Abstract
Donor Ag-reactive CD4 and CD8 T cell production of IFN-γ is a principal effector mechanism promoting tissue injury during allograft rejection. The CXCR3-binding chemokines CXCL9 and CXCL10 recruit donor-reactive T cells to the allograft, but their role during the priming of donor-reactive T cells to effector function is unknown. Using a murine model of MHC-mismatched cardiac transplantation, we investigated the influence of CXCL9 and CXCL10 during donor-reactive T cell priming. In allograft recipient spleens, CXCL9 and CXCL10 were expressed as early as 24 h posttransplant and increased with similar kinetics, concurrently with CXCR3 expression on T cells. CXCL9, but not CXCL10, expression required NK cell production of IFN-γ. The absence of CXCL9 in donor allografts, recipients, or both significantly decreased the frequency of donor-reactive CD8 T cells producing IFN-γ and increased the frequency of donor-reactive CD8 T cells producing IL-17A. In contrast, the absence of CXCL10 increased the frequency of IFN-γ–producing CD8 T cells in a CXCL9-dependent manner. These data provide novel evidence that donor-reactive CD8 T cells use the CXCR3 chemokine axis as a costimulation pathway during priming to allografts where CXCL9 promotes the development of IFN-γ–producing CD8 T cells, and CXCL10 antagonizes this skewing.
Solid organ transplantation is the sole treatment option for patients with end-stage organ failure. MHC-mismatched allografts induce a vigorous antidonor T cell response that requires aggressive immunosuppression to prevent rejection. Acute rejection of allografts is initiated by the emigration of passenger dendritic cells (DCs) from the transplanted organ to the recipient spleen where they prime donor Ag-specific T cells to express the effector functions, including cytolytic activities and cytokine production, that mediate graft tissue injury. IFN-γ is the principal cytokine produced by CD4 and CD8 effector T cells in response to allografts (1–4). The mechanisms that lead to a preferential skewing of the donor-reactive CD4 and CD8 T cell repertoire to predominantly IFN-γ–producing effectors following allograft transplantation is unknown. Although CD4 T cell development to an IFN-γ–producing phenotype requires APC production of IL-12 (5), CD8 T cell development to IFN-γ–producing cells often occurs independently of IL-12 (6).
Following priming in the spleen, allograft-reactive T cells migrate through the recipient blood stream to the graft, where they are activated to express the effector functions, including IFN-γ production, that mediate graft injury. Extensive studies have established that the CXCR3- and CCR5-binding chemokines play prominent roles in the recruitment of effector T cells into allografts (7–10), and correlates to these findings have been found in clinical transplantation (11–13). Studies from this and other laboratories have shown a role for CXCL9/monokine induced by IFN-γ, CXCL10/IFN-γ–inducible protein 10 (IP-10), and CXCR3 expression in accelerating acute rejection of MHC-mismatched allografts (7, 10, 14–16). In cardiac allografts, graft vascular endothelial cells and infiltrating neutrophils and macrophages produce these T cell chemoattractants (14, 17). CXCL9 and CXCL10 are also produced by DCs, B cells, and macrophages (18) and bind the G-protein–coupled receptor CXCR3, which is expressed on multiple cell types but predominantly on memory phenotype cells and primed effector T cells producing IFN-γ (19).
In addition to directing leukocyte trafficking to inflammatory sites, many chemokines are produced at sites of T and B cell activation in primary and secondary lymphoid tissues (20). A recent in vitro study suggested that CXCL9 might influence the proliferation and development of alloantigen-reactive T cells in MLCs (21). Several models of inflammation have also suggested that CXCR3-binding chemokines may influence the functional development of T cells during Ag priming (22–24). CD4 and CD8 T cells express CXCR3 early during priming in response to MHC-disparate allografts, and CXCR3 expression is most pronounced on effector cells that produce IFN-γ. This raises the possibility that downstream signaling from CXCR3 early during CD8 T cell priming may promote preferential polarization of donor-reactive T cells to an IFN-γ–producing phenotype.
In this study, we investigated a potential role for CXCL9 during recipient T cell priming to MHC-mismatched cardiac allografts. We demonstrate that CXCL9 and CXCR3 are coincidentally expressed in the graft-draining lymphoid tissue as early as 24 h following transplant and that CXCL9 is induced by NK cell-derived IFN-γ. The absence of CXCL9 depresses the number of IFN-γ–producing, donor-reactive CD8 and CD4 T cells, but this does not prolong graft survival, potentially as the result of increased frequencies of donor-specific CD8 T cells producing IL-17A. Finally, we provide evidence that CXCL9 and CXCL10 antagonize each other as costimulatory molecules during T cell priming to alloantigen. Taken together, these data implicate an important and unrecognized role for CXCR3-binding chemokines in the acute rejection process. Moreover, we propose that the predominant effector mechanism in complete MHC-mismatched allograft rejection, CD8 T cell production of IFN-γ, is the consequence of high levels of CXCL9 produced in the allograft recipient spleen during priming of these T cells, and this is regulated, in part, by the coincident production of CXCL10 in the priming site.
