Persistence of Memory CD8 T Cells in MHC Class I-Deficient Mice

See allHide authors and affiliations

Science  12 Nov 1999:
Vol. 286, Issue 5443, pp. 1377-1381
DOI: 10.1126/science.286.5443.1377


An understanding of how T cell memory is maintained is crucial for the rational design of vaccines. Memory T cells were shown to persist indefinitely in major histocompatibility complex (MHC) class I–deficient mice and retained the ability to make rapid cytokine responses upon reencounter with antigen. In addition, memory CD8 T cells, unlike naı̈ve cells, divided without MHC–T cell receptor interactions. This “homeostatic” proliferation is likely to be important in maintaining memory T cell numbers in the periphery. Thus, after naı̈ve CD8 T cells differentiate into memory cells, they evolve an MHC class I–independent “life-style” and do not require further stimulation with specific or cross-reactive antigen for their maintenance.

Immunological memory is the ability of the immune system to respond with greater vigor upon reencounter with the same pathogen. Many currently used vaccines and most natural infections induce long-term T cell memory as assessed by rapid anamnestic responses (1–3). These accelerated recall responses are due to increased numbers of antigen-specific T cells and to qualitative changes in memory T cells that allow them to develop into effector cells more rapidly than naı̈ve T cells (2–4). There has been considerable interest in determining whether continued presentation of specific or cross-reactive antigens by MHC molecules is necessary for maintenance of memory T cells and for retaining their “response ready” mode (1–6). We have addressed this issue by analyzing the survival, proliferation, and functional characteristics of memory CD8 T cells under conditions of MHC class I deficiency.

Adult mice resolve an acute lymphocytic choriomeningitis virus (LCMV) infection within 2 weeks and then exhibit long-term CD8 T cell memory (2, 3). To determine the requirement of MHC class I molecules in maintaining CD8 memory, we obtained fluorescence-activated cell sorter (FACS)–purified CD8 T cells (>99% pure) from Thy1.1+ LCMV immune mice and transferred the cells into either MHC class I–positive (β2M+/+) or MHC class I– deficient (β2M−/−) congeneic Thy1.2+mice. Before adoptive transfer, the FACS-purified CD8 T cells were tested for the presence of virus and found to be free of any detectable viral material (7). The transferred CD8 T cells initially expanded in the irradiated recipient mice, reaching a plateau at around day 20, and were then maintained indefinitely in both β2M+/+ and β2M−/−mice (Fig. 1A). Although the initial expansion of CD8 T cells was less in β2M−/−mice, by day 22 the numbers were similar in both groups of mice. The LCMV-specific memory CD8 T cells were CD44hi, Ly6Chi, CD25lo, and CD69lo before transfer, and this phenotype was retained in MHC class I–deficient mice. The total Thy1.1+ CD8 T cell population initially consisted of equal numbers of CD44lo (naı̈ve) and CD44hi (memory) cells. However, by day 50 nearly all of the donor Thy1.1+ CD8 T cells were CD44hi, suggesting preferential survival of memory cells in β2M−/− mice (8).

