Differential use of FasL- and perforin-mediated cytolytic mechanisms by T-cell subsets involved in graft-versus-myeloid leukemia responses

Graft-versus-leukemia (GVL) responses are antitumor effects exerted by donor cells following hematopoietic stem cell (HSC) transplantation (HCT). GVL responses have been observed in patients with myeloid and lymphoid leukemias,1,2 and similar antitumor effects have been found in other malignancies such as multiple myeloma2 and breast cancer.3 However, myeloid leukemias seem particularly susceptible to GVL effects. T cells play a major role in GVL activity; this is based on findings that T-cell depletion of HSC grafts results in higher rates of leukemic relapse,4-6 the isolation of patient-derived T-cell clones that can kill leukemia cells in vitro,7-10 and the observations of successful reversal of leukemia relapse after delayed infusion of T cells.11 

The target antigens to which a GVL response are directed remain controversial and may likely involve both shared host alloantigens, which can induce graft-versus-host disease (GVHD), and unique tumor-specific antigens. Evidence for the alloreactive responses comes from the clinical observation that leukemic relapse is less frequent in patients suffering from acute or chronic forms of GVHD.12,13 In regard to leukemia antigen-specific responses, investigators have reported cohorts of allogeneic HCT patients who have experienced GVL without apparent GVHD.13In addition, autologous and syngeneic HSC grafts can be manipulated to induce GVL with or without a concomitant autoimmune-like GVHD.14-19 

T cells can mediate GVL effects through several means, depending upon their subset functional capabilities, and can involve either cytokine induction of other effector cells (eg, via inflammatory molecules such as interferon-γ [IFN-γ]) or direct cytolytic activity. In regard to the latter, both CD4⁺ and CD8⁺ cytotoxic T lymphocytes (CTLs) can mediate tumor cell apoptosis through several mechanisms including the exocytosis of cytotoxic granules containing perforin and granzymes, the binding of the Fas ligand (FasL) to target cell Fas, and the binding of tumor necrosis factor–alpha (TNF-α) to tumor cell receptors. Although in vitro studies have described the cytotoxic effector functions of antileukemic T-cell clones,20 the cytotoxic mechanisms involved in GVL responses in vivo have only begun to be characterized,21-24particularly with regard to myeloid leukemias.

We have previously established a mouse model of syngeneic and, therefore, tumor antigen-specific GVL activity that features CD4⁺ and CD8⁺ T-cell responses against MMB3.19, a myeloid leukemia of C57Bl/6 (B6) origin.25 In the present study, we demonstrate that GVL activity against MMB3.19 is not mediated by tumor necrosis factor–α (TNF-α). Instead, CD4⁺ T-cell GVL responses rely primarily on FasL and secondarily on perforin mechanisms. In contrast, CD8⁺T-cell GVL responses are primarily dependent upon perforin and secondarily on Fas-FasL interactions.

Wild type (wt) C57BL/6J (B6), perforin-deficient C57BL/6-Pfp^tm1Sdz(pfp^o), and FasL-deficient B6Smn.C3H-Fasl^gld (gld) mice, all expressing the B6 background, were purchased directly or derived from breeding stock (The Jackson Laboratory, Bar Harbor, ME). Male wt andpfp^o mice were used as donors between the ages of 7-12 weeks, while male gld mice were used as donors between the ages of 5-6 weeks. Male wt mice were used as recipients between the ages of 9-16 weeks. The mice were kept in a sterile environment in microisolators at all times and were provided with acidified water and autoclaved food ad libitum.

WEHI164 is a methylcholanthrene-induced fibrosarcoma (American Type Culture Collection, Manassas, VA). MMB3.19, amyc-transformed myeloid leukemia line, was cloned from the ascites of a B6 mouse that had been injected with amyc-encoding Moloney murine leukemia virus (MMLV) construct, as previously described.26 Both cell lines were maintained in Roswell Park Memorial Institute medium (RPMI 1640) supplemented with 10% fetal bovine serum (FBS), 5.5 × 10⁻⁵ mol/L 2-mercaptoethanol, 2 mmol/L L-glutamine, 50 IU/mL penicillin, and 50 μg/mL streptomycin.

