Duration of TCR Stimulation Determines Costimulatory Requirement of T Cells

Duration of TCR Stimulation Determines Costimulatory Requirement of T Cells

Immunity, Vol. 5, 41–52, July, 1996, Copyright 1996 by Cell Press Duration of TCR Stimulation Determines Costimulatory Requirement of T Cells Thomas...

470KB Sizes 3 Downloads 26 Views

Immunity, Vol. 5, 41–52, July, 1996, Copyright 1996 by Cell Press

Duration of TCR Stimulation Determines Costimulatory Requirement of T Cells Thomas M. Ku¨ndig,* Arda Shahinian,† Kazuhiro Kawai,* Hans-Willi Mittru¨ cker,* Eric Sebzda,* Martin F. Bachmann,* Tak W. Mak,*† and Pamela S. Ohashi* *Ontario Cancer Institute Departments of Medical Biophysics and Immunology 610 University Avenue Toronto, Ontario Canada, M5G 2M9 † Amgen Institute 620 University Avenue Toronto, Ontario Canada, M5G 2C1

CD41 T cells (Shahinian et al., 1993; Ku¨ ndig et al., 1993c; Harding and Allison, 1993). We examined the costimulatory requirements of CD81 T cells in vivo by using CD28 2/2 mice, independent of CD41 T helper function, either by in vivo depletion of CD41 T cells (Cobbold et al., 1984) or by using class I major histocompatibility complex (MHC) binding peptide antigens. Comparison and analysis of CTL responses to viral antigens revealed that prolonged signal-1 can overcome the requirement for CD28-mediated costimulation and rescue T cells from the induction of anergy in vivo. Results

Summary Current models suggest that T cells that receive only signal-1 through antigenic stimulation of the T cell receptor (TCR) become anergic, but will mount an immune response when a costimulatory signal-2 is provided. Using mice deficient for an important costimulatory molecule, CD28, we show that a transient signal-1 alone, either through infection with an abortively replicating virus, or through injection of viral peptide, anergizes CD81 T cells, demonstrating the biological relevance of T cell anergy in vivo. However, in the absence of CD28, continued presence of signal-1 alone, either through prolonged viral replication or repeated injection of peptide, prevents the induction of anergy and generates a functional T cell response in vivo. Introduction In vitro, several models demonstrate that the generation of maximal T cell responses requires not only signal-1 through the T cell receptor (TCR), but also a costimulatory signal-2 (Schwartz, 1990). The interaction of CD28 on T cells with B7 family molecules on antigen-presenting cells is important for providing signal-2, since it up-regulates interleukin-2 (IL-2) production and T cell proliferation (June et al., 1994; Allison, 1994; Linsley and Ledbetter, 1993; Schwartz, 1992). Stimulation of the TCR in the absence of signal-2 has been shown to induce T cell anergy, a state characterized by unresponsiveness to further antigenic stimulation, which may be reversed by exogenous IL-2 (Schwartz, 1990; Bluestone, 1995). In vivo, this two signal model of lymphocyte activation is controversial. Although aberrant expression of costimulatory molecules or cytokines can enhance T cell activation (Guerder et al., 1994; Harlan et al., 1994; Heath et al., 1992; Chen et al., 1992; Townsend and Allison, 1993; Baskar et al., 1994; Fearon et al., 1990; Gansbacher et al., 1990), the absence of CD28, B7-1, or IL-2 does not appear to impair the immune response to certain antigens in vivo (Shahinian et al., 1993; Ku¨ ndig et al., 1993c; Bluestone, 1995; Freeman et al., 1993). Notably, CD81 T cells seem less dependent on costimulation than

Costimulatory Requirement of CTL Responses Depends on Properties of the Virus As reported earlier, lymphocytic choriomeningitis virus (LCMV) infection of CD28-deficient (CD282/2) and CD281/1 mice generates comparable cytotoxic T lymphocyte (CTL) activity against LCMV–glycoprotein (LCMV–gp) (Figure 1A) (Shahinian et al., 1993), which contains the dominant class I MHC-presented epitope in H-2b mice (Pircher et al., 1990). This CTL response was fully efficient, since it eliminated LCMV with normal kinetics (data not shown) and conferred immunological protection by rendering mice resistant against subsequent challenge infection with vaccinia–LCMV–gp recombinant virus (vv–LCMV–gp) (Figure 1B); such resistance is mediated exclusively by CD81 CTL (Binder and Ku¨ndig, 1991). Antigenic restimulation of spleen cells from LCMV-infected CD282/2 mice readily generated cytotoxicity, without adding exogenous cytokines (Figures 1C and 1D). Limiting dilution analysis confirmed that massive in vivo proliferation of LCMV–gp-specific CD81 T cells occurred in CD28 2/2 mice, since frequencies rose from <1026 to 1/50 (Figure 1E). Taken together, in CD282/2 mice LCMV infection generated cytotoxic T cell activity and the in vivo proliferation of specific CD81 T cells was not impaired. This CD28 independence did not correlate with a particular LCMV epitope, since LCMV infection of CD282/2 mice induced normal CTL responses not only against the immunodominant glycoprotein–epitope, but also against minor epitopes such as the nucleoprotein of LCMV in H-2b mice (Hany et al., 1989). This is demonstrated by the finding that CD282/2 mice immunized with LCMV are resistant not only to subsequent challenge infection with vv–LCMV–gp, but also to vaccinia–LCMV– nucleoprotein recombinant virus (vv–LCMV–np) (Figures 2A and 2B). Furthermore, in CD282/2 H-2k mice, the immunopathological footpad swelling reaction after local infection with LCMV was unaltered (Figure 2C), demonstrating that CD81 CTL responses against LCMV in H-2k mice were fully efficient, despite the fact that H-2 k mice are CTL low responders against LCMV (Hany et al., 1989). Thus, the CD28 independence of the CTL response induced by LCMV infection does not correlate with any particular viral epitope, but is a property inherent to the virus itself.

Immunity 42

Figure 1. Costimulatory Requirement of CTL Responses Depends on the Virus (A–E) LCMV infection: Mice were infected intravenously with LCMV (2,000 pfu) and the LCMV–gp-specific immune response was assessed after 8 days. (A) Primary ex vivo CTL activity measured by 51Cr release assays. (B) Antiviral protection assessed by challenge infection with a vaccinia–LCMV–gp recombinant virus (vv–LCMV–gp): LCMV-immunized mice were challenged intracerebrally with vv–LCMV–gp and titers of this virus were measured in brains 5 days after the challenge infection. (C) Spleen cells were stimulated in vitro with LCMV-infected peritoneal macrophages, without adding ConA supernatant (CAS), or in the presence of CAS (D). The indicated dilutions of responder cultures were tested for specific cytotoxicity on LCMV-infected target cells. (E) Limiting dilution analysis for LCMV–gp-specific CTL precursor cells in the presence of CAS. (F–J) VSV infection: Mice were infected intravenously with VSV and VSV–np-specific CD81 T cell responses were assessed after 6 days. (F) Primary ex vivo CTL activity. (G) Antiviral protection assay using a vaccinia-VSV–np recombinant virus (vv–VSV–np) for challenge infection. (H) Spleen cells were stimulated in vitro with VSV-infected splenic macrophages (Ku¨ ndig et al., 1993a), without adding CAS, or in the presence of CAS (I). The indicated dilutions of responder cultures were tested for specific cytotoxicity on VSV-infected target cells. (J) Limiting dilution analysis for VSV–np-specific CTL precursor cells, in the presence of CAS. Closed triangles, CD282/2; open triangles, CD281/1; open circles, CD281/1 that had been depleted in vivo of CD41 T cells before the priming infection (Cobbold et al., 1984); open diamonds, noninfected CD281/1 control mice. All mice are H-2b/b with more than 6 backcrossings into C57BL/6.