Figure 1

Maintenance of memory CD8 T cells in MHC class I–deficient (β2M−/−) recipients. (A) FACS-purified Thy 1.1+ CD8 T cells (1 × 106) containing 9 × 104 LCMV-specific memory CD8 cells were isolated from LCMV immune mice and transferred into irradiated [5.5 gray (Gy)] Thy1.2+β2M+/+ (•) and β2M−/−(▵) recipients (21). Data show the persistence of LCMV-specific memory CD8 (top) and total Thy1.1+ CD8 (bottom) cells in both the recipients. (Top) Data represent total LCMV-specific memory CD8 T cells functionally responding to all of the five known epitopes by making IFN-γ (2). (B) Cycling of antigen-specific memory CD8 T cells in β2M−/− and β2M+/+ mice as determined by BrdU incorporation. Seventy days after adoptive transfer of CD8 T cells, recipient mice were given BrdU in their drinking water for 1 week before analysis (2, 9). BrdU staining is shown in MHC tetramers DbNP396- and DbG33-positive cells (top) and total Thy1.1+ CD8 T cells (bottom). (C) Maintenance of MHC class I–deficient memory CD8 T cells in class I–deficient recipients. Mixed bone marrow chimeras were made by injecting 6 × 106β2M−/− and 2 × 106β2M+/+ bone marrow cells into natural killer (NK) cell–depleted and lethally irradiated (8.5 Gy) C57BL/6 mice (22). After 1 month the chimeras were immunized with LCMV and rested for 6 months. The frequencies of LCMV-specific memory cells among MHC class I and class I+ CD8 cells in these immune chimeras were 1/15 and 1/14, respectively. FACS-sorted MHC class I–deficient CD8 T cells (5 × 105, left) (containing 3.4 × 104 LCMV-specific memory cells) were transferred into nonirradiated β2M−/−and β2M+/+ recipients. β2M+/+ recipients were treated with antibody to NK1.1 to prevent NK-mediated rejection of class I–deficient donors (22). (Right) Persistence of class I–deficient memory CD8 T cells among recipients. CD8 T cells from recipient spleens were enriched by negative selection by using mouse CD8 subset column kit (R&D Systems, Minneapolis, Minnesota), and LCMV-specific memory CD8 frequency among enriched CD8 was determined by ELISPOT by stimulating with a mixture of five known LCMV CD8 epitope peptides (2).

Peripheral T cells undergo homeostatic proliferation under lymphopenic conditions to reconstitute the empty immune system (5). When lymphoid homeostasis is reached, naı̈ve cells cease to proliferate, whereas memory cells continue to divide, although at a much slower rate, in order to maintain the pool of memory cells (9). We sought to determine whether memory CD8 cells persisting in MHC class I–deficient mice at relatively constant numbers after reaching lymphoid homeostasis (between day 22 and day 310) (Fig. 1A) were undergoing proliferation or surviving without any cell division. The turnover rate of LCMV-specific memory CD8 T cells (identified by staining with MHC class I tetramers) was similar in β2M+/+ and β2M−/−mice (Fig. 1B). Likewise, the total population of donor Thy1.1+ CD8 T cells, which consist predominantly of non-LCMV–specific CD44hi CD8 T cells, proliferated equally well in +/+ and β2M−/− mice. Thus, not only do memory CD8 T cells persist in β2M−/−mice, but their homeostatic proliferation is not compromised in a heavily MHC class I–deficient environment.

In the experiments described above the donor CD8 T cells were derived from β2M+/+ mice. It is unlikely that survival of memory CD8 T cells in β2M−/−mice was due to the transferred CD8 T cells seeing MHC class I on each other because the transferred cells comprised only about 0.5 to 1% of the total splenocytes [∼5 × 105Thy1.1+ CD8 T cells in (5 to 10) × 107β2M−/− spleen cells]. Nevertheless, to test this possibility, we performed adoptive transfer experiments with MHC class I–deficient CD8 T cells obtained from chimeric mice (β2M−/− bone marrow transferred to irradiated β2M+/+ mice) that had been immunized with LCMV. MHC class I–deficient memory CD8 T cells were able to persist in β2M−/− mice (Fig. 1C). The β2M−/− recipient mice were not irradiated before cell transfer because their “own” cells were transferred back into them. Thus, MHC class I– deficient memory CD8 T cells survive in β2M−/− mice, and persistence of memory cells is seen after adoptive transfer into “full” (that is, nonirradiated) mice.