Phosphate-buffered saline (PBS) supplemented with 0.1% bovine serum albumin (BSA) (Sigma Chemical Co, St Louis, MO) was used for all in vitro manipulations of the donor bone marrow and lymphocytes. Immediately prior to injection, the cells were washed and resuspended in PBS alone. For in vitro assays, the cells were cultured in RPMI 1640 supplemented with 10% FBS, 5.5 × 10⁻⁵mol/L 2-mercaptoethanol, 2 mmol/L L-glutamine, 50 IU/mL penicillin, and 50 μg/mL streptomycin.

All recipient mice received an 850- or 950-cGy exposure from a Mark-I-68A cesium 137 (¹³⁷Cs) source (JL Shepherd and Assocs, San Fernando, CA) at 143 cGy/min as indicated. For in vitro assays, MMB3.19 cells received a 30-Gy exposure.

Ascitic fluids containing mAbs (clone names in parentheses; American Type Culture Collection, Manassas, VA) specific for Thy-1.2 (J1j), CD4 (RL172), CD8(3.168), natural killer 1.1 (NK1.1) (PK136), and rat immunoglobulin G (IgG) (MAR18.5; gift of Dr Robert Levy, Univ of Miami, Miami, FL) were used for the preparation of cellular grafts and culture supernatants. Other mAbs used included the supernatant of the rat-antimurine B220 14.8 mAb (gift of Dr Levy); affinity-purified goat antimouse IgG (whole molecule) antibodies (Cappel, Cosa Mesa, CA); guinea pig serum, which was prepared in our laboratory or purchased (Rockland, Boyertown, PA) and used as a source of complement for all mAb treatments; and biotinylated anti-Fas mAb (Jo2; PharMingen, San Diego, CA). For the phenotyping of MMB3.19 cells and the monitoring of cellular subset depletions from transplant grafts, we used fluorescein isothiocyanate (FITC)- and phycoerythrin (PE)-conjugated versions (clone names in parentheses, PharMingen) of the following mAbs that were specific for CD3ε(145-2C11), CD4 (RM4-5), CD8α(53-6.7), B220 (RA3-6B2), 2B4, and an isotype control (R35-95). We also used streptavidin (SA)-PE (Caltag, South San Francisco, CA).

Appropriate mAbs in volumes of 25 μL were incubated with 2-5 × 10⁵ cells in the wells of a 96-well U-bottom microplate at 4°C for 25 minutes, centrifuged at 1500 rpm for 3 minutes, and washed with PBS containing 0.1% BSA and 0.01% sodium azide (wash buffer). When applicable, SA-PE or a secondary antibody was added in a volume of 25 μL at 4°C for 25 minutes, followed by 2 washes with wash buffer. The fluorescence analysis was performed on an EPICS Profile II analyzer (Coulter, Hialeah, FL) in the Kimmel Cancer Institute Flow Cytometry Facility, Philadelphia, PA.

Bone marrow cells were obtained from the femurs and tibias ofwt mice by flushing with PBS. To prepare anti–Thy-1–treated (T-cell–depleted) bone marrow (ATBM), the cells were incubated with J1j mAb (at 1:100 dilution) and complement (at 1:25 dilution) at 37°C for 45 minutes and were washed extensively. T-cell–enriched donor cell populations were prepared by treating pooled spleen and lymph node cells in 2 steps: (1) Gey's balanced salt lysing solution containing 0.7% ammonium chloride (NH₄Cl) was used to remove red blood cells (RBCs), and (2) panning on a plastic petri dish precoated with a 5 μg/mL dilution of goat antimouse IgG for 1 hour at 4°C was used to remove B cells. These treatments resulted in donor populations of 90%-95% CD3⁺ cells, as quantitated by fluorescent flow cytometry. As required, the suspensions of enriched T cells or unfractionated splenocytes were depleted of CD4⁺ or CD8⁺ cells by treatment with the appropriate mAbs and complement at 37°C for 45 minutes. At the same time, B220⁺ B cells and T cells were depleted by the addition of 14.8 culture supernatant and MAR18.5 ascites to the cellular preparations. These treatments reduced the targeted populations down to background levels on flow cytometric analysis.