Materials and Methods
Mice
The following mice were used: C57BL/6 (H-2b) and A/J (H-2a) from Charles River Laboratories (Wilmington, MA); DBA/1 (H-2q), BALB/c (H-2d), and B6.Thy1.1 from The Jackson Laboratory (Bar Harbor, ME); and B6.2C TCR transgenic, B6.CXCL9−/−, A/J.CXCL9−/−, B6.Rag1−/−, B6.CXCL10−/−, and BALB/c.CXCL10−/− (bred at our facility). All experiments used 8–12-wk-old male mice, and the Cleveland Clinic Institutional Animal Care and Use Committee approved all procedures.
Heterotopic cardiac transplantation
Standard methods of murine heterotopic intra-abdominal cardiac transplantation were adapted from the method of Corry et al. (25). Total operative times averaged 30–35 min, and graft survival was monitored by abdominal palpation, with cessation of beating confirmed by laparotomy.
Ab treatments
For CD8 depletion, 0.2 mg a 1:1 mixture of anti-CD8 mAbs YTS169 and TB-105 (BioExpress, West Lebanon, NH) was administered i.p. on days –3, −2, −1, and +4 and every 4 d until rejection. NK cell-depleting anti-NK1.1 mAb (0.25 mg, PK136; BioExpress) was given i.p. on days −3, −2, −1, +2, and +4 and every 2 d. In all cases, equal amounts of normal rat IgG (Sigma-Aldrich, St. Louis MO) were administered to control groups, and cell depletion was confirmed by flow cytometry analysis of peripheral blood samples. Mouse anti-mouse CXCL9 mAb was administered to graft recipients (150 μg, i.p.) on days 1, 3, and 5 posttransplant.
Bone marrow chimera
An established protocol for radiosensitive mice was used (26). A/J.CXCL9−/− mice received a split dose of 2 × 3.5 Gy gamma radiation delivered at 1.21 Gy/min with a 4-h interval between treatments. One day after irradiation, 10–15 × 106 wild-type A/J bone marrow cells were given i.p. Recipients were treated with 0.25 ml gentamicin (Sigma-Aldrich) i.p. on days 0 and +2 after bone marrow transplantation and were maintained on acid water (pH 2.5) for 2 wk post-bone marrow transplantation. After resting for 8 wk, chimerism was confirmed using standard PCR analysis of DNA purified from PBMCs.
Flow cytometry
Graft-infiltrating leukocytes were analyzed using a modified method of Afanasyeva et al. (27). Following harvest, graft tissue was incubated for 1 h at 37°C in RPMI 1640 plus type II collagenase (Sigma-Aldrich). After incubation, graft tissue was gently crushed and passed through a 40-μm filter and washed with RPMI. For flow cytometric analysis of splenocytes, single-cell suspensions were made after RBC lysis, and 5 × 106 cells were stained. In all cases, surface markers were stained using standard methods and commercially available Abs (eBioscience, San Diego, CA and BD Biosciences, San Jose, CA). Flow cytometry was performed using a FACSCalibur (BD Biosciences) cytometer and FlowJo analysis software (Tree Star, Ashland, OR).
Intracellular cytokine staining
For intracellular cytokine staining, purified cells were incubated in complete RPMI 1640 with 0.01 μg/ml PMA and 1 nM ionomycin at 37°C for 2 h. Monensin (2 μM) was added, and the cells were cultured for an additional 2 h. Following restimulation, cells were surface stained, fixed with 4% paraformaldehyde, and then permeabilized with 0.1% saponin (perm buffer). Cells were washed and stained for intracellular cytokines using commercially available Abs in perm buffer.
In vitro culture assays
Naive 2C transgenic CD44loCD8+ T cells were purified by flow sorting and cultured in 48-well plates with T cell-depleted CXCL9−/− syngeneic stimulators pulsed with 10 μM Ld peptide (SIYRYYGL; a generous gift from A. Morelli, University of Pittsburgh) at a 10:1 (responder/stimulator) ratio for 3 d. In some experiments, recombinant murine CXCL9 protein or anti-CXCL10 mAb (both from R&D Systems, Minneapolis MN) was added to the cultures on day 0. Following culture, cells were restimulated with PMA and ionomycin for 4 h, with monensin for the final 2 h as above, and were processed and stained for measurement of intracellular IFN-γ by flow cytometry.
