Murine acute graft-versus-host disease can be prevented by depletion of alloreactive T lymphocytes using activation-induced cell death

Graft-versus-host disease (GvHD) represents a major complication of allogeneic bone marrow transplantation (BMT), which results in significant morbidity and mortality. Donor T lymphocytes present in the BM or peripheral blood stem cell transplant cause GvHD by recognizing major as well as minor histocompatibility antigens (reviewed by Ferrara et al1). Although innovative strategies for the prediction, prevention, or treatment of this caveat in allogeneic BMT have been proposed,2-9 prophylaxis or treatment of GvHD still remains unsatisfactory. Removal of donor-derived T lymphocytes prior to transplantation efficiently ameliorates GvHD. However, ex vivo depletion of T cells has been associated with an increased risk of viral opportunistic infections,10 impaired engraftment,11 and the loss of the graft-versus-leukemia (GvL) effect, resulting in an increased incidence of recurrence of leukemia.12,13 These observations emphasize the clinical importance of allogeneic T cells in conferring a protective antiviral immunity as well as mediating a GvL effect. Several clinical and experimental studies support the idea that GvHD and GvL are not only mediated by the same (host-specific) T cell but also by distinct (leukemia-specific) alloreactive donor T-cell populations.14-16 Moreover, allo–major histocompatibility complex (MHC)–restricted cytotoxic T lymphocytes (CTLs) to antigens frequently expressed at elevated levels on leukemic cells can be cloned from an allorestricted T-cell repertoire and used to mediate a GvL reaction without causing GvHD.17-19 However, immunotherapy with ex vivo–generated specific T cells requires the identification and characterization of tumor-specific antigens to be targeted. To circumvent this prerequisite of cloning leukemia-specific T cells, we aimed to develop a complementing strategy that would allow ex vivo elimination of alloreactive T cells responsible for GvHD while preserving an allorestricted T-cell population able to both mediate GvL and antiviral immunity effects. Because the activation-induced cell death (AICD) mediated by CD95/CD95 ligand (CD95L) (reviewed by Lenardo et al20) represents an important physiologic pathway to control the expansion of antigen-activated T cells at the down phase of an immune response, it therefore might provide a means to selectively deplete activated alloreactive T lymphocytes. Induction of AICD triggered by repetitive T-cell receptor (TCR) re-engagement in the presence of interleukin-2 requires antigen-specific activation of the T cell and appears not to affect bystander T cells.21-23Thus, inducing AICD in alloreactive T lymphocytes by repetitive stimulation with alloantigen ex vivo prior to transfer should allow us to leave T cells unaffected that confer antiviral or GvL effects.

In this study we tested this approach by the induction of apoptosis in activated alloreactive T lymphocytes in an allogeneic mixed lymphocyte culture (MLC) using an agonistic monoclonal antibody (mAb) to CD95 followed by transfer of residual allogeneic T cells in a lethal P to F₁(P→F₁) GvHD model system (H-2^k→H-2^dxk). We report here that proliferative and lytic alloreactive responses of donor T cells can be substantially decreased by AICD specifically to a given alloantigen. In addition, depletion of activated alloreactive T cells prior to adoptive transfer of allogeneic T cells into F₁ recipients prevents the occurrence of lethal GvHD, which invariably occurred in untreated controls.

The results presented suggest that elimination of alloreactive donor T lymphocytes following activation by CD95 death receptor–mediated depletion might provide one approach to modify the T-cell repertoire in allogeneic transplants and complement strategies to include pathogen-specific T lymphocytes and in vitro–generated tumor-specific T cells for immunotherapy in an allogeneic BMT.

C3H/HeJ (H-2^k) mice, BALB/cJ (H-2^d) mice, (C3H/HeJ × BALB/cJ) F₁ (H-2^kxd) hybrid mice, and BALB/cJ OVA^323-339-specific TCR transgenic mice originally obtained from Dr Loh and colleagues24 were bred and maintained at the Central Animal Facility of the University of Mainz. All mice strains were subjected to regular health status controls and were free of commonly tested viruses, parasites, and bacteria.

The animals were housed in sterilized microisolator cages with filter tops, and experimental procedures were performed according to German federal and state regulations and the National Institutes of Health Guide for the Care and Use of Laboratory Animals.

