Differentiating MHC-Dependent and -Independent Mechanisms of Lymph Node Stromal Cell Regulation of Proinsulin-Specific CD8⁺ T Cells in Type 1 Diabetes

CD8⁺ T cells are critical for type 1 diabetes development, responding to islet-specific antigens, including proinsulin (PI), and mediating β-cell death. The G9C8 CD8⁺ T-cell clone was isolated from pancreatic islets of NOD mice (1); it recognizes PI β-chain amino acids 15-23 restricted by MHC class I K^d (2) and induces rapid diabetes after adoptive transfer (1,3,4). T-cell receptor (TCR)-transgenic G9Cα^−/− mice are a valuable source of clonal G9C8 CD8⁺ T cells (hereafter called G9Cα^−/−CD8⁺ T cells) to study selection, activation, and regulation of low-avidity autoreactive cells (3,5,6).

Thymic expression of peripheral tissue antigens (PTAs) has been well described to provide negative selection of developing T cells. However, low-avidity T cells that recognize self-antigen can escape thymic selection and persist in the periphery. Lymph node stromal cells (LNSC) also express PTAs and can induce T-cell tolerance (7–13), providing a potential therapeutic target for peripheral interventions. PI expressed in pancreatic lymph nodes (LNs) is regulated by deformed epidermal autoregulatory factor 1 (Deaf1) (12,14,15). Various LNSC subsets express Deaf1 (11) and Ins2 (16,17) in murine models. We have previously reported that LNSC can suppress G9Cα^−/−CD8⁺ T cells, dependent upon PI expression (5). However, LNSC regulatory mechanisms are poorly described. In this brief report, we identify distinct MHC-dependent and -independent regulatory pathways in LNSC and that the latter is reduced in NOD mice.

All procedures and protocols were approved by the U.K. Home Office. NOD-PI2^tg mice, expressing PI2 on the MHC class II promoter, were provided by Professors L. Harrison and A. Lew (18). TCR-transgenic G9Cα^−/− mice were generated as previously described (3). C57BL/6 (B6), NOD, and MHC-congenic C57BL/6-H2^g7 (B6.H2^g7) mice were bred in-house. Mice were maintained in specific pathogen-free microisolators or scantainers at Cardiff University. Male and female mice were used in all experiments.

LNSC were prepared from NOD, B6, or B6.H2^g7 mice, as previously described (5,19), by pooling the cervical, axillary, brachial, inguinal, popliteal, and pancreatic LNs from six to eight mice. Analyses of individual LN sets required 20–25 mice/group. LNSC were expanded for 7–10 days in RPMI 1640 with 10% FBS at 37°C with 5% CO₂.

LNSC were isolated and expanded as described above, treated with 0.25% trypsin-EDTA (Sigma-Aldrich) for 5 min at 37°C, and harvested. LNSC were seeded at 10⁵ cells/well in six-well plates for 24–48 h. Splenic G9Cα^−/−CD8⁺ T cells were isolated using CD8a microbeads (Miltenyi Biotec), and 10⁶ cells/well were added to LNSC cultures. The CD8⁺ T cells were collected after 48 h of coculture, with or without LNSC, and investigated for effects of LNSC exposure on cytotoxicity and proliferation (Fig. 1A). To determine the contribution of transforming growth factor-β1 (TGF-β1), a blocking antibody (clone 19D8; BioLegend) was added each day to cocultures (Fig. 1B). Additionally, 10 pg/mL recombinant TGF-β1 (rTGF-β1) (R&D Systems) was added to CD8⁺ T cells cultured alone, prior to, or added directly to the cytotoxicity assays (Fig. 1C).

After 48-h conditioning, G9Cα^−/−CD8⁺ T cells were incubated with 5 μg/mL InsB15-23 peptide-loaded P815 cells (mastocytoma cell line expressing MHC class I K^d) for 16 h (Fig. 1B, top). Freshly isolated (nonconditioned) G9Cα^−/− T cells were suspended in supernatants from the 48-h cultures of LNSC alone, CD8⁺ T cells, or LNSC/CD8⁺ T cells and incubated with peptide-loaded P815 targets labeled with PKH-26 (Sigma-Aldrich) (Fig. 1B, bottom). Specific lysis of PKH⁺ P815 targets detected by TO-PRO-3 (Molecular Probes) staining was calculated using flow cytometry (5,20).

