Studies in IFN-γ-deficient mice suggest that the delivery of IFN-γ to CD8+ T cells early in virus infection programs their eventual contraction, thereby reducing the abundance of CD8+ memory T cells. In this study, we show that such mice fail to completely eliminate virus infection and that, when evaluated without the confounding factor of persisting Ag, both CD4+ and CD8+ T cells undergo profound contraction when they are unable to receive IFN-γ signals. Furthermore, the abundance of CD4+ and CD8+ memory cells that express the IFN-γ receptor is ∼100-fold higher than cells lacking this molecule. Thus, direct IFN-γ signaling is not required for T cell contraction during virus infection, and it enhances, rather than suppresses, the development of virus-specific CD4+ and CD8+ T cell memory.
Memory T cells play a central role in protecting the host against disease caused by secondary exposure to the same, or a closely related, agent, but the manner in which these key cells are generated remains unclear. Some studies suggest that memory cells may be a distinct lineage, predestined to enter the memory pool, while other studies indicate that memory cells may be derived, probably stochastically, from the primary effector cell population. It is clear that brief ex vivo exposure to Ag can trigger an existing developmental pathway in naive CD8+ T cells, causing them to divide, express their effector functions, contract, and form a memory cell pool (1, 2); similar effects have been reported for CD4+ T cells. However, virus-specific naive T cells in vivo encounter their cognate Ag during infection, in the presence of inflammatory signals, and a number of recent studies suggest that these inflammatory events can alter the intrinsic developmental pathway. Type I and type II IFNs appear to play important roles in these early regulatory events (3, 4). In the absence of the type I IFN receptor, the acute CD8+ T cell response is dramatically reduced (5); this is true also for CD4+ T cell responses, although the effects seem to vary depending on the nature of the infecting microbe (6). The effects of IFN-γ are more controversial. We have shown that, during an acute lymphocytic choriomeningitis virus (LCMV)3 infection, the most abundant cells are those that most rapidly elaborate IFN-γ, and we proposed that this cytokine might exert a positive effect on the developing primary T cell response (7). Others have demonstrated the importance of IFN-γ in the priming of naive T cells (8, 9, 10), and further analyses from our laboratory showed that the cytokine acts directly on CD4+ and CD8+ T cells to increase their abundance during the primary antiviral response (11, 12). Thus, during the primary immune response to several infections, IFN-γ appears to have a positive influence on T cell responses. However, the converse is thought to be true for subsequent events. Data suggest that IFN-γ is required for normal T cell apoptosis and contraction (13, 14, 15), and that it may “program” the contraction phase, perhaps via its delivery to T cells early in infection (16, 17, 18); these negative effects may explain why memory T cell numbers are increased following infection of mice deficient in IFN-γ signaling, as has been shown by several laboratories. Moreover, a recent intriguing study indicated that IFN-γ might suppress the development of memory T cell generation after dendritic cell immunization (19). Taken together, these findings suggest that the effects of this cytokine on T cell abundance can be either positive (enhancing the primary response) or negative (programming contraction of CD8+ T cells and suppressing the development of CD8+ memory T cells). As noted elsewhere (20), the clearest example of this apparent functional dichotomy occurs in IFN-γ-deficient mice, where effector T cell expansion and elimination are both impaired. The overall goal of the present study was to clarify these apparent opposing effects of IFN-γ on T cell development during a viral infection, by evaluating the effects of direct IFN-γ signaling on the following: 1) CD8+ T cell contraction, 2) the establishment of CD8+ T cell memory, and 3) the contraction and memory phases of the antiviral CD4+ T cell response.