Mice were presensitized with 5 × 10⁶ irradiated (30 Gy) MMB3.19 cells intraperitoneally (i.p.), and splenocytes or lymph node cells were harvested after 2 weeks and restimulated in vitro for 5 days with irradiated MMB3.19 cells and 1:20 of T-STIM culture supplement (Becton Dickinson Labware, Franklin Lakes, NJ). CD4⁺ or CD8⁺ T cells were prepared by depleting splenocytes or lymph node cells of the other T-cell subset and NK cells. The JAM assay was used to measure cytotoxicity.27 Briefly, MMB3.19 cells growing in the log phase were subcultured at 5 × 10⁵ cells per mL in media containing 0.0925 Bq/mL (2.5 μCi/mL) of³H-thymidine (TdR) and allowed to label for 4 hours. The cells were then washed with media and cultured in 96-well U-bottom plates at 1 × 10⁴ cells per well, along with effector cells at varying numbers. After 4 or 8 hours as indicated, the degree of target-cell DNA fragmentation was determined by comparing the radioactive incorporation of wells containing only target cells to wells containing both target and effector cells.

One day prior to transplantation, recipient mice were challenged with an i.p. injection of 0.5 mL PBS with or without 10⁵ MMB3.19 cells. On the following day, recipient mice were lethally irradiated with 8.5 or 9.5 Gy, and approximately 6 hours later, they were injected intravenously (i.v.) with either 2 × 106 donor ATBM cells alone as a negative control or a mixture of ATBM cells plus manipulated donor lymphocyte populations. The mice were checked daily for morbidity and mortality until the experiments were terminated. As indicated, the data were pooled from 2 separate experiments, and median survival times (MSTs) were determined. Statistical comparisons between experimental groups were performed by the nonparametric Wilcoxon signed rank test.

Total cellular RNA was prepared from 10⁶-10⁷MMB3.19 cells by homogenization in 1 mL Ultraspec (Biotecx Laboratories, Houston, TX), separated of cellular DNA and protein by the addition of a 1:5 volume of chloroform, vortexed for 5 seconds, and centrifuged at 12 500 rpm for 15 minutes. The aqueous phase was transferred to a clean Eppendorf tube, and the RNA was precipitated at 4°C by the addition of an equal volume of isopropanol and then centrifuged at 12 500 rpm for 15 minutes. The pellet was washed with 75% ethanol in diethyl pyrocarbonate (DEPC)-treated water and centrifuged again. The RNA pellet was resuspended in 25 μL DEPC water, heated to 55-65°C for 10 minutes, and stored at −20°C. Recovery of RNA was determined by spectrophotometry. For each sample, the reverse transcription and PCR reactions were performed in a 1-tube format using Ready-To-Go RT-PCR Beads (Amersham Pharmacia Biotech, Arlington Heights, IL). The following were added to each reaction: 1 μmol/L sense and antisense primers, 1.5 μg oligo d(T), and 2-4 μg total RNA.

The following TNF-R and glyceraldehyde-3-phosphate dehydrogenase (GADPH) primer sequences (Dr Keith Kelley, University of Illinois, Urbana, IL) have been previously described28: 55-kd TNF-R (expected product length, 368 base pairs [bp]): 5′-CAG TTG CAA GAC ATG TCG GA-3′ and 3′-GAC CTA GCA AGA TAA CCA GG-5′; 75-kd TNF-R (expected product length, 383 bp): 5′-GAG TGT GTG CTT GCG AAG CT-3′ and 3′-CGA TGT AAG GAT GCT TGG AG-5′; and GADPH (expected product length, 225 bp): 5′-GGA AGC TTG TCA TCA ATG G-3′ and 3′-AGA TCT CGT GGT TCA CAC C-5′. The reactions were performed using a Perkin-Elmer Applied Biosystems GeneAmp PCR System 9700 (Perkin-Elmer, Foster City, CA). Cycling conditions were as follows: 30 minutes at 42°C; 2 minutes at 94°C; and 30 cycles each of 30 seconds at 94°C, 45 seconds at 55°C, and 45 seconds at 72°C. The final extension was then performed for 7 minutes at 72°C. The product size was determined by electrophoresing the samples on an agarose gel and staining them with ethidium bromide.

MMB3.19 cells were cultured in 40 mL medium (RPMI 1640 supplemented with 10% FBS, 5.5 × 10⁻⁵ mol/L 2-mercaptoethanol, 2 mmol/L L-glutamine, 50 IU/mL penicillin, and 50 μg/mL streptomycin) in tissue culture flasks at 1 × 10⁵ cells per mL with either 500 ng/mL anti-Fas mAb (Jo2) or hamster anti-trinitrophenol (TNP) mAb as an isotype control. Both antibodies (PharMingen) were obtained in a low-endotoxin, sodium azide–free format. After 24 hours, the MMB3.19 cells were harvested, and an aliquot was analyzed for viability by trypan blue staining. The remainder of the cells were stained and analyzed using the APO-DIRECT Kit (PharMingen), a flow cytometry–based TUNEL method (terminal deoxynucleotidyl transferase–mediated dUTP nickend-labeling method).

