cells are located in intact CXCL10-deficient islets (Fig. 2B, second row). Large differences between intact and destroyed grafts include also CD8 T cells, B cells (CD19+), and macrophages (F4/80) for which substantial numbers are found in destroyed grafts (Fig. 2B). Far fewer CD11c⁺ and NK cells are found in destroyed grafts (data not shown). Importantly, destroyed RIP-LCMV-GP grafts contain extensive numbers of CXCR3⁺ cells, whereas only a few CXCR3⁺ cells are found in the large infiltrates observed in intact as well as destroyed RIP-LCMV-GPxCXCL10^−/− islet grafts (Fig. 2B). To further visualize the composition of the cellular infiltrates into destroyed RIP-LCMV-GP grafts, we performed MELC, which contains automated staining and bleaching cycles and allows for repetitive staining of a single tissue section by multiple fluorescent antibodies (27). Many CD45⁺ lymphocytes, in particular CD4 T cells, are located directly around the remaining islet cells, which still express insulin as well as CXCL10 (Fig. 2C). Many MHC class II–expressing cells, including macrophages (F4/80) and CD11b⁺ cells, are located in the periphery of the cellular infiltrates around the cluster of CD4 T cells and remaining islet cells (Fig. 2C).

Since the absence of the key chemokine CXCL10 might have a differential effect on the migration of diverse T-cell subpopulations that bear different concentrations of CXCR3 on their surface, we further stained the graft sections for the presence of FoxP3⁺ Tregs. Double-immunofluorescence staining for CD3 and FoxP3 cells of normal or RIP-LCMV-GPxCXCL10^−/− isograft sections obtained several times after engraftment revealed that FoxP3⁺ T cells are indeed present in the grafts (Fig. 3A). At days 7 and 12, the number of FoxP3-expressing cells seems independent of the presence of CXCL10. In contrast, at day 40 the FoxP3⁺ T-cell number was markedly increased in CXCL10-deficient isografts and remained elevated until the end of the observation time (day 100) (Fig. 3A). Note that all mice that have received normal RIP-LCMV-GP islets became severely diabetic and were euthanized before day 100. To quantify the visual impression, the cells were counted in 8–18 sections of grafts from two to three mice per group, and the number of FoxP3⁺ T cells per area of infiltrating cells was calculated (Fig. 3B). Indeed, the number of FoxP3-expressing T cells was nearly the same (∼1,000–1,500 cells/mm²) in normal and RIP-LCMV-GPxCXCL10^−/− grafts on days 7 and 12 post-transplantation (Fig. 3B). In contrast, the number of FoxP3⁺ cells increased significantly to ∼3,000 cells/mm² in CXCL10-deficient isografts after 40 days, whereas it remained significantly (P = 0.0002) lower (∼1,000–1,500 cell/mm²) in normal RIP-LCMV-GP isografts (Fig. 3B). These data suggest that, in the absence of CXCL10, the recruitment of Tregs is favored over that of other T-cell populations.

Our data suggest that CXCL10 is a driver of islet isograft destruction in the RIP-LCMV-GP model. Thus, to test a possible therapy, we studied the effect of a pharmacologic CXCL10 blockade. The dose for the neutralizing anti-CXCL10 antibody therapy was selected from previous studies (22). Diabetic RIP-LCMV-GP mice were engrafted with ∼500 islets isolated from RIP-LCMV-GP mice. All recipient mice achieved normal BG levels within 1–3 days after engraftment, confirming the establishment of functional islet isografts (Fig. 4A). Anti-CXCL10 antibody was repeatedly injected at 100 μg/dose every second day for 19 days starting 1 day before engraftment. Although the majority of mice destroyed the RIP-LCMV-GP islet grafts within 5–30 days after engraftment, 2 of 10 anti-CXCL10 antibody–treated mice did not show elevated BG values until the end of the observation time (day 100), indicating the presence of a sufficient mass of intact islet isografts (Fig. 4A). In addition, a third mouse showed a delayed destruction and maintained a nondiabetic BG level for 70 days (Fig. 4A). Hence, a calculation of the mean time of graft destruction revealed a significant (P = 0.038) delay in mice treated with the anti-CXCL10 antibody (Fig. 4A, inset). The cumulative graft survival curves show that the administration of a neutralizing antibody is less effective than a total absence of CXCL10 in the graft (Fig. 4B). However, the overall autoimmune destruction was also reduced in anti-CXCL10 antibody–treated mice. Statistical evaluation using the log-rank (Mantel-Cox) test reveals a significant difference in the graft survival curves between normal and RIP-LCMV-GPxCXCL10^−/− grafts (P = 0.0009), whereas CXCL10 blockade was not significant (P = 0.27). To ensure that the administered anti-CXCL10 antibody reaches the islet isograft, we performed immunohistochemistry using an anti-Armenian hamster IgG antibody and found that this “secondary” antibody indeed localized to the engrafted islets even 21 days after administration of the last dose of the anti-CXCL10 antibody (Supplementary Fig. 4). This is definitely possible since it has been shown that the plasma half-life of the anti-CXCL10 antibody (clone 1F11) is longer than 10 days (28). The anti-Armenian hamster IgG antibody localized to some, but not all, regions of CXCL10 production as well as to some of the infiltrating immune cells (Supplementary Fig. 4). Thus, by day 21 after the last administration the anti-CXCL10 antibody might be bound to the CXCL10/CXCR3 complex at the cell surface, in addition to free CXCL10 (compare with Fig. 1F). It has been demonstrated before (28) that the anti-CXCL10 antibody (clone 1F11) does not bind to glucosamine glycan–bound CXCL10.