The age of mice to be used for experiments ranged from 8 to 14 weeks. For adoptive transfer experiments, only female F₁ mice were taken as recipients.

The A20 tumor cell line represents a B-cell lymphoma derived from BALB/c mice. The thymoma cell line EL-4 was derived from C57BL/6 (H-2^b) mice.

The tumor cell lines A20 and EL-4 were propagated in RPMI 1640 medium (Biomed, Berlin, Germany) supplemented with 5% fetal calf serum (FCS) (Gibco, Ichinnan, United Kingdom), 5 mM/Ll-glutamine, 0.1 mM β-mercaptoethanol, 1% antibiotic-antimycotic (Gibco) solution consisting of penicillin (10 000 IU)/streptomycin (10 mg/mL)/amphotericin B (25 μg/mL), 1% sodium bicarbonate, 1 mM/L sodium pyruvate, and 0.8% nonessential amino acids, and maintained in flasks during culture. Preparation of cell suspensions made from ex vivo–isolated cells was conducted in Dulbecco modified Eagle medium (DMEM) (Biomed) containing 2% FCS or horse serum (Gibco) and 1% antibiotic-antimycotic (Gibco). Culture medium for mixed lymphocyte reactions (MLC medium) consisted of RPMI 1640 (Biomed) supplemented with 10% FCS (Gibco) and 5% NCTC (Gibco) in addition to supplements used for cell culture medium.

Biotin-conjugated antibodies to H-2^d and H-2^k, fluorescein isothiocyanate (FITC)–labeled anti-CD3 and anti-CD4 antibodies, phycoerythrin-conjugated anti-CD8 antibody, and appropriate isotype controls were purchased from BD Biosciences (San Diego, CA). FITC- and phycoerythrin-labeled streptavidin was from Jackson ImmunoResearch (West Grove, PA). MicroBeads used for immunomagnetic separation of T cells were from Miltenyi (Bergisch Gladbach, Germany). The anti CD3 (145-2C11) mAb and the anti-CD95 antibody (Jo2) were obtained in a low endotoxin and culture grade from BD Biosciences; isotype-matched hamster immunoglobulin (Ig) and polyclonal goat anti–hamster Ig were purchased from Serotec (Oxford, United Kingdom). Protein G used for cross-linking of Jo2 mAb was from Sigma (Deisenhofen, Germany). The ovalbumin (OVA) peptide (amino acids 323-339 from chicken ovalbumin) was obtained from the Department of Immunohematology, Leiden University Medical Center, The Netherlands.

Single-cell suspensions from spleen or a mixture of spleen and lymph node cells derived from inguinal, axillary, and submandibullary lymph nodes were prepared in complete DMEM followed by red cell lysis using NH₄Cl treatment. Purification of CD90⁺, CD4⁺, or CD8⁺ responder T cells from C3H/HeJ or BALB/cJ mice was carried out using a magnetic cell separation system (MACS) according to the manufacturer's instructions (Miltenyi). Cells were incubated with either anti-CD90–, anti-CD4–, or anti-CD8–conjugated MicroBeads in MACS buffer (phosphate-buffered saline [PBS] containing 0.5% FCS) for 20 minutes at 4°C. After removal of excess beads by washing, the cell suspension was passed through a prewashed VS⁺ separation column placed in a magnetic field, and cells coated with MicroBeads were selected to be transferred into MLC medium. Preparations of T cells were periodically stained with anti-CD90, anti-CD4, or anti-CD8 antibody to be analyzed by flow cytometry. Purity of the enriched T cells always exceeded 95%.