Materials and Methods
Mice and virus
C57BL/6 mice were purchased from The Scripps Research Institute (TSRI) breeding facility. IFN-γRKO (B6.129S7-Ifngr1tm1Agt/J) and IFN-γKO mice (B6.129S7-Ifngtm1Ts/J), both strains backcrossed 10 generations to C57BL/6, and C57BL/6 mice congenic for Thy1.1 (B6.PL-Thy1a/CyJ) were purchased from The Jackson Laboratory. SMARTA TCR-transgenic (Tg) mice (21) specific for the I-Ab LCMV epitope GP61–80 were crossed to C57BL/6.Ly5a mice (B6.SJL-PtprcaPep3b/BoyJ) to generate SMARTA.Ly5a mice and to IFN-γRKO mice to generate SMARTA/IFN-γRKO mice (11). P14 TCR-Tg mice specific for the LCMV epitope GP33–41 on the H-2b background were crossed to B6.Ly5a mice to generate the P14.Ly5a stain or to IFN-γRKO mice to generate the P14/IFN-γRKO strain. Mice were infected by i.p. administration of 2 × 105 PFU of LCMV (Armstrong strain). Quantitation of virus in the tissues was done by plaque assay on Vero cell monolayers. To assay for virus RNA, harvested tissues were homogenized in the presence of TRIzol reagent (Invitrogen Life Technologies), and total RNA was isolated from the aqueous phase after chloroform and isopropanol extractions. The presence of viral RNA was measured by RT-PCR using primers specific for sequences within the glycoprotein. All animal experiments were approved by TSRI Animal Care and Use Committee.
Flow cytometry was used to determine the frequency of Tg CD4+ T cells (Vα2+Vβ8.3+) among all spleen cells in SMARTA mice or the frequency of Tg CD8+ T cells (Vα2+Vβ8.1/2+) among all spleen cells in P14 mice. In the dual adoptive transfer experiments, the donor cells from P14/IFN-γR+ and P14/IFN-γRKO mice were mixed so that equal numbers of Tg CD8 T cells (2 × 104 each) were given i.v. to recipient mice. Other recipient mice received a mix containing 2 × 104 SMARTA/IFN-γR+ and 2 × 104 SMARTA/IFN-γRKO Tg CD4 T cells. The recipient mice were infected 4–7 days after transfer. As a control, some mice were given a mix with 1 × 106 wild-type (WT) and IFN-γRKO Tg cells and then left uninfected. These control mice show that the proportion of cells given to the mice was equal, and this proportion did not change over time in the absence of infection.
In vivo test for viral Ag persistence
Naive WT P14 or SMARTA cells were labeled with 5 mM CFSE and 1–2 × 106 cells were transferred i.v. into LCMV-immune mice. After 1 wk, the amount of CFSE dye retained by the donor T cells was measured by flow cytometry; the loss of fluorescence indicates that the T cells have seen Ag in sufficient quantities to undergo cell division.
Spleen cells were stained directly ex vivo with anti-CD4 (clone RM4–5), anti-CD8 (clone 53-6.7), anti-Thy1.2 (CD90.2, clone 53-2.1), anti-CD44 (clone IM7), anti-Ly5a (Ly5.1, clone A20), or anti-Ly5b (Ly5.2, clone 104) all purchased from eBioscience. Anti-IFN-γR1 (clone GR20) and the corresponding isotype control Ab were purchased from BD Pharmingen. The intracellular staining assay was performed as described previously using anti-IFN-γ (clone XMG1.2), anti-TNF (clone MP6-XT22), and anti-IL-2 (clone JES6-5H4) from eBioscience. Cell staining was analyzed by four-color flow cytometry at TSRI core facility using a BD Biosciences FACScalibur and FloJo software (Tree Star).
