It is generally believed that inflammatory cues can attract noncognate, “bystander” T-cell specificities to sites of inflammation. We have shown that recruitment of naive and in vitro activated autoreactive CD8+ T cells into endogenous islets requires local autoantigen expression. Here, we demonstrate that absence of an autoantigen in syngeneic extrapancreatic islet grafts in diabetic hosts renders the grafts “invisible” to cognate memory (and naive) T cells. We monitored the recruitment of islet-specific glucose-6-phosphatase catalytic subunit-related protein (IGRP)206–214-reactive CD8+ T cells into IGRP206–214-competent and IGRP206–214-deficient islet grafts in diabetic wild-type or IGRP206–214−/− nonobese diabetic hosts (harboring either naive and memory T cells or only naive IGRP206–214-specific T-cells, respectively). All four host–donor combinations had development of recurrent diabetes within 2 weeks. Wild-type hosts recruited IGRP206–214-specific T cells into IGRP206–214+/+ but not IGRP206–214−/− grafts. In IGRP206–214−/− hosts, there was no recruitment of IGRP206–214-specific T cells, regardless of donor type. Graft-derived IGRP206–214 activated naive IGRP206–214-specific T cells, but graft destruction invariably predated their recruitment. These results indicate that recurrent diabetes is exclusively driven by autoreactive T cells primed during the primary autoimmune response, and demonstrate that local antigen expression is a sine qua non requirement for accumulation of memory T cells into islet grafts. These findings underscore the importance of tackling autoreactive T-cell memory after β-cell replacement therapy.
Nonobese diabetic (NOD) mice have development of a form of type 1 diabetes that results from destruction of β cells by CD4+ and CD8+ T cells recognizing many autoantigenic peptides (1). A significant fraction of islet-associated CD8+ cells recognize the mimotope NRP-V7 in the context of the major histocompatibility complex (MHC) molecule Kd (2). These cells are a significant component of the earliest NOD islet CD8+ infiltrates (2,3), are diabetogenic (4,5), and target residues 206–214 of islet-specific glucose-6-phosphatase catalytic subunit-related protein (IGRP) (6). The peripheral IGRP206–214-reactive CD8+ T-cell pool is sizeable (7) and, on recruitment into islets, undergoes a local avidity maturation process that contributes to disease progression (8).
Studies in infection and autoimmune disease models have suggested that recruitment of T cells into sites of extralymphoid inflammation does not require local expression of cognate peptide–MHC (pMHC) (9–11). However, we recently have shown that cues emanating from pancreatic islets undergoing spontaneous autoimmune inflammation in NOD mice cannot recruit naive or newly activated bystander T-cell specificities. This was established by monitoring the recruitment of naive or in vitro activated IGRP206–214-specific CD8+ T cells in gene-targeted NOD mice expressing a T-cell “invisible” IGRP206–214 sequence. These mice had development of diabetes with normal incidence, but their insulitic lesions could not recruit either cell type. These results indicated that recruitment of naive T cells or effector cytotoxic T lymphocytes to a site of autoimmune inflammation results from an active process that is strictly dependent on local display of cognate pMHC (12).
Here, we asked whether this revised paradigm also applies to recruitment of memory (autoantigen-experienced) autoreactive T cells and/or recruitment of naive and memory T cells to syngeneic islet grafts. We reasoned that the “nonphysiological” lymphatic and vascular anatomy of islets grafts transplanted under the kidney capsule (13–15), coupled with a high rate of graft cell death (16), should allow recruitment of “graft-irrelevant” (i.e., nonautoreactive) memory T cells to the site in response to local inflammatory cues, including those caused by grafting. We demonstrate that recruitment of CD8+ T cells to islet grafts during disease recurrence exclusively involves autoantigen-specific T cells from the memory pool, excluding a role for bystander T-cell specificities or graft antigen-activated autoreactive T cells.
RESEARCH DESIGN AND METHODS
NOD.IGRPK209A/F213AKI/KI mice, encoding an immunologically silent IGRP206–214 epitope, have been described (12). These studies were approved by the local Animal Care Committee.
Diabetes was monitored twice per week by measuring urine glucose levels and was confirmed by tail vein blood glucose measurements. All recipient mice had at least two successive blood glucose measurements >22.2 mmol/L and underwent transplantation within 1–2 weeks of diabetes onset.
Peptides and tetramers.
The peptides IGRP206–214, NRP-V7, and TUM, and the corresponding tetramers (phycoerythrin -labeled), were prepared as described (17).
Cell suspensions were stained with pMHC tetramers and FITC-conjugated or peridinin chlorophyll protein (PerCP)-conjugated anti-CD8α and anti-CD4 mAbs (BD Pharmingen) for 60 min at 4°C, fixed in 1% paraformaldehyde/PBS, and analyzed by fluorescence-activated cell sorting.