Highly purified ex vivo–isolated naive CD4⁺, CD8⁺, or CD90⁺ responder T cells were mixed with irradiated (30 Gy [3000 cGy]) splenic stimulator cells from BALB/c mice (H-2^d) or F₁ mice (H-2^dxk) at concentrations of 1 × 10⁶/mL and 5 × 10⁶/mL, respectively. They were cultured for 5 to 6 days in MLC medium in 96-well round-bottom plates in the absence or presence of various concentrations of soluble agonistic anti-CD95 mAb (Jo2). Depletion of alloreactive T cells for adoptive transfer studies was performed in a bulk culture in 24-well plates generally culturing 2.5 × 10⁶ responder cells with 7.5 × 10⁶irradiated stimulator cells. To enhance its proapoptotic effect, microtiter plate wells were also coated with anti–hamster Ig at a concentration of 5 μg/mL in PBS prior to addition of anti-CD95 antibody to facilitate cross-linking. Alternatively, protein G was used at 2 μg/mL. In case of a secondary MLC, responder T cells were either prestimulated in vitro with irradiated allogeneic splenic stimulators or immobilized with anti-CD3 antibody in bulk cultures for 3 days or in vivo by intraperitoneal injection of 5 × 10⁶ irradiated BALB/c spleen cells 10 to 14 days prior to isolation. T-cell blasts were then harvested from culture or isolated from spleen and lymph node suspension by immunomagnetic separation using MACS as described above to be restimulated in a bulk culture MLC.

When OVA-specific T cells (H-2^d) were used to evaluate the specificity of depletion, CD4⁺ cells were isolated from OVA-TCR transgenic mice by MACS technology as described above and mixed with BALB/c T-cell responder at ratios of 1:1, 1:5, or 1:10 to be cultured with irradiated allogeneic F₁ splenic stimulators in the presence of Jo2 mAb. Residual T cells were then not only challenged with F₁ cells but also stimulated with these F₁ cells in the presence of OVA peptide (0.5 μg/mL).

Depending on a primary or secondary MLC or T-cell stimulation by peptide antigen, 2 or 3 days after beginning the culture the cells were harvested every day following an overnight pulse with 1 μCi (0.037 MBq) per well of ³H-thymidine. Radioactivity incorporated into DNA was measured on a β-plate liquid scintillation counter (Wallace, Turku, Finland).

A total of 2 × 10⁶ A20 (H-2^d) or EL-4 (H-2^b) tumor cells was labeled with 100 μCi (3.7 MBq) of⁵¹Cr for 4 hours. After washing 3 times in PBS, labeled targets were plated at 1 × 10⁴ cells per well in U-bottom microtiter plates (Nunc, Wiesbaden, Germany). Allogeneic CD90⁺ responder T cells (H-2^k), preactivated in a primary MLC to H-2^d spleen cells for 5 days, were washed 3 times and added to triplicate wells at varying effector-target ratios to be incubated for 5 hours. Maximum and spontaneous release were determined by incubation of targets with Triton X-100 (Sigma) or addition of media, respectively. ⁵¹Cr activity in the supernatants taken 5 hours later was determined in an auto γ counter (Packard Instruments, Dreieich, Germany), and lysis was expressed as a percentage of maximum release.

Acute and lethal GvHD was induced in a classical P→F₁ model using C3H/HeJ mice (H-2^k) or BALB/cJ mice (H-2^d) as donor of T cells and (C3H/He × BALB/c) F₁ mice (H-2^kxd) as recipients. Following depletion of alloreactive T cells in bulk culture MLC in the presence of agonistic anti-CD95 antibody for 5 to 6 days as described above, remaining T cells were harvested to be purified by immunomagnetic separation on a MACS column. Single suspensions of highly purified in vitro–activated viable (trypan blue–excluding) CD4⁺ or CD8⁺ T cells were then injected at various concentrations through the tail vein into female F₁ recipients on day 0. Mice were sublethally total body¹³⁷Cs-irradiated (6.5 Gy [650 cGy]) 4 to 5 hours prior to adoptive transfer of T cells. Control mice consisted of age- and sex-matched F₁ mice that received syngeneic spleen cells or untreated allogeneic T cells, respectively.

For BMT studies, BM was obtained aseptically by flushing femurs and tibias of C3H/HeJ donor mice with complete DMEM, and T-cell depletion was carried out by immunomagnetic removal of CD90⁺ cells using MACS (Miltenyi) as described above. T-cell–depleted BM cells (2 × 10⁷ cells per mouse) were transplanted with highly purified CD4⁺ or CD8⁺ donor T cells or T cells depleted of alloreactivity by AICD (2.5 × 10⁶ per animal), respectively, by tail-vein infusion into lethally irradiated F₁ recipients on day 0. F₁ mice received 13 Gy (1300 cGy) total body irradiation as a split dose (with 3 hours between the 2 doses to reduce gastrointestinal toxicity) 3 to 4 hours prior to BMT.