In contrast, infection of CD28 2/2 mice with vesicular stomatitis virus (VSV) did not induce any directly measurable CTL activity in vitro (see Figure 1F). Induction of strong CTL activity after VSV infection in CD4-depleted mice indicated that the lack of CTL response in the absence of CD28 was not due to impaired T help (see Figure 1F). In vivo, VSV-immunized CD28 2/2 mice were not resistant against challenge infection with vaccinia– VSV recombinant virus expressing VSV–nucleoprotein (np), containing the dominant CTL epitope against VSV in H-2b mice (Puddington et al., 1986; Ku¨ndig et al., 1993b) (see Figure 1G). However, using spleen cells from VSV-infected mice, in vitro restimulation with antigen could generate VSV–np-specific CTL activity, but only if exogenous cytokines were added (see Figures 1H and 1I). Such in vitro stimulation does not generate cytotoxicity in unprimed mice, thus indicating that VSV infection of CD282/2 mice was not simply a null event for specific CD81 T cells, but that these T cells had been activated. In fact, limiting dilution analysis for VSV–np-specific CD81 T cells, which is performed in the presence of exogenous cytokines, revealed that proliferation had taken place in vivo: frequencies rose from <1/106 in uninfected mice up to 1/50,000 (see Figure 1J). This proliferation was, however, 50–100 times less than in CD28 1/1 mice. These data demonstrate that VSV infection primed

specific CD81 T cells and these CD81 T cells proliferated minimally, followed by unresponsiveness to further antigenic stimulation in vivo and in vitro, unless exogenous cytokines were added. Requirement for CD28 Costimulation Inversely Correlates with Viral Replication How can this discrepancy between LCMV and VSV be explained? There are several differences between LCMV and VSV. A very striking difference is the extent and duration of viral replication in the host. LCMV is a natural mouse pathogen and after intravenous infection replicates widely and extensively in several mouse organs (Lehmann-Grube, 1971; Buchmeier et al., 1980). In the spleen, infectious virus can be recovered at any timepoint from day 1 through days 8–10 after infection. On day 4–5, high maximal titers around 106 pfu per gram spleen are reached. In contrast, VSV is not a natural mouse pathogen, and after intravenous infection replicates poorly in the spleen, where only low titers of live virus (10–100 pfu per gram spleen) can be recovered after 1 day, but not 2 days after infection (Wagner, 1987). Thus, compared with VSV, live LCMV titers are around 104–105 times higher and live LCMV can be recovered roughly 10 times longer. Thus, LCMV peptide will be presented in higher doses and for a longer duration.

Extended Signal-1 Mimics Costimulation 43

Figure 2. The Requirement for CD28 Depends Upon the Virulence of the Pathogen and Not Properties Associated with the Viral Epitope or Haplotype (A–C) CTL responses against several LCMV epitopes are CD28 independent. (A and B) H-2b/b mice depleted of CD41 T cells were immunized intravenously with LCMV and, after 12 days, challenged intracerebrally with (A) vv–LCMV–gp or (B) vv–LCMV–np. Vaccinia virus titers in brains were determined 5 days after challenge infection. Closed triangles, CD282/2 mice; open circles, CD281/1 mice; open diamonds, unprimed CD281/1 mice. (C) H-2k/k mice depleted of CD41 T cells were infected with LCMV infection into their hind footpad. Closed triangles, CD282/2 mice; open circles, CD281/1 mice. (D–E) Vaccinia virus virulence determines CD28 dependence. Mice were infected with vaccinia virus and primary ex vivo CTL activity was tested on vaccinia virus infected target cells. (D) infection with highly virulent vaccinia virus strain (WR); (E) infection with low virulence vaccinia virus strain (thymidine kinase negative). Spontaneous 51 Cr-release was <20% and nonspecific lysis of uninfected control target cells was <18% for all effectors. Closed triangles, CD282/2; open triangles, CD281/1; open circles, CD281/1 that had been depleted in vivo of CD41 T cells before the priming infection (Cobbold et al., 1984); open diamonds, noninfected CD281/1 control mice. (F–G) Infection of CD282/2 mice with vv–VSV–np partially anergizes CD81 T cells. CD282/2 (closed triangles) and CD281/1 mice (open triangles) were infected with the low virulence vaccinia virus strain vv–VSV–np. After 6 days, spleen cells were restimulated in vitro using VSV-infected splenic macrophages (Ku¨ ndig et al., 1993a), either without (F) or with (G) ConA supernatant. The indicated dilution of responder cultures were tested for specific cytotoxicity on VSV–np-transfected syngeneic target cells (Puddington et al., 1986). Open diamonds represent unprimed CD281/1 control mice restimulated with the same protocol.

The following experiment was designed to confirm that viral replication per se, and not other differences between viruses, can explain the observed difference in CD28 dependence. Two vaccinia virus strains were used to infect CD28 1/1 and CD282/2 mice. These vaccinia virus strains are identical except for one gene coding for the virulence factor thymidine kinase (Buller et al., 1985): CTL responses against the highly virulent strain, which replicates in lymphatic organs of the mouse for several days, were CD28 independent (Figure 2D), whereas CTL responses against the low virulence stain,

which replicates in lymphatic organs of the mouse only abortively, were CD28 dependent (Figure 2E). In CD282/2 mice infected with low virulence vaccinia virus expressing recombinant VSV–np, in vitro restimulation assays revealed that CD81 T cells specific for this defined antigen mounted a weak but significant response (Figure 2F), also in absence of exogenous cytokines. Addition of exogenous IL-2, however, was necessary to increase in vitro responsiveness (Figure 2G). Thus, presentation of VSV–np to the immune system in the form of the vaccinia recombinant virus, instead of VSV wild type, could induce a weak CD81 T cell response, which became detectable only after in vitro restimulation, but not by direct ex vivo measurement of CTL activity. This in vitro response was significantly, but not absolutely, dependent on exogenous cytokines. The low virulence vaccinia virus strain used here may measurably replicate in certain mouse organs (Binder and Ku¨ndig, 1991; Karupiah et al., 1990) and may therefore be considered more virulent than VSV wild type. These data confirm that virulence appears to be the parameter determining T cell responsiveness in absence of CD28. Using these viruses, a correlation between viral replication and CD28 dependence could be made and further experiments were done to examine whether the higher antigen dose or the continued presence of antigen could render the T cell response independent of CD28. We therefore immunized CD282/2 TCR-transgenic mice (TCR1) specific for LCMV–gp (Pircher et al., 1989) with the dominant LCMV–gp peptide (amino acids 33–41), and varied either the peptide dose or the duration of antigen presentation by continued peptide injections. In all the following experiments, peptide was injected in saline to limit the duration of peptide presentation in vivo (Widmann et al., 1991; Ku¨ ndig et al., 1992). CD28 Deficiency Does Not Alter Antigen Dose Required for T Cell Activation The importance of the dose was evaluated by injecting TCR1 mice with different amounts of LCMV–gp peptide. Blast formation of CD81 T cells, which increases the forward scatter in FCM analysis, was taken as a marker of transgenic T cell activation. Single parameter FCM analyses are shown, since roughly 80%–95% of all CD81 T cells express the transgenic receptor, and this percentage even increases after specific immunization. In both TCR1 CD28 2/2 and TCR1 CD281/1 mice, the dose required for CD81 T cell blast formation, occurring within 24 hr after peptide injection, was 0.01–0.1 mg (Figures 3A and 3B). Peptide Induces Unresponsiveness of CD282/2 T Cells In both TCR1 CD282/2 and TCR1 CD281/1 mice, 3 days after peptide injection, CD81 T cell blasts were no longer detectable by FCM analysis, regardless of the peptide dose injected (data not shown). The CD81 T cell population had been expanded comparably in both groups (data not shown; similar to Figures 4A–4F below). However, peptide-primed TCR1 CD282/2 CD81 T cells no longer responded to antigenic stimulation in vitro, unless exogenous cytokines were added (Figures 3C and 3D),