Both the expansion and contraction of CD8+ T cells are reduced in LCMV-infected mice lacking IFN-γ signaling
First, we sought to confirm published observations, by enumerating epitope-specific CD8+ T cell numbers in mice that are deficient in IFN-γ signaling. WT C57BL/6 mice, and mice lacking IFN-γ (GKO mice), were infected with LCMV and, at the time points indicated in Fig. 1,A, were sacrificed, splenocytes were harvested, and the numbers of CD8+ T cells specific for each of four epitopes were determined. All four epitope-specific populations showed similar kinetics: 1) during the acute phase of infection, CD8+ T cells were 6- to 10-fold more abundant in WT mice than in GKO mice, confirming our recent observation that IFN-γ exerts a positive effect on T cells during acute infection (12); 2) a reduction in T cell numbers over time (T cell contraction) was clearly demonstrable for all epitope-specific populations in WT mice (▪), but was much less pronounced in GKO mice (□); and, presumably as a result of a combination of these factors, 3) the final number of CD8+ memory T cells was marginally higher in IFN-γKO mice than in WT mice. This last observation was broadly confirmed after LCMV infection of mice lacking the IFN-γ receptor 1 (IFN-γRKO mice, Fig. 1,B). Analyses of the four epitope-specific CD8+ T cell responses over the course of infection (Fig. 1,B, left) revealed that contraction was reduced in IFN-γRKO mice (dashed lines/open symbols) compared with WT mice (solid lines/closed symbols); and, for three of the four epitopes (Fig. 1 B, right), CD8+ memory T cells were more abundant in IFN-γRKO mice (□) than in WT mice (▪). In summary, mice deficient in IFN-γ signaling have similar, or slightly elevated, numbers of epitope-specific CD8+ memory T cells following LCMV infection, as reported by others (18, 22, 23); and this may result from a combination of decreased primary expansion (12) followed by defective contraction (16).
Interpretation of T cell kinetics in LCMV-infected “γ-defective” mice is complicated by Ag persistence
Many studies of chronic/persistent virus infections have shown that persistent Ag can effect dramatic changes in T cell quantity and quality. Consequently, it is critical to determine whether or not LCMV Armstrong is cleared in mice with defective IFN-γ signaling, because the retention of low levels of Ag (even in the absence of detectable infectious virus) would greatly complicate the interpretation of T cell abundance; residual virus materials might continuously stimulate virus-induced T cells, and/or recruit new naive cells into the antiviral pool, as recently reported for LCMV (24) and for other viruses (25). LCMV Armstrong strain causes an acute infection in WT mice, and the virus is cleared within 7–12 days of infection; the establishment and maintenance of memory T cells occurs in the absence of any detectable viral Ag (26, 27). Some reports indicate that GKO mice and IFN-γRKO mice completely clear LCMV Armstrong infection (13, 28), but one study, using a different detection method, indicated that infectious virus was retained in GKO mice for at least 150 days postinfection (p.i.), and that, perhaps as a result of Ag persistence, the number of virus-specific CD8+ T cells present in the memory phase (>3 mo p.i.) was elevated compared with WT mice (23). Therefore, at various times after LCMV Armstrong infection of WT or IFN-γ deficient mice, RT-PCR was performed on RNA extracted from mouse spleens (Fig. 2,A). In WT mice, viral RNA was detected at 4 and 8 days post infection, but was undetectable by day 20. In contrast, viral RNA was readily detectable in GKO mice at 20 days p.i., and detectable (albeit at low level) in IFN-γRKO mice. Next, we evaluated whether these mice contained sufficient residual Ag to activate T cells. Naive P14 CD8+ T cells, which carry a TCR specific for the LCMV GP33 epitope, were labeled with CFSE and were transferred into WT or IFN-γ signaling-deficient mice that had been infected with LCMV Armstrong. As a negative control, CFSE-labeled P14 cells were transferred into uninfected mice and, as a positive control, into WT mice that were infected the following day. In all cases, recipient mice were sacrificed 9 days after cell transfer. As shown in Fig. 2,B (panel 2), the donor CD8+ T cells in day 8 WT mice had lost their CFSE signal, indicating that they had undergone extensive proliferation, as expected. However, cells transferred into WT mice in the memory phase (20 days p.i.) showed negligible proliferation when harvested at day 29 p.i., consistent with those mice having eradicated virus and virus Ag (Fig. 2,B, panel 3). In contrast, cells transferred into IFN-γRKO or GKO mice at >20 days p.i. invariably proliferated extensively, indicating that these animals retained sufficient virus Ag to drive T cell expansion (Fig. 2,B, panels 4–7). Thus, we suggest that the apparent reduction in T cell contraction that is observed in LCMV-infected IFN-γ-signaling-deficient mice (Fig. 1) is caused by the abnormal retention of viral Ag, which prevents T cell numbers from declining in the usual fashion (23). Such Ag retention also may contribute to the altered immunodominance hierarchy in memory responses that has been reported for LCMV-infected GKO mice (18, 22).
