CD8+ T cells stimulated in vitro with anti-TCR mAb and B7-1 or ICAM-1 produce IL-2 and clonally expand. Effector function is acquired within 3 days, but proliferation ceases and the cells begin to die by apoptosis. Stimulation in vivo with B7-1-expressing allogeneic tumor results in the same sequence of events with a comparable time course. In both cases, the cells become anergic within 3 or 4 days of responding; they can no longer respond by producing IL-2 and proliferating, but can still be stimulated to proliferate in response to exogenous IL-2. This activation-induced nonresponsiveness (AINR) is not simply a consequence of ongoing cell death; cytokines that promote survival (IL-7 or IFN-α) or proliferation (human IL-2) do not restore the ability to produce IL-2 in response to costimulation. Although similar to the anergy described for CD4+ T cell clones, AINR differs in that it results from an initial stimulation with both signal 1 and signal 2. AINR appears to be an aspect of the normal differentiation of fully stimulated CD8+ T cells. It is probably important in regulating CTL responses; it limits the initial T helper-independent response and converts it to a response that requires T cell help to be sustained and further expanded. When the initial helper-independent response is not sufficient to clear Ag, and if help is not available, AINR likely results in tolerance to the Ag.
Generation of CD8+ T cell responses can either depend upon help from CD4+ T cells or can be helper-independent. In some systems, CD8+ T cells can make an initial, limited response in the absence of CD4+ T cell help, but the response then declines rapidly unless T cell help is available (1, 2, 3, 4, 5). A helper-independent phase of the CD8+ T cell response is consistent with the ability of CD8+ T cells to support their own expansion by producing IL-2 in response to costimulation provided by CD28 binding to B7 ligands (6, 7, 8), LFA-1 binding to ICAM-1 (9, 10), and possibly other costimulatory receptors binding to their ligands. Thus, a helper-independent response can be initiated provided that class I Ag is presented on APC expressing costimulatory ligands.
It is unclear why this autocrine IL-2 production is not sufficient to support a sustained CD8+ T cell response, and that, instead, the response often becomes dependent on help from CD4+ T cells (1, 2, 3, 4, 5). One possibility is that, as the CD8+ T cells develop lytic effector function, they eliminate the APC so that effective Ag presentation is no longer occurring. The decline that occurs following the initial response involves substantial deletion of the activated CD8+ T cells through death and could be consistent with growth factor deprivation in the absence of continued stimulation. It has also been shown, however, that some of the CD8+ T cells that are present during the decline appear to be anergic, unable to respond to the same stimulus that was effective in initially activating the response. In vivo induction of anergy in CD8+ T cells has been found in response to virus (11), HY Ag (12, 13), and superantigens (14).
The origin and status of these anergic CD8+ T cells is not well understood. Classical anergy, defined originally for CD4+ T cell clones, is the nonresponsive state that results when the cells receive a signal 1 stimulus via the TCR in the absence of signal 2 via a costimulatory receptor, and are thus rendered unable to subsequently produce IL-2 in response to full stimulation (15, 16, 17). Thus, the anergic CD8+ T cells present following an initial in vivo response could potentially be a subset of the Ag-specific cells that recognized Ag in a context where costimulation was lacking; it has been demonstrated that cloned CTL lines can be rendered anergic by exposure to just signal 1 (18). The role of anergy in limiting the helper-independent CTL response is also not understood. The anergic state might not be relevant if it simply represents a stage at which cells have lost responsiveness on their way to deletion. Alternatively, it might be essential for preventing further autocrine IL-2 production and converting the response to one that is now under the regulation of CD4+ T cells.
When CD8+ T cells are stimulated in vitro, proliferation and clonal expansion peak on day 3 or 4 and then decline, and this decline has been attributed to killing and elimination of the APC as the CD8+ T cells become lytically active by day 3 (19). However, when the stimulus was anti-TCR mAb and purified B7-1 coimmobilized on latex microspheres, a comparable time course was observed, even though the inert beads could not be “killed,” and the stimulus thus could not be eliminated by the effector cells (8). This raised the possibility that CD8+ T cells may become anergic following a response to full stimulation with both signal 1 and signal 2, i.e., they may become unable to produce IL-2 and proliferate in response to subsequent stimulation. The results described here demonstrate that this is the case for cells responding both in vitro and in vivo, and this activation-induced nonresponsiveness (AINR)3 has important implications for the regulation of CTL responses.