NOD-PI2^tg bone marrow–derived dendritic cells (DCs) were prepared as previously described (5). G9Cα^−/−CD8⁺ T cells were purified using MACS CD8⁺ T-cell Isolation Kit II (Miltenyi Biotec). Lipopolysaccharide-activated DCs and 0.5 µmol/L carboxyfluorescein diacetate succinimidyl ester (CFDA-SE)–labeled T cells were cultured with LNSC (5) or separated by a transwell with 0.4-µm pore membrane (Corning). Proliferation and activation after 72 h were evaluated by flow cytometry using BD LSRFortessa and analyzed with FlowJo (Tree Star, Inc.). LNSC-conditioned G9Cα^−/−CD8⁺ T cells were labeled with 0.5 µmol/L carboxyfluorescein diacetate (Invitrogen) and then stimulated with 10 μL/10⁶ cells of anti-CD3/CD28 Dynabeads (Thermo Fisher Scientific) or with NOD-PI2^tg-DCs in the continued presence of or after the removal of LNSC (Fig. 1A). After 72 h, G9Cα^−/−CD8⁺ T cells were stained with anti-CD8, anti-Vβ6, and live/dead exclusion (eBioscience). Cells were labeled with anti–T-bet and anti-Zap70 (BioLegend) after fixation/permeabilization (eBioscience) following the manufacturer’s directions. Cytokine synthesis of interferon-γ (IFN-γ), MIP-1β, interleukin-10, and free-active TGF-β1 was measured using a LEGENDplex premixed kit (BioLegend) (standard curve range 2–10,000 pg/mL), following the manufacturer’s directions, and analyzed with LEGENDplex V8.0.

Data are presented as mean ± SD. One-way ANOVA with multiple-comparisons test was used to compare experimental groups to controls using Prism V6.07 (GraphPad Software). P < 0.05 was considered significant.

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

We incubated G9Cα^−/−CD8⁺ T cells with LNSC to determine if LNSC could directly affect T-cell activation and whether the effect was maintained after LNSC removal. G9Cα^−/−CD8⁺ T cells were cultured alone or with LNSC for 48 h (Fig. 1A). Cytotoxicity, without LNSC present (Fig. 2), was assessed. Specific lysis of targets by cytotoxic T lymphocytes previously cultured with NOD-LNSC (Fig. 2A, left), B6-LNSC (Fig. 2A, second panel from left), B6.H2^g7-LNSC (Fig. 2A, third panel from left), or alone (Fig. 2A, right) was compared. Preincubation with either K^d-MHC–matched (NOD and B6.H2^g7) or K^b-MHC–mismatched (B6) LNSC reduced the subsequent cytotoxicity of G9Cα^−/−CD8⁺ T cells (Fig. 2B), with the greatest reduction after NOD-LNSC exposure.

We tested the influence of supernatant harvested from the LNSC/T-cell precultures on cytotoxic function of freshly isolated splenic G9Cα^−/−CD8⁺ T cells (Fig. 1A). Control cytotoxicity in each experiment was conducted in supernatant from T cells incubated alone for 48 h (baseline 100%). Specific lysis was significantly reduced by supernatants from LNSC/T-cell precultures (Fig. 2C), with NOD-LNSC supernatants again inducing the most significant suppression. The supernatants collected from LNSC that were incubated alone did not confer suppression, suggesting that LNSC/T-cell contact is required to release a soluble factor capable of inhibiting cytotoxicity. Analyses of supernatants identified TGF-β1 only in samples collected from LNSC/T-cell precultures (Fig. 2D), but not in supernatants from cultures of T cells or LNSC alone (data not shown). No interleukin-10 was detected (data not shown). This supports the hypothesis that LNSC have a direct effect on CD8⁺ T cells and regulate subsequent T-cell activation, even when LNSC are no longer present. Blocking TGF-β1 with anti–TGF-β1 during LNSC/T-cell cultures (Fig. 1B) neither prevented subsequent inhibition of preconditioned CD8⁺ T-cell cytotoxicity (Fig. 2E), nor did it override the suppressed phenotype resulting from LNSC pre-exposure (Fig. 2F). Anti–TGF-β1 blocking during the LNSC/T-cell preculture did not prevent subsequent suppression by the supernatants (Fig. 2G). Anti–TGF-β1 added together with the supernatants in the cytotoxic assay partially reduced the suppressive capacity of supernatants, particularly the B6.H2^g7/T-cell culture supernatant (Fig. 2H). Addition of rTGF-β1 to T cells cultured alone (Fig. 1C) did not confer subsequent suppression (Fig. 2E), although rTGF-β1 suppressed cytotoxicity directly in the cytotoxic assays (Fig. 2F), suggesting that preconditioned LNSC-mediated suppression is TGF-β1 independent.

