CD2-mediated IL-12–dependent signals render human γδ-T cells resistant to mitogen-induced apoptosis, permitting the large-scale ex vivo expansion of functionally distinct lymphocytes: implications for the development of adoptive immunotherapy strategies

Human T lymphocytes recognize and respond to antigens via a clonally expressed T-cell receptor (TCR). Whereas most mature T cells express an αβ-TCR heterodimer, a few express an alternative γδ-TCR heterodimer.1-5 Although the physiologic role of human γδ-T cells remains unclear, evidence continues to accumulate to suggest that γδ-T cells are involved in a number of important physiologic and disease-related processes. For example, both murine and human γδ-T cells have been shown6-10 to exhibit major histocompatibility complex (MHC)-unrestricted cytotoxicity against some tumors, in vitro and in vivo. In addition, γδ-T cells have been shown11-13 to exert antiviral activity against a number of human pathogens, including the human immunodeficiency virus. It has also been proposed14 that γδ-T cells may play a role in wound healing or tissue repair through the elaboration of a number of growth factors, including keratinocyte growth factor. Recently, in both human clinical studies and experimental animal models of allogeneic bone marrow transplantation, it has been recognized that donor-derived γδ-T cells may serve as facilitating cells, promoting the engraftment of donor hematopoietic stem cells across varying degrees of MHC disparity.15 16 

However, the exploitation of γδ-T cells for specific therapeutic ends remains largely unrealized, largely because of the extreme difficulty of obtaining sufficient numbers of viable γδ-T cells given their relative infrequency in peripheral blood (PB) or other readily available tissues. Simply isolating γδ-T cells from fresh PB or bone marrow is likely to prove impractical. Expanding γδ-T cells ex vivo using a variety of mitogenic stimuli, including anti-CD3 or anti-TCRγδ antibodies, is an attractive alternative means by which to obtain sufficient numbers of these cells. However, for reasons that are not entirely clear, human γδ-T cells, when compared to αβ-T cells, appear to undergo apoptosis or activation-induced cell death more readily upon TCR/CD3 engagement, especially in the presence of IL-2.17 Human γδ–T-cell clones have also been shown18 19 to readily undergo apoptosis when stimulated simultaneously by anti-TCR monoclonal antibody (mAb) plus exogenous IL-2, leading some to propose that the induction of programmed cell death upon repeated mitogenic stimulation might serve as a regulatory mechanism whereby excessive in vivo γδ–T-cell proliferation is prevented. In any event, the fact that γδ-T cells may simply die upon ex vivo expansion may represent a serious obstacle to developing approaches to incorporate γδ-T cells into any form of adoptive cellular immunotherapy.

Here, we identify a CD2-mediated, IL-12–dependent signal that results in the selective expansion of human γδ-T cells in mitogen-stimulated human peripheral blood mononuclear cell (PBMC) cultures. Using 4-color flow cytometry integrating surface staining with annexin V and propidium iodide (PI) uptake, we first confirm in PBMC cultures the findings of others that human γδ-T cells are indeed exquisitely sensitive to apoptosis induced by T-cell mitogens. Using these same methods, we then establish that the CD2-mediated, IL-12–dependent signals that lead to the observed expansion of γδ-T cells do so by selectively protecting a subset of human γδ-T cells from programmed cell death induced by mitogenic stimulation, in particular IL-2. Finally, we demonstrate that highly purified apoptosis-resistant human γδ-T cells can mediate antitumor activity against a variety of human tumor cell lines in vitro. Both the biologic and the practical implications of these findings are considered and discussed.

PBMC were isolated by Ficoll gradient centrifugation of whole blood obtained from healthy human volunteers. Where indicated, PBMC were depleted of monocytes by the removal of plastic-adherent cells, as described.20 