Recipient mice were monitored for clinical signs of GvHD, including diarrhea, alopecia, and hunched posture, and their weight loss was determined. At least 4 recipient mice were included in each experimental group, and adoptive transfer was repeated at least 3 times.

Liver, spleen, lymph node, small intestine, kidney, and ear skin sections obtained from mice on various days after adoptive transfer of untreated donor T cells or donor T lymphocytes depleted of alloreactive responders by AICD were either fixed in 10% buffered formalin and embedded in paraffin to be stained with hematoxylin and eosin for microscopic examination or embedded with OCT medium (Tissue-Tek, Zoeterwoude, The Netherlands) to be stored at −80°C for immunohistochemistry. Slides of cryostat-sliced (5 μm) samples were first incubated with rat anti-CD8 mAb at 5 μg/mL (BD Biosciences) or rat anti-CD4 mAb at 2.5 μg/mL (BD Biosciences) overnight at 4°C, followed by rinsing in PBS and incubation with a secondary antibody (0.01 mg/mL biotinylated anti–rat IgG) (Dako, Hamburg, Germany) at room temperature for 30 minutes. The antigen-antibody binding was detected using an avidin-biotin alkaline phosphatase (AP) system (Dako). The slides were immersed briefly in hematoxylin for counterstaining and evaluated under light microscopy.

Generally, 1 × 10⁶ viable T cells per sample, isolated from spleen and lymph node preparations by immunomagnetic separation using CD4-, CD8-, or CD90-conjugated MicroBeads (Miltenyi), were incubated for 15 minutes at 4°C with the indicated antibodies. The antibodies were either directly conjugated or biotinylated to be incubated with streptavidin-conjugated FITC or phycoerythrin as second-step reagents, followed by 2 washes (PBS containing 2.5% FCS and 1% sodium azide) between incubations. Viable lymphocytes were gated either by forward and side scatter or on a life gate using 7–aminoactinomycin D (Molecular Probes, Leiden, The Netherlands), and 2-color fluorescence analysis was performed on 10 000 cells using an Epics Altra instrument (Beckman Coulter, Stanford, CA) with Expo 32 software.

All values are expressed as mean ± SD. Statistical analysis of survival data was performed by the Mantel-Cox log-rank test.P < .05 was considered statistically significant.

Because both CD4⁺ and CD8⁺ T cells undergo CD95-mediated activation-induced apoptosis after prolonged activation,20-23 we investigated the possibility of deploying AICD as a method for selective depletion of alloreactive T cells in an allogeneic mixed lymphocyte reaction. Naive T cells isolated ex vivo from C3H/He (H-2^k) mice were highly purified into CD4⁺ or CD8⁺ T-cell subpopulations by immunomagnetic separation using the MACS technology and cultured together with irradiated unfractionated splenic stimulator cells from BALB/c (H-2^d) mice in the presence of increasing concentrations of soluble agonistic mAb (Jo2) to CD95 for 5 to 6 days. Jo2 has previously been shown to induce apoptosis in lymphocytes.25 Following stimulation to the H-2^d alloantigen, the proliferative responses of both CD4⁺ or CD8⁺ alloreactive T lymphocytes were reduced by up to 90% in the presence of anti-CD95 mAb in a dose-dependent fashion when compared with the proliferation of untreated allogeneic controls (Figure 1). Reduction of proliferative responses was specific for anti-CD95–mediated apoptosis of T cells, because responder cells stimulated in the presence of an anti-CD95 mAb-matched hamster Ig isotype control showed comparable responses to untreated allogeneic controls (Figure 1). Flow cytometry analyses of CD90⁺responders isolated at daily time points during stimulation with T-cell–depleted allogeneic stimulators (H-2^d) in a 6-day bulk culture by immunomagnetic separation revealed an increased CD95 expression within the first 36 to 48 hours following activation and an increase of apoptotic CD4⁺ and CD8⁺ T lymphocytes upon treatment with agonistic anti-CD95 mAb as measured by annexin V and 7–aminoactinomycin D viability staining.

Because T cells have been reported to be initially resistant to AICD upon primary activation but elicit increasing susceptibility to CD95-mediated induction of apoptosis 4 to 5 days20-23 upon sustained TCR re-engagement, we further investigated the induction of AICD on naive versus preactivated T cells.