Immunity 44

Figure 3. Induction of T Cell Unresponsiveness by Injection with LCMV–Peptide (A–B) CD28 deficiency does not alter the antigen dose required for T cell activation. (A) TCR1 CD281/1 and (B) TCR1 CD282/2 mice were intravenously injected with the indicated doses of LCMV–gp peptide (amino acids 33–42). After 24 hr, the forward scatter, correlating to cell size, of CD81 T cells was analyzed by FCM. (C–D) After peptide injection, TCR1 CD282/2 CD81 T cells no longer generate cytotoxicity in vitro. (C) TCR1 CD281/1 and (D) TCR1 CD282/2 mice were intravenously injected with the indicated doses of LCMV–gp peptide. After 3 days, mice were sacrificed and spleen cells in vitro stimulated with the same peptide. The in vitro generation of LCMV–gp-specific cytotoxicity was tested on LCMV–gp peptide-labeled target cells, at the indicated dilution of the responder cultures. Open triangles, restimulation in absence of CAS; closed triangles, restimulation in presence of CAS. Spontaneous release was <24% and nonspecific lysis of control target cells (not labeled with viral peptide) was <20% for all effectors. (E–F) After peptide injection, TCR1 CD282/2 CD81 T cells no longer proliferate in vitro. (E) TCR1 CD281/1 and (F) TCR1 CD282/2 mice were intravenously injected with LCMV–gp peptide. After 3 days, mice were sacrificed and spleen cells in vitro stimulated with the same peptide (left two columns) or media without peptide (right two columns). T cell proliferation was assessed by determination of [3 H]thymidine incorporation. Open bars, [3H]thymidine uptake without CAS in cultures. Closed bars, [ 3H]thymidine uptake with CAS.

whereas such antigenic stimulation of unprimed CD81 T cells in TCR 1 CD28 2/2 mice readily generated cytotoxicity without exogenous cytokines. Corresponding to the lack of cytotoxicity, peptide-primed TCR1 CD28 2/2

CD81 T cells also did not proliferate in vitro, unless exogenous cytokines were added (Figures 3E and 3F), whereas naive TCR1 CD282/2 T cells proliferated normally, at least during the first 2 days. (At later timepoints,

Extended Signal-1 Mimics Costimulation 45

Figure 4. Analysis of Activation, Peripheral Expansion, and Deletion in the Absence of CD28 (A–F) Adult thymectomized TCR1 CD281/1 (A) and TCR1 CD282/2 (B) mice were intravenously injected with the indicated LCMV–gp peptide doses. Percentage of CD81 T cells was measured among peripheral blood leukocytes (6SD, n 5 5 per group). (G–L) Phenotypic analysis of CD81 T cell blasts. FCM analysis for surface expression of CD69 (G, J), or IL-2Ra (H, K) 1 day after peptide injection and for surface expression of the transgenic TCR (Vb8) (I, L) 3 days after peptide injection are shown. Profiles of peptide-induced CD81 T cell blasts are marked “pep” and compared with profiles of CD81 T cells of unprimed mice (unmarked).

similar to an earlier report on allo-specific proliferation [Green et al., 1994, Kawai et al., 1996], we found reduced proliferation also in unprimed CD282/2 mice.) Thus, these data confirm the inverse correlation of pathogen replication with the CD28 requirement within the LCMV system. Immunization of CD282/2 mice with a nonreplicating LCMV antigen, such as LCMV peptide, induced T cell unresponsiveness, whereas immunization with LCMV virus induced normal CTL responses. The unresponsiveness induced by LCMV peptide was reversible by the presence of exogenous cytokines, similar to the unresponsiveness induced after VSV infection. It is important to note here that this unresponsiveness was not overcome even when mice were injected with high antigen doses, such as 100 mg of peptide. T Cell Unresponsiveness Does Not Correlate with T Cell Deletion The use of the transgenic mouse system allowed us to analyze the fate of T cells rendered unresponsive by antigen contact. Adult thymectomized TCR1 CD28 2/2 and TCR1 CD281/1 mice were injected with different peptide doses. In both groups, the CD81 T cell population first expanded comparably for all three peptide doses used (10 mg, 1 mg, 0.1 mg) (Figure 4). The CD81 T cell population reached a maximum on day 3, followed by a rapid decline on day 4. In TCR1 CD282/2 mice, injection with 10 mg was followed by deletion of roughly half the

CD81 T cells, whereas injection of 1 mg and 0.1 mg, which readily induced T cell unresponsiveness (see Figure 3), did not lead to deletion. In TCR1 CD281/1 mice, injection with 10 mg similarly deleted roughly half of the CD81 T cells, but evidence of deletion was also seen in TCR1 CD281/1 mice injected with only 1 mg and 0.1 mg. Taken together, the observed T cell unresponsiveness did not correlate with T cell deletion. In contrast, CD28 deficiency rather protected T cells from deletion. Interestingly, more than 3 weeks after peptide injection, in thymectomized mice, T cells could again be restimulated in vitro, even in the absence of exogenous cytokines, indicating T cell unresponsiveness in TCR1 CD28 2/2 mice was transient (data not shown). T Cell Unresponsiveness Is Biologically Significant During Viral Challenge In Vivo We next examined, whether the peptide-induced T cell unresponsiveness observed in vitro in TCR1 CD282/2 mice, i.e., the need for exogenous IL-2, was significant during viral challenge infection in vivo. TCR 1 CD281/1 and TCR1 CD282/2 mice were injected with peptide and after 3 days challenged with a vv–LCMV–gp recombinant virus (Bachmann and Ku¨ ndig, 1994). While the CD81 T cells from CD28 1/1 mice could control viral replication, CD282/2 mice could not (Figure 5), and visibly succumbed to disease. Together with the results from VSVinfected CD282/2 mice, which also were not protected

Immunity 46

activation markers, such as CD69 and IL-2Ra (see Figures 4G, 4H, 4J, and 4K), as well as CD44 and the transferrin receptor (data not shown), were expressed at markedly lower levels in TCR1 CD28 2/2 mice. However, the T cell unresponsiveness observed in CD282/2 mice 3 days after peptide injection did not correlate with downregulation of the TCR (see Figures 4I and 4L), since this event was less pronounced in TCR1 CD28 2/2 mice, than in TCR 1 CD281/1 mice.