Islet isografts of nondiabetic mice administered with anti-CXCL10 antibody display patches of intact islets that still produce insulin and are largely free of infiltrates, whereas only small clusters of very few β-cells remained functional in destroyed grafts of diabetic mice (Fig. 4C). No obvious difference was found in the overall pattern of cellular infiltration of intact and destroyed grafts (Fig. 4C). Cellular infiltrations are dominated by CD4 T cells (Fig. 4C). No differences were detected for CD11b⁺, CD11c⁺, CD19⁺, and F4/80⁺ cells (data not shown). Interestingly, intact grafts showed fewer CXCR3⁺ cells, which might be a possible reason for long-term resistance to destruction. However, intact grafts also contained more CD8 T cells than destroyed grafts (Fig. 4C). Maybe the majority of CD8 T cells have already left the destroyed grafts after most of the target cells have been eliminated. For the evaluation of a differential Treg recruitment in destroyed versus intact grafts after anti-CXCL10 antibody therapy, we performed double-immunofluorescence staining for FoxP3 and CD3. No differences were found in the pattern (Fig. 4D) and number (Fig. 4E) of Treg between intact and destroyed RIP-LCMV-GP islet isografts. Thus, the blockade of CXCL10 has an inferior effect on graft survival than a total lack of CXCL10 expression in the islet isografts.

Our finding that CXCL10 has an influence on the destruction of RIP-LCMV-GP islet isografts in LCMV-infected RIP-LCMV-GP mice is restricted to fully autoimmune destruction in a syngeneic setting. Therefore, we wanted to evaluate whether the observed protection of grafts from destruction in the absence of CXCL10 also holds true in an allogeneic setting. However, in our hands C57BL/6 islet allografts were not rejected in STZ-treated BALB/c or FVB mice, regardless of the presence of CXCL10 in the islet grafts (Fig. 5A). Indeed, it has been shown before that C57BL/6 islets are less likely to be rejected in BALB/c recipients than vice versa (29,30), supporting the assumption that C57BL/6 mice are strong responders that are able to generate vigorous and quick alloreactive immune responses, whereas BALB/c mice are low responders or nonresponders (30). Therefore, we generated normal and RIP-LCMV-GPxCXCL10^−/− mice with a mixed C57BL/6xBALB/c background and used their islets for transplantation into diabetic RIP-LCMV-GP (C57BL/6). Unfortunately, all grafts were rejected (Fig. 5B). In addition, treatment with the neutralizing anti-CXCL10 antibody had no effect on the rejection frequency (Fig. 5B). These data indicate that CXCL10 plays an important role in the autoimmune destruction of the islet isografts and only a minor role in an allogeneic transplantation setting.

It has been shown that both CXCL10 and CXCR3 are expressed in pancreatic lesions of T1D patients (31). CXCL10 is still produced in remaining β-cells of T1D patients, and the infiltrating T cells indeed express CXCR3 (32). Further, CXCL10 expression and extensive infiltration of CXCR3⁺ T cells was also found, together with enterovirus capsid expression in pancreata of patients who died of fulminant T1D and ketoacidosis (33). We have previously demonstrated in the RIP-LCMV model that CXCL10 plays an important role in the migration of aggressive T cells into the islets (21). In addition, the blockade of CXCL10 with a neutralizing antibody in combination with anti-CD3 therapy resulted in a persistent remission of T1D in the RIP-LCMV and the NOD mouse models (22). Here we demonstrate that isolated islets indeed express functional CXCL10. At day 2 postengraftment, CXCL10 expression colocalized with insulin production and even after 100 days CXCL10 was still detected in the grafts. However, even though CXCL10 was expressed and some cellular infiltration was detected, such syngeneic grafts were not destroyed in the absence of a target autoantigen. In contrast, normal RIP-LCMV-GP islets engrafted into LCMV-infected RIP-LCMV-GP mice were rapidly destroyed. In the absence of CXCL10, however, 58% of mice retained a sufficient mass of intact islet isografts to maintain a nondiabetic BG level for at least 100 days.