Memory cells generated in IFN-γRKO mice have functional defects
Studies in a number of chronic virus infections of humans and other animals have shown that continual exposure to high levels of persisting Ag can cause T cells to become functionally altered over time (29, 30, 31, 32). Therefore, we determined whether or not the CD8+ T cells present 50–80 days after LCMV infection of IFN-γ signaling-deficient mice were functionally normal. The first criterion was their capacity to elaborate both IFN-γ and TNF immediately after peptide stimulation, a phenotype found in the majority of normal CD8+ memory T cells (33, 34). As shown in Fig. 3,A, this was true for the three epitope-specific populations evaluated in WT mice, in which double-positive (IFN-γ+TNF+) always outnumbered single-positive (IFN-γ+TNF−) cells. In contrast, double-positive cells were in the minority in IFN-γRKO mice, often representing only ∼10–20% of the total population. Secondly, the mean quantity of TNF produced was reduced in memory cells of all epitope specificities, taken from both IFN-γKO and IFN-γRKO mice (Fig. 3,B); and reduced production of IFN-γ also was observed in CD8+ memory T cells from IFN-γRKO mice (Fig. 3,C). Thus, both the quantity (Fig. 1) and quality (Fig. 3) of virus-specific CD8+ memory T cells is altered in mice lacking IFN-γ signaling, most probably due to Ag persistence, and all data should be interpreted in this light.
IFN-γ does not directly program CD8+ T cell contraction, but dramatically enhances the development of CD8+ T cell memory
To circumvent the problem caused by LCMV Ag persistence in IFN-γ signaling-deficient mice, a dual adoptive transfer model (12, 35) was used wherein similar numbers of naive T cells of identical Ag specificity, but differing in their expression of the IFN-γR, were cotransferred into a WT host animal, which then was infected with LCMV. The recipient mice mount normal T cell responses, and eradicate the virus, thereby removing the complicating factor of Ag persistence. Using this approach, the responses of the two donor T cell populations, which differ only in their expression of the IFN-γR, can be compared in the same animal over the course of a normal antiviral immune response. P14 mice (36), which express a TCR transgene specific for LCMVGP33–41, were bred to IFN-γRKO mice to generate P14/IFN-γRKO mice (12). Equal numbers of WT (IFN-γR+) P14 cells (Thy1.2+Ly5a+) and P14/IFN-γRKO cells (Thy1.2+Ly5a−) were mixed and injected into B6.PL recipient mice (Thy1.1+). Four to seven days later, the recipient mice were infected, and the responses made by the donor cells (Thy1.2+) were followed over time in individual mice. Three time points (one each from the expansion, contraction, and memory phases) from a single representative mouse are shown in Fig. 4,A; the cumulative results from five individual recipient mice are shown in Fig. 4,B, left panel; and the ratios of IFN-γR+ to IFN-γR− P14 cells over time in these mice are shown in Fig. 4 B, right panel. Three conclusions may be drawn from these data. First, cells expressing the IFN-γR are at an advantage during the primary immune response, as we have previously shown; second, when the complicating factor of persistent Ag is avoided, CD8+ T cells lacking the IFN-γ receptor undergo contraction, indicating that this cytokine does not directly program the contraction phase of the CD8+ T cell response; and, third, the beneficial effect of IFN-γR expression is even greater in the memory phase (∼100-fold increase) than during acute infection (∼3- to 5-fold increase).