Figure 5. T Cell Unresponsive to Viral Challenge In Vivo. Antiviral protection was assessed 3 days after injection with 10 mg of LCMV–gp peptide by challenge infection with vv–LCMV–gp either intracerebrally or intraperitoneally. Titers of vv–LCMV–gp were measured in brains or ovaries 5 days after the challenge infection. Closed triangles, CD282/2; open triangles, CD281/1 mice.

against subsequent challenge infection with a vv–VSV– np recombinant virus in vivo, these studies demonstrate the biological relevance of the CD81 T cell unresponsiveness observed in vitro. T Cell Unresponsiveness Does Not Correlate with Down-Regulation of the TCR Phenotypic comparison of transgenic CD81 T cell blasts 24 hr after peptide injection revealed that early T cell

Short-Lived CTL Response in CD282/2 Mice We next assessed whether the CD81 T cell blasts in CD282/2 mice observed 24 hr after peptide injection, although expressing reduced levels of activation markers, possessed cytotoxic function. Interestingly, they exhibited normal CTL activity when tested directly ex vivo (Figure 6A). When such peptide-primed TCR1 CD282/2 mice were immediately challenged with vv– LCMV–gp recombinant virus, they remained healthy and no viral replication could be measured, whereas unprimed control mice succumed to vv–LCMV–gp-induced disease (data not shown). This was expected, since such in vivo assessment of CTL activity is more sensitive than the direct ex vivo measurement of CTL activity using 51 Cr release assays (Bachmann and Ku¨ndig, 1994). Furthermore, at this early timepoint, in vitro antigenic stimulation generated specific cytotoxicity (Figures 6B and 6C) and proliferation (data not shown) of CD81 T cells even in the absence of exogenous cytokines. In contrast, 3 days after peptide injection, when CD81 T cells were unresponsive in vitro (similar to Figure 3; data not shown), CTL activity was completely absent in TCR1 CD282/2 mice (Figure 6D). At this timepoint, TCR1 CD281/1 mice still exhibited ex vivo CTL activity and normal in vitro reactivity (data not shown; similar to Figures 3C and 3E). Taken together, very early after antigen

Figure 6. Peptide Injection Induces Short-Lived CTL Response in TCR1 CD282/2 Mice (A–C) CD81 T cell response 1 day after peptide injection. (A) Ex vivo CTL activity on targets labeled with LCMV–gp peptide. (B) spleen cells were in vitro stimulated with LCMV–gp peptide and the generation of cytotoxicity was assessed on LCMV–gp peptide-labeled target cells, at the indicated dilutions of responder cultures. No CAS was added to cultures. (C) CAS was present in cultures. (D) CD81 T cell response 3 days after peptide injection: ex vivo CTL activity on H-2b targets labeled with LCMV–gp peptide. In all 51Cr release assays shown spontaneous release was <26% and nonspecific lysis of uninfected control target cells was <21% for all effectors. All animals used were TCR1 mice. Closed triangles, peptide-primed CD282/2 mice; open triangles, peptide-primed CD281/1 mice; closed diamonds, unprimed CD282/2 mice; open diamonds, unprimed CD281/1 mice. Circles, unprimed CD282/2 (open circles) or unprimed CD281/1 (closed circles) mice tested on control targets without peptide.

Extended Signal-1 Mimics Costimulation 47

Figure 7. Prolonged Signal-1 Makes CTL Response CD28 Independent (A–B) Mice were intravenously injected with 1 mg of peptide every 12 hr (7 injections) or left unprimed. LCMV–gp-specific CTL activity was measured 12 hr after the last peptide injection in spleens by a primary ex vivo 51Cr release assay (A). Forward scatter of these CD81 T cells, peptide-primed as well as unprimed TCR 1 CD282/2 and TCR1 CD281/1, are virtually identical in size so that curves overlap (B). (C–D) Mice were intravenously injected with 1 mg of peptide every 12 hr (7 injections) or left unprimed. These mice were tested in the same assay, but the time interval after the last peptide injection was 24 hr. (C) Forward scatter of CD81 T cells is shown (D) as in (B). In all 51Cr release assays shown, target cells have been prepulsed with the LCMV–gp peptide. Spontaneous release was <17% and nonspecific lysis of uninfected control target cells was <22% for all effectors. All animals used were TCR1 mice. Closed triangles, peptide-primed CD282/2 mice; open triangles, peptide-primed CD281/1 mice; closed diamonds, unprimed CD282/2 mice; open diamonds, unprimed CD281/1 mice.

contact CD81 T cells in TCR1 CD282/2 mice were functional, exhibiting efficient cytotoxicity, and these T cells could be normally restimulated in vitro. However, this response was not sustained and followed by T cell unresponsiveness. Prolonged Signal-1 Makes T Cell Responses CD28 Independent Since high antigen doses could not overcome the CD28 requirement of the CTL response (see Figure 3), we evaluated whether the duration of signal-1 determined CTL responsiveness in CD28 2/2 mice. Note that in all above experiments transgenic mice received only a single injection of viral peptide without adjuvant, so that the peptide has a very short half-life in vivo (Widmann et al., 1991; Ku¨ndig et al., 1992). As shown in Figure 3, injection with 1 mg of viral peptide activates all transgenic CD81 T cells in the spleen. In TCR1 CD28 2/2 mice, CD81 T cells then become unresponsive within 3 days. During these 3 days, if TCR 1 CD282/2 mice were continuously injected with additional peptide (1 mg every 12 hr), high levels of CTL activity, which could be measured directly ex vivo, were maintained (Figure 7A). In a similar series of experiments, we confirmed that mice injected with peptide repeatedly were resistant to challenge infection with vv–LCMV–gp (data not shown). This sustained CTL activity measured on day 3 after repeated peptide injections was due to continued restimulation of the same CD81 T cell population and not due to a small fraction of CD81 T cells,

which might not have been primed and therefore remained naive, nor was it due to naive new thymic emigrants, which would then be freshly activated with every additional peptide injection. First, CTL activity was maintained only if peptide was injected every 12 hr, but not by injection every 24 hr (data not shown). If the response was due to naive cells, then prolongation of the time interval between injections would allow for more naive cells to accumulate and should therefore increase the response. Second, 3 days after a single peptide injection, secondary in vitro restimulation with antigen in bulk cultures did not generate cytotoxicity (see Figure 3). Such bulk culture restimulation is highly sensitive and detects specific CD81 T cells at frequencies down to 1026 (Bachmann and Ku¨ndig, 1994). Thus, naive cells must be at frequencies lower than that, which is far below the detection level of the primary ex vivo cytotoxicity assays performed in this set of experiments. To detect primary ex vivo cytotoxicity, specific CD81 T cells have to be at frequencies of >1023. Third, after 3 days of repeated peptide injections, FCM analysis showed a markedly larger CD81 T cell population in the spleen than after a single peptide injection (data not shown), again indicating that the same population of CD81 T cells is being further restimulated with additional peptide injections. However, already 24 hr after withdrawal of this continued signal-1, CTL activity completely collapsed in TCR1 CD282/2 mice, while remaining readily detectable in TCR1 CD28 1/1 mice (Figure 7C), although in both TCR1

Immunity 48

CD282/2 and TCR1 CD28 1/1 mice, CD81 T cells were still blasts (Figure 7D). Thus, in contrast with a single high dose injection of viral peptide, repeated injections with low doses could mimic extensive viral replication and overcome the CD28 requirement. However, T cell responses were maintained only as long as signal-1 was provided. Discussion Our data demonstrate that infection of CD282/2 mice with pathogens that do not replicate extensively, such as VSV (Wagner, 1987) or low virulence vaccinia virus (Buller et al., 1985), or a single injection of TCR 1 CD282/2 mice with short-lived viral peptides (Ku¨ndig et al., 1992; Widmann et al., 1991), initially induced T cells. VSVspecific CD81 T cells were primed and minimal proliferation was observed in vivo. Similarly, peptide injection in vivo induced CD81 T cell blasts exhibiting cytotoxic effector function and expanded the CD81 transgenic T cell population. However, this initial T cell response was transient and followed by T cell unresponsiveness in vitro and in vivo, which could be reversed by exogenous cytokines. Antigen-induced unresponsiveness of mature peripheral T cells may be due to several mechanisms: clonal deletion or exhaustion, down-regulation of the TCR, or anergy (Zinkernagel et al., 1993; Kisielow et al., 1991; Ha¨mmerling et al., 1993; MacDonald et al., 1993; Marrack et al., 1993; Sprent et al., 1993; Rammensee et al., 1989; Ramsdell and Fowlkes, 1990). Given high doses of viral peptide, TCR1 CD281/1 and TCR1 CD28 2/2 mice showed clonal expansion followed by down-regulation of the TCR T cell deletion, possibly due to exhaustion. However, both the first and second mechanisms could not explain the unresponsiveness observed in CD28deficient mice, since both events were even more pronounced in CD281/1 mice, which still retained antiviral effector function. The unresponsiveness observed here was induced only in absence of CD28-mediated costimulation and could be reversed by exogenous IL-2. It thus fits the in vitro characteristics of the third mechanism, anergy (Schwartz, 1990). Interestingly, after injection with high doses of peptide several of the above tolerance mechanisms were observed in CD282/2 mice, but after injection with lower doses anergy could occur by itself: e.g., injection with 0.1 mg of peptide induced blast formation of all specific T cells followed by anergy, but only minimal clonal expansion without deletion or down-regulation of the TCR (data not shown). Also, the observed T cell anergy was of only transient nature, as reported for other situations where the antigen does not persist (Rocha et al., 1993; Migita and Ochi, 1993; Ramsdell and Fowlkes, 1992). This and other studies (reviewed by Sprent and Webb, 1995) examining peripheral tolerance suggest that the degree of T cell activation influences the outcome of tolerance induction. A strong activation signal will lead to significant expansion, detectable immune function, followed by deletion and anergy (Kyburz et al., 1993; Kearney et al., 1994; Aichele et al., 1995). An intermediate activation signal will lead to moderate expansion,

