|Home | About | Journals | Submit | Contact Us | Français|
Cytotoxicity is a key effector function of CD8 T cells. However, what proportion of antigen-specific CD8 T cells in vivo exert cytotoxic activity during a functional CD8 T-cell response to infection still remains unknown. We used the Lysispot assay to directly enumerate cytotoxic CD8 T cells from the spleen ex vivo during the immune response to infection with the intracellular bacterium Listeria monocytogenes. We demonstrate that not all antigen-responsive gamma interferon (IFN-γ)-secreting T cells display cytotoxic activity. Most CD8 T cells detected at early time points of the response were cytotoxic. This percentage continuously declined during both the expansion and contraction phases to about 50% at the peak and to <10% of IFN-γ-producing cells in the memory phase. As described for clonal expansion, this elaboration of a program of differentiation after an initial stimulus was not affected by antigen or CD4 help but, like proliferation, could be influenced by later reinfection. These data indicate that cytotoxic effector function during the response to infection is regulated independently from IFN-γ secretion or expansion or contraction of the overall CD8 T-cell response.
CD8 T cells play an essential role in containing intracellular infections and tumor growth. After priming, mice mount a strong CD8 T-cell response that peaks at around day 7 to 10 after infection and then moves into a contraction phase, with about 10% of the antigen-specific CD8 T cells surviving and remaining as long-term memory cells (6, 7, 14, 19). After priming, the expansion of the CD8 T-cell response progresses in a semiautonomous fashion, in part independent from the presence of the antigen (3, 13, 18, 22, 28, 33). When the contraction phase is initiated is not entirely understood, but CD8 T cells that show the typical preapoptotic annexin V binding can already be found days before the peak of the response (30).
Upon T-cell receptor engagement, primed CD8 T cells exert immediate cytotoxicity on infected cells by secreting perforin and granzyme B (4) and thereby efficiently restrict the spread of the pathogen (23). While it is generally assumed that CD8 T cells in the acute response to an infection are cytotoxic, it is also known that even strong CD8 T-cell responses may under certain conditions fail to exert cytotoxic effector functions (16). Also, the transgenic overexpression of CD70 enhanced the efficiency of cytolytic T cells at the peak of the response to an influenza infection, suggesting that, in wild-type mice, not all CD8 T cells are cytotoxic, even during strong CD8 T-cell responses induced by viral infections (1).
Cytotoxic memory CD8 T cells are confined to specific subpopulations (10, 31) that preferentially localize in peripheral tissues and not in the spleen (17). In contrast, during the acute phase the peripheral and splenic T-cell populations exerted similar cytotoxic activities, although the proportion of cytotoxic CD8 T cells within each population has not been determined (17).
To design improved T-cell-based vaccines, it is important to know precisely how many CD8 T cells are actually cytotoxic over the course of a response. The single-cell Lysispot assay measures the number of CD8 T cells that exert antigen-specific cytotoxicity ex vivo by detecting lysis of individual target cells in an enzyme-linked immunospot (ELISPOT) assay (24). This contrasts with other measures of cytotoxicity that measure bulk cytotoxicity or the expression of a surrogate marker on single cells. We used the Lysispot and ELISPOT assays to determine the proportion of the CD8 T-cell population with immediate ex vivo cytolytic or gamma interferon (IFN-γ)-secreting activity at the different stages of the immune response to an infection. We chose the intracellular bacterium Listeria monocytogenes as a model system, since infection raises strong CD8 T-cell responses, which play an important role in protection, and since the immune response against this pathogen is well characterized and has already been used extensively to decipher the different factors that influence CD8 T-cell responses (20).
Phoenix cells (ATCC, Manassas, VA) were transfected with the retrovirus packaging vector pCL-10A1 and the β-galactosidase (β-Gal)-encoding plasmid pCLMFG-LacZ (Imgenex, San Diego, CA) using calcium phosphate precipitation. Thirty-six hours after transfection, exponentially growing P815 cells were added to the culture to allow retroviral transduction. Twelve hours later the P815 cells were cloned by limiting dilution. Clones expressing β-Gal at levels high enough to detect every P815 cell lysed in a Lysispot assay were frozen in liquid nitrogen. Prior to each Lysispot assay, one aliquot of P815-LacZ cells was thawed and cells were grown exponentially in α20-medium (minimal essential alpha medium; Invitrogen, Carlsbad, CA) containing 20% fetal calf serum (FCS) for several days. After each assay the cells were fixed and tested for homogeneous β-Gal expression.