Next, we evaluated the expression of six memory-related markers, on IFN-γR+ and IFN-γRKO memory cells, before and 6 days after secondary virus infection. As shown in Fig. 4, C and D, both populations of CD8+ memory cells showed similar responses to secondary infection, suggesting that direct IFN-γ signaling does not substantially alter these qualitative criteria. Finally, the capacity of these two populations of P14 memory cells to respond to Ag was determined using intracellular cytokine staining. As shown in Fig. 4,E, both before and after secondary infection, IFN-γR+ and IFN-γRKO memory cells produced IFN-γ in response to GP33 peptide stimulation; as expected, the absolute number of responding receptor-positive cells far exceeded the number of receptor-deficient cells, but within each population the proportion of responding cells was almost identical, regardless of IFN-γR expression (Fig. 4,F). Similar conclusions were drawn for IL-2 production, although the proportion of responding cells was far lower in both populations (Fig. 4 G). Thus, IFN-γR receptor expression on CD8+ T cells greatly enhances the cells’ abundance at all stages of the antiviral immune response, but it appears to have little effect on their quality.
The effects of direct IFN-γ signaling on CD4+ T cell contraction and memory are similar to those observed for CD8+ T cells
Two LCMV CD4+ T cell epitopes have been identified in C57BL/6 mice (37), and analyses similar to those reported above were conducted to gauge the impact of direct IFN-γ signals on CD4+ T cells specific for the dominant epitope, GP61–80. WT and IFN-γRKO mice (5 per group) were infected with LCMV, and the numbers of splenic GP61–80-specific CD4+ memory T cells were measured at >50 days p.i. As shown in Fig. 5,A, comparable numbers of CD4+ memory T cells were produced in both mouse strains. However, we show above that viral Ag persists in mice deficient in IFN-γ signaling, and this Ag can drive the expansion not only of CD8+ T cells (Fig. 2), but also of CFSE-labeled SMARTA cells (data not shown). Therefore, to more accurately determine the effects of IFN-γ on CD4+ T cell contraction and memory, SMARTA mice (which express a TCR transgene specific for LCMV GP61–80 (21)) were mated to IFN-γRKO mice to generate SMARTA/IFN-γRKO mice (11). Equal numbers of WT (Ly5a+) and IFN-γRKO (Ly5a−) CD4+ SMARTA T cells were mixed and transferred into WT hosts, which were then infected with LCMV. The number of SMARTA cells in peripheral blood was determined in each animal at days 8, 18, and 46 p.i. Data from a single representative mouse show that the SMARTA cells expanded vigorously, underwent contraction, and remained detectable well into the memory phase (Fig. 5,B, top row), and cells bearing the IFN-γR were at a numerical advantage throughout (Fig. 5,B, bottom row). Accumulated data from four mice are displayed in Fig. 5,C, and suggest that CD4+ T cells lacking the IFN-γR behave in a manner similar to their CD8+ cousins: 1) the cells are less abundant during acute infection; 2) they contract well, indicating that direct IFN-γ signaling is not required; and 3) late p.i., they are outnumbered by IFN-γR+ cells, by >100:1. Next, the surface marker phenotypes and effector responses of IFN-γR+ and IFN-γRKO SMARTA memory cells were compared (Fig. 5, D–I). Recipient mice were infected with LCMV, and were subjected to further analyses when they were at least 50 days p.i. (i.e., long-term immune). Some mice were analyzed 6 days after secondary infection, while others were evaluated without reexposure to virus. Virus infection resulted in expansion of the Thy1.2+ (SMARTA) cell population (Fig. 5,D), and both IFN-γR+ cells and IFN-γRKO cells increased in number (for example, see dotplots in Fig. 5,F). The expression of four surface markers (CD44, CD62L, IL-7R, and Ly6C) were similar after reinfection, regardless of IFN-γR status (Fig. 5,E), and almost identical proportions of both populations made IFN-γ in response to in vitro peptide stimulation (Fig. 5, F and G). IL-2 production was similar (Fig. 5,H), and the mean fluorescence intensity for three cytokines varied only marginally (Fig. 5 I). Taken together, these data suggest that direct IFN-γ signaling is not a key regulator of CD4+ T cell quality. However, as is true for CD8+ T cells, direct IFN-γ signaling strongly enhances the abundance of virus-specific memory CD4+ cells, and is not required for programming the contraction phase.