Pfp^o, gld, and wt mice were injected with 5 × 10⁶ irradiated (30 Gy) MMB3.19 cells, and their splenocytes were harvested 2 weeks later. The cells were cultured in triplicate at 2 × 10⁵ cells per well in 96-well flat-bottom plates, and irradiated (30 Gy) MMB3.19 cells were added at 100 cells per well to half the samples. After 3, 4, and 5 days, the cells were harvested following an overnight pulse with 0.037 MBq (1 μCi) per well of TdR, and the radioactivity incorporation into DNA was measured. Stimulation indices were calculated by dividing the mean cpm of splenocytes cultured with MMB3.19 cells by the mean cpm of splenocytes cultured alone.

Assay conditions to determine the sensitivity of MMB3.19 cells to exogenous TNF-mediated cytotoxicity were based on previous studies.28,29 Recombinant murine TNF-α was used (Peprotech, Rocky Hill, NJ). MMB3.19 and WEHI164 cells were cultured overnight in quadruplicate in 96-well flat-bottom plates at 2 × 10⁴ cells per well in a total volume of 100 μL. We then added 100 μL media, containing either actinomycin D alone (0.5 μg/mL final concentration) or actinomycin D and titrated concentrations of TNF-α, to each well. After 22 hours, 70 μL was aspirated from each well prior to the addition of 50 μL 3-(4,5-Dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) (final concentration 1 mg/mL). After a 3.5- to 4-hour incubation at 37°C, 50 μL media was aspirated, and 150 μL acidified isopropanol (0.04 N hydrochloric acid [HCl]) was added to each well. The media in each well was then extensively run through a pipette to dissolve crystals prior to the measurement of each well's absorbance at 540 nm.

Since FasL-Fas interactions are a major mechanism by which T cells mediate target cell death, it was first determined whether MMB3.19 cells had Fas molecules on their surface. Flow cytometric analysis revealed a high level of Fas expression, with 86.6% of the cells staining positive for the marker (relative mean fluorescence intensity of 14.85; Figure 1). Next, to determine whether the Fas molecules expressed by MMB3.19 cells were competent to receive death signals, the cells were incubated with anti-Fas mAb for 24 hours, and viability was measured by trypan blue exclusion. Anti-Fas mAb-treated tumor cells exhibited 22.2% nonviability compared to only 3.2% nonviability of control mAb-treated cells (Figure 2A). Furthermore, there were approximately 42% fewer viable cells remaining in the anti-Fas mAb-treated group (4.14 × 10⁶ cells vs 7.13 × 10⁶ cells). TUNEL analysis of the same cultures revealed that anti-Fas mAb treatment increased the percentage of apoptotic cells from 10.6% to 38.7% at the 24-hour time-point (Figure2B). Collectively, this data indicated that MMB3.19 cells displayed sufficient levels of cell-surface Fas molecules that were capable of transducing apoptotic signals.

To determine whether MMB3.19 cells expressed TNF receptors, which could potentially be used as death-signaling molecules and are often expressed at low levels, RT-PCR analyses of both TNF receptor type I and II were performed (Figure 3). Synthesis of both TNF receptors was detectable in the MMB3.19 cells, and expression of each was comparable to that found for the TNF-sensitive cell line, WEHI164. However, despite the presence of the TNF receptors, the MMB3.19 cells were resistant to exogenous TNF-mediated cytotoxicity throughout a concentration range of 0.02-20 ng/mL. This is in contrast to WEHI164 cells, which were highly susceptible at all concentrations (Figure 4). In addition, RT-PCR analysis also revealed that MMB3.19 cells produced endogenous TNF-α (data not shown), which is consistent with the function of this cytokine as a growth and/or activation factor. These combined results suggested that MMB3.19 cells do not receive death signals via ligation of their TNF receptors with exogenous TNF-α.