The Listeria strain 10403S was grown in brain heart infusion broth and harvested while in the exponential-growth phase. Bacterial density was determined by measuring absorption at 600 nm (an optical density at 600 nm of 0.1 is equivalent of 2 × 108 Listeria cells per ml). Female BALB/c mice were purchased from Taconic (Germantown, NY) and infected with various amounts of L. monocytogenes in 100 μl phosphate-buffered saline (PBS) intravenously (i.v.) into the lateral tail vein at an age of 6 to 12 weeks. To determine numbers of bacteria in spleens and livers of infected mice, the tissues were pressed through a cell strainer using antibiotic-free medium. Single-cell suspensions were diluted in PBS and plated onto brain heart infusion agar plates, which were incubated for 24 to 36 h at 37°C.
Thirty-six hours after infection, mice were injected subcutaneously with 1 μg ampicillin in 100 μl PBS, and 2 mg/ml ampicillin (EM Science, Gibbstown, NJ) was added to the drinking water for the following 3 days. Efficiency of bacterial clearance was determined the following day by colony counting as described above.
For IFN-γ ELISPOT and Lysispot assays, 96-well MAIP ELISPOT plates (Millipore, Bedford, MA) were coated with 2 μg/ml AN18 (anti-mouse IFN-γ; ATCC, Manassas, VA) or 4 μg/ml D19-2F3-2 (anti-β-Gal antibody; Roche, Indianapolis, IN) in PBS, respectively, for at least 2 h at room temperature and then washed and blocked twice with RP8 medium (RPMI medium containing 8% FCS). Peptide-pulsed (at least 1 h with 2 μg/ml of a peptide comprising listeriolysin amino acids 91 to 99 [LLO91-99] or p60 amino acids 217 to 225 [p60217-225]; Invitrogen, Carlsbad, CA) and untreated P815-LacZ cells were added (60,000/well) in 50 μl α20 medium. Splenocyte dilutions, starting with 1 × 106 cells per well, were added in 50 μl IMDM10 (Iscove's modified Dulbecco's medium; HyClone, Logan, UT) supplemented with 10% FCS, 1,000 IU penicillin, 100 μg/ml streptomycin, and 100 μg/ml gentamicin (Invitrogen), and the assay mixtures were incubated for 4 h at 37°C and 7% CO2. Twenty minutes prior to the end of the Lysispot assays, 50 μl IMDM10 containing a 1:1,000 dilution of biotinylated anti-β-Gal antibody GAL13 (Sigma, St. Louis, MO) was added gently, without disturbing the cell layer. To stop the assays, plates were thoroughly washed with PBST (PBS plus 0.01% Tween 20), and 2 μg/ml biotinylated XMG1.2 (BD PharMingen, San Diego, CA) was used for detection of IFN-γ spots. Both spot assays were developed with 1 μg/ml alkaline phosphatase-conjugated streptavidin (Jackson Immunobiology, Bar Harbor, ME) in PBST plus 2% bovine serum albumin. The plates were then washed, dried, and analyzed using the CTL ImmunoSpot scanner and software (Cellular Technology, Ltd., Cleveland, OH). Values within the linear dilution ranges were counted.
At days 3 and 4 after infection with Listeria monocytogenes, mice were injected intraperitoneally with 200 μg purified GK1.5 in 100 μl PBS. Efficiency of CD4 T-cell clearance in the spleen was measured 24 h after the last injection by flow cytometry using clone RM4-4 (BD PharMingen), which binds to a nonoverlapping determinant. CD4+ T cells were less than 0.03% of Thy1.2+ spleen cells.
To verify that the ELISPOT reliably detected IFN-γ-secreting CD8 T cells, we compared the numbers of Listeria-specific CD8 T cells in the spleens of infected mice ex vivo by either IFN-γ ELISPOT or major histocompatibility complex/peptide-tetramer staining. Consistent with previous reports (27), we detected similar numbers of responding CD8 T cells using either IFN-γ ELISPOT or major histocompatibility complex/LLO91-99-tetramer staining (data not shown). Since our measured responses were generally within the expected range of 104 (5) to 106 (6) per infected spleen for primary infections (see Fig. Fig.2)2) and 105 to 107 for recall infections (5, 6) (data not shown) and since the LLO91-99-specific IFN-γ ELISPOT results were about 103 spot-forming units/1 × 106 splenocytes at the peak of the response, similar to results described before (29), we felt confident that our assay reliably detected the activated antigen-responsive CD8 T cells ex vivo. Therefore, we used the IFN-γ ELISPOT assay to quantify the total CD8 response.