Quantitative real-time PCR
Snap-frozen graft pieces were crushed, homogenized using Qiashredders, and RNA was isolated using Fibrous Tissue kits (both from Qiagen, Valencia, CA), according to the manufacturer’s instructions. For analysis of mRNA expression in recipient spleens, spleen pieces were snap-frozen in liquid nitrogen, homogenized as above, and RNA was isolated using RNeasy Mini kits (Qiagen). For RNA purification from flow-sorted cells, cell pellets were homogenized using Qiashredders, and RNA purification was performed using RNeasy Plus Micro kits (Qiagen).
Commercially available reagents and probes were used for reverse transcription, and real-time PCR was performed on a 7500 Fast Real-Time thermocycler, all from Applied Biosystems (Foster City, CA). For quantification of message expression, target gene expression was normalized to Mrpl32 gene expression. All samples were plated in triplicate, and the results are expressed as mean fold increase ± SEM in gene expression over controls.
ELISPOT
Responder CD8 or CD4 T cells were column-purified (R&D Systems) from recipient spleens; self, donor, and third-party stimulator splenocytes were depleted of T cells using magnetic beads (Invitrogen, Carlsbad, CA). Stimulator and responder cell populations were cocultured in serum-free HL-1 medium (BioWhittaker, Walkersville, MD) supplemented with 1 mM l-glutamine and 1 mM antibiotic for 24 h at 37°C in 96-well plates coated with anti–IFN-γ (R4-6A2) or anti–IL-17 (TC11-18H10, BD Biosciences) capture Abs. Cells were washed from the plate, and biotinylated anti–IFN-γ (XMG1.2) or biotinylated anti–IL-17 (TC11-8H4.1, BD Biosciences) was added, followed by antibiotin alkaline phosphatase. Spots were developed with 5-bromo-4-chloro-3-indolyl phosphate and NBT (both from Bio-Rad, Hercules, CA), and total spots per well were quantified using an ImmunoSpot Series 2 Analyzer (Cellular Technology, Shaker Heights, OH).
Statistics
All data were analyzed using GraphPad Prism Pro (GraphPad, San Diego, CA). Replicates were used as indicated. Log-rank testing was performed to determine differences in survival data, and the Student t test was used to determine significance throughout; p < 0.05 was considered a significant difference. Error bars throughout indicate SEM.
Results
CXCL9 is required for maximum generation of donor-reactive, IFN-γ–producing T cells
To directly test whether the presence of CXCL9 influences the functional development of donor-reactive T cells in vivo, groups of wild-type or CXCL9-deficient C57BL/6 mice received wild-type or CXCL9-deficient A/J heart allografts. CD8 or CD4 T cells were purified from recipient spleens on day 7 posttransplant, and the numbers of donor-reactive, IFN-γ–producing cells were enumerated by ELISPOT. Wild-type cardiac allografts placed in wild-type recipients induced more than twice the number of donor-reactive, IFN-γ–producing CD8 and CD4 T cells than did CXCL9−/− allografts placed in CXCL9−/− recipients; in both cases, there were ∼10-fold greater numbers of IFN-γ–producing CD8 T cells compared with CD4 T cells (Fig. 1A, 1B).
In concurrence with the decreased frequency of IFN-γ–producing CD8 and CD4 T cells in the spleens of CXCL9−/− allograft recipients, IFN-γ mRNA expression in CXCL9−/− allografts retrieved from CXCL9−/− recipients on day 7 posttransplant was one third of the level of expression in wild-type grafts retrieved from wild-type recipients (Fig. 1C). Additionally, when graft-infiltrating cells were purified and restimulated in vitro with PMA and ionomycin, fewer CD8 T cells produced IFN-γ when recovered from CXCL9−/− allograft combinations compared with wild-type allograft combinations (Fig. 1D). Taken together, these data show that alloimmune responses generated in the complete absence of CXCL9 induce significantly fewer numbers of donor-reactive, IFN-γ–producing T cells compared with alloimmune responses in the presence of CXCL9.
The data also suggest that there are fewer graft-infiltrating CD8 T cells in the CXCL9−/− allografts retrieved from CXCL9−/− recipients compared with wild-type allografts. At day 5 posttransplant, the number of CD8 T cells was significantly reduced in CXCL9-deficient allografts compared with wild-type allografts; the trend continued at day 7 posttransplant, but it was not statistically significant (Fig. 1E). The number of graft-infiltrating CD4 T cells in CXCL9-deficient allograft combinations was only marginally reduced at days 5 and 7 posttransplant compared with wild-type allograft combinations (data not shown).