Materials and Methods
Cells, cell culture, and reagents
T cells were obtained from 6- to 12-wk-old female mice of either the C57BL/6 strain (Charles River Breeding Laboratories, Wilmington, MA) or the OT-1 strain having a transgenic TCR specific for Kb/OVA ( (20); a kind gift from Dr. Frank Carbone, Monash Medical School, Victoria, Australia). Mice were maintained in the University of Minnesota (Minneapolis, MN) specific pathogen-free animal facilities according to National Institutes of Health guidelines. Lymph nodes were harvested, and CD8+ responders were purified as described (8). Responders were routinely assessed for purity by flow cytometry and were always >90% CD8+ with <1% CD4+ contamination.
A soluble form of ICAM-1 was purified, as described by Kuhlman et.al. (21). Purification of B7-1 and preparation of microspheres bearing optimal levels of the anti-TCR mAb F23.1, B7-1, and ICAM-1 have been described in detail (8, 22). Purified CD8+ C57BL/6 responders were always found to be 20–23% positive for Vβ8 TCR expression. For experiments using responder cells from OT-1 mice, anti-Vα2 mAb B20 (PharMingen, San Diego, CA) was used. Triplicate cultures were set up as described (8) using Falcon (Franklin Lakes, NJ) flat-bottom 96-well tissue culture plates . Primary cultures were initiated with 5 × 104 purified cells, and cultures of activated cells were initiated with 2 × 104 viable cells, or as indicated. F23.1/B7-1/ICAM-1 microspheres (F23/B/I) were added to cultures at 1 × 105 per well, and the final volume was 0.2 ml. For the experiments shown in Figs. 5–8, activated responders were pooled from multiple microwell cultures and, viable cells were purified over Lympholyte-M (Cedarlane, Hornby, Ontario, Canada), according to the instructions provided by the supplier. For assessment of proliferation, cells were pulsed with 1 μCi of [3H]TdR for the last 6–8 h of culture, lysed with distilled H2O, and the amount of [3H]TdR incorporated into DNA was determined by liquid scintillation counting. Results are expressed as the average ± SD of triplicate samples. For quantitation of clonal expansion, cells were counted using a hemacytometer, and viability was determined based on trypan blue exclusion.
Final concentrations in cultures of cytokines used were 50 U/ml for murine IL-2 (PharMingen), 1000 U/ml for human IL-2 (R&D Systems, Minneapolis, MN), 5 ng/ml for recombinant murine IL-7 (R&D Systems), and 5000 U/ml for IFN-α A/D (Genzyme, Cambridge, MA). Anti-IL-2R mAb, PC61.5.3, was purified from culture supernatant and added to cultures at a final concentration of 10 μg/ml in blocking studies. CTLA-4 blocking was performed using purified anti-CTLA-4 mAb, clone 9H10 (PharMingen), at 50 μg/ml.
In vivo CTL activation and ex vivo functional analysis
The procedure for adoptive transfer, stimulation, and analysis of 2C TCR transgenic CD8+ T cells has been described in detail (23, 24). Briefly, 5 × 106 adherence-depleted spleen and lymph node (LN) cells from 2C transgenic mice (25) (a kind gift of Dr. Dennis Loh Hoffman-La Roche, Nutley, NJ) were injected i.v. into C57BL/6 sex-matched recipients. After 2 days, recipient mice were challenged i.p. with 5 × 107 B7-1-transfected P815 tumor cells (26) (a kind gift from Dr. Thomas Malek, University of Miami School of Medicine, Miami, FL). After 6 days, spleen and peritoneal exudate cells were harvested from challenged mice, and the CD8+ T cells were purified. Resting CD8+ 2C T cells were similarly purified from the spleen of transgenic mice. Ex vivo restimulation for detection of proliferation by [3H]TdR uptake was performed essentially as described above for stimulation of C57BL/6 cells, with the exception that 2.5 × 104 responders were used and microspheres had the 1B2 anti-clonotypic mAb specific for the 2C receptor (1 μg/ml) immobilized instead of F23.1 mAb.