Following polyclonal activation in vitro by TCR cross-linking through coated anti-CD3 mAb (145-2C11) for 3 days, CD4⁺ T-cell blasts as well as freshly isolated resting CD4⁺ T lymphocyte mice were stimulated with alloantigen (H-2^d) in a bulk culture MLC in the presence of agonistic mAb to CD95 for 5 days to be tested for residual response to the stimulating alloantigen in a mixed lymphocyte reaction. Upon induction of AICD, both primary stimulated (Figure 2A) or restimulated (Figure 2B) T-cell responders showed a marked decrease in their capacity to proliferate to allogeneic splenic stimulators regardless of their preactivation state (note the reduced proliferation of 85% both in Figure 2A and 2B on day 3) in contrast to their untreated responder controls, although with an earlier kinetic in the response to H-2^d for anti-CD3–preactivated cells.

These results suggested that induction of AICD at least in an allogeneic situation upon repetitive stimulation of T lymphocytes with MHC alloantigen was not dependent on the preactivation state of the alloreactive T cells.

To evaluate whether alloantigen-specific T-cell depletion by AICD would retain T-cell specificities to defined nonalloantigens, we chose to add I-A^d-restricted OVA^323-339-specific CD4⁺ T cells isolated from BALB/c OVA-TCR transgenic mice24 to allogeneic responder T cells from BALB/c mice (H-2^d) at different ratios. These T cells should not be activated in the absence of OVA and thus not be susceptible to AICD. Following culture with splenic stimulators from F₁ mice (H-2^dxk) in the presence of cross-linked anti-CD95 mAb, remaining responder T cells were then challenged with F₁stimulators as well as with BALB/c-derived spleen cells presenting OVA peptide.

As expected from the above findings, restimulation to H-2^kalloantigen in vitro resulted in an impaired proliferative response by these T cells compared with untreated controls (Figure3A, in vitro). The proliferation of untreated T cells to H-2^k observed in Figure 3 was truly allospecific because T cells from the OVA transgenic mice were shown not to cross-react with H-2^kxd (data not shown). In contrast, proliferation of the responder cells to OVA was preserved in both anti-CD95–treated and untreated T-cell responder populations (Figure 3A, in vitro).

In addition, challenge of purified CD4⁺ and CD8⁺ T cells isolated ex vivo from F₁recipients 6 days after adoptive transfer of a mixture of OVA-specific T cells and allogeneic T-cell responders of BALB/c donors either left untreated or depleted of alloreactive cells by AICD into irradiated F₁ mice confirmed the findings obtained in vitro. Again, the response of pretreated responders to H-2^k alloantigen was reduced in contrast to untreated controls (Figure 3B, ex vivo), but stimulation of OVA-specific T cells by OVA peptide–presenting BALB/c spleen cells appeared not to be significantly different (Figure 3B, ex vivo) in both responder populations.

These results strongly suggested that induction of AICD using the agonistic mAb Jo2 to CD95 was specific in depleting alloreactive T cells to a given alloantigen while retaining T cells specific for non-MHC antigens like OVA.

We further tested whether induction of AICD by anti-CD95 mAb would also reduce the cytolytic responses of alloreactive T cells. Following stimulation of purified CD90⁺ responder T cells from C3H/He mice with irradiated allogeneic splenic stimulators from BALB/c mice in the presence of anti-CD95 mAb (1 μg/mL) in a bulk culture MLC for 5 days, remaining T lymphocytes were recovered from culture after treatment and tested in a cytolytic assay using chromium-labeled A20 cells as targets. The capacity of residual CTLs to kill A20 targets was reduced by at least 80% compared with untreated alloreactive control T cells, which exhibited 50% to 60% specific lysis at an effector-target ratio of 3:1 (Figure 4). This reduction in alloreactive cytoxicity after induction of AICD was again specific for the alloantigen H-2^d because there was no significant lysis of EL-4 target cells expressing H-2^b(Figure 4). Taken together, these findings and the results presented above strongly suggested that the induction of alloantigen-dependent apoptosis in an activated allogeneic T-cell pool by administration of a CD95 agonistic mAb results in marked depletion of alloreactive T cells to a given alloantigen without grossly affecting T-cell specificities that have not previously been activated in an alloantigen-dependent fashion.