Donor- and recipient-derived CXCL9 contribute to CD8 T cell priming
Because the complete absence of CXCL9 resulted in depressed frequencies of donor-reactive T cells producing IFN-γ, the relative contribution of donor- and recipient-derived CXCL9 was tested. Groups of reciprocal allograft combinations were performed with wild-type and CXCL9−/− donors and recipients, and the number of IFN-γ–producing, donor-reactive CD8 T cells was enumerated by ELISPOT on day 7 posttransplant. Although CXCL9−/− recipients of wild-type allografts demonstrated significantly decreased numbers of IFN-γ–producing CD8 T cells compared with wild-type recipients (3,105 ± 951 versus 12,248 ± 1377 spots/106 responders; Fig. 2A), CXCL9−/− grafts placed in wild-type recipients induced a 32% decrease in IFN-γ–producing CD8 T cells compared with wild-type grafts. These data indicate that donor- and recipient-derived CXCL9 influence priming of CD8 T cells, but the primary contribution comes from recipient-derived CXCL9.
CXCL9 can be produced by the cardiac graft endothelium, interstitial/passenger APCs, and infiltrating leukocytes (14, 18). To determine which population of graft cells (endothelium or passenger APCs) is responsible for producing the CXCL9 that influences CD8 T cell priming in the recipient spleen, bone marrow chimeras were generated in which CXCL9−/− A/J mice were lethally irradiated and received wild-type A/J bone marrow. After resting for 8 wk, chimerism was confirmed using PCR to assess CXCL9 expression in PBMCs (Fig. 2B). B6 recipients of CXCL9−/− allografts had ∼26% fewer CD8 T cells producing IFN-γ compared with recipients of wild-type grafts, but the use of a chimeric CXCL9−/− donor graft containing wild-type bone marrow-derived passenger leukocytes restored the number of IFN-γ–producing CD8 T cells to wild-type levels (Fig. 2C).
CXCL9 and CXCR3 are expressed in the graft recipient spleen with similar kinetics
Because CXCL9 plays a clear role in the functional development of donor-reactive T cells to IFN-γ–producing effector cells within the allograft recipient spleen, CXCL9 and its receptor must be expressed in the spleen during this priming. The expression profiles of CXCL9 and its receptor, CXCR3, were measured by quantitative real-time PCR (qRT-PCR) and flow cytometry in wild-type cardiac allograft recipients. CXCL9 mRNA expression in the allograft recipient spleen was biphasic, with increased levels as early as day 2–3 posttransplant and greatly increased levels at 5–6 d posttransplant; CXCL9 mRNA was expressed in isograft recipient spleens 24 h posttransplant but declined thereafter (Fig. 3A). The early expression of CXCL9 in the allograft recipient spleen was due to constitutively high expression by CD11c+ DCs (Fig. 3B), and the contribution of donor-derived CXCL9 was evident because the use of a CXCL9-deficient donor reduced the level of CXCL9 mRNA in the recipient spleen by 50% at 24 and 48 h posttransplant (Fig. 3C). The late burst in CXCL9 expression was due to a significant increase in the mRNA expression by splenic CD19+ B cells, which outnumbered DCs by a ratio of 40:1 (Fig. 3B).
When Ld-reactive 2C transgenic CD8 T cells were adoptively transferred to wild-type recipients 1 d prior to transplant, and CXCR3 expression on the transferred cells was measured at various times posttransplant, CXCR3 expression on the graft-reactive 2C cells in the spleen closely approximated the kinetics of CXCL9 mRNA expression (Fig. 4A). CXCR3 expression on polyclonal CD4 and CD8 T cells in the recipient spleens of wild-type and CXCL9-deficient allograft combinations was measured by flow cytometry. Although the percentage of CD4+ and CD8+ T cells expressing CXCR3 in the wild-type allograft recipient spleen was relatively constant (20–25% of CD8 T cells and 10–15% of CD4 T cells; Fig. 4B) over time posttransplant, the percentage of CD4 and CD8 T cells expressing CXCR3 in CXCL9-deficient allograft recipients continuously increased with time posttransplant (40–45% of CD8 T cells and 20–25% of CD4 T cells on day 5 posttransplant). Because CXCR3 is internalized upon ligand binding, these data suggest that in the wild-type allograft recipient spleen, CXCL9 binds to CXCR3 on CD4 and CD8 T cells, maintaining a constant frequency of CXCR3-expressing T cells in the spleen.
In the allograft recipient spleen, ∼50–60% of divided CD8 and CD4 T cells expressed CXCR3, whereas only 2% of undivided T cells expressed CXCR3 (Fig. 4C). A significant majority of CD8+CD44hi effector T cells recovered from wild-type allograft recipient spleens on day 5 posttransplant expressed CXCR3, but very few naive CD44lo T cells expressed CXCR3 (Fig. 4D).