Cultures were initiated at a cell density of 1 × 10⁶ cells/mL in 24-well flat-bottom tissue culture trays (Costar, Cambridge, MA) and maintained in 5% CO₂ at 37°C in RPMI-1640 with 10% fetal bovine serum (HyClone, Logan, UT), 2 mmol/L L-glutamine, 100 U/mL penicillin, 100 U/mL streptomycin, and 50 μmol/L 2-ME. On the day of culture initiation (day 0), human recombinant interferon (rIFN)-γ (1000 U/mL; Boehringer Mannheim, Indianapolis, IN); human rIL-12 (10 U/mL; R&D Systems, Minneapolis, MN), and mouse antihuman CD2 mAb clone S5.2 (1-10 μg/mL, mouse IgG2_a; Becton Dickinson, San Jose, CA) were added. Twenty-four hours later (day 1), cultures were stimulated with 10 ng/mL anti-CD3 mAb OKT3 (mouse IgG2_a; Orthobiotec, Raritan, NJ) and 300 U/mL human rIL-2 (Boehringer Mannheim). Where indicated, neutralizing monoclonal antihuman IL-12 antibody or an irrelevant isotype control antibody (R&D Systems) was added as a single dose at a final concentration of 25 μg/mL. Neutralizing monoclonal antihuman CD58 mAb clone L306 (mouse IgG2_a; Becton Dickinson) or IgG2_aisotype control antibody were added to cultures where indicated (5 μg/mL). Fresh medium with 10 U/mL IL-2 was added every 7 days. Where indicated, mAb OKT3 and mAb S5.2 were bound to plastic tissue culture plates as described.21 

Proliferation assays were performed using standard methods as described in figure text. Assays were performed in triplicate with data presented as mean counts per minute (cpm) (± SD).

Cells were stained using fluorescein isothiocyanate (FITC)-, phycoerythrin (PE)-, or allophycocyanin (APC)-directly conjugated mAbs recognizing CD3, CD5, TCR-γδ, or TCR-αβ or directly conjugated isotype-matched irrelevant control antibodies (Becton Dickinson). Cells were stained for 30 minutes at 4°C in Hank's buffered saline solution (Mediatech, Herndon, VA) containing 2% fetal bovine serum (FBS) and immediately analyzed using a FACScalibur flow cytometer or sorted using a FACS Vantage cell sorter (Becton Dickinson). PI uptake was used to exclude nonviable cells. Data analysis was performed using CellQuest software (Becton Dickinson).

Cells were first surface stained (1 × 10⁵ total cells in 100 μL) using anti-CD3–APC and anti-TCR-γδ–PE mAbs. Cells were washed twice with cold phosphate-buffered saline (PBS), washed twice again with annexin binding buffer (Apoptosis Detection Kit; R&D Systems), and then resuspended in 100 μL binding buffer. After the addition of annexin V-FITC and PI, cells were incubated for 15 minutes at room temperature in the dark. At this point, cells were kept on ice to prevent the capping and internalization of surface-bound mAbs. Calibration and compensation of all fluorescence detectors was performed using cells stained with individual positive and negative control reagents in the presence or absence of annexin V-FITC, PI, or both.

Tumor target cells (2.5 × 10⁴ in 500 μL complete RPMI) were cultured alone or cocultured with either human αβ- or γδ-T cells at varying effector–target (E:T) ratios (1:1 to 20:1) in sterile 5-mL round-bottom polystyrene tubes (Falcon, BD Labware, Franklin Lakes, NJ). Cells were incubated for 4 hours at 37°C in 5% CO₂, after which they were washed with PBS and resuspended in 100 μL annexin binding buffer to which annexin V-FITC was added. After 15 minutes at room temperature in the dark, cocultured cells were gently vortexed to disrupt any tumor–lymphocyte aggregates, and they immediately were analyzed by flow cytometry. Voltages in the forward-scatter (FSC) and side-scatter (SSC) detectors were set to permit the discrimination of tumor cells (high FSC and high SSC) from lymphocytes (low FSC and low SSC) on the basis of light scatter. Electronically gated tumor cells were subsequently analyzed for the binding of annexin V-FITC using the FL1 detector.

We modified a previously described method to distinguish living from dead cells.22 Briefly, target cells (2 × 10³ in 100 μL complete RPMI) were cultured alone or cocultured with either human αβ-or γδ-T cells at various E:T ratios in 96-well flat-bottom tissue culture trays (Corning Glassware, Corning, NY). Cells were incubated for 18 hours at 37°C in 5% CO₂, after which 100 μL of a solution containing 10 μg/mL ethidium bromide and 3 μg/mL acridine orange (Sigma, St Louis, MO) was added. Cells were immediately viewed using an inverted fluorescence microscope illuminated with a 300-W xenon light source (Intracellular Imaging, Cincinnati, OH). By using a blue filter set configured to excite for fluorescein (470 ± 20 nm excitation filter, 500 nm dichroic/beamsplitter filter, 515 nm emitter filter; Chroma Technology, Brattleboro, VT), live cells were observed to fluoresce green (acridine orange) whereas dead cells fluoresced orange (ethidium bromide).

