Disruption of Transplant Tolerance by an ‘Incognito’ Form of CD8+ T Cell-Dependent Memory

Abstract

Several approaches successfully achieve allograft tolerance in preclinical models but are challenging to translate into clinical practice. Many clinically relevant factors can attenuate allograft tolerance induction including intrinsic genetic resistance, peri-transplant infection, inflammation, and pre-existing anti-donor immunity. The prevailing view for immune memory as a tolerance barrier is that the host harbors memory cells that spontaneously cross-react to donor MHC antigens. Such pre-existing ‘heterologous’ memory cells have direct reactivity to donor cells and resist most tolerance regimens. In this study, we developed a model system to determine if an alternative form of immune memory could also block tolerance. We posited that host memory T cells could potentially respond to donor-derived non-MHC antigens, such as latent viral antigens or autoantigens, to which the host is immune. Results show that immunity to a model non-self antigen, ovalbumin (OVA), can dramatically disrupt tolerance despite undetectable initial reactivity to donor MHC antigens. Importantly, this blockade of tolerance was CD8 T cell-dependent and required linked antigen presentation of alloantigens with the test OVA antigen. As such, this pathway represents an unapparent, or ‘incognito,’ form of immunity that is sufficient to prevent tolerance and that can be an unforeseen additional immune barrier to clinical transplant tolerance.

INTRODUCTION

Clinical applications of tolerance-inducing therapeutics that were developed in preclinical transplantation models (1–4) remain challenging to translate into practice (5, 6). Intrinsic genetic resistance (7–10), pathogen exposure (11, 12), non-specific immune stimulation (13, 14), and pre-existing immune memory (15, 16) each can impede the tolerance process. Alloreactive T cell memory can also block transplant tolerance (15, 17–22), in part because prior autoimmunity or exposure to pathogens or vaccines can generate populations of memory cells that cross-react to any given unrelated MHC allele. Since memory cells resist many tolerance-inducing treatments, this burden of donor MHC-reactive ‘heterologous’ immunity represents an important clinical dilemma.

Here, we explored an alternative pathway for tolerance disruption by immune memory reactive to donor-derived non-MHC antigens. In clinical transplantation, donors often harbor latent infections with any number of different pathogens, such as Epstein-Barr virus (EBV) and cytomegalovirus (CMV) (23–26). Recipients can have corresponding immune memory to non-self antigens, either through microbial exposure or by immunization. Alternatively, a subset of transplant recipients have underlying autoimmune diseases that generate immune memory to non-MHC antigens expressed in donor tissues. The consequence of such pre-existing host immunity in allograft outcomes is often unclear. Specifically, it is not clear whether this form of immune memory is sufficient to disrupt tolerance induction.

We hypothesized that antigen-specific immune memory to donor-derived antigens could be sufficient to impair tolerance independent of anti-donor MHC reactivity or host inflammation. To address this issue, we developed a generic model system in which immune memory was generated to a model non-self antigen, ovalbumin (OVA), and tested for potential impact on tolerance induction. Results demonstrate that anti-OVA memory could disrupt tolerance when OVA antigens were presented in association with donor cells. Importantly, tolerance disruption was CD8 T cell-dependent, required linked presentation of alloantigens and the memory-directed antigen (OVA), and could occur independently of anti-donor MHC alloreactivity within the memory CD8+ T cell pool. As such, this unapparent, or ‘incognito,’ immune memory represents an unanticipated and alternate form of host immunity capable of disrupting transplant tolerance.

MATERIALS and METHODS

Mice

C57BL/6 (B6), B6-Tg(CAG-OVA)916Jen/J (B6-OVA), B6-CD45.1, B6.129S7-Rag1tm1Mom/J (B6-Rag1−/−), and BALB/cByJ (BALB/c) mice were purchased from The Jackson Laboratory. B6-OVA and BALB/c mice were intercrossed and OVA-transgene-positive offspring were backcrossed with BALB/c mice for 5–6 generations to generate OVA-expressing transgenic BALB/c (BALB/c-OVA) mice. B6-CD45.2 OT-1 Rag1−/− transgenic mice were purchased from Taconic Biosciences and intercrossed with B6-CD45.1 mice to generate B6-CD45.1 OT-1 Rag1−/− mice. All mice were maintained under specific pathogen-free conditions at the University of Colorado Anschutz Medical Campus (AMC). Except for BALB/c pancreatic islet donors, all experiments used mice that were bred in-house.

Vaccinations

To generate antigen-specific immunity, randomized groups of 8–12 week old age-matched, co-housed B6 male littermates were injected intraperitoneally (i.p.) with one of three vaccinations. Ovalbumin/adjuvant-primed (OVA/Adj′) mice received 200 μg endotoxin-free ovalbumin (Sigma-Aldrich) plus a subunit adjuvant of 50 μg polyinosinic-polycytidylic acid sodium salt (Sigma-Aldrich) and 50 μg agonistic CD40-specific antibody (FGK45, BioXCell) in phosphate-buffered saline, as previously described (27). Adjuvant-primed (Adj′) mice received the subunit adjuvant alone. BALB/c-primed (BALB/c′) mice received 2 x10⁷ BALB/c splenocytes in Hanks balanced salt solution. Where indicated, CD8+ or CD4+ T cells were depleted during vaccination with CD8α-specific (2.43, BioXCell) or CD4-specific (GK1.5, BioXCell) monoclonal antibodies (10 mg/kg, days −4, −2, 3, and 7 relative to vaccination). Vaccinated B6 mice were used in experiments within 30 to 150 days post-immunization (median 56 days).