NK cell IFN-γ production induces early CXCL9 production in the recipient spleen
In murine and human models of inflammation and disease, CXCL9 is specifically induced by IFN-γ (28–30). Thus, because CXCL9 was expressed in the allograft recipient spleen and we have shown that its presence influences the functional phenotype of donor-reactive T cells, the IFN-γ requirements for induction were investigated. In isograft and allograft recipient spleens, IFN-γ mRNA was expressed at elevated levels during the first 3 d posttransplant (Fig. 5A). Given that splenic IFN-γ mRNA levels were elevated in isograft and allograft recipients and that use of an IFN-γ−/− donor did not alter recipient splenic IFN-γ mRNA expression (data not shown), it was hypothesized that an innate immune cell of recipient origin was responsible for IFN-γ production after being activated by danger signals or the systemic mediators of ischemia/reperfusion injury. When isograft and allograft recipients were treated with anti–NK1.1-depleting mAb, spleen levels of IFN-γ and CXCL9 at 24 and 48 h posttransplant were significantly reduced compared with controls (Fig. 5B), and these levels were also reduced in NK1.1-depleted RAG1−/− allograft recipients (Fig. 5C), indicating that the early IFN-γ driving early CXCL9 production was produced by NK cells and not NK T cells.
Because the presence of CXCL9 was important for optimum generation of donor-reactive, IFN-γ–producing T cells, and NK cells were required to induce early CXCL9 in the recipient spleen, the effect of NK cell depletion on donor-reactive T cell development was tested. When wild-type cardiac allograft recipients were depleted of NK cells, the frequency of IFN-γ–producing CD8 T cells in the recipient spleen at day 7 posttransplant was significantly lower (∼50%) than in IgG-treated controls (Fig. 5D). Taken together, these data indicate that NK cells produce IFN-γ in the allograft recipient spleen within 2 d posttransplant, and this subsequently induces the CXCL9 production that influences donor-reactive T cell development in the spleen.
CXCL10 antagonizes CXCL9 during T cell priming
CXCL9, CXCL10, and CXCL11 all bind the same receptor, CXCR3. Because the presence of CXCL9 in the allograft recipient spleen clearly influenced the functional phenotype of donor-reactive T cells, the role of CXCL10/IP10 in this process was tested. Similar to CXCL9 expression in the wild-type allograft recipient spleen, CXCL10 mRNA is expressed at increasing levels with time posttransplant, principally by DCs early and by B cells only by day 5 posttransplant (Fig. 6A, 6B). Interestingly, CXCL10 mRNA expression was relatively insensitive to NK cell depletion with anti-NK1.1 mAb (Fig. 6C).
Because CXCL10 is expressed in the allograft recipient spleen in a similar manner as CXCL9, and CXCL9 influences the functional development of donor-reactive T cells, the effect of CXCL10 deficiency on T cell priming was tested. Unexpectedly, when CD8 and CD4 T cells were purified from CXCL10−/− recipients of CXCL10−/− complete MHC-mismatched cardiac allografts on day 7 posttransplant, there was a marked increase in the number of IFN-γ–producing CD8 T cells compared with wild-type allograft recipients (Fig. 7A).
To further investigate whether and how CXCL9 and CXCL10 strike a balance as antagonistic costimulatory molecules during the functional polarization of donor-reactive CD8 T cells, we attempted to recapitulate our in vivo findings using an in vitro approach. When increasing amounts of rCXCL9 was added to 2C CD8 T cells stimulated with CXCL9−/− syngeneic splenocytes pulsed with Ld peptide (SIYRYYGL), the percentage of IFN-γ–producing cells increased concomitantly (Fig. 7B). The magnitude of increase over baseline was heightened when anti–IP-10 mAb was added to the cultures (Fig. 7B).
To extend these findings in vivo, the number of donor-reactive CD4 and CD8 T cells producing IFN-γ was determined at day 7 posttransplant in wild-type allograft recipients, CXCL10-deficient allograft recipients, and CXCL10-deficient allograft recipients treated with anti-CXCL9 mAb. In agreement with previous findings, the number of donor-reactive, IFN-γ–producing CD8 T cells was significantly higher in CXCL10-deficient recipients of CXCL10−/− allografts compared with wild-type recipients; however, the treatment of CXCL10−/− recipients with anti-CXCL9 mAb reduced the frequency of IFN-γ–producing CD8 T cells to wild-type levels (Fig. 7C, upper panel). These data suggest that CXCL9 promotes the development of IFN-γ–producing CD8 T cell effectors and that CXCL10 impedes this process. Of note, CD4 T cells were relatively insensitive to the presence or absence of CXCL10 (Fig. 7C, lower panel). Taken together with the findings in Fig. 1, these data suggest that CXCL9 and CXCL10 are antagonistic in providing costimulation to influence the effector phenotype of donor-reactive CD8 T cells to IFN-γ–producing cells during priming.