The cytotoxic effector mechanisms by which T-cell subsets mediate syngeneic GVL responses to MMB3.19 leukemia challenge were investigated. Donor T cells from FasL- and perforin-deficient, mice along with 2 × 106 ATBM cells, were transplanted into lethally irradiated (8.5 Gy) B6 mice that were preinoculated with 1 × 10⁵ leukemia cells. Mice infused with 2 × 10⁶ CD4⁺ T cells from MMB3.19-presensitized B6 wt donors exhibited GVL activity with a significant (P ≤ .03) extension of MST to 43 days. This contrasts with the 22-day MST for ATBM control mice that were challenged with MMB3.19 cells but did not receive any donor T cells (Figure 5A). Animals that were administered 2 × 10⁶ CD4⁺ T cells from MMB3.19-presensitized B6 pfp^o donor mice exhibited intermediate and marginally significant (P ≤ .075) GVL activity, with an MST of 33 days. Furthermore, the survival pattern of the pfp^o group was significantly less (P ≤ .02) than the wt control. In contrast, 2 × 106 CD4⁺ T cells from MMB3.19-presensitized B6 gld donors did not appear to mediate any GVL activity because recipients of these cells had an MST of 22 days, which is equivalent to the recipients of ATBM plus leukemia cell challenge alone (P ≥ .38). The levels of GVL activity also correlated with the proportion of surviving mice at the termination of experiments on day 46; ie, 40% for mice receiving wt CD4⁺ T cells, 20% for recipients ofpfp^o cells, and 0% for mice administered gld cells.

In these experiments, pfp^o CD4⁺T cells mediated GVL activity at a level that was less effective thanwt donor cells, which could still suggest a role for perforin cytolytic mechanisms. Because the initial experiment with 2 × 10⁶ donor gld CD4⁺ T cells failed to demonstrate any GVL activity mediated by non-FasL pathways, it was imperative to test this notion with higher donor cell doses. The B6 mice were challenged with an i.p. dose of 1 × 105MMB3.19 cells, lethally irradiated with 9.5 Gy the next day, and administered either 2 × 10⁶ ATBM cells alone or in combination with CD4⁺ T-cell–enriched (CD8⁺T-cell–depleted) splenocytes (Figure 5B). One group of mice received 10⁷ CD4-enriched splenocytes from MMB3.19-primed wtdonors, while another group received 5 × 10⁷CD4-enriched splenocytes from MMB3.19-primed gld donors. Flow cytometric analysis demonstrated that CD4⁺ T cells represented approximately 25% of the splenocyte populations (data not shown), which corresponds to a CD4⁺ T-cell dose of 2.5 × 10⁶ cells and 1.25 × 10⁷cells for recipients of wt and gld cells, respectively.

As expected, the mice that were challenged with MMB3.19 cells and received ATBM alone died rapidly, with only 14% surviving long-term (MST of 30 days). When the mice were challenged with MMB3.19 cells and administered wt CD4⁺ T-cell–enriched splenocytes, 57% survived beyond 60 days (P ≤ .03), with an MST of more than 60 days. Of most importance, 100% of the MMB3.19-challenged mice receiving gld CD4-enriched splenocytes survived for longer than 60 days (P ≤ .03). This demonstrated that a 5-fold increase of the dose of gld cells relative to wt cells could result in comparable if not superior GVL activity (P ≤ .10). Considered together with the earlier lack of GVL activity mediated by low-dose gld cells (Figure 5A), these results suggested that given enough effector cells, the perforin-mediated cytotoxic pathway could participate in the CD4⁺ T-cell GVL response, but that the FasL-Fas pathway was clearly dominant.

We infused 5 × 106 CD8⁺ T cells from MMB3.19-presensitized wt donors into lethally irradiated (8.5 Gy) mice challenged with 1 × 10⁵leukemia cells. After this infusion, GVL activity was clearly demonstrated by the survival of 80% of the recipients for more than 46 days (at which time the experiments were terminated). This compares (P ≤ .02) to 100% lethality with a MST of 22 days for MMB3.19-challenged mice that received only ATBM cells (Figure6). The administration of 5 × 106 CD8⁺ T cells from MMB3.19-presensitized gld donors to leukemia-challenged recipients resulted in intermediate GVL activity, with 54.5% long-term survival and an MST of 34 days. This compared (P ≤ .02) to the ATBM plus MMB3.19 control, but it was not significantly different (P > .20) from the wt group. However, cells from the pfp^o donors mediated very weak responses, with 13.3% long-term survival and an MST of 24 days. This was significantly greater (P ≤ .02) than the ATBM plus MMB3.19 control, but also significantly less (P ≤ .02) than the gld donors. Thus, CD8⁺ T cells mediating GVL activity seemed to rely heavily upon the use of perforin, with only minimal dependence upon FasL-mediated functions.

