Division-linked generation of death-intermediates regulates the numerical stability of memory CD8 T cells

Abstract

Infection or successful vaccination results in the formation of long-lived memory CD8 T-cell populations. Despite their numerical stability, memory CD8 T-cell populations are thought to completely turn over through proliferation within a 2- to 3-mo period. Therefore, steady-state memory cell proliferation must be balanced by a precisely regulated and equivalent death rate. However, the mechanisms regulating this balancing process remain completely undefined. Herein, we provide evidence for “death-intermediate memory cells” (T_DIM) within memory CD8 T-cell populations generated by infection. Importantly, CD62L^Lo/CD27^Lo T_DIMs are functionally characterized by an inability to produce cytokines, the failure to internalize T-cell receptor following antigenic stimulation, and signatures of apoptotic death. Furthermore, we demonstrate that, mechanistically, T_DIM are directly generated from dividing “central memory” T-cell populations undergoing memory turnover in vivo. Collectively, these results demonstrate that as central memory CD8 T cells proliferate, they continuously generate a population of CD8 T cells that are nonfunctional and apoptotic; thus, our data support a model wherein division-linked generation of T_DIM contributes to numerically stable CD8 T-cell memory.

Following infection or successful vaccination, naive CD8 T cells undergo a programmed series of biological events that ultimately result in the formation of long-lived memory populations (1–4). Independent of further antigenic stimulation (5, 6), memory CD8 T cells undergo modest rates of cellular proliferation driven by cytokines, such as IL-7 and IL-15 (7–12), and this “memory turnover” is critically important for maintaining stable memory CD8 T-cell numbers (1, 11). Despite their numerical stability, memory CD8 T-cell populations are thought to completely turn over through proliferation within a 2- to 3-mo period (13, 14). Thus, memory cell proliferation must be balanced by a precisely regulated and equivalent death rate. Although the relevance of stable memory maintenance for vaccination and immunity to infection is clear, the mechanisms and cellular intermediates in this balancing process remain completely undefined.

Multiple studies of memory CD8 T cells reveal that some cells within the population proliferate one or more times over an interval of several weeks (4, 13, 14). From a reductionist perspective, it is apparent that two basic models might account for the balancing act between proliferation and death that must occur to maintain stable CD8 T-cell numbers. In the first model (Fig. 1A, Left), one daughter cell generated by cytokine-driven division of a memory CD8 T cell lives, whereas the other daughter cell is destined to die. In the second model (Fig. 1A, Right), both daughter cells generated by division of a memory CD8 T cell live and one nondividing memory cell is destined to die. Thus, to begin to discriminate between these models, we sought to identify the CD8 T cells that are destined to die within a memory CD8 T-cell population.

Fig. 1.

Identification of functionally impaired CD8 T cells within a population of TCR-tg memory CD8 T cells. (A) Proposed models for the numerical stability of memory CD8 T-cell populations during steady-state proliferation. (B) B6 or BALB/c mice (Thy1.2) that had received a physiological number (500–5,000) of the indicated TCR-tg CD8 T cells (Thy1.1) were analyzed for TCR-tg T-cell production of IFN-γ following peptide stimulation. Analysis was performed at least 80 d following the indicated infection. [B6, P14 CD8 T cells, LCMV–Armstrong infection (Top); B6, OT-I CD8 T cells, Listeria monocytogenes (LM)-OVA infection (Middle); BALB/c, CS₂₈₀-tg, LM-CS infection (Bottom)]. (C) Expression of CD62L and CD27 on total Thy1.1 memory P14 CD8 T cells 120 d after LCMV–Armstrong infection. (D) Ex vivo stimulation with GP_33–41 peptide was performed as in B, with the exception that N-(R)-(2-(Hydroxyaminocarbonyl)methyl)-4-methylpentanoyl-l-t-butyl-glycine-l-alanine 2-aminoethyl amide was added to inhibit CD62L cleavage. Expression of CD62L and CD27 was then determined on IFN-γ⁺ and IFN-γ⁻ memory P14 CD8 T cells. (E) Same as in D, with the exception that cells were analyzed for their ability to degranulate through surface expression of CD107a. (F) Same as in D, with the exception that surface expression of Vα2 or Vβ8.1 TCR levels on P14 CD8 T cells was determined following ex vivo stimulation with GP_33–41 peptide. (G) CD62L^Hi/CD27^Hi T_CM and CD62L^Lo/CD27^Lo memory P14 CD8 T cells were purified by FACS from spleens of LCMV-infected animals and transferred into naive B6 recipients. Secondary expansion of the transferred cell populations was analyzed in the blood on day 7 postinfection with LCMV. PBMC, peripheral blood mononuclear cell. (H) Production of IFN-γ was analyzed in T_CM and CD62L^Lo/CD27^Lo memory P14 CD8 T cells following stimulation with PMA and ionomycin (Iono). (I) Analysis of apoptosis of T_CM and CD62L^Lo/CD27^Lo memory P14 CD8 T cells by activation of caspase-3/7, Annexin V staining, and TUNEL. All data are representative of multiple mice analyzed in independent experiments performed three or more times.