Antibodies, tetramer reagents, and flow cytometry

T cells were analyzed for antigen-specific production of intracellular cytokines by stimulating peripheral blood mononuclear cells or spleen cells for 4 hours with gamma-irradiated B6-OVA or BALB/c cells in the presence of Brefeldin A. After blockade of crystallizable fragment (Fc) receptors with anti-FcγRII/III, cells were stained with anti-CD8α (53–6.7) and OVA_257–264/Kb tetramers conjugated to Brilliant Violet 421 (NIH Tetramer Core Facility) (1:100 dilution) at 37°C for 15 minutes. Surface staining used anti-CD45 (30-F11), anti-CD4 (GK1.5), anti-CD44 (IM7), anti-CD62L (MEL-14), anti-CD122 (5H4), anti-CD11α (2D7), anti-CD3ε (145-2C11), anti-CD19 (eBio1D3), and anti-CD45.1 (A20). Intracellular staining used anti-interferon-gamma (IFN-γ) (XMG1.2) with a FoxP3/transcription factor staining kit (eBioscience). All antibodies were from BioLegend, BD Biosciences, or eBioscience. Dead cells were excluded using Fixable Viability Dye (eBiosciences). FACS data was collected using an LSRII and FACSDiva software (BD Biosciences), and analyzed using FlowJo version 9.7.7 software (TreeStar).

Detection of OVA-specific Ig

Immunolon II plates (ThermoScientific) were coated with 100 μg/ml ovalbumin (Sigma-Aldrich) and blocked with 1% bovine serum albumin and 0.05% NaN₃. Sera were serially diluted from 1:50 dilution and assayed for anti-OVA reactivity against the plates by incubation for 2 hours at room temperature. Bound Ig was detected with a goat polyclonal anti-mouse Ig detection antibody conjugated to alkaline phosphatase (SouthernBiotech) and visualized at 405 nm using a para-nitrophenylphosphate substrate (Sigma-Aldrich) and Infinite 200 PRO plate reader (Tecan).

Detection of donor-specific T cells

Purified splenic T cells were obtained using negative selection magnetic enrichment (STEMCELL Technologies). T cells were cultured for 18 hours in serum-free medium at 1 x10³ to 2 x10⁵ cells/well and stimulated with 3 x10⁵ gamma-irradiated BALB/c, B6-OVA, or B6 splenocytes. IFN-γ-secreting cells were detected by enzyme-linked immunospot (ELISPOT) silver staining (U-Cytech biosciences) and counted by a Bioreader 4000 Pro-X plate reader (Bio-Sys Laboratories).

Alloantibody detection

Flow cytometric assessment of donor-reactive IgG antibodies was performed as previously described (28) with modifications. BALB/c, B6-OVA, and B6 thymocytes were blocked with 1% fetal bovine serum, 0.01% NaN₃, and FcγRII/III. Sera (1:25 dilution) from vaccinated mice were incubated with 1 x10⁵ donor thymocytes for 20 min at 4°C, then washed. Bound IgG was detected by staining samples with fluorescein isothiocyanate (FITC)-conjugated goat polyclonal anti-mouse IgG for 20 min at 4°C. After washing, T cells were stained with anti-CD3. The alloantibody production of each serum sample was measured as the percent donor T cells that stained positive with FITC-anti-mouse IgG.

Pancreatic islet cell transplantation

Islet isolation and transplantation was performed as previously described (29). Briefly, female BALB/c donor pancreata were injected with collagenase (Sigma-Aldrich Type V or VitaCyte CIzyme RI), digested by static incubation at 37°C, and purified over Histopaque (Sigma-Aldrich) or Lympholyte 1.1 (Cedarlane Labs) gradients. 400–450 hand-picked islets were grafted in the left renal subcapsular space of streptozotocin-induced diabetic B6 recipient mice. Rejection was defined as the first day of consecutive hyperglycemic blood-glucose values of >270 mg/dl. In long-term euglycemic hosts, confirmation of graft-dependent blood glucose control was determined by nephrectomy of the graft-bearing kidney followed by return to hyperglycemia.

Tolerance-promoting regimens

To induce allograft tolerance using either anti-CD154 monotherapy or donor-specific transfusion (DST), transplant recipients were injected i.p. with CD154-specific monoclonal antibodies (MR-1, BioXCell) at 250 μg/dose either on days −1, 2, 7, and 9 (relative to day 0 transplant) or on days −7, −4, 0, and 4 with intravenous (i.v.) administration of 10⁷ T cell-depleted donor splenocytes on day −7. Spleen cells from OVA-transgene-positive BALB/c donors or transgene-negative littermates were used for BALB/c-OVA DST or BALB/c DST, respectively. For B6-OVA + BALB/c DST, spleen cells from OVA-transgene-positive B6 donors and OVA-transgene-negative BALB/c donors were used and separately injected 1 hour apart.

Memory OT-1 T cells

To generate antigen-experienced OVA-reactive OT-1 T cells, B6-Rag1−/− hosts were adoptively transferred with 1 x10⁷ B6-CD45.1 OT-1 Rag1−/− spleen cells and immunized with ovalbumin/adjuvant vaccine. Within 18 days, transferred T cells were purified from spleen and lymph nodes by magnetic negative selection, checked by flow cytometry to be CD44hi, and 2–4 x10⁶ purified OT-1 cells (73–84% purity) were then injected i.v. into naïve 8–12 week old male B6 mice. All B6 mice that were adoptively transferred with memory OT-1 T cells (OT1_M mice) were bled and analyzed by flow cytometry for OT-1 engraftment. Those with OT-1 cells engrafting at ≥0.2% of peripheral CD8+ CD44hi T cells were used in experiments within 30 to 60 days after adoptive T cell transfer.