To determine the capacity of the CD4⁺ T-cell subset to mediate in vitro cytotoxic activity against MMB3.19 target cells, CD4⁺ T cells were prepared from CD8-cell– and NK-cell–depleted lymph nodes of MMB3.19-presensitized mice. The cells were then restimulated with leukemia cells in vitro and used as effectors in an 8-hour JAM assay using ³H-TdR–labeled MMB3.19 target cells (Figure 7A). As reflected in vivo, wt cells exhibited the best, albeit modest, cytotoxic activity at all effector:target (E:T) ratios tested, with a maximum specific killing of 15% at a 50:1 ratio. The gld cells yielded less cytolysis, 8.3%, although they were statistically insignificant from the wt group (P ≥ .12). However, the pfp^o cells were significantly less capable of lysing MMB3.19 cells (1.43%; P ≤ .01).

In a similar experiment to examine the antileukemic activity of CD8⁺ CTLs, splenocytes from MMB3.19-challenged wt,pfp^o, and gld mice were depleted of CD4⁺ and NK cells prior to use as effectors against³H-TdR–labeled MMB3.19 target cells (Figure 7B). CD8⁺ CTLs exerted the most cytotoxic activity when derived from wt or gld mice (for both groups, peak specific killing of 54% at an E:T of 10:1), whilepfp^o CTLs mediated virtually no lysis of MMB3.19 target cells (4.4% peak killing).

To ensure that pfp^o and gld T cells still responded to the MMB3.19 cells, despite their cytotoxic defects, MMB3.19-primed wt, pfp^o, and gld splenocytes were cultured with irradiated (30 Gy) MMB3.19 cells and assayed for their proliferative capability (Figure8). The stimulation indices of all 3 types of splenocytes were not significantly different from each other at the peak of the response on day 5 (P ≥ .20 for all comparisons). This data implied that T cells from pfp^oand gld mice can respond to MMB3.19 antigens, even though they are incapable of killing the tumor cells via their individual deficient cytotoxic functions.

GVL responses can potentially be mediated by T cells present in both autologous and allogeneic HSC grafts. Interestingly, the myeloid leukemias seem most sensitive to GVL activity, although very little is known about these responses. Investigations into murine GVL myeloid models have been hampered in the past by the limited availability of myeloid cell lines, particularly those that do not involve the expression of retroviral antigens. Our model for GVL responses features a myeloid leukemia that was transformed through transduction with a c-myc construct that does not result in any detectable virus-related protein expression.25 In addition, c-myc is a frequently mutated gene in hematologic malignancies, and therefore in contrast to other cell lines,30-32 the leukemia-specific antigen(s) expressed by the MMB3.19 cells may reflect the consequences of oncogenesis rather than retroviral infection.

Although our syngeneic model is highly suited for the study of leukemia antigen-restricted GVL responses, clinical GVL has been noted to occur more frequently in patients receiving allogeneic rather than autologous or syngeneic HSC grafts.12,13 However, it is unclear to what degree GVL reflects genuine leukemia antigen-specific responses versus overlapping GVHD responses mounted against alloantigens shared by the host tissues and leukemia cells. Yet, isolated clinical studies have described allogeneic HSC patients who have experienced GVL without apparent GVHD and vice versa.13 Similar separation of allogeneic GVHD and GVL has also been achieved in several animal models.22-25,33-35 Furthermore, multiple experimental systems and clinical protocols have demonstrated that syngeneic or autologous GVL can be induced without a concomitant autoimmune-like GVHD.14-19,25 Finally, several studies have successfully used leukemia-specific antigens, such as bcr-abl and proteinase-3, to stimulate T-cell clonal responses both in vitro and in vivo.8,36-39 

The dependence of GVL on T cells has been inferred through observations that leukemic relapse occurs less frequently in patients receiving T-cell–replete HSC grafts than in patients receiving grafts that have been T-cell depleted.4-6 In addition to these clinical data, reports that human T-cell clones can kill leukemic cells in vitro7-10 strongly suggest that T cells are the primary mediators of GVL activity.