Human cervical carcinoma cell line HeLa (ATCC); human melanoma cell lines SK-MEL-3, SK-MEL-5, and SK-MEL-28 (ATCC and kindly provided by Dr B. McAlpine, Emory University, Atlanta, GA); human T-lymphoblastoid cell line CCRF-HSB-2 (ATCC); human peripheral blood lymphoblast cell line NC-37 (ATCC); human myeloid leukemia cell line K-562 (ATCC), human ovarian cancer cell line SK-OV-3 (ATCC), and human B-cell lymphoma line OCI-Ly823 were used as targets for chromium release assays. Target cells were labeled with 100 μCi Na₂⁵¹CrO₄ (Amersham Pharmacia Biotech, Piscataway, NJ) from 2 hours to overnight at 37° C, after which cells were washed, trypsinized, and resuspended in RPMI containing 10% FBS. Cells were then plated (2 × 10³/well) in 96-well V-bottom microtiter trays. Purified αβ- or γδ-T cells in varying numbers were added to target cells in a final volume of 150 μL. Trays were briefly centrifuged and then incubated for 4 hours at 37° C, after which 50 μL supernatant was removed to determine ⁵¹Cr release in cpm. Percentage specific target cell lysis was calculated as [(experimental release−spontaneous release)/(maximum release−spontaneous release)]×100. Maximum and spontaneous release were respectively determined by adding either 0.1% Triton X-100 or culture medium alone to labeled target cells in the absence of effector cells. Data are presented as the mean (± SD) of triplicate samples.

Previously, we described the expansion of human CD56⁺αβ-T cells arising in OKT3/IL-2-stimulated PBMC cultures, particularly if these cultures were first primed with IFN-γ 24 hours before stimulation with mitogens.24 25 In the process of examining the role of various surface antigens involved in CD56⁺ αβ–T-cell expansion, the inclusion of one particular mouse antihuman CD2 mAb (S5.2, IgG_2a), but not its isotype control, resulted in a large increase in the percentage (Figure 1A) and absolute number of γδ-T cells (Figure 1B). Table 1 shows the results of experiments performed using PBMC obtained from several additional healthy donors. Data are presented as the mean fold expansion of these cultures (± SD, n = 5) determined after equivalent-length short-term cultures.

The importance of IL-12 in the mAb S5.2-mediated expansion of γδ-T cell is shown in Figure 2, where the greatest percentage and absolute numbers of γδ-T cells are found in cultures initiated in the presence of anti-CD2 mAb S5.2 and exogenous IL-12. Importantly, if a neutralizing mAb to human IL-12 (but not its isotype control, not shown) is added to cultures initiated in the presence of mAb S5.2, both the percentage of γδ-T cells (Figure2A, lower histogram) and the absolute number of γδ-T cells (Figure2B, right column, anti–IL-12) are significantly diminished. Furthermore, as indicated in Figure 2, panel C, γδ–T-cell expansion induced by the addition of S5.2 and exogenous IL-12 does not occur as a consequence of the inhibition of αβ–T-cell growth or expansion. In addition, from these data we conclude that 5 μg/mL is the optimum mAb S5.2 concentration to promote γδ–T-cell expansion because in most instances, at higher concentrations (>10 μg/mL), culture growth is often globally inhibited in the presence ofeither the anti-CD2 mAb S5.2 or the corresponding isotype control IgG_2a antibody (not shown). We attribute this finding to possibly the nonspecific inhibitory effects of the low concentration of sodium azide present in the mAb S5.2 preparation.