Clonal expansion of 2C transgenic T cells resulting from ex vivo stimulation was determined by flow cytometry. Purified responders were stained with anti-CD8 and 1B2 mAbs and quantitated as described (23) to determine input responder frequency. A total of 2.5 × 105 cells were then cultured in 2 ml of media in 24-well plates for 72 h. At the end of the culture period, cells were harvested, viability determined by trypan blue exclusion, and the frequency of 1B2+/CD8+ 2C cells determined by flow cytometry. For this analysis, viable lymphocytes were gated based on forward and side scatter characteristics and exclusion of 7-amino-actinomycin D (7-AAD).
Cells and microspheres were stained using identical protocols using the relevant Ab for 20 min at 4°C in 0.1 ml of HBSS with 2% FCS and 0.02% NaN3 (flow buffer) in the dark. Two washes (before analysis and between multistep stains) were performed using flow buffer. Flow cytometric analysis was performed on a Becton Dickinson (Mountain View, CA) FACScan using CellQuest software. Ligand densities on all microsphere preparations were quantitated as described elsewhere (22). To quantitate apoptosis, cells were stained with 1 μg of F23.1 mAb, washed, then stained with 1 μg FITC-conjugated goat anti-mouse Ig (Jackson ImmunoResearch, West Grove, PA) and washed. The cells were then fixed with 70% ethanol for 60 min on ice, washed with PBS, and resuspended in 1 ml PBS with RNase I-A at 0.5 μg/ml. An amount of 1 ml of propidium iodide at 100 μg/ml was then added and the cells incubated with periodic agitation for 1 h at room temperature in the dark. Vβ8+ cells were gated, and the percent of F23+ cells having hypodiploid DNA content was determined based on the propidium iodide staining.
For the phenotypic analysis shown in Fig. 6, cells were stained first with 1 μg of F23.1 then with 1 μg PE-conjugated goat anti-mouse Ig (Jackson ImmunoResearch) and either nothing or the indicated FITC-conjugated mAbs. Anti-CD8-FITC (clone CT-CD8α) and anti-CD28-FITC (clone 37.51.1) were from Caltag (Burlingame, CA), and anti-LFA-1-FITC (clone 2D7) was from PharMingen. A three-step procedure was used for detection of CTLA-4 expression. Approximately 5 × 105 cells were stained first with anti-CTLA-4 (clone 9H10; PharMingen), followed by biotin-conjugated goat anti-hamster Ig (Jackson ImmunoResearch) and then PE-conjugated streptavidin (PharMingen). Control staining was performed with second- and third-step reagents only. Viable lymphocyte responders were gated based on forward and side scatter characteristics (22) and exclusion of 7-AAD.
Sandwich ELISAs were used to detect murine IL-2 and IFN-γ in 0.05 ml supernatants removed from cultures at the times indicated, using mAbs obtained from PharMingen, according to the protocol provided by the supplier. HRP-conjugated streptavidin was from Sigma (St. Louis, MO). Standard curves using recombinant murine IL-2 or IFN-γ (PharMingen) were used to quantitate production. For both cytokines, the detection limit of the assay was ∼0.04 ng/ml. Results are expressed as the average ± SD of triplicate samples.