The Lysispot assay uses the same principle as the IFN-γ ELISPOT assay; however, it detects CD8 T cells with cytotoxic activity at the single-cell level (24). In this assay, peptide-loaded target cells are infected with β-Gal-encoding herpes simplex virus amplicons and β-Gal release due to cell lysis is measured in an ELISPOT assay. We modified the original Lysispot assay (24) by using P815 cells stably transduced with β-Gal as targets (see Materials and Methods). To verify that all P815-LacZ cells lysed were detectable in our assay, we added increasing numbers of P815-LacZ cells to a monolayer of resting 2C cells, which are transgenic for a T-cell receptor recognizing H2-Ld expressed on the P815-LacZ cells. The number of induced IFN-γ spots above background was approximately equal to the number of Lysispots measured (Fig. (Fig.1A),1A), indicating that all P815-LacZ cells that induced IFN-γ secretion by the 2C cells had also been lysed and that all cells that had been lysed were also detected by the Lysispot assay (Fig. (Fig.1A1A).
By titration, we determined that the optimal level for reliably detecting antigen-specific CD8 T cells ex vivo with a minimum of background caused by spontaneous lysis was 60,000 P815-LacZ cells/well (Fig. (Fig.1B).1B). The Lysispot was then used to measure the frequency of CD8 T cells that exerted immediate ex vivo cytolytic effector function.
To quantify immediate cytotoxic CD8 T cells during the course of an infection, we infected BALB/c mice i.v. with the intracellular pathogen Listeria monocytogenes and monitored the CD8 T-cell response to the immunodominant CD8 T-cell epitope LLO91-99 by IFN-γ ELISPOT and Lysispot. The numbers of immediately ex vivo cytolytic LLO91-99-specific CD8 T cells and IFN-γ-secreting cells expanded and contracted with generally similar kinetics, increasing over the first 8 days and then declining (Fig. 2A and D). At early time points after infection similar numbers of cells secreted IFN-γ and were cytolytic. However, the subsequent expansion of the IFN-γ-secreting population in the spleen was faster, so, even early in the response, a population of antigen-responsive but noncytotoxic CD8 T cells appeared. Consequently, at the peak of the immune response, there were roughly twice as many IFN-γ-secreting as cytolytic CD8 T cells (Fig. 2A and D). During the contraction phase, the frequency of cytolytic cells declined even more rapidly than the IFN-γ-secreting population. The proportion of cytotoxic T cells stabilized at around 10% of antigen-specific IFN-γ-secreting CD8 T cells in the spleen. This ratio remained stable even in long-term memory, 169 days after the recall infection (data not shown). Similar patterns of expansion and contraction of cytotoxic versus IFN-γ-secreting cell populations were seen in two independent experiments (Fig. 1A to C and D to F), with substantially different (up to 10-fold-higher) frequencies of CD8 T cells specific for LLO91-99 and for a second, subdominant CD8 Listeria epitope, p60217-225 (data not shown). In addition, 28 days after the primary infection, a secondary infection was performed and expansion and contraction of cytotoxic versus IFN-γ-secreting cell populations were analyzed. Similar patterns as in the primary response were seen, although, as expected, with faster kinetics of the CD8 T-cell response (data not shown).
Figure 2B and E show the total numbers of ex vivo cytotoxic versus IFN-γ-secreting CD8 T cells in the spleen of each mouse. Over the course of the infection, the antigen-specific CD8 T-cell response followed a consistent pattern. Until the peak of the response (day 8 after infection), the ratio of cytotoxic to IFN-γ-secreting cells deviated progressively more from the approximately 1:1 ratio that is found in the mice early after infection. After the peak of the response (day 10 after infection), the cytotoxic population decreased more rapidly than the IFN-γ-secreting population. Overall, this led to a continuous decline of the fraction of cytotoxic cells within the IFN-γ-secreting population (Fig. 2C and F), regardless of whether the population was still in its expansion phase or its contraction phase.
Taken together, these data suggest that the cytotoxic effector function is regulated independently from IFN-γ secretion.
After priming, the CD8 T-cell response goes through a clonal expansion phase that can be antigen independent (18, 22). To determine whether the presence of the pathogen may influence the selective loss of cytotoxicity of the responding CD8 T-cell population, we infected mice with L. monocytogenes and eliminated the bacteria 36 h later by treatment with ampicillin. If the overall presence of the pathogen, and therefore antigen (3), had a direct influence on the effector phenotype of the CD8 T-cell response, then the ratio of cytotoxic to IFN-γ-secreting CD8 T cells should be altered in the ampicillin-treated mouse group due to the artificial shortening of antigen exposure.