Results

Nonfunctional Apoptotic Cells Are Found Within Memory CD8 T-Cell Populations.

Memory CD8 T cells can rapidly produce a variety of cytokines, such as IFN-γ and TNF-α, following T-cell receptor (TCR) stimulation. In fact, production of IFN-γ following stimulation with peptide is often used to identify antigen-specific CD8 T cells following infection or vaccination (15). However, when splenic memory P14 TCR-transgenic (tg) CD8 T cells (Thy1.1⁺, Vα2⁺Vβ8.1⁺) that were generated following lymphocytic choriomeningitis virus (LCMV) infection were stimulated with saturating amounts of GP_33–41 peptide (Fig. S1), a substantial proportion of these cells (∼20–25%) failed to produce IFN-γ (Fig. 1B). This was not attributable to incomplete recruitment of the naive P14 CD8 T cells into the immune response or contamination with Thy1.1⁺ cells expressing incorrect TCR, because low numbers of P14 cells were initially transferred, high numbers were recovered at memory time points, and all the Thy1.1⁺ cells expressed the appropriate TCR (Vα2/Vβ8.1) (Fig. S1). We also found a similar percentage of memory CD8 T cells that failed to produce IFN-γ within two other TCR-tg T-cell populations (OT-I and CS₂₈₀-tg in C57BL/6 and BALB/c mice, respectively) following infection with Listeria monocytogenes expressing either the ovalbumin epitope, OVA_257–265, or the Plasmodium yoelii circumsporozoite protein epitope, CS_280–288, respectively (Fig. 1B). Costaining revealed that these IFN-γ⁻ populations also failed to produce TNF-α and IL-2 (Fig. S2). Thus, the failure of a substantial proportion of cells to produce IFN-γ following peptide stimulation is a general feature of memory TCR-tg CD8 T-cell populations.

We next determined whether memory CD8 T cells that failed to produce IFN-γ exhibited a unique phenotype. Indeed, surface marker analysis revealed that >90% of IFN-γ–nonproducing memory P14 cells expressed low levels of both CD62L and CD27 (Fig. 1 C and D). Of note, the CD62L^Lo/CD27^Lo phenotype does not conclusively identify only cytokine nonproducing memory CD8 T cells, especially at early time points (i.e., day 40) following infection, but identifies the majority of these cells by day 150 postinfection with LCMV (Fig. S3). In addition to the failure to produce IFN-γ and the CD62L^Lo/CD27^Lo phenotype, this cell population was unable to undergo degranulation or internalize TCR following peptide stimulation (Fig. 1 E and F). Furthermore, this CD62L^Lo/CD27^Lo memory CD8 T-cell population did not undergo robust secondary expansion in vivo following LCMV infection, compared with CD62L^Hi/CD27^Hi “central memory” T cells (T_CM) [the dominant memory cell subset found at late time points (16)] (Fig. 1G). Therefore, these data demonstrate that the majority of CD62L^Lo/CD27^Lo memory CD8 T cells are nonfunctional at late time points following infection.