B7-1 and ICAM-1 costimulate transient IL-2 production and clonal expansion of CD8 T cells
Stimulation of resting C7BL/6 CD8+ T cells with microspheres having just F23.1 anti-TCR mAb immobilized on the surface results in only marginal proliferation, as measured by [3H]TdR incorporation (Fig. 1,A). In contrast, beads having purified B7-1 coimmobilized with the F23.1 (F23/B) stimulate quite effectively, and the most potent stimulation is obtained when both B7-1 and ICAM-1 are coimmobilized with the anti-TCR mAb (F23/B/I) (Fig. 1,A). The cells initially produce IL-2 in response to costimulation with B7-1 and ICAM-1, but levels decline dramatically by day 4 (Fig. 1,B). Proliferation is paralleled by clonal expansion through days 3 and 4, but at longer times, IL-2 levels decline, [3H]TdR incorporation ceases, total cell numbers no longer increase, and cell viability rapidly declines (Fig. 1 C) as cells die by apoptosis. Apoptotic death during this period was demonstrated by analysis of subdiploid DNA content using propidium iodide, and confirmed using 7-AAD staining to detect apoptotic cells (data not shown). Candidates for mediating the death of the activated T cells include TNF that is produced by activated CTL, and Fas-dependent killing, since activated CD8+ T cells express both Fas and Fas ligand. However, neither neutralizing anti-TNF-β mAb nor anti-Fas ligand mAb inhibited the apoptotic death (data not shown). Death due to growth factor withdrawal also occurs by apoptosis (27), and disappearance of IL-2 from the cultures appears likely to be responsible for the death that occurs at later times, since addition of exogenous IL-2 can prolong survival (see below).
CD8 T cells exhibit AINR to costimulation following response to an initial activating stimulus
The decline in the CD8+ T cell response that begins after day 3 could not be attributed to elimination of the APC, since the microspheres cannot be killed. However, it was possible that activation ceased because the anti-TCR mAb and/or costimulatory ligands were no longer available on the microspheres due to loss or degradation. Analysis of the beads by flow cytometry after 4 days indicated that this was probably not the case, since only small decreases in the density of the surface ligands were found (data not shown). This raised the possibility that the cells were becoming nonresponsive as a result of activation, even though signals 1 and 2 were both being delivered, and experiments were done to directly examine this.
CD8+ T cell cultures were stimulated with microspheres having F23.1, B7-1, and ICAM-1 on the surface, and the cells responded with the expected time course of proliferation, clonal expansion, and decline in viable cell numbers (Fig. 2,A). Cells stimulated in parallel were harvested from culture on day 4 (90 h), washed, and viable cell number determined. The cells were then placed in culture in fresh medium at a cell density comparable to that used at the initiation of the cultures on day 0, stimulated in various ways, and both proliferation and viable cell number determined over the next several days (Fig. 2, B and C). In the absence of stimulation, minimal [3H]TdR incorporation was found and the number of viable cells declined. In contrast, exogenous IL-2 provided potent stimulation of proliferation, and the number of viable cells expanded greatly over days 5–7. Although IL-2 could support proliferation and clonal expansion, microspheres having F23.1, B7-1, and ICAM-1 on the surface (F23/B/I beads) could not. Thus, F23/B/I beads that can provide costimulatory activation to induce IL-2 production and proliferation by resting CD8+ T cells could not support continued expansion of previously activated cells. The experiment shown in Fig. 2 is representative of at least five similar experiments. Additional experiments have been done with cells stimulated 3 days (72 h) before washing and restimulation, and similar results have been obtained (data not shown, and see below).
CD8+ LN cells from normal C57BL/6 mice include both naive and memory cells, as assessed by CD44 levels and other phenotypic markers (28), with 10–15% of the cells having a memory phenotype. In other work, we have isolated naive (CD44low) and memory (CD44high) CD8+ cells from C57BL/6 by cell sorting and shown that both populations respond to stimulation with F23.1 and either B7.1 or ICAM-1 as the costimulatory ligand (29). To further confirm that naive CD44low cells could indeed respond to anti-TCR mAb and costimulatory ligands on microspheres, and directly demonstrate that they develop nonresponsiveness following an initial response, naive cells were isolated from OT-1 mice. These mice have a transgenic TCR specific for Kb/OVA, and LN cells from OT-1 include only a small CD44high population. CD8 cells were purified, and naive CD44low cells were further purified by flow cytometry, sorting for CD8+CD44low cells (Fig. 3,A). As expected from the experiments examining C57BL/6 CD8+ T cells that are predominantly naive, the naive OT-1 cells proliferated strongly in response to microspheres having anti-TCR mAb along with B7.1 and ICAM-1 on the surface (Fig. 3,B), but, by day 3, had beome nonresponsive to costimulation, although they could still proliferate in response to added IL-2 (Fig. 3 C).