Statistical analysis

Statistical analyses were performed using Prism version 6.05 for Windows (GraphPad Software). Results for cell numbers, frequencies, sera Ig titers, and representative ELISPOT data are shown as the mean ± standard deviation (SD); results for pooled ELISPOT data are shown as the mean ± standard error of the mean (SEM) of averaged quadruplicate measures. Differences among groups were analyzed using either an unpaired two-tailed Student t test (2 groups) with Welch’s correction or a two-way ANOVA (>2 groups) with posttest multiple t test comparisons and Holm-Sidak correction. Allograft survival data were analyzed using the Kaplan-Meier method with statistical differences determined by the log-rank test and Cox regression model. P values of less than 0.05 were considered significant.

Study approval

All animal care and experiments conformed to National Institutes of Health guidelines and were approved by the Institutional Animal Care and Use Committee at AMC.

RESULTS

Vaccination generates robust OVA-reactive cellular and humoral immunity

We generated pre-existing host immunity to a model xenogeneic antigen in transplant recipients by vaccinating B6 mice with ovalbumin plus an adjuvant of anti-CD40 and polyinosinic-polycytidylic acid (27). Control B6 mice were immunized with the adjuvant alone. Ovalbumin/adjuvant-primed (OVA/Adj′) mice and control adjuvant-primed (Adj′) mice were analyzed ranging from 4 to 19 weeks post-vaccination for OVA-specific cellular and humoral immunity (Figure 1). Compared with Adj′ mice, OVA/Adj′ mice showed pronounced generation of OVA_257–264/Kb-tetramer specific CD8+ T cells that expressed markers of central memory (Figures 1A and B, and Figure S1). OVA/Adj′ mice also developed tenfold higher titers of serum OVA-specific total Ig than Adj′ mice (Figure 1C). Thus, OVA-vaccinated mice generated robust anti-OVA cellular and humoral immunity.

Figure 1. Vaccination generates robust OVA-specific immunity.

Male C57BL/6 mice were intraperitoneally vaccinated with ovalbumin (200 μg) and adjuvant (50 μg anti-CD40 and 50 μg polyinosinic-polycytidylic acid) (ovalbumin/adjuvant-primed, OVA/Adj′), or for controls, with the adjuvant alone (adjuvant-primed, Adj′). (A) Representative FACS plots showing the percentage of OVA_257–264/Kb-tetramer-specific CD44hi cells among splenic CD8+ T cells from one OVA/Adj′ mouse compared with one Adj′ mouse 55 days after vaccination. (B) Frequencies (left) and absolute numbers (right) of OVA_257–264-specific T cells among CD8+ CD44hi spleen cells from OVA/Adj′ mice (n=12) and Adj′ mice (n=7). (C) Serum titers of OVA-specific antibodies (total Ig) from OVA/Adj′ mice (n=7) compared with Adj′ mice (n=5) as determined by ELISA. Data were pooled from 12 independent experiments. Each symbol in (B) and (C) indicates the data from a single mouse. End-point titers in (C) were measured relative to a baseline normal value taken as the mean ± 3 SD of sera pooled from 6 non-vaccinated B6 mice. Bars represent mean ± SD. ***p<0.001, unpaired two-tailed t statistical test.

OVA-vaccinated B6 mice develop limited cross-reactivity to donor BALB/c MHC and are amenable to allograft tolerance induction

Previous reports indicated that vaccinations or pathogen infections can generate heterologous immunity toward donor MHC (18–21, 30) that can impair tolerance induction in preclinical transplant models (15, 31, 32). We therefore assessed whether the OVA-immunization in B6 mice generated elevated pre-transplant heterologous immunity to the intended allograft donor strain, BALB/c. While purified splenic T cells from OVA/Adj′ mice made a vigorous and rapid IFN-γ response to transgenic B6 cells that constitutively express OVA as a trans-membrane protein (B6-OVA), they did not show elevated reactivity to allogeneic BALB/c cells relative to responses by splenic T cells from control Adj′ mice (Figure 2A).

Figure 2. OVA-vaccinated B6 mice harbor limited detectable cross-reactivity to BALB/c and do not resist CD154-specific antibody-induced allograft tolerance.

(A) Frequencies of IFN-γ producing OVA-specific or donor-specific T cells from vaccinated mice were detected by ELISPOT after stimulating them for 18 hours with OVA-expressing B6 cells (B6-OVA) or allogeneic BALB/c cells. Bars represent mean ± SD of quadruplicate measurements from one OVA/Adj′ mouse and one Adj′ mouse. These results are representative of 5 independent experiments, each with similar results. (B) Detection of OVA-specific and donor-specific antibodies by FACS analysis for percent of B6-OVA and BALB/c target cells staining positive with FITC-conjugated goat anti-mouse IgG secondary antibody that was added after incubating target cells with serum from vaccinated mice. Results from OVA/Adj′ mice (n=6) are compared with results from Adj′ mice (n=5) and mice that were intentionally pre-sensitized to BALB/c (BALB/c′, n=3). Bars represent mean ± SD. (C) Untreated survival of BALB/c islets transplanted into OVA/Adj′ recipients (n=7) compared with Adj′ recipients (n=7) (no significant difference), and BALB/c′ recipients (n=7). (D) CD154-specific monoclonal antibody-induced survival of BALB/c islets transplanted into OVA-vaccinated mice (n=16) compared with control-vaccinated recipients (n=9) (no significant difference), and BALB/c′ recipients (n=7). Each symbol represents an individual mouse. ***p<0.001, 2-way ANOVA statistical test with multiple unpaired t statistical posttests and Sidak correction (A, B), Kaplan-Meier method with log-rank test and Cox regression model (C, D).

Given the potential importance of pre-existing alloantibody in transplantation (33, 34), we also determined whether the OVA-vaccination generated detectable anti-BALB/c antibodies. Relative to sera from control Adj′ mice, sera from OVA/Adj′ mice did not significantly bind BALB/c target cells, despite high reactivity to OVA-expressing B6 target cells (B6-OVA) (Figure 2B). As a positive control, sera from B6 mice intentionally immunized with BALB/c cells showed strong staining of BALB/c target cells (Figure 2B). Taken together, pre-transplant evaluations showed limited detectable pre-existing anti-donor immunity in OVA-vaccinated mice.