Failure of this cytokine “nonproducing” memory CD8 T-cell population to internalize TCR following antigenic stimulation suggested the possibility that insufficient proximal TCR signaling resulted in the inability to produce IFN-γ. However, although stimulation with phorbol 12-myristate 13-acetone (PMA)/ionomycin resulted in IFN-γ production by >97% of T_CM CD8 T cells, only a small fraction of CD62L^Lo/CD27^Lo memory cells made detectable cytokine, suggesting that the inability to produce cytokine was not attributable to impaired proximal TCR signaling (Fig. 1H). Because this unique cell population failed to exhibit major functional characteristics of memory CD8 T cells (cytokine production, degranulation, and secondary responses), we next determined whether the CD62L^Lo/CD27^Lo memory cells exhibited hallmark features of apoptosis. Indeed, compared with T_CM in the same samples, CD62L^Lo/CD27^Lo memory CD8 T cells displayed substantially more activated caspase-3 and caspase-7, surface expression of phosphatidylserine, and DNA fragmentation (Fig. 1I). Collectively, these data suggest that the CD62L^Lo/CD27^Lo memory CD8 T cells that cannot produce IFN-γ are highly susceptible to apoptosis.

As shown in Fig. 1F, this unique cell population fails to internalize TCR; thus, it can still be detected using H2-D^b-GP₃₃ tetramer staining following peptide stimulation (Fig. S4A). Indeed, following stimulation with GP_33–41 peptide, memory P14 CD8 T cells that remained tetramer-positive were CD62L^Lo/CD27^Lo, did not make IFN-γ, and did not internalize TCR (Fig. S4 B–D). Thus, we used this approach to determine whether a population of nonfunctional apoptotic memory CD8 T cells was present in an endogenous memory population. We infected B6 mice with LCMV; at a memory time point, we stimulated their splenocytes with saturating amounts of GP_33–41 peptide. This resulted in robust IFN-γ production and a substantial decrease in the percentage of CD8 T cells that could be detected with H2-D^b-GP₃₃ tetramer (Fig. 2A). Strikingly, cells that failed to internalize TCR (remained tetramer-positive) following antigen stimulation exhibited a uniform CD62L^Lo/CD27^Lo phenotype identical to that observed in the cytokine nonproducing TCR-tg memory CD8 T-cell populations (Fig. 2 A and B). In contrast, the T_CM population could no longer be detected, suggesting that this cell population efficiently internalized TCR following peptide stimulation. Similar results were found for memory populations specific for the LCMV NP₃₉₆ and GP₂₇₆ epitopes (Fig. 2B). In addition, a large percentage of CD62L^Lo/CD27^Lo endogenous memory CD8 T cells displayed activation of caspase-3 and caspase-7 (Fig. 2C). Thus, these functionally impaired apoptotic CD62L^Lo/CD27^Lo CD8 T cells are found in both TCR-tg and endogenous memory CD8 T-cell populations.

Fig. 2.

Functionally impaired CD62L^Lo/CD27^Lo cells can be detected in the endogenous repertoire of memory CD8 T-cell populations. (A) IFN-γ production and the ability to bind H2-D^b-GP₃₃ tetramer were determined following stimulation with GP_33–41 peptide in day 150 memory CD8 T cells from LCMV-infected B6 mice. Expression of CD62L and CD27 was then analyzed on H2-D^b-GP₃₃ tetramer-positive CD8 T cells. (B) Quantification of the percentage of H2-D^b-GP₃₃, H2-D^b-NP₃₉₆, and H2-D^b-GP₂₇₆ tetramer-positive CD8 T cells that expressed CD62L and CD27 following no stimulation or stimulation with the corresponding peptide. (C) Analysis of activation of caspase-3/7 on LCMV-specific tetramer-positive cells. Data are representative of two independent experiments.

Memory CD8 T Cells That Cannot Produce IFN-γ Are Phenotypically and Functionally Distinct from Effector Memory Cells.