Untreated animals cleared the pathogen in the spleen precipitously at day 6, although infected cells were still detectable in the liver up to 7 days after infection. In ampicillin-treated animals the pathogen was undetectable at 60 h after infection, i.e., 24 h after initiation of treatment (data not shown). In agreement with previous publications (18, 22), the overall expansion of antigen-specific CD8 T cells was diminished (Fig. (Fig.3A),3A), while the frequency of IFN-γ-secreting T cells per million splenocytes at day 8 after infection remained similar (Fig. (Fig.3B).3B). The ratios between ex vivo cytotoxic and IFN-γ-secreting LLO91-99-responsive CD8 T cells, however, for ampicillin-treated and nontreated groups were similar (Fig. (Fig.3C).3C). These data suggest that the proportion of CD8 T cells with cytotoxic effector function is not directly influenced by the presence or absence of the pathogen. Thus, similar to CD8 T-cell proliferation, the maintenance of cytotoxic effector function does not directly depend on sustained antigen exposure after priming.
CD4 T cells can positively and negatively influence CD8 T-cell responses. In addition to providing help for durable CD8 T-cell memory responses at early time points (11, 25), depletion of CD4+ (15) or CD25+ (26) cells prior to recall or primary infection, respectively, led to extended expansion of CD8 T cells, suggesting that an expanding antigen-specific regulatory CD4 T-cell population could control CD8 T cells at the peak of the response (26). To determine whether such a CD4+ T-cell population could be responsible for the rapid loss of ex vivo cytotoxic function of the pathogen-responsive CD8 T-cell population, CD4 cells were depleted in vivo by anti-CD4 treatment 3 days after infection with L. monocytogenes, when CD8 T-cell priming should be completed (34).
CD4-depleted and untreated mice cleared the pathogen with similar kinetics and had similar frequencies of LLO91-99-responsive CD8 T cells in the spleen at the peak of the CD8 T-cell response (data not shown), and also the ratios of cytotoxic to IFN-γ-secreting cells were similar in the two groups (Fig. (Fig.4A).4A). These data indicate that the observed selective loss of ex vivo cytolytic activity is unlikely to be mediated in a substantial way by the suppressive activity of a CD4 T-cell population.
Busch et al. have shown that the reinfection of mice at day 6 after the original infection leads to an enhanced expansion of the pathogen-specific CD8 T cells, even if the antigen is absent in the bacteria used for reinfection (5). To test whether such a general inflammatory environment could also influence the cytotoxic effector function of the CD8 T cells, we reinfected BALB/c mice with a 500-fold-higher dose of pathogen at a time point when the initial infection had been cleared, 6 days after primary infection. In accordance with previous reports, this treatment led to an extended CD8 T-cell expansion (data not shown). When we tested the phenotype of responding CD8 T cells 2 days after the reinfection, the ex vivo cytolytic capacity of the CD8 T-cell response was significantly restored (Fig. (Fig.4B).4B). Thus, similar to results for long-term memory cells, present 28 days after primary infection, which rapidly regain effector function upon reinfection (data not shown), these results show a reactivation of cytotoxicity already during the initial response.
These results, taken together, show that, similar to the effects described for proliferation, an initial stimulus leads to a program of differentiation that is not affected by antigen or CD4 help. Furthermore, the progressive loss of cytotoxic effector function can be reversed by late reinfection.
We have directly enumerated ex vivo cytotoxic T cells responding to infection with Listeria monocytogenes. We show that early after infection with live bacteria a very high percentage of responding antigen-specific CD8 T cells gain cytotoxic effector functions. This portion of cytotoxic, antigen-responsive CD8 T cells, in relation to the IFN-γ-secreting population, steadily declines over the course of the response. These findings follow the same trends as the expression of surrogate markers of cytotoxicity, such as perforin and granzyme A/B, at the protein (31) and mRNA levels (12, 21), although differences in sensitivities of particularly the mRNA detection methods may account for quantitative differences in the frequencies observed. Thus the early regulation of cytotoxicity that we describe may be mediated at the level of effector molecule expression, although we cannot exclude the possibility that changes in the threshold required to trigger degranulation also occur.