We next determined whether anti-OVA memory impacted acute rejection or tolerance induction to BALB/c islet allografts. Acute allograft rejection was comparable between untreated OVA/Adj′ mice and control Adj′ mice (Figure 2C). Conversely, OVA-vaccination did not prevent allograft tolerance induction, as BALB/c graft survival was not significantly different between OVA/Adj′ mice and Adj′ mice treated with CD154-specific monoclonal antibodies (Figure 2D). In stark contrast, allosensitized BALB/c′ recipients demonstrated accelerated acute allograft rejection (Figure 2C) and complete disruption of allograft prolongation following anti-CD154 treatment (Figure 2D). Taken together, data indicate that robust vaccine-induced anti-OVA immunity that is largely devoid of pre-existing anti-donor immunity can be innocuous for acute allograft rejection and tolerance.

Peri-transplant bystander reactivation of OVA-specific memory does not impair allograft tolerance

Considering that a variety of peri-transplant events, including de novo microbial infections and reactivation of latent infections, can stimulate antigen-specific immunity and impair allograft survival and tolerance (11, 12, 35–37), we explored whether bystander peri-transplant reactivation of memory could impact tolerance induction. OVA/Adj′ mice were challenged with B6-OVA cells 7 days prior to receiving BALB/c islet transplants and tolerance-promoting anti-CD154 (Figure 3A). The B6-OVA cell challenge expanded CD8+ CD44hi tetramer+ T cells even in the presence of anti-CD154 treatment (Figure 3B). Despite pronounced bystander reactivation and expansion of OVA-specific memory CD8+ T cells, recipients could nevertheless be tolerized to BALB/c islet allografts (Figure 3C).

Figure 3. Bystander reactivation of vaccine-induced immunity does not block tolerance induction.

(A) Scheme for rechallenging OVA-vaccinated mice using an intraperitoneal injection of 10⁷ OVA-expressing B6 cells (B6-OVA) 7 days before transplanting BALB/c islets and inducing tolerance with CD154-specific monoclonal antibody treatment. (B) FACS plots showing the percentages of splenic OVA-specific CD8+ T cells from OVA-vaccinated mice 7 days after rechallenge with B6-OVA cells and treatment or no treatment with CD154-specific antibodies as described in (A), compared with unmanipulated OVA-vaccinated mice. Results are representative of 3 independent experiments. (C) Survival of BALB/c islets transplanted into OVA-vaccinated mice (n=10) that were rechallenged with B6-OVA cells and treated with CD154-specific antibodies as described in (A), compared with allograft survival data previously shown in Figure 2D of OVA/Adj′ mice (n=16) treated with CD154-specific antibodies and not re-challenged with B6-OVA (no significant difference). NS, not significant. NS≥0.05, Kaplan-Meier method with the log-rank test and Cox regression model.

‘Linked’ presentation of donor antigens and vaccine-directed antigens disrupts tolerance induction

In contrast to tolerance disruption by alloreactive memory (31, 38–40), at this point it would appear that OVA-specific immune memory is essentially irrelevant to allograft immunity and tolerance. However, donors can often express non-MHC antigens, such as pathogen-derived antigens or autoantigens, to which the host is immune. Host reactivity to donor-associated non-MHC antigens could be innocuous, or could represent an alternate means through which immune memory can impair tolerance. Therefore, using our vaccination model, we explored conditions whereby donor-derived cells expressing memory-directed antigens might block tolerance induction in immune recipients without apparent pre-existing immunity to donor MHC.

For proof-of-principle studies, we utilized a standard tolerization protocol (41, 42) in which recipients were treated pre-transplant with anti-CD154 therapy plus varied forms of donor-specific transfusion (DST) (Figure 4A). Following treatment with anti-CD154 therapy and BALB/c DST cells that did not express OVA antigens, OVA-vaccinated mice and control-vaccinated mice showed similar survival of transplanted BALB/c allografts (Figure 4B). Next, to model the scenario of recipient exposure to donor-derived cells that also express non-MHC memory-directed antigens, OVA-immune mice were treated with BALB/c DST cells that transgenically express OVA (BALB/c-OVA). When treated with BALB/c-OVA DST plus anti-CD154, control Adj′ recipients still accepted transplanted BALB/c islets. In striking contrast, OVA-vaccinated mice displayed a profound disruption of allograft prolongation with 14/16 animals showing acute allograft rejection (Figure 4C).

Figure 4. ‘Linked’ DST presentation of BALB/c antigens with vaccine-directed OVA antigens results in tolerance disruption in OVA-vaccinated mice with a gradual difference in alloreactivity.

(A) Scheme for inducing tolerance to transplanted BALB/c islets in vaccinated mice using CD154-specific monoclonal antibodies and an intravascular injection of 10⁷ donor-derived spleen cells for donor-specific transfusion (DST) treatment. (B) Survival of BALB/c islets transplanted into OVA/Adj′ mice (n=9) compared with Adj′ mice (n=8) that were treated as described in (A) with BALB/c DST (no significant difference). (C) Survival of BALB/c islets transplanted into OVA/Adj′ mice (n=16) compared with Adj′ mice (n=18) that were treated as described in (A) with OVA-expressing BALB/c DST (BALB/c-OVA) to expose recipients to the “linked antigens” of donor BALB/c MHC and vaccine-directed OVA antigens expressed on the same cells. (D) Survival of BALB/c islets transplanted into OVA/Adj′ mice (n=7) compared with Adj′ mice (n=7) that were treated as described in (A) with OVA-expressing B6 (B6-OVA) cells plus BALB/c DST to expose recipients to the “unlinked antigens” of donor BALB/c MHC and vaccine-directed OVA antigens on different cells (no significant difference). (E–G) Frequencies of IFN-γ producing donor-specific (anti-BALB/c) splenic T cells from OVA-vaccinated mice were detected by ELISPOT at (E) 7 days following treatment with tolerance-promoting therapies (BALB/c-OVA DST and anti-CD154, given as described in (A)), (F) 7 days after transplantation with BALB/c islets, or (G) the time of allograft rejection by OVA/Adj′ mice (control paired Adj′ mice were non-rejectors). Data were pooled from 3–9 independent experiments. Bars represent mean ± SEM of quadruplicate measurements from OVA/Adj′ mice (n=3–9 per treatment) and Adj′ mice (n=3–9 per treatment). ***p<0.001, Kaplan-Meier method with the log-rank test and Cox regression model (A–D), *p<0.05, multiple unpaired t statistical tests for individual treatments and time points (E–G).