Experiments were also done using Ag-bearing cells as APC for in vitro stimulation, and again, a nonresponsive state was observed by day 3 (data not shown). This was extended to examine the responsiveness of CD8+ T cells following an initial in vivo response by using adoptive transfer of TCR transgenic cells into naive recipients to allow tracking and examination of the Ag-specific cells. When 2C TCR transgenic CD8+ T cells specific for Ld alloantigen are adoptively transferred into naive C57BL/6 recipients, they rapidly equilibrate to form a stable population in the spleen and lymph nodes, where they can be easily detected using the 1B2 mAb that recognizes the transgenic 2C TCR (23). Upon challenge of the mice by i.p. injection of live allogeneic P815 tumor (H-2d), the 2C cells undergo clonal expansion in the peritoneal cavity at the site of tumor growth and develop cytolytic activity. As tumor is eliminated, the 2C cells migrate from the peritoneal cavity to the spleen and lymph nodes, and total numbers decline. With normal P815 tumor that does not express B7 ligands, the response in the peritoneal cavity peaks at about day 8 (23). When challenge is with P815 tumor that expresses B7-1, the 2C response is essentially the same with the exception that it peaks on day 6 (R. M. Kedl and M. F. Mescher, manuscript in preparation). Thus, the transgenic CD8+ T cells in the spleen and peritoneal cavity on day 6 after challenge with P815-B7-1 are at a stage in their development similar to that of responders after 3–4 days of in vitro stimulation, since 2–3 days are required for the Ag-specific cells to traffic to the peritoneal cavity and begin responding (23).
Following adoptive transfer of 2C cells and challenge with P815-B7-1 tumor, cells were isolated from the spleen and peritoneal cavity on day 6 and stimulated in vitro. Microspheres used for in vitro stimulation had the 1B2 anti-TCR mAb on the surface to insure that observed responses were mediated by the transgenic T cells. Resting 2C cells were examined in parallel and showed the expected pattern of response (Fig. 4,A). Beads with anti-TCR, B7-1, and ICAM-1 stimulated proliferation, and the response did not increase significantly when IL-2 was added, indicating that the cells produce sufficient IL-2 upon costimulation to support a maximal response. In contrast, IL-2 alone stimulated no response. Responses by the 2C cells that had undergone an in vivo response to P815-B7-1 were quite different. The 1B2, B7-1, ICAM-1 beads (1B2/B/I) stimulated no response, while IL-2 alone stimulated a strong proliferative response (Fig. 4,A). Addition of both beads and IL-2 resulted in a response essentially identical to that of IL-2 alone. When [3H]TdR incorporation is normalized to the number of 2C cells in the different populations, it is readily apparent that responses by resting cells to anti-TCR and costimulation are much stronger than those of the cells that have responded in vivo (Fig. 4 B).
Alloreactive endogenous host cells are also present in the populations being examined in these experiments, and will presumably contribute to the proliferation (as measured by [3H]TdR incorporation) that occurs in response to exogenous IL-2. Responses of the 2C cells were therefore also examined by flow cytometric analysis to determine the number of 1B2+ cells and quantitate clonal expansion upon restimulation. Again, resting cells are found to respond well to IB2/B/I beads and not IL-2 alone, while cells previously activated in vivo respond to IL-2 but not to 1B2/B/I beads (Fig. 4 C). In fact, it appears that stimulation of previously activated cells with the beads in the absence of exogenous IL-2 may increase the rate at which these cells die.
The results shown in Fig. 4 are representative of two experiments examining responsiveness of cells from P815-B7-1-challenged mice. Similar results were obtained when irradiated P815-B7-1 tumor cells were used instead of microspheres as the in vitro stimulus, and in experiments examining cells from mice challenged with normal P815 tumor cells lacking B7-1 (data not shown). Thus, full activation (with costimulation) results in subsequent nonresponsiveness of CD8+ T cells, whether the cells are responding in vitro or in vivo.