Our data further suggest that IFN-γ secretion, cytotoxicity, and proliferation are all regulated independently during the acute phase of CD8 T-cell responses to infection. In a set of very elegant experiments, it was shown before that proliferation and differentiation of CD8 T cells are regulated in different ways (16). Immunization of mice with heat-killed L. monocytogenes induced the expansion of antigen-specific CD8 T cells, which, however, did not acquire cytotoxic effector function. Most remarkably, immunization with a mixture of live bacteria and heat-killed L. monocytogenes also induced a mixed CD8 T-cell response, indicating that (i) the stimulus for differentiation is given already at a very early time point during priming and (ii) this stimulus must be given at a compartmentalized level, perhaps even at the single-cell level, during interaction between the priming dendritic cells and the naive CD8 T cells. Further, it has been shown that the early inflammatory environment can influence clonal expansion (2). However, additional signals later during the response, as created, for example, by extended pathogen presence (18, 22) or by reinfection with the pathogen (5), can further influence the magnitude of the response. Our results suggest a similar form of regulation for the cytotoxic effector function. At early time points of infection, the responding CD8 T cells gain cytotoxic effector function. However, they rapidly lose this again with a characteristic dynamic and independent of the presence of antigen (Fig. (Fig.3)3) or CD4 T cells (Fig. (Fig.4A).4A). At the same time, strong stimuli can increase the portion of cytotoxic T cells (Fig. (Fig.4B4B).
Most interestingly, the percentage of cytotoxic cells followed a consistent pattern during the immune response and was well coordinated with T-cell expansion: the further the antigen-specific response had expanded within an individual mouse, the further the decline of the cytotoxic proportion had progressed (Fig. (Fig.2B).2B). This trend occurred even in groups with considerable heterogeneity at the peak of the response. This suggests that in each mouse the CD8 T-cell response unfolds in a consistent manner regardless of the precise kinetics of the response within individual mice.
Taken together, our data suggest that, similar to the clonal expansion of the CD8 T-cell response after priming, the dynamics of cytotoxic effector function are determined at a very early time point and then follow a specific program.
These findings raise the question of why the cytotoxic effector function is restrained already when the overall antigen-specific CD8 T-cell response is still expanding. One obvious advantage can be seen in the fact that the cytotoxic effector function causes direct damage to the host. In situations in which a very high percentage of the tissue is infected, tight regulation of this effector function could potentially prevent excessive, life-threatening tissue damage to the host. An additional advantage might be related to preservation of professional antigen-presenting cell activity. The expanding CD8 T-cell response limits the priming of novel pathogen-specific T-cell responses by a feedback loop that kills professional antigen-presenting cells that present the original epitope (34). Thus, the early loss of cytotoxic effector function might allow an accelerated renewed priming of potentially more protective CD8 T-cell responses in case the original response failed to do so.
How cytotoxic effector function in vivo is regulated precisely, on a molecular level, remains unknown. In order to kill a target cell, a CD8 T cell has to express multiple different effector molecules (12, 21) and focus these in the immunological synapse (8). It is possible that there are two independent CD8 T-cell populations, one cytotoxic and one not, and that the observed selective loss of cytotoxic effector function is explained by different rates of expansion or contraction of these two populations within the spleen. Nevertheless, we consider it far more likely that a change in phenotype of individual T cells occurs within the entire CD8 T-cell population, which may explain the observed selective reduction in cytotoxicity during the immune response. Loss of cytotoxicity could precede the depletion of the CD8 T cells in the contraction phase. This could also explain why cells retain increased cytotoxicity levels in this phase if they receive additional survival signals, such as through interleukin-7/interleukin-7 receptor interaction (10), or activation signals, such as through CD70/CD27 interaction during an influenza virus infection (1). These reports fit well with our observation that reinfection with a very high load of pathogen, which is likely to have provided a novel pulse of additional antiapoptotic, inflammatory stimuli, just before the onset of the contraction phase led to an increased proportion of antigen-specific CD8 T cells expressing cytotoxic effector function (Fig. (Fig.3C3C).
In conclusion, our data shed new light on the different regulations to which different CD8 T-cell effector functions are subjected during an immune response. Our data indicate that, upon infection, individual CD8 T cells rapidly gain cytotoxic effector functions and then progressively loose them, according to a set internal regulation.
This work was supported by NIH grants AI48604 (to T.R.M.) and AI064576 (to A.J.A.M.S.). D.M.W.Z. was supported in part by a fellowship of the German Research Council (DFG Za 280).
We thank Alexandra Livingstone for commenting on the manuscript.
The authors have no conflicting financial interests.
Editor: J. L. Flynn
Published ahead of print on 4 August 2008.