To verify the efficacy of depleting alloreactive T cells by AICD in vitro, we investigated the potential of residual T cells to evoke a GvHD in the parent C3H/He (H-2^k)→F₁(BALB/c × C3H/He, H-2^dxk) murine GvHD model.

Naive allogeneic CD90⁺ donor T lymphocytes or CD90⁺ T cells primed to H-2^d BALB/c spleen cells in vivo 14 days prior to isolation from donor mice were cultured with irradiated BALB/c-derived splenic stimulators in bulk MLC in the presence of Jo2 mAb for 5 to 6 days. Remaining responders were separated into CD4⁺ or CD8⁺ donor T cells by immunomagnetic separation to be injected into the tail vein of F₁ recipients irradiated (6.5 Gy [650 cGy]) within 6 to 8 hours prior to transfer. During the preparation of T-cell responders, extensive care was taken to wash out excessive anti-CD95 mAb. Furthermore, viability and purity of cell preparations was assessed by flow cytometry as described in “Materials and methods.” A dose response of donor T cells depleted of alloreactivity or left untreated was performed, and F₁ mice that received 1 × 10⁶ to 5 × 10⁶ in vivo–primed and in vitro–restimulated donor T lymphocytes invariably developed an acute GvHD between days 9 to 18 after transfer depending on the amount of T cells injected. GvHD was clinically characterized by rapid weight loss, severe diarrhea, and hunched posture, and all mice died within 20 days after T-cell transfer, as shown for injection of 2.5 × 10⁶ T cells per recipient in Figure5.

Surprisingly, adoptive transfer of donor T lymphocytes depleted of alloreactive donor cells by soluble mAb to CD95 alone also led to the development of lethal GvHD (Figure 5). However, depletion of activated alloreactive T cells using anti-CD95 mAb cross-linked by polyclonal anti–hamster Ig prior to adoptive transfer of purified CD4⁺ (Figure 5A) or CD8⁺ (Figure 5B) allogeneic T cells resulted in the amelioration of acute GvHD in F₁recipients. In this experimental group all recipient mice survived and were free of clinical GvHD, as did control mice that received syngeneic spleen cells. Prevention of GvHD was stable and lasted until day 120, when the monitoring was terminated.

In recent experiments we also used this P→F₁ model for allogeneic BMT in which 2 × 10⁷ T-cell–depleted BM cells of H-2^k origin were transplanted together with 2.5 × 10⁶ purified allogeneic (H-2^k) CD90⁺ T cells or T cells subjected to anti-CD95 treatment prior to transfer. These studies appear to confirm the results described above because F₁ recipients that received pretreated donor T lymphocytes did not develop GvHD, in contrast to F₁ BMT controls.

Histopathological examination of sections taken from skin, small intestine, and spleen at different time points after adoptive transfer confirmed the survival data: apoptotic cells in the basal layer of epidermis and inflammation in the cryptic epithelium within the small intestine due to infiltrating lymphocytes were clearly detectable in mice injected with unmodified donor T cells, in contrast to F₁ recipients that had received T cells depleted of alloreactive lymphocytes (Figure 6). Moreover, immunohistologic staining of spleen showed expansion of CD4⁺ T cells and, albeit less extensive, of CD8⁺ T cells in F₁ mice after transfer of untreated donor T cells compared with recipients injected with donor T lymphocytes treated by AICD (Figure 6).

To monitor homing and GvH reactions of alloantigen-activated T cells after adoptive transfer into F₁ hosts, the presence and expansion of adoptively transferred donor T cells in vivo was evaluated. Following immunomagnetic separation of CD90⁺cells from single spleen and lymph node cell suspensions of F₁ recipients at various time points after adoptive transfer, the expansion of H-2^k–positive T cells versus recipient-derived T lymphocytes (H-2^kxd) isolated from the pool of CD90⁺ cells was analyzed by 2-color flow cytometry (Figure 7). Proliferation of allogeneic T cells (H-2^k) depleted from alloreactive responders in vitro by AICD was very low in F₁ recipients at day 10 after initial transfer of 2.5 × 10⁶ cells (Figure 7B; 6.4% for CD4⁺ and 4.5% for CD8⁺, respectively). In contrast, untreated allogeneic T lymphocytes from C3H/He mice were significantly expanded (Figure 7A; 33.4% for CD4⁺ and 22.8% for CD8⁺, respectively) from the same number of cells injected.