We hypothesized that tolerance disruption required ‘linkage’ of donor MHC and memory-directed antigens on the same donor-derived cells. ‘Linked’ recognition is a well-known property in which T cells can impact either activation (43, 44) or inhibition (45–48) of other T cells interacting with the same APC. To address this issue, we determined the impact of delivering BALB/c and OVA antigens on separate sets of DST cells as a form of ‘unlinked’ antigen presentation. Following sequential treatment with B6-OVA then BALB/c DST plus anti-CD154 therapy, OVA/Adj′ mice and control Adj′ mice predominantly accepted BALB/c allografts (Figure 4D). These results indicate that tolerance disruption requires physical linkage of donor and memory-directed antigens on the same donor cells.

The BALB/c-OVA DST resulted in tolerance blockade in the absence of the memory-directed antigen (OVA) on the islet transplant itself, implying that the recognition of OVA was inciting a response to the linked BALB/c alloantigens despite anti-CD154 treatment. We posited that OVA-specific memory T cells interacted with host APCs that co-presented alloantigens during the initial phase of tolerance induction. Therefore, we next determined the degree and timing of donor-specific alloreactivity potentially triggered by anti-OVA immunity. While one might expect that anti-OVA memory would rapidly license anti-donor reactivity in response to BALB/c-OVA DST, this was not the case. ELISPOT analysis for splenic anti-donor IFN-γ production indicated that BALB/c-OVA DST alone (without anti-CD154) triggered robust alloimmunity within 7 days after DST (day 0 of transplant) in both OVA/Adj′ and Adj′ mice (Figure 4E). In contrast, treatment with BALB/c-OVA DST plus anti-CD154 strongly inhibited initial peri-transplant alloimmunity in both OVA/Adj′ and Adj′ animals during the same time period (Figure 4E). Following islet transplantation, host splenic anti-donor reactivity continued to be restrained in OVA/Adj′ recipients relative to Adj′ mice at 7 days post-transplant (Figure 4F, p=NS). However, a clear demarcation emerged between Adj′ and OVA/Adj′ animals as the latter group progressed toward allograft rejection. Graft rejection in OVA/Adj′ animals (over a period of 8–47 days) was associated with significant anti-donor IFN-γ production by splenic T cells, while paired, non-rejecting Adj′ control animals showed continued restraint of anti-donor reactivity (Figure 4G). Interestingly, a concomitant production of alloreactive antibodies did not occur in either vaccination group after treatment with DST plus anti-CD154 and transplantation (data not shown). Thus, OVA-specific memory did not trigger an immediate allogeneic T cell response following BALB/c-OVA DST and anti-CD154 therapy, but rather this response developed more slowly and was correlated with eventual graft rejection.

To illustrate that DST treatment is not an essential component for this model, we conducted ancillary studies using heart allografts from bone marrow chimeras in which BALB/c donors harbored OVA-expressing hematopoietic cells (denoted as BALB/c-OVA-BALB/c chimeras; Figure S2A). We found that BALB/c-OVA-BALB/c chimeric heart allografts readily disrupted anti-CD154-specific prolongation of graft survival in OVA-vaccinated animals but not in control-vaccinated recipients (Figure S2B). These results demonstrate that the donor graft itself is capable of introducing tolerance-disrupting antigens without a requirement for DST.

Taken together, we find that when donor cells express memory-directed antigens, it is possible for host pre-existing immunity to incite alloreactivity and markedly disrupt allograft tolerance induction. We refer to this kind of immune memory as ‘incognito’ memory, because conventional assays for detecting initial anti-donor reactivity would fail to identify this potential barrier for transplantation, yet it can covertly and unexpectedly disrupt tolerance induction during host exposure to donor and memory-directed linked antigens.

CD8-dependent ‘incognito’ memory blocks tolerance induction

Because OVA-vaccinated recipients harbor both OVA-specific cellular and humoral immunity, we next determined whether vaccine-induced memory T cells, antibodies, or both, were necessary for ‘incognito’ memory to block tolerance induction. To this end, we selectively prevented the initial development of CD4- or CD8-dependent immune memory by depleting either CD4+ T cells or CD8+ T cells during initial vaccination (Figure 5A). Four weeks after T cell depletion and vaccination, T cell populations in CD4-depleted mice but not CD8-depleted mice had reconstituted to nearly normal levels (Figure S3A). Consistent with previous results indicting that our vaccination protocol generates antigen-specific memory CD8+ T cells without CD4+ T cell help (49), CD4-depleted OVA/Adj′ mice developed splenic CD8+ CD44hi tetramer+ T cells that rapidly produced IFN-γ in response to OVA-expressing B6 cells (B6-OVA) (Figure 5B). Moreover, CD4-depleted OVA-vaccinated mice did not develop detectable OVA-specific antibodies (Figure 5C). Conversely, CD8-depleted OVA/Adj′ mice generated high titers of OVA-specific antibodies without developing memory CD8+ T cells (Figures 5B and C). Thus, vaccinated recipients developed selective CD8+ cellular memory or humoral immune memory in CD4-depleted or CD8-depleted mice, respectively.