The presence of intact CXCL10-deficient islet isografts was associated with a massive decrease of infiltrating T cells. In contrast, the FoxP3⁺ Treg fraction increased over time. By day 40 after transfer, the Treg number per area of cellular infiltrate was approximately three times higher in CXCL10-deficient isografts than in normal RIP-LCMV-GP islet isografts and remained stable to the end of the observation time. Our data suggest that CXCL10 influences the composition of the cellular infiltration into islet isografts. A correlation of islet protection with an increased presence of Tregs has also been demonstrated in allotransplanation settings (34–36). Further, it has been demonstrated on several occasions that Tregs are indeed actively involved in the prevention of graft rejection. Wu et al. (37) have demonstrated that Tregs migrate to the lymphoid tissue draining the graft site and inhibit the priming of both CD4 and CD8 T cells, thus preventing the generation of potential effector T cells that could elicit graft rejection (37). Similarly, in an islet allograft model Tregs migrated first to the allograft site, where they were activated and subsequently migrated to the draining lymph nodes (38). The migration of Tregs was thereby also dependent on the chemokine receptors CCR2, CCR4, and CCR5 (38).

The CXCL10 receptor CXCR3 is predominantly expressed on activated T-helper type 1 T cells (39,40). However, the existence of hybrid Tregs expressing CXCR3 has been demonstrated as well (41–43). In splenocytes of C57BL/6 mice, such CXCR3⁺ CD4 Tregs comprise ∼24% of all CD4 Tregs, whereas 76% are CXCR3⁻ (42). Interestingly, T1D was accelerated in CXCR3-deficient NOD mice, and splenocytes of such mice displayed a higher diabetogenicity (44). Further transfer of wild-type, but not CXCR3-deficient, CD4⁺CD25⁺ T cells slightly delayed T1D onset in cyclophosphamide-treated NOD mice, indicating that hybrid CXCR3⁺ Tregs diminish T1D severity (44). The situation in the virus-induced RIP-LCMV model is different, since we found previously that, after LCMV infection, CXCR3 is expressed on ∼80% of CD8 T cells and on >90% of LCMV-specific T-helper type 1 CD8 T cells (21). Therefore, blocking CXCL10 might affect the vast majority of specific and unspecific CD8 T cells, but only a much smaller fraction of Tregs.

Hence, to further quantify the number of antigen-specific effector T cells and Tregs, and to characterize the differential expression of CXCR3, we dissected the islet isografts at day 40 after engraftment and performed flow cytometry. However, although we found more lymphocytes in material obtained from engrafted kidneys than from control kidneys, we found no clear evidence for a Treg enrichment in CXCL10-deficient isografts (data not shown). One reason for this discrepancy might be that by the time of graft removal the islets had already nestled tightly in the kidney parenchyma, and thus the grafts had to be scraped out of the neighboring parenchyma. Thereby a considerable contamination by remaining peripheral blood and kidney resident lymphocytes cannot be excluded.

The protection from autoimmune destruction in the absence of CXCL10 was not absolute since 42% of recipients receiving CXCL10^−/− islet isografts were hyperglycemic after a short period of normoglycemia. One reason for the incomplete protection might be differences in the islet isografts themselves, including factors such as initial islet fitness as well as the degree of local trauma due to the surgery. Another reason might be the overall magnitude of inflammatory factors and infiltrating immune cells locally in the graft that might have reached a critical threshold in mice with destroyed grafts. In the future, a detailed analysis of laser-dissected islet graft tissue sections might bring up alternative inflammatory factors that are critical for graft destruction. Mice that did not destroy the islet graft remained normoglycemic until the end of the observation time (day 100). By then, a clear production of insulin and the presence of newly developed microvessels associated with the islets were detected. Such long-term survival is rather remarkable since, without bone marrow transplantation, not many treatment protocols have resulted in similar graft outcomes (45,46). However, it should be noted that most of the islet transplantation studies were performed using allogeneic mouse models (47) and that our results originate from a syngeneic engraftment model in which the reactivity to the islet isograft occurs in an autoimmune fashion. Since T1D patients receive islets from non–HLA-matched donors, syngeneic engraftment models do not reflect the human situation accurately. Similarly, allogeneic models, using β-cell toxins (STZ) are lacking the autoimmune component to the human islet graft destruction process. Therefore, we used an allogeneic approach in which normal or RIP-LCMV-GPxCXCL10^−/− islets of mixed background (F1 generation of C57BL/6xBALB/c) RIP-LCMV-GP mice were transplanted into RIP-LCMV-GP mice (pure C57BL/6). Regardless of the presence of CXCL10, all islet grafts were rejected. These data indicate that CXCL10 might be important predominantly for the attraction of islet antigen-specific T cells and for the “autoimmune rejection” of the grafts.