detectable immune function, followed by minimal deletion and anergy. A weak activation may lead to detectable phenotypic activation, a smaller “window” of immune function, undetectable expansion and deletion, but once functional anergy has been established, it will persist in a manner that reflects persistence of antigen. In this report, we show that the phenotype of T cell activation is influenced by CD28-mediated costimulation in vivo. In absence of CD28, T cell activation markers are reduced, CD81 T cell blasts exhibit measurable cytotoxic activity only for a short period of time, T cell deletion is less marked, and T cell anergy is observed at an early timepoint. Thus, in the absence of CD28, the induction of peripheral tolerance may be biased toward anergy, since normal T cell activation signal is altered in these animals. These studies analyzed the CD28 requirement for the in vivo generation of CD81 T cell–mediated cytotoxicity against a variety of antigens that differ in their kinetics and extent of antigen presentation in the host. After intravenous infection of mice with VSV or with the low virulence vaccinia virus strain, low titers of live virus can be recovered from spleen for only 1 or 2 days (Wagner, 1987; Buller et al., 1985). In contrast with high virulence vaccinia virus (Buller et al., 1985) and LCMV (LehmannGrube, 1971; Buchmeier et al., 1980), higher virus titers can be recovered for a prolonged period of time, e.g., very high LCMV titers are recovered around day 4, and LCMV can be detected in the spleen for 8–10 days. The CD28 dependence of the CD81 T cell response inversely correlated with the virulence of the above viruses; in the absence of CD28, infection with VSV and the low virulence vaccinia virus did not generate cytotoxicity, whereas infection with the high virulence vaccinia virus or with LCMV generated normal cytotoxicity. We further dissected whether the extent of viral replication, i.e., the antigen dose, or duration of viral replication, i.e., the duration of antigen presentation, rendered the CD81 T cell response CD28 independent. Analysis using TCR1 CD281/1 and TCR1 CD28 2/2 mice showed that a 104-fold increase of the specific virus peptide dose could not mimic viral replication (Figure 3). However, repeated injections of low peptide doses could sustain the CTL response for the duration that peptide was given (Figure 7). The use of viral peptides demonstrated that sustained antigen presentation, rather than the high dose of antigen, rendered the T cell response independent of CD28. However, while differences in the duration of viral replication may sufficiently explain the differential CD28 dependence observed, this may not be the only explanation for the CD28 independence of the CD81 T cell response against LCMV. It is quite possible that additional factors, such as the induction of costimulatory pathways other than via CD28, enhance T cell responsiveness against LCMV. Our demonstration that the duration of signal-1 determines the costimulatory requirement for T cells may reconcile the controversy surrounding the two-signal model of lymphocyte activation in vivo (Bluestone, 1995). The discrepancy between in vitro and in vivo studies, i.e., that in vitro T cell responses largely depend on costimulatory molecules or cytokines (Schwartz,

Extended Signal-1 Mimics Costimulation 49

1992), but CD28-, B7-1, and IL-2-deficient mice can generate immune responses (Shahinian et al., 1993; Ku¨ndig et al., 1993c; Freeman et al., 1993), may readily be explained by differences in the duration of antigen presentation. In vivo, CTL responses have been demonstrated in IL-2- and CD28-deficient mice using LCMV (Ku¨ndig et al., 1993c; Shahinian et al., 1993) or allo-grafts (Kawai et al., 1996) as antigens, which both provide signal-1 for an extended period of time. In contrast, in vitro assays generally use inactivated antigen-presenting cells, by irradiation or chemically, which are present for only 1 or 2 days. This may be the reason why in vitro allospecific or ovalbumin-specific CD282/2 T cells initially proliferate, but cannot sustain this response (Green et al., 1994; Lucas et al., 1995, Kawai et al. 1996). In vitro results demonstrated that high antigen density or high concentrations of anti-CD3 antibody and therefore supraoptimal TCR signaling can replace signal-2 (Harding et al., 1992; Green et al., 1994). Although we observed no such dose effect with viral peptides in vivo (Figure 3), this does not rule out that the dose of other antigens may influence in vivo costimulatory requirements. Several antigens might provide a stronger signal-1 than that achieved here by exogenously loading MHC molecules. Also, since viral peptides are unstable in serum, not retained by kidney membranes and therefore immediately excreted (Widmann et al., 1991; Ku¨ndig et al., 1992), injection of higher doses cannot significantly prolong the duration of antigen presentation, whereas increasing doses of other antigens may crucially prolong antigen presentation. Our results do not completely follow the basic twosignal model of peripheral self-tolerance (Bretscher and Cohn, 1970; Lafferty and Woolnough, 1977; Janeway, 1989), since self-antigens would always be present to deliver signal-1 and thus may induce an immune response. However, T cells that may have the potential to react with a self-antigen do not generally extravasate and interact with self-ligands in a particular organ. In addition, MHC expression on most tissues is relatively low, therefore limiting self-antigen presentation and avoiding efficient interaction with autoreactive T cells (Boehme et al., 1989; Ohashi et al., 1991, 1993; von Herrath et al., 1994; Ando et al., 1994). Under conditions where inflammation is induced or aberrant expression of self-ligand presentation exists, sustained signal-1 through the TCR may provide an explanation for the direct induction of T cell function by nonprofessional APCs. In conclusion, these studies using CD28-deficient mice demonstrate that in absence of the CD28-costimulatory signal-2, anergy results from transmission of signal-1 alone, either through low virulence virus infection or injection of viral peptide. This is in accordance with the traditional two-signal model of lymphocyte activation. In addition, the biological relevance of CD81 T cell anergy was confirmed in vivo, by a lack of protection to subsequent viral infection. Our studies, however, modify the two-signal model by demonstrating that the continued presence of signal-1, either through high virulence infection, or repeated infusion of peptide, generates a functional T cell response.