Figure 5. Host OVA-specific memory CD8+ T cells block tolerance induction in response to linked antigens.

(A) Scheme for depleting CD8+ or CD4+ T cells near the day of vaccination, and then approximately 60 days later, treating the vaccinated mice with BALB/c-OVA DST and CD154-specific antibodies to induce tolerance to transplanted BALB/c islets. (B) FACS plots showing the percentages of IFN-γ producing, CD44hi OVA-specific T cells among CD8+ spleen cells detected by intracellular cytokine staining after 4 hours of stimulation with OVA-expressing B6 cells (B6-OVA). Results are compared with one non-depleted OVA-vaccinated mouse, one CD4-depleted control-vaccinated mouse, and one CD8-depleted control-vaccinated mouse. These results are representative of 3 independent experiments. (C) Serum titers of OVA-specific antibodies (total Ig) from CD4- or CD8-depleted vaccinated mice as determined by ELISA as described in Figure 1C, and compared with non-depleted OVA-vaccinated mice. Each symbol indicates ELISA analysis data from a single mouse and these results contain 5 to 6 mice per group. Bars represent mean ± SD. (D) Survival of BALB/c islets transplanted into OVA/Adj′ mice (n=7) compared with Adj′ mice (n=6) that were CD4-depleted before the day of vaccination and treated as described in (A) with OVA-expressing BALB/c (BALB/c-OVA) DST and CD154-specific antibodies. (E) Survival of BALB/c islets transplanted into OVA/Adj′ mice (n=6) compared with Adj′ mice (n=6) that were CD8-depleted near the day of vaccination and treated as described in (A) (no significant difference). **p<0.01, unpaired two-tailed t statistical test with Welch’s correction (C), Kaplan-Meier method with the log-rank test and Cox regression model (D, E).

Mice with selective CD4-dependent or CD8-dependent anti-OVA memory were then tested for their respective propensity for allograft tolerance induction following treatment with BALB/c-OVA DST/anti-CD154 treatment. Vaccine-induced memory generated in the absence of CD4+ T cells still markedly disrupted tolerance in 7/7 OVA/Adj′ animals despite the absence of detectable anti-OVA antibodies (Figure 5D). Conversely, memory generated in the absence of CD8+ T cells failed to disrupt tolerance (Figure 5E), despite both the presence of high-titer anti-OVA antibodies (Figure 5C) and the ability of these mice to acutely reject BALB/c islets when not treated with a tolerance-promoting therapy (Figure S3B). Therefore, CD8-dependent and not CD4-dependent components of vaccine-induced immunity were essential for ‘incognito’ memory to disrupt tolerance.

‘Incognito’ memory CD8+ T cells do not require heterologous immunity to block tolerance

We next determined whether memory CD8+ T cells that definitively lacked the capacity for anti-donor heterologous immunity could be sufficient to disrupt allograft tolerance. To achieve this goal, we utilized OT-1 Rag1−/− TCR transgenic mice, which generate OVA_257–264-specific CD8+ T cells that are not cross-reactive to BALB/c targets in vitro and do not reject BALB/c skin allografts in vivo (data not shown). Antigen-experienced OT-1 Rag1−/− cells were adoptively transferred into naïve B6 recipients to generate mice with OVA-reactive memory CD8+ T cells lacking donor cross-reactivity (Figure 6A). Memory OT-1 cells engrafted into host B6 mice at similar or somewhat lower frequencies compared to endogenous CD8+ tetramer+ T cells detected in the blood of OVA/Adj′ mice (Figure 6B) and had similar phenotypic profiles to vaccine-induced memory cells (Figure S4).

Figure 6. OVA-specific memory CD8+ T cells without donor-reactivity are sufficient to block tolerance.

(A) Scheme for generating recipients with OVA-specific memory CD8+ T cells devoid of BALB/c-reactivity by adoptively transferring 2–4 x10⁶ antigen-experienced OT-1 Rag1−/− cells into B6 mice (OT1_M), then treating them ~40 days later with donor-specific transfusion (DST) and CD154-specific antibodies to induce tolerance to transplanted BALB/c islets. (B) Composite data showing percentages of engrafted adoptively transferred OVA_257–264-specific OT-1 T cells among CD8+ CD44hi cells in the blood of OT1_M mice (n=11), compared with frequencies of vaccine-induced OVA_257–264-specific T cells in the blood of OVA/Adj′ mice (n=8) and control Adj′ mice (n=8) (no significant difference). Each symbol represents an individual mouse, and bars represent mean ± SD of data pooled from 4 independent experiments. (C) Survival of BALB/c islets transplanted into OT1_M mice treated with CD154-specific antibodies and BALB/c-OVA DST (n=9) compared with control OT1_M mice treated with BALB/c DST (n=9) and control naïve B6 mice treated with BALB/c-OVA DST (n=9). **p<0.01, unpaired t statistical test with Welch’s correction (B), Kaplan-Meier method with the log-rank test and Cox regression model (C).

Animals bearing transferred memory OT-1 cells (denoted as OT1_M mice) were treated with either control BALB/c DST or OVA-expressing BALB/c (BALB/c-OVA) DST plus anti-CD154 and subsequently transplanted with BALB/c islets. Transferred memory OT-1 T cells did not impact BALB/c allograft survival when recipients were treated with BALB/c DST plus anti-CD154 therapy (Figure 6C). However, when recipients were treated with BALB/c-OVA DST plus anti-CD154 therapy, memory OT-1 T cells disrupted long-term allograft survival in 9/9 OT1_M mice (Figure 6C). Memory OT-1 T cells were required for these results as B6 mice accepted the majority of their grafts when they were not adoptively transferred with OT-1 T cells but were treated with BALB/c-OVA DST plus anti-CD154 therapy and transplanted with BALB/c islets (Figure 6C). Taken together, these data demonstrate that memory CD8+ T cells incapable of heterologous donor reactivity are nevertheless sufficient to block tolerance induction in this model system.