“Effector” and “effector memory” CD8 T cells (T_EM) do not express CD62L and are often CD27^Lo (16–20). In fact, as shown in Fig. S3, a substantial proportion of CD62L^Lo/CD27^Lo memory CD8 T cells produce IFN-γ, especially at early memory time points following infection. Therefore, we next determined whether there were additional phenotypic differences that could discriminate between the CD62L^Lo/CD27^Lo populations based on their ability to produce cytokines following stimulation with peptide (Fig. 3A). Low expression of CD127 and high expression of KLRG1 are often used to identify effector CD8 T-cell populations following infection (21–23). Indeed, CD62L^Lo/CD27^Lo memory CD8 T cells that produced IFN-γ following stimulation with peptide expressed high levels of KLRG1 and low levels of CD127 (Fig. 3B). In contrast, CD62L^Lo/CD27^Lo memory CD8 T cells that did not produce IFN-γ exhibited the opposite phenotype and expressed low levels of KLRG1 and high levels of CD127, similar to T_CM (Fig. 3B). Thus, these data suggest that memory CD8 T cells that cannot produce IFN-γ are CD62L^Lo/CD27^Lo/CD127^Hi/KLRG1^Lo, demonstrating that they are phenotypically distinct from effector and T_EM CD8 T cells.

Fig. 3.

Functionally impaired CD62L^Lo/CD27^Lo memory CD8 T cells are found in lymphoid organs and are phenotypically distinct from “effector memory.” (A) 5 × 10³ naive P14 CD8 T cells were transferred into B6 mice and subsequently infected with LCMV–Armstrong. On day 50 postinfection, CD62L^Lo/CD27^Lo memory P14 CD8 T cells from the spleen were analyzed for the ability to produce IFN-γ following GP_33–41 stimulation. (B) Using the gating strategy shown in A, expression of CD127 (Left) and KLRG1 (Right) was analyzed on IFN-γ⁺ and IFN-γ⁻ cells within the CD62L^Lo/CD27^Lo population. (C) Same as in A, with the exception that memory P14 CD8 T cells were analyzed from the indicated tissue. The percentage of IFN-γ⁻ cells following GP_33–41 stimulation within the CD62L^Lo/CD27^Lo population was quantified. Representative dot plots are shown in Fig. S5. Data are representative of two independent experiments.

As a result of changes in homing receptors, T_CM and T_EM exhibit different localization patterns within host tissues. To address the localization of the cytokine nonproducing memory population, we analyzed the ability for the CD62L^Lo/CD27^Lo memory P14 CD8 T cells from blood, lymphoid organs, and peripheral tissues to produce IFN-γ. On day 60 postinfection with LCMV, ∼75% of the CD62L^Lo/CD27^Lo memory CD8 T-cell population in the spleen did not make IFN-γ following stimulation with GP₃₃ peptide (Fig. 3C and Fig. S5). Furthermore, nearly all the CD62L^Lo/CD27^Lo memory CD8 T cells in the lymph node failed to produce IFN-γ. In contrast, in both the blood and peripheral tissues (liver and lung), >90% of the memory CD8 T cells that were CD62L^Lo/CD27^Lo produced IFN-γ following stimulation with peptide. Overall, these data demonstrate that this unique CD62L^Lo/CD27^Lo memory CD8 T-cell population is found predominantly within lymphoid tissues, such as the spleen and lymph nodes.

CD62L^Lo/CD27^Lo Memory CD8 T Cells Do Not Undergo Homeostatic Proliferation.

Sustained but relatively slow proliferation is thought to be critical in maintaining the long-term numerical stability in memory CD8 T-cell populations observed in both humans and laboratory mice (24–26). Because cytokine nonproducing memory CD8 T cells exhibited the molecular signatures of apoptosis, we next determined whether these functionally impaired memory CD8 T cells were undergoing proliferation in vivo at late time points (>150 d postinfection with LCMV). Surprisingly, a similar fraction of CD62L^Lo/CD27^Lo memory cells incorporated BrdU over a 2-wk period compared with the highly proliferative T_CM subpopulation. Thus, the CD62L^Lo/CD27^Lo memory cells had gone through at least one cell division during that time interval (Fig. 4A). However, when T_CM and CD62L^Lo/CD27^Lo memory CD8 T cells were purified and transferred into separate groups of lymphopenic Rag1^−/− recipients (to drive rapid homeostatic proliferation), only T_CM, and not the CD62L^Lo/CD27^Lo cells, expanded in the new hosts (Fig. 4B). In fact, the few CD62L^Lo/CD27^Lo memory CD8 T cells found at day 3 after transfer could no longer be detected in the Rag1^−/− recipients on day 9 after transfer. In addition, CD62L^Lo/CD27^Lo memory cells failed to undergo proliferation in vitro in response to IL-15 stimulation, whereas T_CM readily underwent cell division (Fig. 4C). Therefore, these data suggest that consistent with their nonfunctional and apoptotic phenotype, this unique population of CD62L^Lo/CD27^Lo memory CD8 T cells does not undergo proliferation and does not survive in vivo. Instead, these data suggest that the CD62L^Lo/CD27^Lo memory CD8 T-cell population may be generated as a byproduct of T_CM turnover.