As a more stringent test for survival of memory T cells in the absence of MHC class I molecules, we next used Db−/−× Kb−/− × β2M−/− mice as recipients (10). LCMV-specific memory CD8 T cells (as well as total CD8 T cells) persisted indefinitely (>240 days) in Db−/− × Kb−/− × β2M−/− recipients. Similar to results seen with β2M−/− mice (Fig. 1A), the initial expansion of LCMV-specific memory CD8 T cells was less in Db−/− × Kb−/− × β2M−/− mice, but by day 20 there were similar numbers of memory CD8 T cells in both +/+ and Db−/− × Kb−/− × β2M−/− mice. Figure 2B shows that LCMV-specific CD8 T cells can be identified by using MHC class I tetramers, and Fig. 2C documents proliferation of antigen-specific memory CD8 T cells and proliferation of total donor Thy1.1+ CD8 T cells in Db−/− × Kb−/− × β2M−/− mice. Taken together, these results show that even in mice completely lacking the classical MHC class I molecules (Kb and Db) and β2M, memory CD8 T cells can proliferate and survive for extended periods.

Figure 2

Persistence of memory CD8 T cells in Db−/− × Kb−/− × β2M−/− recipients. (A) Maintenance of antigen-specific memory CD8 T cells (top) and Thy1.1+ CD8 T cells (bottom) in irradiated Db−/− × Kb−/− × β2M−/− recipients. Adoptive transfers were done as described (Fig. 1A), with the exception that in most of these mice a mixture of sorted CD8 and CD4 [(7 to 8) × 105each] cells from LCMV immune mice were transferred to control for possible rejection. In some experiments only sorted CD8 T cells were transferred into Db−/− × Kb−/−× β2M−/− mice to ensure that memory CD8 T cells could persist on their own (as seen in Fig. 1) after transfer into β2M−/− mice. (B) Visualization of antigen-specific CD8 T cells in +/+ and Db−/− × Kb−/− × β2M−/− recipients at 100 days after transfer. The numbers indicate the percentage of Thy1.1+CD8 that stain with the indicated MHC class I tetramer. (C) Cycling of antigen-specific (MHC tetramer DbGP33+), Thy1.1+, and endogenous CD8 cells in +/+ (top) and Db−/− × Kb−/− × β2M−/− (bottom) recipients at 240 days after transfer. Three mice in each group were given BrdU-containing water for 1 week before analysis, and their spleens were pooled and analyzed. Dotted line represents staining of the same cells with isotype control for anti-BrdU. (D) Rapid functional response to different epitopes by memory CD8 T cells obtained from +/+ and Db−/− × Kb−/− × β2M−/− recipients at 100 days after transfer. Spleen cells from +/+ and Db−/− × Kb−/− × β2M−/−recipients were stimulated with different peptides as described (21). Data represent events gated on Thy1.1+ cells. Numbers indicate the percentage of Thy1.1+ CD8 that score positive for intracellular IFN-γ.

Db−/− × Kb−/− × β2M−/− mice contain a few (∼20 times lower than in +/+ mice) of their own CD8 T cells in the periphery (11), all of which are CD44hi (8). We compared the turnover rate of these endogenous MHC class I– negative CD44hi CD8 T cells with the cycling of the transferred memory CD8 T cells (Fig. 2C). The transferred memory CD8 T cells exhibited a pattern of cycling in Db−/− × Kb−/− × β2M−/− mice [17 to 18% bromodeoxyuridine-positive (BrdU+) cells] similar to that in +/+ mice (17 to 20% BrdU+ cells), whereas the endogenous CD8 T cells of the −/− mice showed a distinctive pattern of hyperproliferation (61% BrdU+ cells), with a large proportion of blasting cells (Fig. 2C) (12). These results show that proliferation of each of these populations is independently regulated, and that the cycling pattern of the “true” antigen-specific memory cells (generated in +/+ mice) remains the same in the presence or absence of MHC class I molecules.