DISCUSSION

Conventional screening assays test for cellular and humoral host reactivity to donor MHC molecules but may not predict host immune responses to donor-associated non-MHC antigens, such as microbe-derived antigens or autoantigens. In this study, we determined the impact of immunity to donor-derived non-MHC antigens on tolerance induction. Specifically, we tested the hypothesis that host immunity to non-MHC antigens can disrupt allograft tolerance independently of donor MHC-reactive heterologous immunity. It was essential that our study design distinguish between antigen-specific reactivity and non-specific host inflammation in attenuating allograft tolerance. While peri-transplant host immune stimulation by either pathogens (11, 12) or TLR agonists (13, 14) can prevent tolerance, the ability of pre-existing immune memory itself to impair tolerance may be more variable and less apparent, so we studied its impact independently of tolerance-disrupting pathogen-associated stimuli. We modeled a common clinical scenario in which donor cells harbor non-MHC antigens to which the recipient is immune. By linking a model xenogeneic antigen (OVA) to donor cells, we found that host anti-OVA T cell immunity impaired tolerance and incited the activation of alloreactive T cells despite tolerance-inducing treatments. This type of unapparent, or ‘incognito,’ antigen-specific immune memory was sufficient to dramatically disrupt allograft tolerance.

The observed capacity for host anti-OVA immunity to disrupt allograft tolerance had a number of specific properties. First, the vaccination that generated anti-OVA immunity did not generate detectable donor-specific heterologous immunity and had no discernable impact on either immunity or tolerance to an allograft. Even secondary exposure to OVA-expressing cells during the peri-transplant period was insufficient to disrupt tolerance induction by anti-CD154 therapy. These results illustrate the potentially innocuous impact of antigen-specific immune memory on tolerance induction, because even a simultaneous bystander response to OVA-bearing cells was insufficient to impair allograft tolerance provided that OVA antigens were not expressed on donor cells.

Second, seemingly benign immune memory can have severe consequences on tolerance induction if tolerance-inducing donor cells harbor the memory-directed antigen. This inhibition of tolerance required ‘linked’ expression of donor MHC and the immunizing antigen on the same DST inoculums. While other transplant tolerance studies indicate a role for linked suppression of differing antigens on the same APC for promoting tolerance (45, 48, 50), the current results illustrate the key converse property of linked immune activation (44, 51, 52). Importantly, the activating property of the memory cell population appears dominant to the tolerizing property of anti-CD154 towards the linked, unrelated alloantigens. Thus, we propose that pre-existing immunity to OVA results in the dominant disruption of tolerance to other antigens in the microenvironment during initial alloantigen encounter.

Third, an essential feature of our model involved the expression of the test antigen (OVA) on the initial tolerogen (DST) but not on the subsequent islet transplant. If the donor graft expressed OVA, then tolerance disruption could be the result of tolerance-resistant memory cells directly contributing to allograft rejection. The DST model demonstrated the principle that OVA immunity could incite gradual immunity to unrelated alloantigens akin to epitope spreading (53). Ancillary studies using heart allografts from bone marrow chimeric donors showed that the expression of OVA antigens by donor hematopoietic cells could incite tolerance disruption in OVA/Adj′ animals. Thus, a donor graft itself can introduce tolerance-disrupting antigens without a requirement for DST.

Surprisingly, CD8+ T cells and not CD4+ T cells or antibodies were necessary for this particular form of tolerance blockade. Prior studies showed that alloreactive memory CD4+ T cells could confer tolerance resistance (16, 54). Moreover, in animal models using a similar costimulation blockade approach, memory B cells and anti-donor MHC antibodies can block cardiac allograft tolerance (33, 34). In those studies, humoral immunity was directed toward donor MHC molecules. Importantly, Burns et al. demonstrated that anti-donor MHC antibodies could disrupt tolerance via linked recognition with other alloantigens, similar to the process we observed in the current study (34). In contrast, we found that tolerance could be induced despite high anti-OVA antibody titers. The difference in the specificity of memory T cells and pre-formed antibodies for donor non-MHC antigens versus MHC antigens could be an important distinction between these differing studies. Moreover, CD8+ T cell dependent memory may not be a universal feature of how donor non-MHC-directed immunity can impair tolerance. The mechanisms through which donor expression of microbial antigens or autoantigens activate specific humoral or cellular immunity in the host might dictate which immune memory pathways dominate in disrupting tolerance.

A key question centers on how this type of immune memory results in tolerance blockade. Developing or expanding cross-reactive cells from populations of memory CD8+ T cells was not required to prevent tolerance, because OVA-specific OT-1 Rag1−/− CD8+ memory T cells that do not exhibit detectable donor reactivity were sufficient to confer tolerance resistance to naïve B6 hosts. Given that memory OT-1 T cells could not react to the donor MHC to directly mediate graft rejection, the ability of this type of ‘incognito’ immune memory to prevent tolerance is best explained by classical linked recognition, in which memory cells respond to APCs that simultaneously present OVA antigens and donor antigens (Figure 7). Analogous to APC licensing during CD4+ T cell help for CD8+ T cells, we propose that memory cells alter initial alloantigen presentation and subsequently trigger the activation of donor-reactive T cells that are pathogenic to the graft. We envision that memory blockade interferes with an as yet unidentified checkpoint during fate decision toward tolerance. ‘Incognito’ memory does not immediately license the activation of donor-reactive T cells following DST in the presence of anti-CD154 treatment (Figure 4E) or result in accelerated allograft rejection (Figure 4C). Instead, the licensing of donor reactivity occurs more gradually, indicating a change in ongoing T cell fate decision from tolerance to immunity (Figure 7). We must emphasize that ‘incognito’ memory is not a new phenotype or subset of memory T cells. We use the term ‘incognito’ simply to connote the pathway and consequences of memory cell responses to donor non-MHC antigens that can be unapparent pre-transplant and yet still disrupt transplant tolerance induction.