The existence of accessory or alternative CD2 signaling pathways triggered by mAbs to CD2, which function exclusively in γδ-T cells, has previously been suggested by several investigators.18,26 Although most anti-CD2 mAbs capable of delivering proliferative signals to either αβ- or γδ-T cells appear to do so only if combined with a second anti-CD2 mAb recognizing a separate CD2 epitope, single epitope-binding anti-CD2 mAbs have been reported that appear to preferentially stimulate γδ-T cells.18,26 27 We performed the following experiments to show that mAb S5.2 functions in an agonistic and not a blocking capacity, thereby initiating rather than inhibiting CD2 signaling events that contribute to IL-12–dependent γδ–T-cell expansion.

In mice and humans, both CD58 (LFA-3) and CD48 have been shown to serve as ligands for CD2; in humans, however, only CD58 has been shown to interact with CD2 on T cells in a functionally significant manner.28-31 We reasoned, therefore, that if anti-CD2 mAb S5.2 were inducing γδ–T-cell expansion by blockinginteractions between CD2 on γδ-T cells and CD58 expressed on other cells in culture, then the effect of a neutralizing anti-CD58 mAb would be the same—the enhancement of γδ–T-cell expansion. This is clearly not the case, as is shown in Figure3, panel A. To further demonstrate that mAb S5.2 is acting in an agonistic rather than an inhibitory manner, we next compared the capacity of both soluble and immobilized mAb S5.2 to induce γδ–T-cell expansion. Antibodies that bind to specific cell surface receptors usually cannot trigger signaling through these receptors unless immobilized or cross-linked. Consistent with this, previous reports18 26 have shown that the single anti-CD2 mAbs known to induce proliferative responses in γδ–T-cell clones appear to do so only if immobilized or if accessory cells that can cross-link the mAb through Fc receptors (FcR) are present. Because CD14⁺ cells (monocytes) present in our mitogen-stimulated cultures express FcR capable of cross-linking mAb S5.2 (mouse IgG_2a), we performed the following experiments using PBMC first depleted of monocytes (<0.1% CD14⁺ cells; not shown). Figure 3, panel B shows that in mitogen-stimulated cultures, γδ-T cells can be induced to expand significantly by immobilized, but not soluble, mAb S5.2.

Especially in the presence of IL-2, γδ-T cells rapidly undergo apoptosis after receiving mitogenic stimuli through the TCR.17 Thus, one possible interpretation of our findings is that CD2 engagement by mAb S5.2 in the presence of IL-12 provides a signal to a subset of γδ-T cells that renders them resistant to activation-induced cell death caused by mitogenic OKT3 and IL-2.

Annexin V binds with high affinity to phosphatidylserine (PS), which is normally confined to the inner plasma membrane leaflet of nonapoptotic cells; the appearance of PS on the outer plasma membrane leaflet is an early event associated with apoptosis. These findings have been exploited to allow the examination of apoptosis by flow cytometric means.32 33 Thus, annexin V-FITC, in combination with directly conjugated antibodies, can be used to detect apoptosis occurring in phenotypically defined subpopulations of cells within heterogeneous cell cultures.

To demonstrate that CD2 engagement by mAb S5.2 in the presence of IL-12 protects γδ-T cells from activation-induced cell death, we performed the following experiment. By convention, we designated day 0 stimuli (IFN-γ, IL-12, and anti-CD2 mAb S5.2) as protective signals. PBMC cultures were initiated as described above. Those receiving day 0 stimuli were defined as protected; those not receiving day 0 stimuli (PBS only) were defined as unprotected. All cultures received OKT3 and IL-2 on day 1. After receiving the day 1 mitogenic signals, γδ- and αβ–T-cell populations within protected and unprotected cultures were assessed for apoptosis using 4-color flow cytometry 3 days after mitogen stimulation. As shown in Figure4, in the absence of protective day 0 signals, mitogen stimulation induces apoptosis in most γδ-T cells. In contrast, apoptosis occurs to a far lesser extent in γδ-T cells receiving day 0 protective signals. These results also show that apoptosis occurring in αβ-T cells in response to mitogenic stimulation is negligible under either of these conditions. In this regard, αβ-T cells serve as a control and support in part our argument that it is the γδ–T-cell compartment in cultures that is most affected by our manipulations.