A hallmark of memory T cells is their ability to exhibit rapid functional responses after reexposure to antigen (2,4). LCMV-specific memory CD8 T cells make interferon-γ (IFN-γ) within 4 hours after stimulation with the cognate peptide (2). In contrast, naı̈ve CD8 T cells make IFN-γ after more than 24 hours, yet produce substantially lower amounts (2, 4, 8). We therefore examined whether memory CD8 T cells persisting in Db−/− × Kb−/− × β2M−/− mice retain the ability to make rapid cytokine responses after reexposure to antigen and MHC class I, or revert to a naı̈ve T cell phenotype. We found that this functional characteristic of memory CD8 T cells was retained indefinitely after transfer into Db−/− × Kb−/− × β2M−/− mice (Fig. 2D). The responsiveness of memory CD8 T cells transferred into −/− mice is identical to that of memory cells transferred into +/+ mice. Thus, after memory T cells are generated, they no longer need contact with MHC class I molecules to maintain their “response ready” mode. In addition, memory responses to multiple epitopes are maintained in the absence of MHC class I, and the epitope hierarchy is the same in +/+ and −/− mice (Fig. 2D).

Naı̈ve CD8 and CD4 T cells require contact with MHC class I and class II molecules, respectively, for their maintenance in the periphery (5, 6). Because our results (Figs. 1 and 2) showed that memory CD8 T cells can persist in MHC class I −/− mice, we directly compared MHC class I requirements of naı̈ve and memory cells. Two series of experiments were done: one with monoclonal naı̈ve and memory CD8 T cells expressing the T cell receptor (TCR) specific for LCMV peptide GP 33-41 (13), and the other with polyclonal CD44lo(naı̈ve) and CD44hi (memory) CD8 T cells. Naı̈ve transgenic CD8 T cells were unable to persist after transfer into Db−/− × Kb−/− × β2M−/− mice (Fig. 3A, bottom), and by day 40 there were 50- to 100-fold fewer cells in −/− mice compared with +/+ mice. T cells expand after transfer into lymphopenic hosts (that is, irradiated, severe combined immunodeficient or RAG −/− mice) (5,14). This “emptiness-induced” proliferation of naı̈ve CD8 T cells was dependent on MHC class I molecules (Fig. 3A, top); although some proliferation was observed in −/− mice at day 5 and 9, this was considerably less than that seen in +/+ mice and was not sustained. Our data also suggest that survival of naı̈ve T cells is affected in the absence of MHC class I; a precipitous decline in cell numbers was observed between day 17 and 40. Maintenance is the sum of proliferation and survival, and taken together our results show that naı̈ve CD8 T cells require interaction with MHC class I molecules not only for their proliferation after transfer into lymphopenic mice, but also for their long-term survival. In marked contrast to the behavior of naı̈ve CD8 T cells, memory transgenic CD8 T cells persisted in −/− mice, and at day 40 there was less than a twofold difference in the number of memory cells in +/+ versus −/− mice (Fig. 3B). In addition, the proliferation of memory CD8 T cells after transfer into empty mice was only minimally affected, and by day 17 all of the memory cells had divided multiple times in Db−/− × Kb−/− × β2M−/− mice (Fig. 3B). A similar pattern was seen with polyclonal CD44lo (naı̈ve) and CD44hi (memory) CD8 T cells isolated from uninfected mice; proliferation of CD44lo cells was compromised in −/− mice, whereas proliferation of CD44hicells was independent of MHC class I molecules (Fig. 4). Memory cells divided at a faster rate than naı̈ve cells even in +/+ mice.