AINR is not a consequence of ongoing cell death
The above experiments demonstrated that previously activated CD8+ T cells were “anergic” in that they could no longer proliferate in response to costimulation. However, continued death of the previously activated cells was occurring during the period of restimulation with microspheres, raising the possibility that the inability to proliferate was simply a consequence of the induction of death. This appeared unlikely, since cells that could not respond to F23/B/I bead restimulation on day 3 still remained viable in sufficient numbers by day 4 to give a good response to exogenous IL-2 (data not shown). To more directly examine this issue, we took advantage of the fact that murine T cells can respond to human IL-2. Thus, human IL-2 could be added to cultures to support survival and proliferation, and, at the same time, allow the production of murine IL-2 to be examined by ELISA using a species-specific mAb for detection.
Activation of resting cells with F23/B/I beads results in clonal expansion by day 3, and this is blocked by addition of anti-IL-2R mAb, confirming that the response is IL-2-dependent (Fig. 5,A). The CD8+ T cells are stimulated to produce IL-2 (Fig. 5,B) in sufficient amounts to fully support the response, since addition of exogenous human IL-2 does not increase clonal expansion (Fig. 5,A) or alter the level of murine IL-2 that is produced (Fig. 5,B). In contrast, for cells that had been activated 3 days earlier by F23/B/I beads, restimulation with F23/B/I beads did not result in clonal expansion (Fig. 5,D) or significant IL-2 production (Fig. 5,E). When both F23/B/I beads and human IL-2 were added, the cells remained viable and expanded in number (Fig. 5,D), but even under these conditions, production of murine IL-2 was minimal (Fig. 5 E). Thus, even when previously activated cells remain viable, they are unable to produce IL-2 in response to costimulation by B7-1 and ICAM-1. When the same experiment is done in the presence of anti-IL-2R mAb to block consumption of murine IL-2, IL-2 levels in cultures of resting cells are comparable at day 2 and remain about the same at day 3, indicating that the majority of IL-2 production occurs in the first 2 days following costimulation. Murine IL-2 production remains minimal in cultures of previously activated cells, even when anti-IL-2R mAb is added to prevent consumption (data not shown).
Stimulated CD8+ T cells acquire cytolytic activity by day 3, and this activity usually reaches maximal levels by day 4 (8). Thus, the cells can receive signals via the TCR to mediate degranulation and cytolysis during the time that they are anergic with respect to costimulation of IL-2 production. A similar “split anergy” is seen when production of IFN-γ is examined. Stimulation of resting cells results in production of IFN-γ by days 2 and 3 (Fig. 5,C). When the cells are washed and placed back into culture, no release of IFN-γ is detected in the absence of further stimulation, indicating that production requires ongoing signaling (Fig. 5,F). In contrast to IL-2 production, restimulation with F23/B/I beads results in IFN-γ production by the previously activated cells that begins more rapidly and occurs to higher levels than for resting cells (Fig. 5,F). More IFN-γ is detected when exogenous IL-2 is added to cultures of the previously activated cells, since more cells remain viable for longer times to produce the cytokine (Fig. 5 F). In contrast, addition of just IL-2 in the absence of a TCR stimulus does not cause IFN-γ production (data not shown). Thus, although previously activated cells are anergic with respect to IL-2 production, they can be stimulated via the TCR to produce IFN-γ.
The effects of IL-7 and IFN-α on previously activated cells were also examined, since both can support prolonged survival of the cells but do not stimulate proliferation to as great an extent as IL-2. When cells were stimulated with F23/B/I beads, harvested on day 3, and restimulated, F23/B/I beads alone caused no proliferation (Fig. 6,A), and the number of viable cells present was low by day 3 (Fig. 6,B). As shown above, addition of human IL-2 along with the beads stimulated proliferation and clonal expansion (Fig. 6, A and B) but did not allow production of significant amounts of murine IL-2 (Fig. 6,C). IL-7 supported less proliferation or clonal expansion than IL-2, and IFN-α allowed the cells to survive without significant expansion (Fig. 6, A and B). In neither case were significant amounts of murine IL-2 produced in response to the F23/B/I beads, while IFN-γ production was stimulated (Fig. 6, C and D). Thus, previously activated CD8+ T cells remain anergic with respect to IL-2 production when cell viability is maintained with or without further proliferation.