Pancreatic islets were isolated by hand-picking after collagenase P digestion of the pancreas and cultured overnight at 37°C in 5% CO2.
Islet transplantation and graft harvest.
Five hundred islets were transplanted under the left kidney capsule. Successful engraftment was defined as restoration of glycemic control for >48 h. Graft failure was defined as nonfasting blood glucose >15 mmol/L.
Specificity of islet-associated CD8+ T cells.
The grafts of recurrent diabetic hosts were cut into ∼2-mm3 fragments and cultured for 1 week in 0.5 units/mL rIL-2. T cells were analyzed by fluorescence-activated cell sorter as described. Measurements of interferon-γ secretion by graft-associated T cells (2 × 104/well) in response to peptide-pulsed irradiated NOD splenocytes (105/well) were determined by enzyme-linked immunosorbent assay (R&D Systems) and normalized to values obtained with TUM.
Purified splenic CD8+ T cells were labeled with carboxyfluorescein succinimidyl ester (CFSE) (2.5 μmol/L) and injected intravenously (5 × 106) 24 h after transplantation. Mice were killed 7 days later, and the grafted and nongrafted kidney-draining lymph nodes, spleens, pancreatic lymph nodes, and mesenteric lymph nodes were examined for dilution of CFSE in the CFSE+CD8+ gate.
Data were compared by Mann-Whitney U or χ2 log-rank tests. Statistical significance was assumed at P < 0.05.
Online supplementary materials.
The online supplementary materials provide supporting data on the recruitment of IGRP206–214-reactive or InsB15–23–reactive CD8+ T cells to islet grafts, as well as representative fluorescence-activated cell sorter profiles.
NOD.IGRPK209A/F213AKI/KI mice (referred to as IGRP206–214−/− or “epitope-deficient” mice) have development of diabetes with the same incidence and kinetics as wild-type NOD (“epitope-competent”) mice but cannot trigger the activation or recruitment of naive or in vitro activated IGRP206–214-reactive CD8+ T cells (12). Here, we investigated if the naive IGRP206–214-reactive CD8+ T cells of epitope-deficient hosts and/or their memory counterparts arising in epitope-expressing hosts are recruited into epitope-competent or epitope-deficient islet grafts (from NOD.scid and NOD.rag2−/−.IGRPK209A/F213AKI/KI donors, respectively).
We first tracked the recruitment of IGRP206–214-reactive CD8+ T cells from diabetic IGRP206–214+ hosts (harboring both naive and memory IGRP206–214-reactive CD8+ T cells) into IGRP206–214+ or IGRP206–214−/− grafts. The presence of IGRP206–214-reactive CD8+ T cells recruited into the graft was analyzed by flow cytometry using pMHC tetramers. We also measured the amount of interferon-γ that graft-infiltrating T cells secreted in response to peptide-pulsed irradiated splenocytes as an additional read-out of T-cell recruitment.
Diabetic IGRP206–214+ hosts receiving IGRP206–214−/− islets had development of recurrence of disease, but did so a few days later than those receiving IGRP206–214+ islets (12.7 ± 3.5 vs. 5.9 ± 0.7 days; Figs. 1A top, B, and C left). This indicated that recruitment of naive and/or preactivated IGRP206–214-reactive CD8+ T cells contributes to, but is dispensable for, graft destruction in diabetic IGRP206–214+ hosts. Importantly, however, whereas IGRP206–214-reactive T cells accounted for a significant fraction of IGRP206–214+ graft-associated CD8+ T cells (18.2% ± 4.7%), they were undetectable in IGRP206–214−/− grafts (Fig. 2A–C and Supplementary Fig. 1A). Furthermore, the lymph nodes draining the grafted (left) kidney in mice receiving IGRP-206–214–expressing islet grafts harbored more IGRP206–214-reactive CD8+ T cells than those draining the contralateral (nongrafted) kidney, and this was not seen in diabetic hosts grafted with IGRP206–214−/− islets (Fig. 3A). In addition, the pancreatic lymph nodes and the spleen, and to a lesser extent the mesenteric lymph nodes, of mice grafted with IGRP206–214+ islets contained more IGRP206–214-reactive CD8+ T cells than those from mice grafted with IGRP206–214−/− islets (Figs. 3B and C). This suggests that graft-derived IGRP206–214 induces the activation and retention of host naive and/or preactivated IGRP206–214-reactive CD8+ T cells in graft-proximal lymphoid organs.