In addition to T cells playing a crucial role, studies have also identified NK cells as mediators of GVL activity, either by themselves or as an effector arm dependent upon CD4⁺ T-cell cytokine induction.40,41 However, in the MMB3.19 model, there was no evidence of NK-cell involvement in GVL activity; weekly injections of depleting anti-NK1.1 mAb did not alter the survival of MMB3.19-challenged recipients of naive wt CD4⁺ T cells. In addition, in vitro NK-cell–mediated cytotoxicity by naive splenocytes could be demonstrated against susceptible YAC–1 cells but not against MMB3.19 cells (R.K. and Townsend, unpublished observations, 1996).

The findings presented here constitute one of the few reports that CD4⁺ T cells are capable of mediating GVL effects in vivo (Figure 5A). Specifically, we found that this subset of CD4⁺ T cells relied primarily on the use of FasL and, surprisingly, secondarily on perforin in order to mediate GVL effects against the MMB3.19 leukemia. The finding that higher doses of gldCD4⁺ T cells could mount effective GVL activity (Figure5B) was consistent with perforin acting as a secondary mechanism involved in this particular CD4⁺ T-cell–mediated GVL response. Interestingly, despite the obvious dominance of the FasL GVL pathways in vivo, in the in vitro CD4⁺ CTL response to MMB3.19 target cells (albeit a weak response, with only 15% specific lysis at a 50:1 E:T ratio using CTLs from wt mice; Figure 7A), CTLs from pfp^o mice displayed only minimal cytolysis of MMB3.19 cells. By comparison, CD4⁺ CTLs fromgld mice proved much more effective in vitro. This paradox may be accounted for by the observations made elsewhere that restimulating CD4⁺ T cells in the absence of CD8⁺ T cells allows development of CD4⁺ CTLs, but the restimulation may give rise to a skewed population that is more reliant on perforin than cells restimulated in vivo.42 

Clearly, CD8⁺ CTLs dominate the cytolytic response after in vivo priming and in vitro restimulation with MMB3.19 cells. The proportion of cytolytic precursor cells within the CD8⁺population is normally very high to begin with (90%-95%), whereas only a relatively small portion (less than 10%) of the CD4⁺ T cells have perforin granules and cytolytic capability.42 Of course, these proportions will vary with individual responses to antigens, but in any event, the findings are consistent with a lower frequency of CD4⁺ effector cells. This proportional difference in vitro is also consistent with the in vivo result, whereby the CD4⁺gld population had to be increased significantly to exhibit GVL activity.

In contrast to the CD4⁺ responses, the CD8⁺ T cells involved in GVL responses to MMB3.19 challenge relied primarily on perforin and only secondarily on FasL (Figure 6). CTL assays using CD8⁺ effector cells were consistent with the GVL experiments (Figure 7B) in that CD8⁺ CTLs from wtor gld mice could efficiently lyse the MMB3.19 cells, whereas the pfp^o CTLs were ineffective.

Of note, a previous study43 cautioned that FasL-deficientgld T cells may be unable to mediate cytotoxicity because of an inability to expand in vivo and not just because of a lack of functional Fas-FasL interactions between T cells and their targets. To exclude this possibility, we used gld mice at 5-6 weeks of age and depleted B220⁺ T cells from inocula derived from these animals.44 Both of these measures have been shown to exclude the anergized B220⁺CD4⁻CD8⁻ T-cell subset that develops in older gld mice and features aberrant functions.45,46 With regard to similar concerns in the use of pfp^o mice, careful studies have confirmed that these animals are appropriate for examining antigen-specific responses.47 We have also demonstrated that tumor-primed wt, gld, andpfp^o splenocytes were all capable of comparable proliferation in response to restimulation with irradiated MMB3.19 cells (Figure 8). This strongly suggested that at least on a gross level, T cells from all 3 types of mice were equivalent by one classic definition of T-lymphocyte function.

In another study of cytotoxic effector mechanisms involved in GVL, Tsukada and colleagues23 concluded that CD8⁺GVL activity could be preserved while alleviating GVH through the use of gld donors but not through the use ofpfp^o donors. These experiments were performed using the MHC Class I⁺II⁻ L1210 T-cell leukemia and P815 mastocytoma cell lines. Because L1210 is impervious to TNF- and Fas-mediated apoptosis in vitro, it is not surprising that perforin, the only other significant cytotoxic molecule available for antileukemic effects, expressed by glddonors is sufficient to induce GVL effects. It is more interesting that gld lymphocytes can induce GVL against P815 because this leukemia is sensitive to apoptosis induced by FasL-transfected effector cells, but it is resistant to TNF-mediated cell death.23Analogous to GVL activity against L1210 and to MMB3.19 cells in our own study, this result implies that the perforin expressed by gldCD8⁺ T cells is adequate to eliminate P815 cells.