Experimental Procedures Mice The generation of CD28-deficient mice by targeted gene disruption has been described in detail previously (Shahinian et al., 1993). Unless indicated otherwise, all mice used were H-2b/b, backcrossed into C57BL/6 for more than 6 generations. C57BL/6 mice (Jackson Laboratories, Bar Harbor, Maine) were used as CD281/1 controls in the experiments depicted. All experiments have also been done, at least once, in CD281/2 littermate controls and comparable results were obtained. The CD282/2 H-2k/k mice were obtained through breeding into CBA/J mice. TCR1 CD282/2 mice were obtained through breeding of TCR1 mice (Pircher et al., 1990) in the C57BL/ 6 background with CD282/2 mice in the C57BL/6 background. All experimental mice were between 8–12 weeks of age. Viruses VSV serotype Indiana was obtained from Dr. L. Prevec (McMaster University, Hamilton, Ontario, Canada). Seeds were grown on BHK cells and plaqued on Vero cells following standard protocols (Wagner, 1987). LCMV (Armstrong isolate) was originally obtained from Dr. M. B. A. Oldstone (Scripps Clinics and Research Foundation, LaJolla, San Diego, California) (Buchmeier et al., 1980). Seeds were grown on BHK cells and plaqued on MC57 cells using an immunological focus assay, as described previously (Battegay et al., 1991). vv–VSV–np was a gift from Dr. B. Moss (National Institutes of Health, Bethesda, Maryland) (Mackett et al., 1985). vv–LCMV–np was provided by Dr. D. H. L. Bishop (Oxford University, Oxford, Great Britain) (Hany et al., 1989). The characterization of vv–LCMV–np has been previously described (Hany et al., 1989). vv–LCMV–np was used to present a low virulence vaccinia strain, whereas vaccinia strain WR, from which vv–LCMV–np is derived, was used as the high virulence thymidine kinase–positive strain. Depletion of CD41 T cells The rat monoclonal antibody YTS191.1 was used for the in vivo depletion of CD41 T cells (Cobbold et al., 1984; Leist et al., 1987). It was produced in rat ascites and purified as described previously (Cobbold et al., 1984). Mice were given 1 mg on day 23 and 1 mg on day 21 before beginning the experiment. In control mice, this treatment depleted CD41 T cells below detection levels by FCM analysis when stained with GK1.5 (Dialynas et al., 1983) and completely abrogated CD41 T helper cell function: after immunization with VSV the T helper–independent neutralizing immunoglobulin M (IgM) response was not followed by any detectable immunoglobulin class switch to IgG (Leist et al., 1987). In Vivo Protection Assays of CTL Activity The in vivo assay for the detection of CTL activity by challenge infections with vaccinia recombinant viruses has been described in detail previously (Binder and Ku¨ndig, 1991). In brief, mice that had been immunized with either VSV or LCMV wild type are challenged intracerebrally by infection with vaccinia recombinant viruses expressing VSV or LCMV proteins, respectively (5 3 10 3 pfu in 30 ml). The titers of the vaccinia recombinant virus were determined in brain homogenates 5 days after the challenge infection, using a standard virus plaque assay (Karupiah et al., 1990). If VSV- or LCMV-specific memory CTL were present, then the respective vaccinia recombinant virus is usually eliminated below detection levels at this timepoint. Conversely, if replication of the vaccinia recombinant viruses in the experimental animals is comparable to replication in unprimed control mice, this indicates the functional absence of in vivo VSVor LCMV-specific CTL activity. Primary Ex Vivo Cytotoxicity Against VSV and LCMV Mice were infected intravenously with VSV (2 3 106 pfu on day 26) or with LCMV (2 3 10 3 pfu on day 28) or with vaccinia viruses (5 3 104 pfu intravenously on day 26). On day 0, spleen cells were coincubated for 5 hr with 51Cr-labeled MC57 (H-2b) target cells that were either uninfected, or infected with VSV (15 pfu per target cell for 2 hr), with LCMV (infection at low multiplicity of infection 48 hr before assay), or with vaccinia virus (strain WR 5 pfu per target cell for 2 hr). Specific lysis was calculated as (experimental 51Cr release 2

Immunity 50

spontaneous 51Cr release) / (total 51Cr release 2 spontaneous release) 3 100%.

51

Cr

LCMV-Induced Footpad Swelling Reaction Mice were infected with LCMV Armstrong by intradermal injection into the hind footpad (500 pfu in 30 ml). Footpad thickness was measured daily with a spring-loaded caliper. Footpad swelling is calculated as (measured thickness 2 thickness before injection) / (thickness before injection). In Vitro Generation of Cytotoxicity Restimulation of CD81 T cells in bulk cultures using CD282/2 or CD281/1 TCR2 mice: 4 3 10 6 responder cells were restimulated either with VSV- or LCMV infected irradiated macrophages (2,000 rads g) in 24-well plates, following standard protocols (Ku¨ ndig et al., 1993a; Oehen et al., 1992). Limiting dilution analysis: the indicated numbers of responder cells were restimulated in 96-well plates using 24 wells per dilution step, with VSV- or LCMV-infected irradiated macrophages (irradiated with 2000 rads g), as described in detail previously (Ciavarra, 1990; Moskophidis et al., 1987; Oehen et al., 1992). Irradiated (2000 rads g) spleen cells from unprimed C57BL/6 mice were used as feeder cells for a total cell density of 5 3 10 5 cells per well. Bulk and limiting dilution cultures were incubated for 5 days, in the presence or absence of rat concanavalin A (ConA) supernatant (10% v/v) and then tested for specific cytotoxicity by 51Cr release assays on MC57 (H-2b) target cells infected with VSV (15 pfu per cell for 2 hr) or with LCMV (infection with low multiplicity 48 hr before assay). Wells in the limiting dilution analysis were considered positive when specific lysis was higher than x 1 3SD of values obtained in unprimed mice. Restimulation of CD81 T cells in bulk cultures using CD282/2 or CD281/1 TCR 1 mice. Spleen cells (2 3 106 ) from TCR1 mice were antigenically restimulated in 24-well plates by adding LCMV–gp peptide to the culture medium (10 26 M). After 3 days, LCMV–gp-specific cytotoxicity was tested on LCMV–gp-labeled syngeneic target cells (EL-4, an H-2b thymoma derived from C57BL/6 mice, and MC57, an H-2b fibrosarcoma derived from C57BL/6 mice) or the same target cells without peptide. Results obtained for MC57 and EL-4 target cells were similar; we depict lysis of EL-4 target cells. Antibodies Antibodies specific for CD69, IL-2Ra, Vb8.1, and CD8a were obtained from Pharmingen (La Jolla, California). FCM analysis for surface expression of CD69, or IL-2Ra 1 day after peptide injection and for surface expression of the transgenic TCR (Vb8.1) 3 days after peptide injection was done using double stainings for CD8a (conjugated with phycoerythrin) and the respective other antibodies conjugated with fluorescein isothiocyanate. T Cell Proliferation Assays Spleen cells (2 3 10 5) from TCR1 mice were in vitro stimulated in 96-well flat-bottomed plates by adding LCMV–gp peptide (amino acids 33–41, 1025 M) to the cultures. Iscove’s modified Dulbecco’s medium supplemented with 10% fetal calf serum and 1025 M b-2mercaptoethanol was used as the culture medium. [3H]thymidine was added after 36 hr of culture and [3H]thymidine incorporation was assessed after another 12 hr. Received March 15, 1996; revised June 3, 1996. References Aichele, P., Brduscha-Riem, K., Zinkernagel, R.M., Hengartner, H., and Pircher, H.P. (1995). T cell priming versus T cell tolerance induced by synthetic peptides. J. Exp. Med. 182, 261–266. Allison, J.P. (1994). CD28–B7 interactions in T-cell activation. Curr. Opin. Immunol. 6, 414–419.