To demonstrate that the combination of CD2-mediated signals and IL-12 signaling promotes the expansion of apoptosis-resistant γδ-T cells, the following experiment was performed. Separate PBMC cultures were prepared (Figure 5A-E) receiving on day 0 as indicated, IFN-γ, IL-12, or anti-CD2 mAb S5.2. After 24 hours (day 1), all cultures received mitogenic stimulation with OKT3 and IL-2; after 21 days, γδ-T cells in each culture were analyzed for apoptosis. The percentages of viable and apoptotic cells in each dot plot are indicated in the corresponding quadrants and in the mean fold expansion (±SD) of apoptosis-resistant γδ-T cells in these cultures. As shown in Figure 5, panel E, the smallest percentage of apoptotic γδ-T cells and the greatest fold expansion of nonapoptotic γδ-T cells is found in cultures that received protective signals, including both exogenous IL-12 and anti-CD2 mAb S5.2. In contrast, a greater percentage of apoptotic γδ-T cells (annexin⁺/PI⁻) and a significantly lower expansion of viable γδ-T cells are noted in cultures to which no mAb S5.2 was added (Figure 5A-C). Interestingly, as shown in Figure 5, panel D, in cultures to which only mAb S5.2 has been added (no exogenous IL-12 or IFN-γ), it appears that though a significant proportion of γδ-T cells in these cultures remains viable, a significantly reduced expansion of viable γδ-T cell occurs. We interpret this to indicate that whereas engagement of CD2 with mAb S5.2 may generate a critical signal that induces resistance to apoptosis in mitogen-stimulated γδ-T cells, these signals in the absence of IL-12 are not sufficient to induce a significant expansion of apoptosis-resistant γδ-T cells. Thus, in conjunction with CD2-mediated signals, IL-12 appears to act synergistically to induce the greatest degree of expansion of apoptosis-resistant γδ-T cells. It is especially important to emphasize that day 0 signals alone (IFN-γ, IL-12, and anti-CD2 mAb S5.2), without day 1 signals (OKT3 and IL-2), cause no significant αβ- or γδ–T-cell proliferation (not shown).

We have postulated that signaling through CD2 in the presence of IL-12 can protect γδ-T cells from mitogen-induced apoptosis. Alternatively, these signals might be leading to enhanced γδ–T-cell expansion by simply providing an early proliferative advantage to γδ-T cells compared with αβ-T cells. The following experiments were performed to show that this is not the case. By measuring [³H]-thymidine incorporation, we compared proliferative capacities of γδ- and αβ-T cells at bothearly time points (culture initiation, Figure6A) and late time points (3 week cultures, Figure 6B). These data show that γδ-T cells isolated early from protected cultures do not proliferate to a greater degree than αβ-T cells isolated from identical cultures. This is in contrast to γδ-T cells isolated later from protected cultures, which clearly manifest enhanced proliferative capacities compared to αβ-T cells. These data do not support a model where overrepresentation of γδ-T cells in longer-term S5.2-treated cultures occurs as a consequence of an early γδ–T-cell proliferative advantage.

Despite the significant early mitogen-induced apoptosis occurring in unprotected γδ-T cells, after 7 days, surviving γδ-T cells are present in these cultures, though to a lesser extent than in protected cultures (Figure 7). Nonetheless, 1 day after the subsequent addition of IL-2 (day 8), a significantly greater percentage of unprotected but not protected γδ-T cells are induced to undergo apoptosis. This indicates that compared to unprotected γδ-T cells, protected γδ-T cells remain relatively resistant to IL-2–induced apoptosis. Furthermore, the ability of agonistic mouse antihuman CD95/Fas mAb CH11 to induce apoptosis in both protected and unprotected γδ-T cells suggests that the greater resistance to apoptosis of unprotected γδ-T cells does not result from a simple loss of CD95/Fas expression and is supported by the findings that CD95/Fas expression determined by FACS does not differ between protected and unprotected γδ-T cells (data not shown).

We next examined whether apoptosis-resistant γδ-T cells exert measurable antitumor activity against human tumor cells in vitro. We explored this question using 3 distinct methods. In all experiments, apoptosis-resistant γδ-T cells and control αβ-T cells were expanded and isolated simultaneously from a given individual. In virtually all instances, control αβ-T cells derived fromeither protected (day 0 plus day 1 signals) or unprotected cultures (day 1 mitogenic stimulation alone) were indistinguishable with regard to antitumor activity (not shown). Thus, αβ-T cells derived from protected or unprotected cultures were used interchangeably as controls for MHC-restricted alloreactivity.