We have previously demonstrated that pancreatic LNSC induce peptide/MHC-dependent suppression of G9Cα^−/−CD8⁺ T-cell proliferation (5). NOD-LNSC derived from draining (pancreatic) and nondraining (inguinal and cervical) sites all inhibited NOD-PI2^tg-DC–induced T-cell proliferation (Supplementary Fig. 1A), suggesting that PI is systemically expressed in LNs as part of a tolerogenic self-antigen repertoire. Transwell separation of the G9Cα^−/−CD8⁺ T cells eliminated the suppressive effect of MHC-matched NOD-LNSC, supporting a contact-dependent mechanism (Supplementary Fig. 1B). Preincubation with LNSC did not induce suppression; rather, the continued presence of K^d-matched NOD-LNSC was required for reduced CD8⁺ T-cell proliferation (Fig. 3A and B). There was no difference in the production of IFN-γ (Fig. 3C) or MIP-1β (Fig. 3D) with LNSC. Analyses of T-bet (Fig. 3E and Supplementary Fig. 2A) and Zap70 (Fig. 3F and Supplementary Fig. 2B) expression did not show any statistically significant modification associated with LNSC exposure or suppressed proliferation.

We were surprised to find B6-LNSC reduction of cytotoxicity (Fig. 2) but not DC-stimulated proliferation (Fig. 3), suggesting that MHC-independent mechanisms also play a role in driving LNSC-mediated suppression. To delineate the effects of peptide/MHC-dependent (NOD/B6.H2^g7) and -independent (B6) mechanisms, as well as investigate whether there were strain differences (NOD vs. B6 genetic background), G9Cα^−/−CD8⁺ T cells were cultured with LNSC for 48 h. LNSC-conditioned CD8⁺ T cells were then stimulated with anti-CD3/CD28 Dynabeads (Fig. 1A), following removal of the LNSC or in the continued presence of LNSC for 3 days, to provide robust antigen-independent stimulation. Proliferation (Fig. 4A and B) and TCR-Vβ6 (Fig. 4C), and CD8 (Fig. 4D) expression were analyzed by flow cytometry. NOD-LNSC only suppressed proliferation when continuously present (Fig. 4A and B, in the continued presence of LNSC); however, B6- and B6.H2^g7-LNSC induced more significant suppression when T cells were stimulated by direct TCR engagement using anti-CD3/CD28 (Fig. 4B). Reduced proliferation was associated with decreased upregulation of surface TCR and CD8 (Fig. 4C and D, respectively) compared with CD8⁺ T cells not exposed to LNSC. Synthesis of IFN-γ and MIP-1β was significantly reduced, particularly when LNSC were present, further demonstrating a loss of proinflammatory signals (Fig. 4E and F, respectively).

Exposure to B6- or B6.H2^g7-LNSC significantly reduced G9Cα^−/−CD8⁺ T-cell intracellular expression of T-bet (Fig. 4G and Supplementary Fig. 3A) after anti-CD3/CD28 antibody stimulation, irrespective of whether LNSC were still present in the culture; Zap70 was only reduced in continuous culture (Fig. 4H and Supplementary Fig. 3B). These data suggest that there are distinct signaling pathways involved in direct, non–antigen-specific TCR engagement (anti-CD3/CD28) compared with low-avidity peptide/MHC stimulation and that there are strain differences in regulation of these pathways. However, in contrast to the B6-LNSC, NOD-LNSC only reduced T-bet in continuous culture and did not significantly reduce Zap70 in G9Cα^−/−CD8⁺ T cells, suggesting a defect in NOD-LNSC MHC-independent suppressive mechanisms.

We have identified distinct MHC-dependent and -independent mechanisms of peripheral regulation by LNSC, resulting in decreased proliferation and effector function of PI-specific CD8⁺ T cells. Not only do LNSC have critical structural and chemotactic roles, but they also modulate and regulate T-cell responses (5,7–9,11,21–23), through expression of PTAs (8,11,15,24). We have previously demonstrated that pancreatic LNSC suppress PI-specific G9Cα^−/−CD8⁺ T cells (5). In this study, we found that nondraining LNSC also induce contact-dependent suppression (Supplementary Fig. 1A). Furthermore, LNSC regulation includes both MHC-dependent and -independent components. These data suggest that LN PI expression is widespread and regulates T-cell activity, supporting a role for PI in peripheral self-tolerance. Studies in other transgenic mouse models suggested that tolerance mechanisms were intrinsic to LNSC (7–9). The identification of LNSC endogenously expressing PTAs confirmed this essential compartment for maintaining T-cell tolerance and immune homeostasis (12,13). Nichols et al. (13) found that lymphatic endothelial cells expressing tyrosinase directly presented antigen and induced deletion of high-affinity CD8⁺ T cells. However, low-avidity tyrosinase-specific cells persist in the periphery and contribute to restricting melanoma outgrowth (25,26). Low-avidity recognition of PI/MHC class I K^d allows G9Cα^−/−CD8⁺ T cells to escape both thymic selection and peripheral deletion, as these cells persist in the periphery (4,5). Therefore, continued regulation of these cells is required to prevent aberrant activation and autoreactivity.