for the assessment of T- and B-cell function. Curr. Opin. Immunol. 6, 320–326. Baskar, S., Ostrand-Rosenberg, S., Nabavi, N., Nadler, L.M., Freeman, G.J., and Glimcher, L.H. (1994). Constitutive expression of B7 restores immunogenicity of tumor cells expressing truncated MHC class II molecules. Proc. Natl. Acad. Sci. USA 90, 5687–5693. Battegay, M., Cooper, S., Althage, A., Baenziger, J., Hengartner, H., and Zinkernagel, R.M. (1991). Quantification of lymphocytic choriomeningitis virus with an immunological focus assay in 24 or 96 well plates. J. Virol. Meth. 33, 191–198. Binder, D., and Ku¨ ndig, T.M. (1991). Antiviral protection by CD81 versus CD41 T cells: CD81 T cells correlating with cytotoxic activity in vitro are more efficient in antivaccinia virus protection than CD4dependent interleukins. J. Immunol. 146, 4301–4307. Bluestone, J.A. (1995). New perspectives of CD28–B7-mediated T cell costimulation. Immunity 2, 555–559. Boehme, J., Haskins, K., Stecha, P., van Ewijk, W., LeMeur, M., Gerlinger, P., Benoist, C., and Mathis, D. (1989). Transgenic mice with I-A on islet cells are normoglycemic but immunologically intolerant. Science 244, 1179–1183. Bretscher, P., and Cohn, M. (1970). A theory of self-nonself discrimination. Science 169, 1042 Buchmeier, M.J., Welsh, R.M., Dutko, F.J., and Oldstone, M.B.A. (1980). The virology and immunobiology of lymphocytic choriomeningitis virus infection. Adv. Immunol. 30, 275–312. Buller, R.M.L., Smith, G.L., Cremer, K., Notkins, A.L., and Moss, B. (1985). Decreased virulence of recombinant vaccinia virus expression vectors is associated with a thymidine-kinase negative phenotype. Nature 317, 813–815. Chen, L., Asche, S., Brady, W.A., Hellstro¨m, I., Hellstro¨ m, E.K., Ledbetter, J.A., McGowan, P., and Linsley, P.S. (1992). Costimulation of antitumor immunity by the B7 counterreceptor for the T lymphocyte molecules CD28 and CTLA-4. Cell 71, 1093–1102. Ciavarra, R.P. (1990). T helper cells in cytotoxic T lymphocyte development: role of L3T41-dependent and -independent T helper cell pathways in virus-specific and alloreactive cytotoxic T lymphocyte responses. Cell. Immunol. 125, 363–379. Cobbold, S.P., Jayasuriya, A., Nash, A., Prospero, T.D., and Waldmann, H. (1984). Therapy with monoclonal antibodies by elimination of T cell subsets in vivo. Nature 312, 548–551. Dialynas, D.P., Wilde, D.B., Marrack, P., Pierres, A., Wall, K.A., Havran, W., Otten, G., Loken, M.R., Pierres, M., Kappler, J.W., and Fitch, F.W. (1983). Characterization of the murine antigenic determinant, designated L3T4a, recognized by monoclonal antibody GK1.5: expression of L3T4a by functional T cell clones appears to correlate primarily with class II MHC antigen reactivity. Immunol. Rev. 74, 29–56. Fearon, E.R., Pardoll, D.M., Itaya, T., Golumbek, P., Levitsky, H.I., Simons, J.W., Karasuyama, H., Vogelstein, B., and Frost, P. (1990). Interleukin-2 production by tumor cells bypasses T helper function in the generation of an antitumor response. Cell 60, 397–403. Freeman, G.J., Boriello, F., Hodes, R.J., Reiser, H., Hathcock, K.S., Laszlo, G., McKnight, A.J., Kim, J., Du, L., Lombard, D.B., Gray, G.S., Nadler, L.M., and Sharpe, A.H. (1993). Uncovering of functional alternative CTLA-4 counter-receptor in B7-deficient mice. Science 262, 907–909. Gansbacher, B., Zier, K., Danials, B., Cronin, K., Bannerji, R., and Gilboa, E. (1990). Interleukin-2 gene transfer into tumor cells abrogates tumorigenicitiy and induces protective immunity. J. Exp. Med. 172, 1217–1224. Green, J.M., Noel, P.J., Sperling, A.I., Walunas, T.L., Gray, G.S., Bluestone, J.A., and Thompson, C.B. (1994). Absence of B7-dependent responses in CD28-deficient mice. Immunity 1, 501–508.

Ando, K., Guidotti, L.G., Cerny, A., Ishikawa, T., and Chisari, F.V. (1994). CTL access to tissue antigen is restricted in vivo. J. Immunol. 153, 482–488.

Guerder, S., Meyerhoff, J., and Flavell, R. (1994). The role of the T cell costimulator B7-1 in autoimmunity and in the induction and maintenance of tolerance to peripheral antigen. Immunity 1, 155–166.

Bachmann, M.F., and Ku¨ndig, T.M. (1994). In vitro vs. in vivo assays

Ha¨mmerling, G., Scho¨nrich, G., Ferber, I., and Arnold, B. (1993).

Extended Signal-1 Mimics Costimulation 51

Peripheral tolerance as a multi-step mechanism. Immunol. Rev. 133, 94–104. Hany, M., Oehen, S., Schulz, M., Hengartner, H., Mackett, M., Bishop, D.H.L., and Zinkernagel, R.M. (1989). Anti-viral protection and prevention of lymphocytic choriomeningitis or of the local footpad swelling reaction in mice by immunisation with vaccinia-recombinant virus expressing LCMV-WE nucleoprotein or glycoprotein. Eur. J. Immunol. 19, 417–424. Harding, F.A., and Allison, J.P. (1993). CD28–B7 interactions allow the induction of CD81 cytotoxic T lymphocytes in the absence of exogenous help. J. Exp. Med. 177, 1791–1796. Harding, F.A., McArthur, J.G., Gross, J.A., Raulet, D.H., and Allison, J.P. (1992). CD28-mediated signalling co-stimulates murine T cells and prevents induction of anergy in T-cell clones. Nature 356, 607–609. Harlan, D.M., Hengartner, H., Huang, M.L., Kang, Y.H., Abe, R., Moreadith, R.W., Pircher, H.P., Gray, G.S., Ohashi, P.S., Freeman, G.J., Nadler, L.M., June, C.H., and Aichele, P. (1994). Mice expressing both B7 and viral glycoprotein on pancreatic beta cells along with glycoprotein-specific transgenic T cells develop diabetes due to a breakdown ot T-lymphocyte unresponsiveness. Proc. Natl. Acad. Sci. USA 91, 3137–3141. Heath, W.R., Allison, J., Hoffmann, M.W., Scho¨nrich, G., Ha¨mmerling, G., Arnold, B., and Miller, J.F.A.P. (1992). Autoimmune disease as a consequence of locally produced IL-2. Nature 359, 547–549. Janeway, C.A., Jr. (1989). Approaching the assymptote?: evolution and revolution in immunology. Cold Spring Harbor Symp. Quant. Biol. 54, 1–13. June, C.H., Bluestone, J.A., Nadler, L.M., and Thompson, C.B. (1994). The B7 and CD28 receptor families. Immunol. Today 15, 321–332. Karupiah, G., Coupar, B.E., Andrew, M.E., Boyle, D.B., Phillips, S.M., Mullbacher, A., Blanden, R.V., and Ramshaw, I.A. (1990). Elevated natural killer cell responses in mice infected with recombinant vaccinia virus encoding murine IL-2. J. Immunol. 144, 290–298. Kawai, K., Shahinian, A., Mak, T.W., and Ohashi, P.S. (1996). Skin allograft rejection in CD28-deficient mice. Transplantation 61, 352–355. Kearney, E.R., Pape, K.A., Loh, D.Y., and Jenkins, M.K. (1994). Visualization of peptide-specific T cell immunity and peripheral tolerance induction in vivo. Immunity 1, 327–339. Kisielow, P., Swat, W., Rocha, B., and von Boehmer, H. (1991). Induction of immunological unresponsiveness in vivo and in vitro by conventional and super-antigens in developing and mature T cells. Immunol. Rev. 122, 70–85. Ku¨ndig, T., Althage, A., Hengartner, H., and Zinkernagel, R.M. (1992). A skin test to assess virus specific cytotoxic T cell activity. Proc. Natl. Acad. Sci. USA 89, 7757–7761. Ku¨ndig, T.M., Bachmann, M.F., Lefrancois, L., Puddington, L., Hengartner, H., and Zinkernagel, R.M. (1993a). Nonimmunogenic tumor cells may efficiently restimulate tumor antigen-specific cytotoxic T cells. J. Immunol. 150, 4450–4456. Ku¨ndig, T.M., Castelmur, I., Bachmann, M.F., Abraham, D., Binder, D., Hengartner, H., and Zinkernagel, R.M. (1993b). Fewer protective cytotoxic T cell epitopes than T-helper cell epitopes on vesicular stomatitis virus. J. Virol. 67, 3680–3683. Ku¨ndig, T.M., Schorle, H., Bachmann, M.F., Hengartner, H., Zinkernagel, R.M., and Horak, I. (1993c). Immune responses in IL-2 deficient mice. Science 262, 1059–1061. Kyburz, D., Aichele, P., Speiser, D.E., Hengartner, H., Zinkernagel, R.M., and Pircher, H.P. (1993). T cell immunity after a viral infection versus T cell tolerance induced by soluble viral peptides. Eur. J. Immunol. 23, 1956–1962.