We first examined the antitumor activity of apoptosis-resistant γδ-T cells against a number of human tumor cell lines using a standard 4-hour ⁵¹Cr-release assay. Labeled tumor cells were incubated with apoptosis-resistant γδ-T cells or control αβ-T cells derived from a given person. Specific tumor lysis was measured, and results obtained from 2 separate persons are shown in Figure 8, panel A and panel B, respectively. Table 2 compiles the results of additional experiments performed using apoptosis-resistant γδ-T cells and control αβ-T cells derived from other healthy donors and tested against the indicated tumor cell lines. To allow comparison of multiple experiments, these data are presented as percentage- specific tumor lysis at an E:T ratio of 20:1.

Although ⁵¹Cr-release assays remain an established means by which to measure the in vitro cytotoxic activity of effector lymphocytes, we performed the following experiment to demonstrate that we could also measure the specific induction of apoptosis in sensitive target cells on coculture with effector γδ-T cells. HeLa cells were initially chosen for further experimentation. HeLa cells were cultured alone or cocultured for 4 hours with apoptosis-resistant γδ-T cells or control αβ-T cells at varying E:T ratios (1:1 to 20:1). As shown in Figure 9, panel A, T-lymphocytes alone (left plot) and HeLa target cells alone (center plot) have characteristic light scatter properties that allow both cell populations to be distinguished, even when mixed together in coculture (right plot). By gating only on HeLa cells, it was then possible to analyze HeLa target cells for their uptake of annexin V-FITC; this served as a measure of the ability of cocultured γδ- or αβ-T cells to induce apoptosis (Figure 9B). These representative data demonstrate that apoptosis-resistant γδ-T cells can induce a significantly greater degree of apoptosis in HeLa cells when compared to control αβ-T cells. Similar results were obtained using other target cells such as human melanoma cell lines (not shown).

Finally, we examined the ability of apoptosis-resistant γδ-T cells to kill tumor targets using acridine orange and ethidium bromide uptake as a means to distinguish live from dead target cells. As shown in Figure 9, panel C, at an E:T ratio as low as 1:1, HeLa tumor cell viability was significantly decreased after 18 hours of coculture with apoptosis-resistant γδ-T cells but not with control αβ-T cells. Although in agreement with the ⁵¹Cr-release and annexin data above, these experiments may be of additional biologic and even clinical significance given that this degree of tumor cell killing was induced at a significantly lower E:T ratio (1:1). Higher E:T ratios (ranging from 5:1 to 20:1, not shown) resulted in a similar induction of target cell death with the addition of γδ-T cells but not control αβ-T cells.

The existence of alternative CD2 signaling pathways that function predominantly, if not exclusively, in γδ- but not αβ-T cells has been established by the important work of others.18,19 34 These pathways were largely revealed by the recognition that certain anti-CD2 mAbs could generate signals in cloned T cells, resulting in proliferation as measured by standard means, such as [³H]-thymidine incorporation. Although our current work is in agreement with these fundamental findings, we are able to extend these findings in several important biologic and possibly clinically relevant ways: our recognition that anti-CD2 mAb S5.2 can generate IL-12–dependent signals that protect human γδ-T cells from mitogen-induced apoptosis now provides us with the biologic basis for obtaining large numbers of viable and, as we show,functional human γδ-T cells.

The relation between CD2 signaling, IL-12, and acquired resistance to apoptosis in γδ-T cells is likely complex, but several observations may help elucidate the mechanisms underlying our observation, beginning with signaling through CD2 itself. In most studies, the capacity of anti-CD2 mAbs to signal through CD2 is almost entirely assessed in terms of proliferation.26,27,35,36 However, engagement of CD2 is not always associated with transduction of proliferative signals. Breitmeyer and Faustman37 have shown that the rosetting of human T cells by sheep red blood cells (which express a functional homologue of human CD58/LFA3, the natural ligand for CD2) results in the paralysis of TCR/CD3-mediated signal transduction and activation. These studies demonstrated that both calcium mobilization and proliferative responses to subsequent mitogenic anti-TCR antibodies were blocked for up to 48 hours after CD2 engagement. Similarly, in a more recent report, Miller et al38 showed that interaction between CD58/LFA-3 and CD2 can lead to T-cell unresponsiveness to antigenic or mitogenic stimuli in vitro. Conversely, it has been reported that the ability of certain anti-CD2 mAbs to stimulate γδ–T-cell clones was significantly diminished by the co-engagement of CD3, suggesting a possible antagonistic interaction between γδ-T cell CD2 and CD3 signaling pathways.26 