Acute rejection in the absence of CXCL9
Graft-survival studies using CXCL9−/− donors, recipients, or both indicated no prolongation in graft survival in these combinations (CXCL9−/− to CXCL9−/− allograft: mean survival time = 8.5 d; wild-type control allograft: mean survival time = 8 d; data not shown), despite the marked decrease in T cell priming to IFN-γ–producing effector cells. These results raised the possibility that, in the absence of CXCL9, donor-reactive T cell development was skewed to another functional phenotype that was effective in mediating allograft rejection. In such a case, the absence of CXCL9 in the T cell priming environment would not be expected to affect the clonal proliferation of the donor-reactive T cells during priming. To determine whether CXCL9 influences proliferation in response to alloantigen in vivo, Ld-reactive 2C CD8 T cells or polyclonal CD90.1+ CD4 T cells were labeled with CFSE and were adoptively transferred to wild-type or CXCL9−/− CD90.2 B6 recipients of wild-type or CXCL9−/− A/J heart grafts 1 d prior to transplant. Grafts and recipient spleens were recovered on day 5 posttransplant, and proliferation of transferred cells was quantified by CFSE dilution using flow cytometry. Equivalent percentages of CD8 and CD4 T cells (Fig. 8A, 8B, respectively) proliferated in recipient spleens and grafts, regardless of the presence or absence of CXCL9. Moreover, NK cell depletion did not affect donor-reactive CD8 T cell proliferation (Fig. 8C). These data indicate that the decrease in the number of donor-reactive, IFN-γ–producing T cells resulting from CXCL9 deficiency is not due to a defect in clonal expansion of donor-reactive cells in allograft recipients.
Taken together, the above findings suggest that graft rejection in a CXCL9-deficient environment may rely on other effector mechanisms. Initial testing by ELISPOT of CD8 T cells recovered from wild-type and CXCL9-deficient allograft recipients demonstrated no difference in the frequency of donor-reactive cells producing granzyme B or IL-4 (data not shown). However, when CD8 T cells were purified from the spleens of CXCL9−/− recipients of CXCL9−/− cardiac allografts, the frequency of IL-17A–producing cells was significantly higher than CD8 T cells recovered from wild-type allograft recipients, but there was no difference in the frequency of CD4 T cells producing IL-17A (Fig. 9A). Consistent with these results, intragraft cytokine mRNA expression in wild-type or CXCL9-deficient allograft combinations at day 7 posttransplant revealed that IL-17A mRNA expression (but not IL-4 or -5, perforin, or granzyme B; data not shown) was substantially increased in CXCL9−/− allografts placed in CXCL9−/− recipients compared with wild-type allografts (Fig. 9B). Thus, donor-specific production of IL-17 by CD8 T cells may be an alternative mechanism leading to graft rejection in the absence of CXCL9.
Discussion
The inflammatory environment within allografts during rejection is composed of a complex organization of infiltrating cells, cytokines, and other mediators contributing to graft tissue injury. In clinical and experimental transplantation, graft-infiltrating T cell production of IFN-γ is a principal mechanism mediating tissue injury. IFN-γ has direct and indirect effects on the donor graft, including stimulation of macrophages and neutrophils to produce enzymes that generate toxic molecular products and upregulated expression of integrins, MHC molecules, and interferon-inducible chemokines and chemokine receptors (importantly CXCL9, CXCL10, and CXCR3), all of which increase the potency of the alloimmune response. Despite the prominent role of IFN-γ in graft rejection, the factors directing donor-reactive T cell development to IFN-γ–producing cells remain poorly defined.
In the current study, the potential role of CXCR3-binding chemokines in the skewing of donor-reactive T cells to IFN-γ–producing effector cells was directly tested in recipients of MHC-mismatched heart allografts. We report the following novel findings: CXCL9 is produced in the allograft recipient spleen within 24 h posttransplant in a NK cell-dependent manner; in the absence of NK cells or CXCL9, the development of donor-reactive, IFN-γ–producing CD4 and CD8 T cells is impaired; the effector CD8 T cell population is skewed to an IL-17A–producing phenotype in the absence of CXCL9; and a balance between CXCL9 and CXCL10 influences the development of the donor-reactive effector T cell repertoire to IFN-γ–producing effector cells. These results are important in light of recent emphasis on CXCR3 blockade for clinical application to prolong allograft survival (31–35), suggesting that altering the functional phenotype of donor-reactive T cells would be an ancillary effect of CXCR3 inhibition. These current findings support our previous studies testing a CXCR3 small-molecule inhibitor that did not affect alloantigen-primed T cell infiltration into cardiac allografts but did decrease the number of donor-reactive CD8 T cells producing IFN-γ and the level of IFN-γ mRNA in the graft tissue (36). The current report extends and confirms these findings by implicating novel roles for CXCL9 and CXCL10 in the development of donor Ag-reactive T cells to IFN-γ–producing cells.