AINR is not due to receptor down-regulation or negative signaling by CTLA-4
Ag or anti-TCR Abs can stimulate internalization of the TCR (30, 31), and, in some in vivo systems, CD8+ T cells that have been rendered anergic following a strong in vivo response have dramatically decreased surface expression of TCR and CD8 (12). This appeared unlikely to be the case for the anergy examined here, since anti-TCR mAb could still trigger IFN-γ production by the cells (Figs. 5 and 6). This was confirmed using flow cytometry to examine surface receptor levels on resting cells in comparison to cells stimulated 3.5 days previously with F23/B/I beads. TCR expression was comparable on both resting and activated cells, while CD8 expression was somewhat increased on the activated cells (Fig. 7,A). Furthermore, both the CD28 and LFA-1 costimulatory receptors were expressed at substantially higher levels on the activated cells (Fig. 7 A).
Activated T cells express CTLA-4, a second receptor for B7-1 and B7-2 ligands, and recent evidence is suggesting that this receptor may deliver a “negative” signal to down-regulate T cell responses (32, 33, 34, 35). Stimulation of CD8+ T cells with F23/B/I beads results in up-regulation of CTLA-4 expression by 48 h, and expression persists at relatively high levels through at least 4 days (Fig. 7,B). Thus, negative signaling by the up-regulated CTLA-4 receptor could potentially provide the mechanism by which the cells are rendered anergic. This does not appear to be the case, however, since addition on day 0 of anti-CTLA-4 mAb to block binding to B7-1 did not prevent the decline in response to F23/B/I beads that occurs after day 3 (Fig. 8,A). Furthermore, addition of anti-CTLA-4 mAb could not reverse the nonresponsiveness of cells that had been activated 3 days previously with F23/B/I beads; no response to restimulation with F23/B/I beads occurred in the absence or presence of anti-CTLA-4 mAb, while, again, the cells responded well to addition of IL-2 (Fig. 8 B).
The results presented here demonstrate that CD8+ T cells can produce IL-2 and undergo autocrine IL-2-dependent proliferation and clonal expansion in response to signal 1 through the TCR and signal 2 through CD28 and/or LFA-1. The response is short-lived, however, with proliferation ceasing after day 3 or 4 as the cells lose the ability to respond to costimulation by producing IL-2. A similar nonresponsive state is induced within a few days of CD8+ T cells responding in vivo, even when the APC express costimulatory ligands, and despite the plethora of accessory cytokines likely to be produced during an allogeneic response. This nonresponsive state bears some similarity to the anergy that was originally described for CD4+ T cell clones (15, 16, 17). However, the CD8 nonresponsiveness described here is strikingly different from the “classical” anergy of CD4+ T cell clones in that it occurs following full stimulation with both signals 1 and 2, while CD4+ T cell anergy results when the cells receive signal 1 only. For this reason, we suggest that the nonresponsive state of CD8+ T cells described here is more appropriately termed “activation-induced nonresponsiveness” (AINR). A similar state of nonresponsiveness may occur for CD4+ T cells following an initial response to both signals 1 and 2 (D. Mueller, unpublished observations, and Refs. 36, 37).
Although activated CD8+ T cells become unresponsive to costimulation with respect to IL-2 production within 3–4 days, they remain responsive to stimulation through the TCR since they can lyse target cells (8) and secrete IFN-γ (Figs. 4 and 5) in a TCR-dependent manner during this time. A phenomenon similar to this has been observed in long-term cloned lines of murine CTL and was termed “split-anergy” (18). The role of costimulatory ligands was not directly assessed in that study, but it was similar to anergy induction in CD4+ T cell clone in that stimulation by fixed APC, but not irradiated APC, induced the hyporesponsiveness (17, 18). Induction of a nonresponsive state has also been described for human CD8+ CTL clones (38). Ag-specific stimulation with B7-expressing EBV-transformed B cells as APC resulted in hyporesponsiveness with respect to IL-2 production and proliferation, but not cytolytic activity. Thus, studies of long-term cloned CTL lines have demonstrated nonresponsive states similar to those shown here for normal resting CD8+ T cells; whether the nonresponsive clones reflect the same physiological state and mechanism(s) of nonresponsiveness as the normal cells is uncertain.