To what extent is the ultimate fate of a naive T cell determined by its initial encounter with cognate Ag? Naive T cells, like most somatic cells, have built-in programming that, to some extent, dictates their responses to external stimuli; in the case of T cells, triggering of this program can cause them to undergo activation, expansion, contraction, and memory formation (1, 2). It is well established that several inflammatory cytokines can improve T cell priming, most probably by optimizing the Ag-presenting and stimulatory capacities of specialized APCs; indeed, this phenomenon is exploited by the addition of nonspecific inflammatory adjuvants to “dead” vaccines. However, it is equally clear that, after appropriate triggering by APCs, the “output” of a naive T cell’s intrinsic program, that is, the ultimate fate of the T cell, can be changed by the environment subsequently encountered by the cell and its progeny. Important questions that are now being addressed by many research groups include: 1) what molecules regulate the quantity and quality of T cell responses?; 2) do these molecules exert their effects on T cells by acting directly, indirectly, or both?; and 3) at what time(s) during T cell development do these interactions take place, for example, there is evidence that cytokine signaling during the initial Ag encounter can reprogram a naive T cell, dictating (perhaps irreversibly) some of the biological attributes of its progeny. These questions have been partially answered for the primary T cell response, during which the type I and type II IFNs act both directly and indirectly to increase T cell abundance (3, 5, 11, 12). We have found that the expression of IFN-γR-1 (the ligand-binding subunit of the receptor) increases substantially on virus-specific CD4+ and CD8+ T cells for at least 8 days p.i., implying that the cells’ sensitivity to IFN-γ signaling may increase over this time period (11, 12). However, an ex vivo evaluation of CD8+ T cell responsiveness to IFN-γ signaling suggested that the cells become resistant to cytokine effects within 24 h of Ag contact, perhaps because of changes in expression of the second (signaling) component of the IFN-γ receptor (38); this suggests that the direct effects of IFN-γ on CD8+ T cells must be imposed very early, perhaps even during the initial encounter with Ag, as has been proposed for some key aspects of CD4+ T cell differentiation (39). Herein, we have addressed some of the above questions for later events in the antiviral T cell response.
The regulation of T cell contraction is poorly understood. Following most microbial infections, ∼90–95% of the primary responder cells are lost from the spleen, and a general rule has emerged linking the size of the primary response to the size of the memory pool. However, this high level of attrition is by no means invariant. For example, DNA vaccination induces a primary CD8+ T cell response that, when compared with the response induced by infectious virus, develops more slowly and reaches a lower peak abundance; but these responses decline only very slightly over months or years postimmunization (40, 41). Therefore, the degree of T cell contraction is, to some extent, dependent on immunization protocol, leading to the important conclusion that the intrinsic CD8+ T cell contraction program is open to change. However, it is unclear whether this change results from reprogramming at the moment of Ag encounter, or from differences in the external milieu in which the T cells develop. The two possibilities are not mutually exclusive, and evidence is available in support of both. Studies in IFN-γ-deficient mice using Listeria monocytogenes and/or LCMV have suggested that, very early in infection, IFN-γ alters the intrinsic program of naive T cells, dictating the extent to which their progeny contract several days later (16, 42, 43), and more recent studies have shown that memory cell generation following dendritic cell immunization can be delayed by the coadministration of proinflammatory molecules (19). Taken together, these findings suggest that IFN-γ signals, delivered to T cells early in infection, may redefine the cells’ intrinsic contraction program (17). However, if this very early reprogramming occurs, it is clear that its outcome can be modified by subsequent events; some time ago, it was shown that T cell contraction and the size of the memory pool could be altered by the administration of IL-2 from days 8–15 p.i., long after the initial Ag encounter (44). We show here that profound T cell contraction occurs in CD4+ and CD8+ T cells that are incapable of receiving direct IFN-γ signals. This argues strongly against the hypothesis that IFN-γ directly modifies T cell contraction, either by reprogramming the cell early in infection, or by triggering apoptosis at a later time. We cannot exclude the possibility that IFN-γ plays an indirect role in regulating contraction, perhaps acting via its effects on APCs; such effects have recently been inferred from studies of peptide-immunized mice (20). Furthermore, the effects of IFN-γ on T cell responses may vary depending on the nature of the infecting organism; others have reported normal CD8+ T cell kinetics during vesicular stomatitis virus infection of IFN-γKO mice (45).