To date, the extrachemotactic activities of the CXCR3-binding chemokines in Ag-specific T cell priming have been relatively unknown. Studies from other laboratories showed that neutralization of CXCL9 can impair the antidonor IFN-γ response by reactive CD4 T cells and that rCXCL9 can stimulate T cell proliferation in vitro, suggesting that CXCL9 may function as a costimulatory molecule during the priming of T cells to various inflammatory stimuli (21, 37). The current article demonstrates that the absence of CXCL9 significantly decreases the frequency of donor-reactive, IFN-γ–producing CD8 and CD4 T cells (Fig. 1A, 1B), and this corresponds to decreased IFN-γ production in the cardiac allograft. We were unable to demonstrate an effect of CXCL9 on CD4 or CD8 T cell proliferation in vivo (Fig. 8), indicating that CXCL9 influences the polarization of expanding effector T cells rather than proliferation in vivo. Additionally, we demonstrate that donor- and recipient-derived CXCL9 can influence T cell priming. In agreement with the paradigm that interstitial DCs migrate out of the graft to the recipient spleen (4), we show that CXCL9 derived from donor cells of hematopoietic origin are responsible for early CXCL9 production in the recipient spleen (Figs. 2, 3C). Furthermore, the dramatic increase in CXCL9 production by day 5–6 posttransplant in allograft recipient spleens is due, in part, to B cell production of CXCL9 (Fig. 3B).
The role of NK cells in allograft rejection has yet to be fully elucidated. Several studies indicate an important role for NK cells at the interface between innate and adaptive immune responses, demonstrating that NK cells assist in CD4 T cell priming in response to inflammatory stimuli (38, 39). Other studies used semiallogenic graft combinations or costimulation blockade to impair T cell responses to uncover a role for NK cells in allograft rejection (40, 41). Interestingly, we showed that in the spleens of recipients of full MHC-mismatched cardiac allografts, NK cells produce IFN-γ early posttransplant; this induces the production of CXCL9 that influences the polarization of donor-reactive CD8 T cells (Fig. 5). It is important to note that NK cells produce IFN-γ during the first 24–48 h posttransplantation of syngeneic or allogenic cardiac grafts, indicating that NK cell activation occurs as a result of the innate, nonspecific systemic inflammatory response to cardiac transplantation. Although it is unclear from our studies how NK cells are activated to produce IFN-γ posttransplant, increased levels of proinflammatory cytokines like TNF-α are likely to play a role, because blockade of TNF-α depresses early IFN-γ in the spleen (data not shown). Alternatively, NK cells may be activated by so-called alarmins or danger signals like high-mobility group box protein 1, which are released by activated endothelium following ischemia-reperfusion injury of syngeneic and allogenic grafts (42–44).
Mounting evidence from several groups has focused on the mechanisms underlying allograft rejection in the absence of IFN-γ. Although donor-specific production of IFN-γ remains a hallmark of clinical and experimental allograft rejection, uncovering alternative effector cytokines becomes important when antirejection therapy is specifically targeted to limiting IFN-γ–induced rejection. In allograft recipients genetically deficient in IFN-γ, CD8-mediated graft rejection is resistant to costimulation blockade and occurs independently of IL-4 (45). These findings suggest that, in the absence of donor-reactive T cells capable of producing IFN-γ, other effector mechanisms are activated to mediate allograft rejection. Recent studies using Th1-deficient T-bet−/− recipients demonstrated that cardiac allografts were rejected in an IL-17–dependent manner and that the rejected grafts had histological findings consistent with the development of allograft vasculopathy (46, 47). Because CXCL9−/− allograft combinations were rejected at the same rate as wild-type grafts, we investigated whether other effector functions were overexpressed in these groups. IL-17A production by CD8 T cells was significantly increased in CXCL9-deficient allograft recipients, and this correlated with increased levels of IL-17A mRNA in the graft tissue. In contrast to recent reports indicating a role for IL-17–mediated accelerated graft loss due to vasculopathy in an IFN-γ–deficient environment, we did not observe an overt vasculopathy in grafts from CXCL9−/− combinations (data not shown).