Figure 7. Proposed model for how ‘incognito’ memory T cells interfere with tolerance induction via linked antigen recognition.

Donor-associated linked antigens (Ag), A and B (such as OVA and BALB/c, respectively), are acquired by host APCs and presented on host MHC. (A) During normal tolerance induction, naïve donor-reactive T cells are controlled by anti-CD154 therapy, which influences fate decision toward tolerance. (B) During ‘incognito’ memory blockade, memory T cells interact with APCs that co-present the memory-directed antigen and alloantigens, and alter the microenvironment of early donor antigen presentation. This disrupts tolerance-promoting processes and alters the fate of naïve donor-reactive T cells from tolerance to immunity.

Although we intentionally developed a generic model to illustrate a principle of tolerance disruption by memory cells, the results raise an important dilemma for the prospect of inducing allograft tolerance in most settings of conventional transplantation. The donor will likely harbor a number of metagenome-encoded antigens (55) to which the recipient may have pre-existing immunity that could be deleterious to future tolerance-promoting regimens. Of course, the tissue distribution of pathogen-associated antigens could influence whether immune memory disrupts tolerance induction. In the current study, we modeled one scenario in which the non-self antigen was restricted to hematopoietic cells, as would occur in EBV infection (25). In another example, recipients may be intentionally vaccinated against donor-derived pathogens, such as CMV, and then transplanted with pathogen-infected donor organs (56). An inadvertent consequence of eliciting protective CMV-reactive immune memory is that ‘linked antigen’ reactivity could incite graft rejection in the immune host. It would be intriguing in future studies to determine if hematopoietic versus parenchymal distributions of donor non-self antigens have differing effects on tolerance induction in the immune host. Finally, in the setting of autoimmunity, pre-existing memory to non-MHC autoantigens might also invoke tolerance disruption and trigger the development of alloimmunity due to co-presented allo- and autoantigens. This process could partly explain the severe resistance of autoimmune recipients to allografts. Taken together, ‘incognito’ memory represents an additional and potentially significant barrier to transplantation tolerance that is characterized by not requiring demonstrable pre-transplant anti-donor MHC heterologous immunity.

Supplementary Material

Supp info

Figure S1: OVA-specific CD8+ T cells from OVA-vaccinated mice express markers of central memory. FACS plots showing relative surface expression of CD44, CD62L, CD122, and CD11α (LFA-1) on splenic OVA257–264-tetramer specific CD8+ T cells (black histograms) compared with tetramer-negative CD8+ T cells (gray histograms) analyzed from the same OVA/Adj′ mouse. Results are representative of 4 independent experiments.

Figure S2: Expression of ‘linked’ antigens by donor hematopoietic cells in heart allografts results in tolerance disruption in OVA-vaccinated mice. (A) FACS analysis of OVAexpression by CD45+ cells in blood 6 weeks following BALB/c or BALB/c-OVA bone marrow injection into irradiated BALB/c recipients. Results show that bone marrow reconstitution in BALB/c-OVA-BALB/c chimeras (n=7) generates OVA-expressing CD45+ cells at levels comparable to intact BALB/c-OVA mice (n=4). (B) OVA-specific memory blocks tolerance induction by anti-CD154 monotherapy following transplantation of heart allografts from OVAexpressing BALB/c bone marrow chimeras (BALB/c-OVA-BALB/c) (n=5). Results are compared with OVA-vaccinated recipients of control BALB/c-BALB/c bone marrow chimeric heart allografts (n=7). *p<0.05, Kaplan-Meier method with the log-rank test and Cox regression model.

Figure S3: Reconstitution of CD4+ or CD8+ T cells in B6 mice depleted near the time of vaccination. (A) FACS analysis 40–60 days following T cell depletion and vaccination showing frequency of CD4+ or CD8+ T cells. Compared to non-depleted mice, depleted CD4+ T cell populations reconstituted to nearly normal levels. Depleted CD8+ T cell populations were not fully reconstituted, however, frequencies were not different between CD8-depleted Adj′ mice and CD8-depleted OVA/Adj′ mice. Results are from n=3 mice per treatment group from 3 independent experiments. (B) Although not fully reconstituted to normal frequencies of CD8+ T cells, untreated CD8-depleted mice retained the ability to reject BALB/c islet allografts.

Figure S4: Primed TCR transgenic OT-1 (OVA-specific) CD8+ T cells are phenotypically similar to endogenous OVA-specific CD8+ T cells from vaccinated OVA/Adj′ mice. FACS plots showing relative surface expression of CD44, CD62L, CD122, and CD11α (LFA-1) on splenic OVA257–264-tetramer specific CD8+ T cells from a B6 host adoptively transferred with primed OT-1 cells (OT1M) or from an OVA/Adj′ mouse. Results are representative of 4 independent experiments.

Abbreviations

Adj′

adjuvant-primed

BALB/cByJ

BALB/c

BALB/c′

BALB/c-primed

C57BL/6

B6

CMV

cytomegalovirus

DST

donor specific transfusion

EBV

Epstein-Barr virus

ELISA

enzyme-linked immunosorbent assay

ELISPOT

enzyme-linked immunospot

FITC

fluorescein isothiocyanate

IFN-γ

interferon gamma

i.p

intraperitoneal

i.v

intravenous

OVA

ovalbumin

OVA/Adj′

ovalbumin/adjuvant-primed

SD

standard deviation

SEM

standard error of the mean