Our observations on the expression of CXCL10 upon engraftment are consistent with those of various studies (10,13,48,49) demonstrating an enhanced CXCL10 production in recipients of transplants of different organs. However, our findings that CXCL10 might be important for the autoimmune destruction of islet isografts, but possibly not for the rejection of islet allografts, stands somewhat in contrast to previous findings where a blockade of CXCL10 reduced the rejection of heart allografts and CXCL10-deficient heart allografts displayed a long-term survival (18). However, the local inflammatory milieu might be different in the microenvironment of the islets under the kidney capsule and the heart. Therefore, CXCL10 might play a more dominant role during allorejection processes of heart allografts than islet allografts.

To demonstrate the important role of CXCL10 in the autoimmune destruction process, we also performed experiments with islets isolated from RIP-CXCL10 mice overexpressing CXCL10 in the β-cells. RIP-CXCL10 mice do not develop T1D spontaneously but display massive insulitis and have a reduced tolerance to challenge with high glucose levels (50). Unfortunately, due to the massive insulitis the isolation of islets from RIP-LCMV mice is very difficult. We isolated islets from pancreata of 6-day-old mice. However, by that time the islets have been already dispersed with infiltrating lymphocytes. Therefore, substantial contamination by lymphocytes that carried over from the donor cannot be excluded. In addition, the islets showed clear signs of necrosis after 2 days in culture. Hence, as expected, engraftment using islets from 6-day-old RIP-CXCL10 never resulted in an immediate reduction of the BG levels, indicating that the RIP-CXCL10 islets are not sufficiently functional for such a setting.

With regard to a possible therapy, we have also treated RIP-LCMV-GP islet–engrafted RIP-LCMV-GP mice with a neutralizing anti-CXCL10 antibody. Although the effect of the CXCL10 neutralization was not as impressive as a complete absence of CXCL10, the mean time of autoimmune destruction was delayed significantly. In other models, therapeutic interference with the CXCL10/CXCR3 axis has been shown to delay cardiac and skin allograft rejection (14–18). We have previously demonstrated that CXCL10 neutralization diminishes T1D in the RIP-LCMV model when given at the time of disease initiation (21). More importantly, when administered to already diabetic RIP-LCMV or NOD mice after an anti-CD3 antibody treatment, CXCL10 neutralization induces a persistent T1D remission by preventing the re-entry of recovered T cells into the islets (22). Thus, anti-CXCL10 antibody treatment blocked the second wave of T-cell migration to the target site. In some way, the islet engraftment results in a similar situation, where, during the ongoing autoimmune β-cell destruction in the pancreas, a novel target site devoid of infiltrating cells appears at the engraftment site. The following second wave of T-cell migration, which now flows toward the islet isograft, seems largely dependent on CXCL10. Therefore, CXCL10 neutralization might attenuate this second wave of T-cell migration and thereby reduces the autoimmune destruction of the islet isograft. Interestingly, the anti-CD3/anti-CXCL10 combination therapy shifted the T-cell repertoire balance locally in the pancreatic lymph nodes and the islets toward a more regulatory phenotype (22). These data support our finding in the current study that CXCL10 differentially influences the migration of T-cell subtypes to the graft site. With a plasma half-life of >10 days, and a confirmed localization to the islet graft (present study) and to the pancreatic islets (anti-CD3/anti-CXCL10 combination therapy), an application of a humanized anti-CXCL10 antibody with similar pharmacokinetic features seems feasible.

In summary, our experiments demonstrate that CXCL10 plays an important role in the autoimmune destruction of islet isografts. Although the absence of CXCL10 only prolonged the survival of islet grafts in an autoimmune, but not an allogeneic, transplantation setting, a blockade of CXCL10 should be considered as an alternative to the current immunosuppressive regimen. CXCL10 neutralization therapy might support current efforts with Treg therapies that do not have sufficient potency as a stand-alone therapy for transplantation (51).

Funding. This study was supported by the German Research Foundation and the Goethe University Hospital Frankfurt.

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

Author Contributions. C.B. contributed to the design of the study, experiments, interpretation of data, and the drafting of the manuscript. S.C. and M.B. contributed to the experiments. K.S. and E.H. contributed to the experiments and to critical revision of article J.M.P. contributed to the critical revision of the article. U.C. contributed to the idea and design of the study, to interpretation of the data, and to the drafting of the manuscript. U.C. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.