We next investigated whether the IGRP206–214-reactive CD8+ T cells recruited to the epitope-expressing grafts include naive T-cells primed by graft-derived IGRP206–214. We followed the fate of IGRP206–214+ and IGRP206–214−/− islet grafts and IGRP206–214-reactive CD8+ T cells in diabetic IGRP206–214−/− hosts, which are unable to generate antigen-experienced IGRP206–214-reactive CD8+ T cells from an otherwise normal pool of naive T-cell precursors. IGRP206–214−/− hosts rejected IGRP206–214+ and IGRP206–214−/− islets with kinetics similar to those seen in NOD mice (Figs. 1A–C). Yet, IGRP206–214-reactive CD8+ T cells were barely detectable in IGRP206–214+ and IGRP206–214−/− grafts implanted into IGRP206–214−/− hosts (Figs. 2A and B), indicating that the grafts do not recruit newly primed IGRP206–214-reactive CD8+ T cells, at least within the first 2 weeks after transplantation. Interestingly, IGRP206–214+ grafts in IGRP206–214−/− hosts recruited slightly more InsB15–23-reactive CD8+ T cells than in IGRP206–214+ hosts (Supplementary Figs. 1B and C). Although these differences were not statistically significant, they suggest that in these mice the IGRP206–214-reactive CD8+ T-cell niche is occupied by other memory T-cell specificities. In addition, the islet graft-associated CD8+ T cells express markers of memory (i.e., are CD44high, CD62L−, and CD127+; data not shown).
In agreement with these data, the proximal lymphoid organs (graft-draining lymph nodes, pancreatic lymph nodes, and spleen) and blood of IGRP206–214−/− hosts transplanted with antigen-expressing islets contained fewer IGRP206–214-reactive CD8+ T cells than their IGRP206–214+ host counterparts, suggesting that graft-derived antigen does not induce a detectable peripheral expansion of naive autoreactive T cells (Fig. 3D). In addition, because the percentages of IGRP206–214-reactive CD8+ T cells in the graft-draining lymph nodes, pancreatic lymph nodes, and spleen of IGRP206–214+ hosts grafted with IGRP206–214-deficient islets were also low (Fig. 3D, right panels), we conclude that the peripheral expansion of IGRP206–214-reactive CD8+ T cells seen in IGRP206–214+ islet-grafted IGRP206–214+ hosts (Figs. 3A–C) largely, if not exclusively, involves antigen-experienced T cells. Interestingly, NOD hosts grafted with IGRP206–214−/− islets accumulated IGRP206–214-reactive CD8+ T cells in the bloodstream (Fig. 3D), suggesting that, in the absence of antigen in the graft and graft-draining lymphoid organs, preactivated IGRP206–214-reactive CD8+ T cells are “trapped” in the bloodstream.
Finally, we asked whether absence of graft-antigen-primed naive IGRP206–214-reactive CD8+ T cells in the epitope-expressing grafts of IGRP206–214−/− hosts was caused by inability of graft-derived IGRP206–214 to activate cognate naive CD8+ T cells, or to protracted recruitment and/or accumulation of these T cells into the graft. This was performed by tracking the proliferation of naive splenic CFSE-labeled IGRP206–214-reactive CD8+ T cells from 8.3 T-cell receptor transgenic mice in NOD hosts grafted with IGRP206–214+ or IGRP206–214−/− islets the previous day. Various host lymphoid organs were examined for dilution of CFSE in the CFSE+CD8+ gate 7 days after T-cell transfer. Naive 8.3-CD8+ T cells proliferated vigorously in the lymph nodes draining IGRP206–214+ (but not IGRP206–214−/−) grafts and, to a lesser extent, in the pancreatic lymph node and spleen, where some of the proliferation appears to be induced by host-derived (residual) IGRP206–214 (Fig. 4A and B). There were very few donor 8.3-CD8+ T cells in IGRP206–214+ or IGRP206–214−/− grafts (0.06% ± 0.03% vs. 0.06% ± 0.016% of CD8+ cells, respectively) (Fig. 4C). These observations indicate that destruction of IGRP206–214+ and IGRP206–214−/− grafts in IGRP206–214+ hosts (Fig. 1A) predates recruitment of newly activated T cells.
The data presented herein challenge a current paradigm stating that nonantigen-specific inflammatory cues can attract and retain noncognate, bystander T-cell specificities to sites of inflammation, including syngeneic islet transplants in diabetic mice. We demonstrate that absence of the cognate autoantigen in a syngeneic extrapancreatic islet graft in a diabetic host renders the graft invisible to cognate memory (and naive) T cells. Local antigen expression, in addition to MHC class I expression (18), is thus a sine qua non requirement for accumulation of autoreactive CD8+ T cells into islet grafts.