The significant incidence of GvHD following transfer of donor T lymphocytes contained in the stem cell graft into recipients remains as a major caveat in allogeneic BMT, although various strategies for the prevention have been proposed.2-9 In particular, T-cell depletion is undoubtedly a very efficient method to prevent GvHD; however, it increases the risk of severe infections caused predominantly by Epstein-Barr virus26 and cytomegalovirus,10 engraftment failure,11 and relapse of leukemia.12,13 An ideal strategy for a specific and individualized immunotherapy would thus aim for selective elimination of host-reactive T lymphocytes but would leave all other T-cell specificities recognizing Epstein-Barr virus and cytomegalovirus or leukemia-restricted (minor) histocompatibility antigens and leukemia antigens unaffected. Various experimental means, including elimination of alloreactive T cells by immunotoxin-conjugated mAbs to CD25 mAb (as first described by Datta et al14 and followed by others8) or via the early T cell activation antigen CD69,27 have been employed. Furthermore, suicide gene transfer into donor T lymphocytes,4,5 induction of anergy by blocking costimulatory signals necessary for optimal T-cell activation,6,7 and the administration of cytokines to modify T-cell effector responses9,28 have been reported. These methods, however, affect either primarily T-cell subopulations6,7 or all T cells regardless of their specificity.4,5 They also appear not to be very effective on the negative selection of alloreactive T cells,14technically quite complex,27 or limited by the development of resistant T lymphocytes transduced with the HSV-Tk suicide gene to ganciclovir.29 Alternatively, the in vitro generation of allorestricted T cells conferring GvL reactivity without causing GvHD has been described.18 19 

We explored the possibility of separating alloreactive T cells from antiviral T-cell specificities and possibly GvL-conferring T lymphocytes using the CD95/CD95L-mediated AICD as a strategy to deplete GVH-reactive T lymphocytes by apoptosis. AICD via the CD95 system has been demonstrated to occur in T-cell hybridomas as well as in human and murine T cells following activation by TCR cross-linking or by antigen and is thought to represent an important mechanism to limit the clonal expansion of activated T cells at the down phase of an immune response.20-22,30,31 AICD also plays a major role in the maintenance of peripheral T cell tolerance.20,32 Mature T cells acquire susceptibility to AICD after prolonged activation.20-22 This process requires repetitive TCR re-engagement by antigen or anti-TCR/CD3 complex antibodies, is dependent on cell cycle progression caused by interleukin-2,23,33 and results in apoptosis of antigen-specific T cells but not of resting T cells.20,22 23 

Using an (MHC) allogeneic model system we first demonstrated the requirements for induction of AICD in vitro. Secondly, we used this physiologic mean to deplete alloreactive T cells from an allogeneic T-cell repertoire. Upon stimulation of purified naive T lymphocytes from C3H/He mice with BALB/c mouse–derived allogeneic splenic stimulators in the presence of an agonistic mAb (Jo2) to CD95 during a primary MLC proliferation of both CD4⁺- or CD8⁺-alloreactive T cell was substantially (≥ 80%) decreased in comparison to control allogeneic responses. Reduction of proliferative responses was mediated by the agonistic anti-CD95 mAb Jo2 25 in a dose-dependent fashion. Seemingly, the capacity of CTLs to lyse allogeneic target cells was also strongly impaired after depletion of alloreactive responders by CD95-mediated AICD. Flow cytometry analysis of responder T cells performed at different time points during MLC with T-cell–depleted stimulator cells revealed that activated alloreactive T lymphocytes up-regulated expression of the activation markers CD69 and CD25 as well as CD95 in a time-dependent fashion and became susceptible for AICD when cultured with agonistic anti-CD95 mAb. Induction of AICD upon activation of alloreactive T cells with one particular alloantigen was specific and did not affect I-A^d–restricted T lymphocytes expressing the transgenic αβ-TCR (DO11.10) specific for the OVA^323–339 peptide upon titration into the experimental system. Moreover, OVA-specific T cells could be restimulated from F₁ recipients after adoptive transfer of a mixture of OVA–T cells and allogeneic responders pretreated by AICD. Thus, depletion of alloreactivity by AICD retained functional T-cell specificities to defined nonalloantigens.