Figure 3

Maintenance of memory but not naı̈ve transgenic CD8 T cells in Db−/− × Kb−/− × β2M−/−recipients. (A) CD44lo CD8 T cells from naı̈ve P-14 transgenic mice expressing the TCR specific to LCMV peptide GP 33-41 were sorted, labeled with fluorescent dye CFSE (23), and transferred into irradiated recipients. The transferred population contained >95% transgenic cells as assessed by MHC tetramer DbGP33 staining, anti-Vα2 and anti-Vβ8.1 staining, or both. (Bottom) The total number of MHC tetramer DbGP33+ CD8 T cells (average of two to four mice at each time point) and (top) the number of cell divisions as seen by fluorescent dye dilution in +/+ and Db−/− × Kb−/− × β2M−/− (−/−) recipients at various time points after transfer. (B) Memory P-14 transgenic CD8 T cells were made as follows: Fifty thousand CD8 T cells from naı̈ve P-14 transgenic mice were transferred intravenously into C57BL/6 mice. After 2 days mice were immunized with 2 × 105 PFU of LCMV and rested for more than 6 months. Forty percent of CD44hi CD8 cells from these immune mice were positive for MHC tetramer Db GP 33-41, Vα2, and Vβ8.1. These cells also exhibit other memory cell characteristics described in (21). CD8 CD44hicells from these immune transgenic chimeras at 6 months after transfer were sorted and transferred into +/+ and Db−/− × Kb−/− × β2M−/−recipients as described above. Data represent the total number of tetramer-positive memory CD8 T cells at various days after transfer (left) (average of two to four mice) and the number of divisions among tetramer-positive cells on day 17 after transfer (right).

Figure 4

Naı̈ve (CD44lo) and memory (CD44hi) polyclonal CD8 T cells from uninfected mice behave similarly to naı̈ve and memory CD8 T cells of known specificity. CD44lo and CD44hi CD8 T cells from Thy1.1+ donors were sorted, labeled with CFSE, and transferred into irradiated +/+, Db−/− × Kb−/− × β2M−/−, and β2M−/− recipients. The CFSE fluorescence levels among Thy1.1+ CD8 T cells are shown on day 5 after transfer. The percentage of cells at each division number is shown in the table.

In agreement with our findings, two recent studies with TCR transgenic CD8 T cells specific for the HY antigen showed that specific antigen is not necessary for memory maintenance. However, these studies reported that MHC class I molecules play a role in maintaining memory CD8 T cells (6). This conclusion differs from our findings and may reflect a unique characteristic of HY-specific T cells, differences in the experimental design, or both (15). It should be emphasized, however, that we have used the most extensive and stringent protocols for examining the requirement of MHC class I molecules in maintaining CD8 T cell memory. We have monitored the survival of several different populations of memory cells, including monoclonal and polyclonal antigen-specific CD8 T cells and total CD8 CD44hi cells from both uninfected and immunized mice. Furthermore, in one series of experiments we used MHC class I–deficient memory CD8 T cells as donors. As recipients we used both irradiated and nonirradiated MHC class I–deficient mice. All of these experiments gave the same result, that is, that maintenance of CD8 T cell memory is independent of MHC class I molecules. This finding is in agreement with the observations of Swain et al.(16), who documented the persistence of memory CD4 T cells in MHC class II−/− mice. Thus, it appears that MHC independence is a property of both CD8 and CD4 memory.

The finding that memory CD8 T cells can maintain their functional characteristics and undergo proliferative renewal in the absence of MHC class I molecules does not necessarily rule out a role for the TCR in cell survival and proliferation. It is possible that intrinsic constitutive signals from the TCR may be necessary for the maintenance of T cells (17). Recent studies with the cre-lox system have shown that the B cell receptor (immunoglobulin) is required for survival of B cells (18). Similar studies with T cells will be necessary to directly assess the contribution of the TCR.

Why do naı̈ve CD8 T cells differ from memory CD8 T cells in terms of their requirements for MHC class I molecules? It is conceivable that the topology of the signaling machinery (for example, TCR, CD8 coreceptor, adhesion molecules, and kinases) is sufficiently different between naı̈ve and memory cells such that memory cells can still receive the necessary signals in the absence of contact with MHC, whereas naı̈ve cells require MHC interaction for their maintenance. Recent studies have shown changes in the organization of molecules at the cell surface, as well as changes in the chromatin organization and gene expression pattern, of memory compared with naı̈ve cells (19). A combination of these changes may be required to generate a memory cell that is independent of MHC class I.

  • * Present address: Department of Microbiology, University of Pennsylvania, Philadelphia, PA 19104, USA.

  • To whom correspondence should be addressed. E-mail: ra{at}


Stay Connected to Science

Navigate This Article