Direct IFN-γ signaling has a profound effect on the abundance of CD4+ and CD8+ memory T cells following LCMV infection (Figs. 4 and 5). When and how does this occur? Are the effects on memory imprinted early in infection, or later? We think that it is more likely to take place later in the immune response than early, for several reasons. Memory T cells are not a single homogeneous population, and can differ by several criteria, including anatomical location, effector function, and surface marker phenotype (46). Although it is possible that such intricacies are imprinted at the moment of Ag contact, it seems to us more likely that these qualities are shaped by later events. Indeed, it would be extraordinarily surprising to us if T cell contraction and memory, both of which are key components of the antimicrobial T cell response, were irrevocably set at, or immediately after, the initial Ag encounter; such inflexibility would surely reduce the biological effectiveness of the adaptive immune response. Furthermore, there is compelling, albeit indirect, evidence favoring later effects of IFN-γ. This cytokine is secreted by cells of the innate immune system, including NK cells, and APCs also may make IFN-γ (47). After LCMV infection, elevated levels of IFN-γ are produced by days 3–4 during the innate immune response (3), and the ability of T cells to produce IFN-γ improves throughout the induction phase of the response, resulting in activated T cells that make robust amounts of IFN-γ more rapidly and with less stimulation than cells found earlier in the response (48, 49). Our laboratory has shown that the rapidity with which virus-specific CD8+ T cells initiate IFN-γ synthesis is directly related to their abundance during the primary response (i.e., to their immunodominance), and that only rapid-onset cells enter the memory phase (7). Consistent with this concept, IFN-γR+ cells begin to outnumber IFN-γRKO cells as early as 5 days p.i. (J. K. Whitmire, N. Benning, and J. L. Whitton, unpublished study), and this difference increases dramatically over the course of infection (Figs. 4 and 5). Moreover, changes in cytokine production continue to occur long after resolution of the infection (7, 32, 50), yielding mature memory T cells that coexpress IFN-γ and TNF and make IL-2; gene-chip analyses of early, activated, and memory T cells in a model of acute virus infection show that genes whose expression is regulated by IFN-γ were preferentially induced among surviving memory CD8 cells (50), also consistent with the idea that IFN-γ continues to act on T cells later in the response. IFN-γ signals may also regulate leukocyte movement, because it induces several chemokines (including RANTES, MIP1α and MIP1β, MCP-1, and IP-10) that affect T cell recruitment to sites of infection; these effects, too, are more likely to be exerted later in the immune response. Perhaps direct IFN-γ signaling improves the homeostatic maintenance of CD4+ and CD8+ memory T cells.
In conclusion, we confirm published work (23) showing that IFN-γ is required to ensure the complete eradication of virus/viral Ag; and that, in the absence of this clearance, the contraction of virus-specific T cells (CD4+ and CD8+) cannot proceed normally. However, our studies using WT mice indicate that IFN-γ does not directly program T cell contraction, and that the direct effects of this key antiviral effector molecule greatly enhance the establishment of virus-specific CD4+ and CD8+ T cell memory following LCMV infection.
We are grateful to Annette Lord for excellent secretarial support. This is manuscript number 18168-MIND from TSRI.
The authors have no financial conflict of interest.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
This work was supported by National Institutes of Health Grants AI-27028 and AI-52351.
Abbreviations used in this paper: LCMV, lymphocytic choriomeningitis virus; TSRI, The Scripps Research Institute; Tg, transgenic; WT, wild type; GKO mice, mice lacking IFN-γ; p.i., postinfection.