We then established a classical parent (C3H/HeJ or BALB/cJ)→F₁ (C3H/HeJ × BALB/c) murine model. P→F₁ models have become an invaluable tool in a number of adoptive transfer and BMT studies on GvHD and GvL because recipients cannot evoke any host-versus-graft effects, in contrast to models using 2 different mice strains. Our model also allows us to use CD95L-deficient C3H/gld donor T cells as controls for current studies on the GvL effect of anti-CD95–pretreated T cells.

In line with the results obtained in vitro, we could show by adoptive transfer of donor-derived T lymphocytes that F₁ recipients that received allogeneic T cells, depleted from alloreactive responders by AICD in vitro, did not develop a lethal GvHD. In contrast, mice receiving untreated allogeneic T-cell blasts displayed all clinical signs of acute GvHD and died within 20 days after transfer. Recent experiments performed on lethally irradiated F₁ mice and receiving BM cells together with pretreated donor T cells also did not yet result in the induction of GvHD compared with controls.

Additionally, histopathological examinations of skin and small intestine sections taken from F₁ mice with GvHD revealed apoptotic cells in the basal lamina of epidermis and infiltration of lymphocytes into cryptic epithelium of small intestine. The monitoring of allogeneic donor cells in the host by flow cytometry also supported the findings described, because 5-fold expansion of alloreactive T cells was exclusively seen in mice receiving untreated allogeneic T lymphocytes. In contrast, very low host-reactive cells were detected in mice that had received allogeneic cells depleted of alloreactivity. Moreover, splenic cells isolated from F₁ recipients 6 days after adoptive transfer of both anti-CD95–pretreated OVA-specific T cells and donor T cells could respond to OVA peptide, whereas proliferation to H-2^k was reduced.

In summary, these results strongly suggest that depletion of alloreactive T lymphocytes by AICD prevents GvHD and retains functional specificities of non-GVH–reactive donor T cells for engraftment.

Although we cannot exclude that a low number of alloreactive T cells might escape induction of AICD because it has been reported that some T cells escape CD95-mediated apoptosis to become memory cells,34 the numbers of remaining alloreactive T cells in our GvHD model appear to be below a clinically relevant threshold. In fact, considering the diverse repertoire of allogeneic T-cell specificities that might confer a GvL effect, it might be conceivable to tolerate a very low number of, for example, defined alloreactive T-cell subpopulations. Because CD4⁺ T helper 1 (Th-1)–type T cells have been described to be more susceptible to CD95-mediated AICD than their Th-2 type counterparts,35 depletion of alloreactive T cells by AICD also polarizes residual CD4⁺GVH-reactive T lymphocytes toward a Th-2 type (data not shown), which additionally might protect against clinical manifestation of acute GvHD.36,37 Moreover, because the CD95/CD95L system has been documented by several groups to play an important role in the pathogenesis of GvHD38,39 and blockade of this pathway appears to ameliorate GvHD without impairing GvL effect,40 41 it remains to be elucidated whether the elimination of alloreactive T lymphocytes by the induction of AICD results in the modifications of allogeneic donor T lymphocytes suitable for preserving a GvL effect in the absence of a clinically manifested GvHD.

In conclusion, we have established an (MHC) allogeneic model system that allows elimination of alloreactive T cells by the induction of CD95-mediated AICD. Transfer of a modified allogeneic donor T-lymphocyte repertoire into a P→F₁ model demonstrated that depletion of alloreactive T cells was sufficient for preventing a lethal GvHD in F₁ recipients. Thus, utilizing AICD for elimination of alloreactivity in allogeneic transplants might represent a suitable strategy for the enrichment of T cells that mediate an antiviral as well as a GvL effect.

Further studies on the phenotypic and functional characterization of the residual allogeneic T lymphocytes as well as their capacity to confer a GvL effect are in progress. They will help us assess the utility of AICD to optimize clinical BMT.

The authors thank K. Kunz, C. Metz, and S. Khan for expert technical assistance and M. Jülch for the fluorescence-activated cell sorter analysis.