We have shown that the effect of LNSC-induced tolerance persists. LNSC conditioning of G9Cα^−/−CD8⁺ T cells (Fig. 1A) significantly reduced subsequent cytolytic potential (Fig. 2), suggesting an intrinsic modification to T cells after LNSC contact. However, supernatants from LNSC/T-cell culture also reduced cytotoxicity of nonconditioned effectors (Fig. 2C), which provides evidence that LNSC/T-cell contact induces suppressive soluble factors, including TGF-β1 (Fig. 2D). Surprisingly, we observed that the MHC-mismatched K^b-expressing B6-LNSC suppressed cytotoxicity (Fig. 2A and B). Supernatants from B6-LNSC/CD8⁺ T-cell incubation also suppressed cytotoxicity (Fig. 2C) and generated TGF-β1 (Fig. 2D), although to a lesser degree than NOD-LNSC/T-cell supernatants. Blocking TGF-β1 (Fig. 1B) did not fully restore cytotoxicity (Fig. 2E–H), nor did the addition of rTGF-β1 during preculture (Fig. 1C) provide subsequent suppression (Fig. 2E). These results suggest that while regulatory, TGF-β1 alone did not induce the intrinsic modification to T-cell function during LNSC/T-cell precultures. These data suggest that MHC-independent mechanisms also contribute to overall regulation of T cells when preconditioned by LNSC.

We investigated the suppressive mechanisms intrinsic to LNSC by stimulating LNSC-conditioned G9Cα^−/−CD8⁺ T cells using antigen-specific NOD-PI2^tg-DCs or non–antigen-specific anti-CD3/CD28 Dynabeads (Fig. 1A) and observed differential effects. LNSC from NOD and B6.H2^g7suppressed antigen-specific NOD-PI2^tg-DC-stimulation of G9Cα^−/−CD8⁺ T cells (Fig. 3A and B). B6- and B6.H2^g7-LNSC suppressed anti-CD3/CD28 T-cell proliferation and activation much more than NOD-LNSC (Fig. 4). Furthermore, suppression induced by exposure to B6- and B6.H2^g7-LNSC was observed even when LNSC were removed; however, continued presence of NOD-LNSC was required for suppression (Fig. 4, following removal of LNSC vs. in the continued presence of the LNSC), suggesting NOD-LNSC did not induce sufficient intrinsic modifications to the T-cell phenotype. These data highlight differential pathways when regulating robust (anti-CD3/CD28) versus weak (insulin antigen) activation. MHC-dependent suppressive mechanisms reduced antigen-dependent cytotoxicity (Fig. 2) and proliferation (Fig. 3) but did not suppress the effects of direct TCR engagement (Fig. 4). The MHC-independent suppression was also associated with decreased expression of T-bet and Zap70 (Fig. 4 and Supplementary Figs. 2 and 3). However, T-bet expression was not affected by low-avidity antigen stimulation, suggesting different pathways are used. Further work is needed to elucidate the differential signaling and regulation of these pathways associated with signal strength and antigen specificity.

Finally, our work suggests NOD-LNSC are less able to provide MHC-independent suppression of antibody stimulation and that there is a defect in this regulatory compartment associated with autoimmunity. Most studies have used models of high-affinity antigen when addressing peripheral T-cell regulation and deletional tolerance; the fate of low-avidity cells is less known but has important implications for regulating autoreactive T cells and monitoring outcomes of therapeutic interventions.

Acknowledgments. The authors thank members of the Joint Biological Services of Cardiff University for animal care and Len Harrison and Andrew Lew, from the Walter and Eliza Hall Institute of Medical Research (Parkville, Victoria, Australia), for the gift of NOD-PI2^tg mice. The authors also thank Daniel S. Thayer (Swansea University, Swansea, U.K.) for editorial assistance.

Funding. The work was supported by a postdoctoral fellowship from JDRF (3-PDF-2014-211-A-N to T.C.T.) and funding from the Medical Research Council (U.K.) (MR/K021141/1 to F.S.W.).

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

Author Contributions. T.C.T. carried out experiments and analyzed the data. S.J.H. assisted with experimental design. J.D. and J.A.P. assisted with experiments. T.C.T and F.S.W. conceived the project and wrote the manuscript. T.C.T., S.J.H., L.W., and F.S.W. edited the manuscript. F.S.W. is the guarantor of this work, and, as such, had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.