Reasoning along these lines, we propose that γδ–T-cell expansion may occur in our system, not simply as a consequence of CD2-mediated proliferative signals per se but rather as a consequence of postreceptor CD2-mediated disruption or moderation of signals that could otherwise induce apoptosis in already apoptosis-prone γδ-T cells. In such a model, it is particularly important to emphasize that mAb S5.2 functions in an agonistic rather than in a blocking capacity,initiating rather than inhibiting CD2 signaling events (Figures 3).

In any event, it is important to note that CD2 signaling in the absence of IL-12 is insufficient to lead to the expansion of apoptosis-resistant γδ-T cells in mitogen-stimulated PBMC cultures, as is established by the neutralization experiments shown in Figure 2, panels A and B. This suggests that CD2 signaling is necessary, but not sufficient, for the expansion of apoptosis-resistant γδ-T cells. Conversely, it is evident that in the absence of CD2 signaling, IL-12 can neither enhance γδ–T-cell expansion (Figure 2C, Table 1) nor inhibit apoptosis (Figure 5C) in mitogen-stimulated γδ-T cells. Taken together, these data suggest that, in our system, γδ-T cells do not optimally respond to IL-12 without first receiving signals through CD2. Such an interpretation would be in keeping with the findings of Gollub et al,29 30 who have previously shown that responsiveness to IL-12 in activated T cells was indeed regulated and dependent on signals delivered through CD2. Thus, the requirement for CD2 signaling in our system might be explained on the basis that these signals simply enhance responsiveness to IL-12 in γδ-T cells, leading to apoptosis resistance. This would be in part supported by data that show IL-12–receptor (IL-12R) is found to be up-regulated on protected but not on unprotected γδ-T cells (unpublished data, manuscript in preparation).

It is unclear how IL-12 might inhibit apoptosis in mitogen-stimulated γδ-T cells, though the observations of Perussia et al39may provide an important clue. In one study, it was shown that whereas IL-12 always acts synergistically with IL-2 in inducing αβ–T-cell proliferation, in contrast, IL-12 could significantly inhibit IL-2–induced proliferation in resting γδ-T cells.39 Thus, it is conceivable that during the first critical 24 hours (day 0) of our protected cultures, responsiveness to IL-12 is first established through CD2-mediated signals delivered by mAb S5.2. Subsequent strong mitogenic signals delivered on day 1,in particular, IL-2, would then be less able to induce apoptosis in γδ-T cells. Thus, γδ-T cells spared this initial IL-2–mediated apoptosis would be those observed to expand in our cultures. That resistance to apoptosis in γδ-T cells might be a consequence of blunted responsiveness to IL-2 is supported by the findings shown in Figure 7, where clearly a differential susceptibility to IL-2–induced apoptosis is noted between protected and unprotected γδ-T cells. This is consistent with our observation that protected γδ-T cells fail to up-regulate the expression of CD25/IL-2Rα when compared to unprotected γδ-T cells (unpublished data, manuscript in preparation) and is further supported by our findings that γδ-T cells isolated early from protected cultures appear to proliferate to a lesser extent in response to IL-2 (Figure 6).