The absolute need for local autoantigen expression is highlighted by two important considerations. First, IGRP206–214/Kd (NRP-V7)-reactive CD8+ T cells are among the most prevalent in NOD islet infiltration (8). Second, the vascular beds irrigating islet grafts, including the subcapsular kidney space, have a porous, fenestrated architecture (13–15) that could conceivably render them permeable to bystander T cells. Therefore, it is remarkable that autoantigen-experienced (i.e., memory) IGRP206–214-reactive T cells, despite their prevalence in the periphery, do not accumulate into IGRP206–214-deficient grafts. Recruitment of memory CD8+ T cells to islet grafts thus follows the same rules that we have described for the recruitment of naive and in vitro activated CD8+ T cells into endogenous islets (i.e., requiring a cognate pMHC interaction in situ). A recent report demonstrates a striking similarity in human insulitis: all the CD8+ T cells found in the inflamed islets of type 1 diabetic patients bound self-pMHC complexes (19).
Our findings further imply that individual autoantigenic specificities, even when prevalent, play relatively minor roles in the anamnestic autoimmune response contributing to graft destruction in autoimmune disease-affected hosts. Our results also clearly indicate that destruction of syngeneic islet grafts in diabetic NOD mice is largely, if not exclusively, effected by autoantigen-experienced T cells primed during the primary autoimmune response. Although graft antigen-loaded antigen-presenting cells residing in the graft-draining lymph nodes can readily induce the activation of naive autoreactive CD8+ T cells, graft destruction precedes recruitment of these T cells into the graft. The high physical and functional pMHC-binding avidities of antigen-experienced T cells coupled with their ability to mount rapid recall responses to limiting amounts of antigen (20) likely afford them a competitive advantage, particularly during the first 2 weeks after transplantation. Differences in the bio-distribution of memory versus naive T cells may be another contributing factor. These considerations, however, do not exclude the likely involvement of graft antigen-primed naive autoreactive T cells in chronic loss of graft function, such as, for example, in the context of partially matched islet allografts.
Recurrent autoimmunity in allogeneic islet cell transplantation has become a topic of growing interest. For example, the pretransplant peripheral frequencies of autoreactive T cells in diabetic recipients are predictive of islet allograft fate, and posttransplant increases are associated with loss of graft function (21–24), suggesting that recurrent autoimmunity may contribute to allograft destruction. Although clinical islet transplantation is a more complex situation, our model has allowed us to dissect the specific roles of bystander immunity versus anamnestic and naive autoimmunity to islet graft rejection. Our observations emphasize the importance of developing therapies capable of preventing priming of naive alloreactive T cells causing allograft rejection, recruitment of memory autoreactive T cells causing anamnestic autoimmunity in the immediate posttransplant period, and the priming of naive autoreactive T cells causing chronic loss of graft function, giving special attention to the pretransplant autoreactivity status of the diabetic host.
G.M.A. is currently affiliated with the Department of Endocrinology, University Medical Center Groningen, Groningen, the Netherlands.
B.-Y.X. is currently affiliated with the Alberta Diabetes Institute, University of Alberta, Edmonton, Alberta, Canada.
The work described here was funded by grants from the Canadian Diabetes Association (CDA) and the Canadian Institutes of Health Research. G.M.A. and B.O.R. are supported by the Dutch Diabetes Research Foundation, the Juvenile Diabetes Research Foundation, and the European Union FP7 Program (BetaCellTherapy and NAIMIT). S.T. and X.C.-C. are supported by studentships from Alberta Innovates–Health Solutions (AIHS, formerly AHFMR) and the AXA Research Fund, respectively. J.R.W. was supported by a fellowship from the CDA. P.S. is a Scientist of the AIHS and a Scholar of the Juvenile Diabetes Research Foundation and Instituto de Investigaciones Sanitarias Carlos III. The Julia McFarlane Diabetes Research Centre is supported by the Diabetes Association (Foothills) and the CDA.
No potential conflicts of interest relevant to this article were reported.
G.M.A., X.C.-C., S.T., and J.R.W. researched data and contributed to the editing of the manuscript. G.M.A., X.C.-C., S.T., J.R.W., and B.O.R. contributed to the discussion. B.-Y.X. and J.W. assisted with experiments. J.W. and B.O.R. reviewed and edited the manuscript. P.S. designed and supervised the study. P.S. is the guarantor of this work and, as such, had full access to all the data and takes responsibility for the integrity of the data and the accuracy of data analysis.
The authors thank A. Shameli, S. Thiessen, J. Luces, and R. Barasi (University of Calgary) for technical assistance and L. Kennedy and L. Robertson (University of Calgary) for flow cytometry.