Leist, T.P., Cobbold, S.P., Waldmann, H., Aguet, M., and Zinkernagel, R.M. (1987). Functional analysis of T lymphocyte subsets in antiviral host defense. J. Immunol. 138, 2278–2281. Linsley, P.S., and Ledbetter, J.A. (1993). The role of the CD28 receptor during T cell responses to antigen. Annu. Rev. Immunol. 11, 191–212. Lucas, P.J., Negishi, I., Nakayama, K., Fields, L.E., and Loh, D.Y. (1995). Naive CD28-deficient T calls can initiate but not sustain an in vitro antigen-specific immune response. J. Immunol. 154, 5757– 5768. MacDonald, H.R., Lees, R.K., Baschieri, S., Herrmann, T., and Lussow, A.R. (1993). Peripheral T-cell reactivity to bacterial superantigens in vivo: the response/anergy paradox. Immunol Rev. 133, 105–118. Mackett, M., Yilma, T., Rose, J.K., and Moss, B. (1985). Vaccinia virus recombinants: expression of VSV genes and protective immunization of mice and cattle. Science 227, 433–435. Marrack, P., Hugo, P., McCormack, J., and Kappler, J. (1993). Death and T cells. Immunol. Rev. 133, 119–129. Migita, K., and Ochi, A. (1993). The fate of anergic T cells in vivo. J. Immunol. 150, 763–770. Moskophidis, D., Assmann Wischer, U., Simon, M.M., and Lehmann Grube, F. (1987). The immune response of the mouse to lymphocytic choriomeningitis virus. V. High numbers of cytolytic T lymphocytes are generated in the spleen during acute infection. Eur. J. Immunol. 17, 937–942. Oehen, S., Waldner, H.P., Ku¨ndig, T.M., Hengartner, H., and Zinkernagel, R. M. (1992). Cytotoxic T cell memory to lymphocytic choriomeningitis virus is governed by elevated CTL-P frequencies and persisting antigen. J. Exp. Med. 176, 1273–1281. Ohashi, P.S., Oehen, S., Bu¨rki, K., Pircher, H.P., Ohashi, C.T., Odermatt, B., Malissen, B., Zinkernagel, R., and Hengartner, H. (1991). Ablation of “tolerance” and induction of diabetes by virus infection in viral antigen transgenic mice. Cell 65, 305–317. Ohashi, P.S., Oehen, S., Aichele, P., Pircher, H., Odermatt, B., Herrera, P., Higuchi, Y., Buerki, K., Hengartner, H., and Zinkernagel, R.M. (1993). Induction of diabetes is influenced by the infectious virus and local expression of MHC class I and tumor necrosis factor alpha. J. Immunol. 150, 5185–5194. Pircher, H.P., Bu¨rki, K., Lang, R., Hengartner, H., and Zinkernagel, R. (1989). Tolerance induction in double specific T-cell receptor transgenic mice varies with antigen. Nature 342, 559–561. Pircher, H.P., Moskophidis, D., Rohrer, U., Bu¨rki, K., Hengartner, H., and Zinkernagel, R.M. (1990). Viral escape by selection of cytotoxic T cell–resistant virus variants in vivo. Nature 346, 629–633. Puddington, L., Bevan, M.J., and Lefranc¸ ois, L. (1986). N protein is the predominant antigen recognized by vesicular stomatitis virusspecific cytotoxic T cells. J. Virol. 60, 708–717. Rammensee, H.G., Kroschewski, R., and Frangoulis, B. (1989). Clonal anergy induced in mature V beta 61 T lymphocytes on immunizing Mls-1b mice with Mls-1a expressing cells. Nature 339, 541–544. Ramsdell, F., and Fowlkes, B.J. (1990). Clonal deletion versus clonal anergy: the role of the thymus in inducing self tolerance. Science 248, 1342–1348. Ramsdell, F., and Fowlkes, B.J. (1992). Maintainance of in vivo tolerance by persistence of antigen. Science 257, 1130–1134. Rocha, B., Tanchot, C., and von Boehmer, H. (1993). Clonal anergy blocks in vivo growth of mature T cells and can be reversed in the absence of antigen. J. Exp. Med. 177, 1517–1521. Schwartz, R.H. (1990). A cell culture model for T lymphocyte clonal anergy. Science 248, 1349–1356.

Lafferty, K., and Woolnough, J. (1977). The origin and mechanism of allograft reaction. Immunol. Rev. 35, 231–262.

Schwartz, R.H. (1992). Costimulation of T lymphocytes: the role of CD28, CTLA-4, and B7/BB1 in interleukin-2 production and immunotherapy. Cell 71, 1065–1068.

Lehmann-Grube, F. (1971). Lymphocytic choriomeningitis virus. Virol. Monogr. 10, 1–173.

Shahinian, A., Pfeffer, K., Lee, K.P., Ku¨ndig, T.M., Kishihara, K., Wakeham, A., Kawai, K., Ohashi, P.S., Thompson, C.B., and Mak,

Immunity 52

T.W. (1993). Differential costimulatory requirements in CD28-deficient mice. Science 261, 609–612. Sprent, J., and Webb, S.R. (1995). Intrathymic and extrathymic clonal deletion of T cells. Curr. Opin. Immunol. 7, 196–205. Sprent, J., Kosaka, H., Gao, E.K., Surh, C.D., and Webb, S.R. (1993). Intrathymic and extrathymic tolerance in bone marrow chimeras. Immunol. Rev. 133, 151–176. Townsend, S.E., and Allison, J.P. (1993). Tumor rejection after direct costimulation of CD81 T cells by B7-transfected melanoma cells. Science 259, 368–372. von Herrath, M.G., Dockter, J., and Oldstone, M.B.A. (1994). How virus induces a rapid or slow onset insulin-dependent diabetes mellitus in a transgenic model. Immunity 1, 231–242. Wagner, R.R. (1987). The Rhabdoviruses (New York: Plenum Press). Widmann, C., Maryanski, J.L., Romero, P., and Corradin, G. (1991). Differential stability of antigenic MHC class I–restricted synthetic peptides. J. Immunol. 147, 3745–3751. Zinkernagel, R.M., Moskophidis, D., Ku¨ndig, T.M., Oehen, S., Pircher, H., and Hengartner, H. (1993). Effector T-cell induction and T-cell memory versus peripheral deletion of T cells. Immunol. Rev. 133, 199–223.