Although the principal tenet of our model is that γδ-T cells in protected cultures preferentially expand as a consequence of acquired resistance to apoptosis, it is important to appreciate that γδ-T cells can and do expand in unprotected cultures as well, albeit to a lesser extent (Figures 1B, 2C, 5, Table 1). This must be taken into account, particularly when the magnitude of γδ–T-cell expansion is compared between various culture conditions, especially after longer periods of expansion (Figure 5). In view of this, it must be emphasized that in protected cultures, not all γδ-T cells are resistant to apoptosis; conversely, in unprotected cultures, not all γδ-T cells are apoptotic (Figures 4, 5, 7). This suggests that protective signals might simply alter the proportion of apoptosis-resistant and -sensitive γδ-T cells present at a given time in culture. Hence, if a significant fraction of γδ-T cells in protected cultures acquire resistance to apoptosis early, even if only transiently, this in itself could account for the greater γδ–T-cell expansion observed at later time points. This is supported by the observation that in both protected and unprotected cultures, the proportions of viable and apoptotic γδ-T cells change in a predictable manner over time: at very early time points (up to culture day 1), protected and unprotected cultures contain comparable total numbers and similar proportions of viable and apoptotic γδ-T cells (not shown). However, by culture day 2 to day 4, though total γδ–T-cell numbers still remain comparable (not shown), protected cultures are routinely found to contain a greater proportion of apoptosis-resistant γδ-T cells (Figure 4). Eventually, as protected and unprotected cultures enter exponential growth phases (usually simultaneously between day 6 and day 10; not shown), protected cultures—already containing a larger proportion of viable γδ-T cells (Figure 7)—invariably proceed to surpass unprotected cultures with regard to expansion of viable γδ-T cells (Figure 1). This survival advantage of γδ-T cells in protected cultures is further accentuated after restimulation with IL-2 (Figure 7) and is ultimately reflected in the larger total numbers of viable γδ-T cells found in protected cultures at even later time points up to 45 days and beyond (not shown). Although we do not propose that protective signals render γδ-T cells permanently resistant to apoptosis, it is interesting to note that by the third week in culture, compared to unprotected cultures, protected cultures still contain a larger proportion of apoptosis-resistant γδ-T cells, though the differences in the magnitude of γδ-T cell–expansion is often no longer as great (Figure 5).

With regard to antitumor activity, the data provided in Figures 8 and 9and in Table 2 establish that apoptosis-resistant γδ-T cells, expanded and isolated from a number of persons, do indeed possess the ability to kill a variety of human tumors in vitro as measured by several methods. Although the mechanism of γδ–T-cell tumor recognition is not addressed here, several points are noteworthy. First, in virtually all instances in which killing is observed, apoptosis-resistant γδ-T cells were found to kill tumor targets to a significantly greater degree than αβ-T cells on a cell-for-cell basis. These data suggest that γδ-T cells recognize target cells by mechanisms distinct from those used by αβ-T cells, as is known.1,2,4 Second, it is interesting to note that a number of tumor cell lines of epithelial origin were found to be relatively sensitive to killing by γδ-T cells. This is especially intriguing given the recent findings that γδ-T cells expressing particular Vδ TCR can recognize an MHC class I-related molecule frequently expressed on tumor cells of epithelial origin.40-42 Third, that the prototypic natural killer (NK)–sensitive target K562 is found to be relatively resistant to γδ-T-cell–mediated killing might also suggest the involvement of mechanisms distinct from NK or lymphokine-activated killer cell–mediated killing. These cytotoxicity data, though not exhaustive—especially with respect to the number of donors tested or the number of tumor targets assessed—nevertheless serve to underscore the practical significance of our findings, namely that an otherwise rare subset of human lymphocytes can now readily be expanded, isolated, and subjected to more rigorous study. This is particularly relevant as one considers the potential for using such cells in various forms of adoptive immunotherapy.

Whether γδ-T cells have therapeutically exploitable biologic properties such as antiviral, antitumor, or hematopoietic stem cell graft-facilitating effects, remains to be determined. Although far larger numbers of apoptosis-resistant γδ-T cells would be required to design adoptive cellular therapy clinical experiments, it should be emphasized that usually only 2 mL PBMC (1 × 10⁶cells/mL), derived from 3 to 5 mL fresh blood, is used as starting material to generate cultures larger than 50 to 100 × 10⁶ T cells containing 40% to 60% γδ-T cells after 2 to 3 weeks or longer. Thus, with proper culture optimization, many more than 1 × 10⁹ viable γδ-T cells could readily be obtained after ex vivo expansion using, as starting materials, safely procurable volumes of fresh autologous or allogeneic peripheral blood. Such γδ-T-cell–enriched products could readily be subjected to positive or negative selection techniques (immunomagnetic columns, high-speed FACS, panning, and so on) to obtain a product essentially “pure” in terms of γδ–T-cell content and, thus, ideal for examining clinical questions, something that was until now not practically possible.

We thank Christopher Ferrigno for his technical contributions to this work.