Given that CXCL9 and CXCL10 bind CXCR3, we sought to determine whether CXCL10 also played a role in the polarization of donor-reactive CD8 T cells. Surprisingly, the absence of CXCL10 in allograft combinations resulted in a significant increase in the number of IFN-γ–producing CD8 T cells compared with wild-type controls (Fig. 7A). Taken in context with subsequent in vitro experiments (Fig. 7B), these data indicate that CXCL9 and CXCL10 antagonize each other in the allograft recipient spleen to promote effector CD8 T cell polarization. When CXCL9 is neutralized, the effect of unopposed CXCL9 in CXCL10-deficient allograft recipients is abrogated (Fig. 7C). We hypothesize that CD8 T cells use the CXCR3 axis as a costimulatory pathway, in which binding of CXCL9 promotes differentiation to an IFN-γ–producing phenotype, and binding of CXCL10 may promote an IL-17–producing phenotype or may simply inhibit generation of an IFN-γ–producing phenotype, with development to an IL-17–producing phenotype being the default pathway. Current studies are aimed at determining the mechanism that provides a differential signaling cascade through CXCR3, depending on which ligand binds. Reports demonstrating that CXCL9 and CXCL10 bind to CXCR3 in different manners and, at the least, induce different downstream signals to mediate chemotaxis suggest that different signaling mechanisms may influence the costimulatory nature of the receptor (48, 49).
Although our data suggest that CXCL9 and CXCL10 act as antagonistic costimulation molecules, it is possible that an alternative mechanism underlies our findings. Elegant work from the Germain laboratory (50, 51) demonstrated that in models of infection and inflammation, chemokines produced by individual cells within the lymph node create a microchemotactic environment and that CD4 T cell interaction with DCs induces these DCs to produce MIP-1α, which recruits naive CCR5+CD8 T cells to the DC. The chemokines may promote CD8 T cell migration to a cluster of APC-CD4 T cell complexes (52), or chemokines may be produced by competent APCs to recruit naive CD8 T cells following licensing by the CD4 T cells (53). Taken in this context, our results raise the possibility that the CXCR3 axis plays a role in the priming of naive T cells by acting as a chemotactic factor. This may provide a rationale as to why CXCL9 and CXCL10 production in the allograft recipient spleen have such variant effects on the differentiation of T cells. An APC that has been licensed to prime a CD8 T cell to an IFN-γ–producing phenotype may produce high levels of CXCL9. In contrast, CXCL10 production may mark an APC that expresses costimulatory ligands that polarize T cells to an IL-17–producing phenotype. Reconciling our disparate results with CXCL9−/− and CXCL10−/− allograft combinations using the microchemotaxis model would require two separate populations of DCs migrating from the allograft: one producing CXCL9 and one producing CXCL10. Although we have no direct evidence to support either model at this point, we believe that the alternate signaling hypothesis outlined above is more likely, and we have focused our current efforts to test this as a mechanism.
Recent evidence has shown that CXCR3 expression on CD4+ regulatory T cells (Tregs) is regulated by T-bet induction following exposure to IFN-γ during a Th1-type immune response (54). Conceivably, this mechanism may serve to bring regulatory cells to the site of inflammation to dampen aberrant immune responses. Could the expression of CXCR3 on Tregs and the differential signaling via binding of CXCL9 or CXCL10 contribute to the results observed in this study? First, there is little evidence that Tregs function effectively in the face of a complete MHC-mismatched alloimmune response, making this possibility unlikely. Second, in the absence of CXCL9, there does not seem to be any effect on antidonor CD8 T cell proliferation (Fig. 8A), again implicating that Tregs are not effective at dampening this immune response. Finally, we demonstrate a measurable difference in CD8 T cell differentiation in the presence and absence of CXCL9 in vitro, indicating that the effect of CXCL9 and CXCL10 on CD8 T cell priming occurs at the level of the CD8 T cell; however, the possibility that Tregs are involved in this process in a limited MHC-mismatch or a minor Ag-mismatched graft combination in vivo cannot be ruled out.
Collectively, this study provides novel evidence that CXCR3-binding chemokines function as costimulation molecules for CD8 T cells during priming to alloantigen because CXCL9 and CXCL10 can skew the balance of effector phenotypes. In addition to these findings providing novel insights into T cell development following alloantigen exposure, they present important implications and targets for the pharmacological prolongation of graft survival in clinical transplantation.
Acknowledgements
We thank the Cleveland Clinic Biological Resources Unit and their staff for the exceptional care and maintenance of the experimental animals used. Adrian Morelli, University of Pittsburgh, generously provided the peptide used in this work.
Disclosures The authors have no financial conflicts of interest.
Footnotes
This work was supported by National Institutes of Health Grants RO1 AI51620 and AI40459 (to R.L.F.). J.M.R. was supported in part by the Case Western Reserve University Medical Scientist Training Program (National Institutes of Health T32 GM07250) and an individual National Research Service Award (National Institutes of Health F30 HL940052).