Investigators have reported that in an allogeneic GVL model, CD8⁺ T cells rely equally on perforin and FasL to mediate antitumor activity.21 Based on the criteria of body weight, spleen size, and the appearance of mice, the authors also reported that the CD8⁺ T cells did not mediate GVHD. This study is very interesting in light of our findings that CD8⁺ T cells mediated GVL activity primarily through perforin and only minimally through FasL (Figure 6). In another system, T cells present in granulocyte colony-stimulating factor (GCSF)–mobilized peripheral blood stem cell transplantation grafts mediated a perforin-dependent allogeneic GVL effect against the P815 mastocytoma.22 As stated above, the reliance of T cells on perforin is particularly intriguing because P815 expresses functional cell-surface Fas. In this model, the exact T- cell subsets responsible for GVL are as of yet unknown.

MMB3.19 cytolysis by CD4⁺ and CD8⁺ CTLs is not likely to be induced through TNF-α–TNF receptor interactions because the MMB3.19 cells were resistant to high concentrations of exogenous TNF-α (Figure 4) and produced endogenous cytokine (data not shown). This may reflect an autocrine growth loop that merits future study. The perforin-utilizing CD4⁺ T cells that are effective at high doses in our model (Figure 5B) are especially intriguing because such cells have only been rarely characterized in vivo.48 This unusual cellular subset has been described in vitro more extensively.49-52 In addition, the importance of the perforin pathway in antitumor responses other than GVL has been underscored by other in vivo models.53 

Cytotoxic effector functions involved in GVL are best discussed in the context of GVHD. Perforin-deficient, unfractionated T cells from fully allogeneic mice or MHC-matched mice differing at minor histocompatibility loci can induce acute GVHD, but with delayed kinetics.43,44 However, FasL-deficient, unfractionated T cells can induce cachexia without cutaneous or hepatic pathology.44 Anti-FasL antibodies administered in vivo have confirmed that FasL on unfractionated T cells seems to mediate cutaneous and hepatic GVHD.54 In addition, through use of FasL-deficient donors, GVHD-associated lymphoid hypoplasia and dysfunction have also been attributed to FasL rather than to perforin expression.43,44 

For other fully allogeneic models, both CD4⁺ and CD8⁺ T cells depend on perforin for induction of optimal GVHD.48 In contrast, other studies have reported that CD4⁺ T-cell–mediated GVHD is heavily dependent on FasL, whereas CD8⁺ T-cell–induced disease is more reliant on perforin-mediated pathways.55 Furthermore, granzyme B has been found to be important for GVHD mediated by CD8⁺ MHC class I–mismatched T cells, but it is not important for GVHD induced by CD4⁺ T cells.56 In addition, through use of neutralizing anti-TNF antibodies in various models of GVHD, several investigations have demonstrated that GVHD can be highly dependent on TNF activity.54,57-59 

Because GVL activity against MMB3.19 cells is as equally affected by an absence of FasL on CD4⁺ T cells as by an absence of perforin on CD8⁺ T cells, we do not believe that MMB3.19 cells are inherently more sensitive to FasL-mediated versus perforin-mediated apoptosis. Instead, we hypothesize that GVL-conferring T-cell subsets use different patterns of effector mechanisms whose development is determined by MHC class I- and class II-restricted tumor antigens. Investigators have characterized several systems in which antigens influence the cytotoxic effector functions used by T cells.60-63 Final proof of this hypothesis will require, at the minimum, identification of at least one leukemia antigen that preferentially induces FasL responses and another antigen that skews T cells to use perforin-containing cytotoxic granules. Another set of tenable hypotheses to account for the particular cytolytic mechanisms used by GVL-mediating T cells in our model include those involving the influence of cytokines.

As shown earlier by others, the balance between Tc1- and Tc2-associated cytokines present during CTL development may dictate the armamentarium employed by antileukemic effector cells.24Regardless of what determines the effector mechanisms used by GVL-mediating T cells, the patterns of cytotoxic functions employed against tumors may differ from those brought to bear against host tissues in GVHD.23 This possibility could form the basis for future therapeutic approaches to minimize GVHD while preserving GVL activity.

We express our gratitude to Jeffrey Vakil and Kristin Naper for technical assistance and to David Dicker for his expertise in flow cytometric analyses.