Although CD8+ T cells begin to undergo apoptotic cell death by days 3 and 4 after stimulation (Fig. 1), the inability to produce IL-2 in response to costimulation cannot be attributed to the fact that the cells are dying. If viability is maintained by the addition of IFN-α, IL-7 (Fig. 5), or human IL-2 (Fig. 4), the cells remain unable to produce murine IL-2. Thus, AINR appears to be a distinct phenomenon from that of activation-induced cell death (39, 40, 41), and this conclusion is further supported by the fact that nonresponsiveness is not prevented by addition of anti-TNF-β or anti-Fas ligand mAbs to either the primary or restimulation cultures (data not shown). AINR is also distinct from the effector T cell down-regulation observed by Liu et al. (42), since addition of blocking anti-IFN-γ mAb also failed to prevent its development (data not shown).
The mechanism(s) by which activated CD8+ T cells become nonresponsive does not appear to involve down-regulation of the receptors needed for delivering signals 1 or 2; TCR and CD8 levels are comparable on resting and activated cells, while CD28 and LFA-1 are expressed at higher levels on the activated cells (Fig. 6). CTLA-4 is a second receptor that binds B7-1 and B7-2 ligands, and evidence is accumulating that suggests that it can have a “negative” signaling role, at least on CD4+ T cells (32, 33, 34, 35). Resting T cells express little or no CTLA-4 on the surface, but its expression is up-regulated upon activation (43). This was confirmed to be the case for the CD8+ T cell responses studied here, with CTLA-4 being readily detected within 48 h and persisting through at least 4 days (Fig. 6). Although expressed, “negative” signaling via CTLA-4 did not account for the nonresponsiveness of activated CD8+ T cell since addition of a blocking anti-CTLA-4 mAb did not prevent or reverse the nonresponsiveness (Fig. 7). This conclusion is further supported by the observation that, while resting cells can respond to microspheres having just F23.1 mAb and ICAM-1 on the surface, activated cells cannot (data not shown). Thus, the cells remain nonresponsive even when no B7 ligand is present. Anergic CD4+ T cell clones have been shown to have a defect in the signaling pathways activated by the CD28 receptor (44, 45), and it appears likely that this may also be the case for activated CD8+ T cells. Costimulation through CD28 and LFA-1 results in activation of some common signaling events (29), and these are likely candidates for being affected by AINR.
AINR may play an important physiological role in converting the CD8+ CTL response from a helper-independent mode to a helper-dependent mode. Thus, a CTL response could be initiated via an autocrine IL-2 pathway, with the cells undergoing clonal expansion and developing effector function. As part of the developmental program, a critical signaling pathway(s) linking costimulatory receptors to IL-2 production may be “disconnected.” Thus, the CTL would be able to continue to carry out their Ag-specific, TCR-dependent effector functions, cytolysis, and cytokine production, but further expansion of the population would be prevented unless CD4+ T cell help became available. This could explain the importance of CD4+ T cell help for generating sustained CTL responses (1, 4, 5) and would be consistent with the observations that acute viral infections can be effectively controlled in the absence of CD4+ T cells (46, 47, 48), but chronic infections cannot (2, 3). AINR would insure that CD8+ cells do not continue to expand in an uncontrolled manner if the effector CTL generated initially are unable to rapidly eliminate Ag. In such cases, it would also lead to tolerance to the Ag if CD4+ T cell help is not available.
We thank Kathryn Eisenmann for excellent technical assistance and Dan Mueller for advice and critical reading of the manuscript.
This work was supported by National Institutes of Health Grants PO1 AI35296 and RO1 AI34824 (to M.F.M.). Support was also provided by U.S. Public Health Service Training Grants AI07421 (to M.J.D.) and AI07313 (to R.M.K.).
Abbreviations used in this paper: AINR, activation-induced nonresponsiveness; LN, lymph node; 7-AAD, 7-amino-actinomycin D.