Fig. 4.

Functionally impaired CD62L^Lo/CD27^Lo cells are generated during homeostatic proliferation of CD62L^Hi/CD27^Hi T_CM CD8 T cells. (A) 5 × 10³ naive P14 CD8 T cells were transferred into B6 mice and subsequently infected with LCMV–Armstrong. At 135 d postinfection, mice were placed on drinking water containing BrdU for 15 d. Cellular proliferation, determined by incorporation of BrdU, was then analyzed on CD62L^Hi/CD27^Hi (T_CM) and CD62L^Lo/CD27^Lo memory P14 CD8 T cells. (B) 1.5 × 10⁵ T_CM or CD62L^Lo/CD27^Lo memory P14 CD8 T cells were FACS-sorted from the spleen and transferred into Rag1^−/− B6 mice. Total cell numbers in the blood were then monitored longitudinally. Error bars indicate SD of three mice per group. LOD, limit of detection. (C) Cells were sorted as in B and labeled with CFSE. Proliferation was then analyzed on the sorted populations following 3 d of stimulation in vitro with 200 ng/mL IL-15. (D) 2.5 × 10⁵ T_CM P14 CD8 T cells were sorted by FACS and transferred into Rag1^−/− B6 mice. Thirty days after transfer, the adoptively transferred population was analyzed for expression of CD62L and CD27. (E) Same as in D, with the exception that incorporation of BrdU was determined following pulse on days 10–25 after transfer. (F and G) Same as in D, with the exception that IFN-γ production was determined following stimulation with GP_33–41 peptide (F) or PMA and ionomycin (Iono) (G). (H) Same as in D, with the exception that cell size was determined based on forward scatter (FSC) on the two cell populations. (I) Same as in D, with the exception that activation of caspase-3 and caspase-7 and annexin V staining were analyzed on the adoptively transferred population. All data are representative of three independent experiments.

Proliferating T_CM CD8 T Cells Continuously Generate Nonfunctional Apoptotic Cells.

Because the overall number of memory CD8 T cells remains constant over long periods of time, the observed continual proliferation of memory CD8 T cells must be accompanied by a nearly equal death rate. Indeed, highly purified (>99%) T_CM transferred into Rag1^−/− mice undergo substantial proliferation (Fig. 4B). Strikingly, when the adoptively transferred T_CM population was analyzed 30 d after transfer for expression of CD62L and CD27, a substantial proportion of the transferred T_CM CD8 T cells (CD62L^Hi/CD27^Hi) now exhibited a CD62L^Lo/CD27^Lo phenotype and all these newly generated cells had proliferated following transfer (Fig. 4 D and E). In addition, the CD62L^Lo/CD27^Lo CD8 T cells that were generated following transfer failed to produce IFN-γ following stimulation with either peptide or PMA/ionomycin, whereas the cells that exhibited the T_CM phenotype uniformly produced this cytokine (Fig. 4 F and G). Finally, the CD62L^Lo/CD27^Lo memory cells that appeared following transfer exhibited hallmarks of apoptosis, including reduced size (as measured by forward scatter), activation of caspase-3 and caspase-7, and surface phosphatidylserine expression (Fig. 4 H and I). Thus, these data would argue that memory CD8 T cells that cannot produce cytokines and exhibit the hallmarks of apoptosis are directly generated from T_CM populations undergoing homeostatic proliferation.

Generation of impaired memory CD8 T cells from a fully functional T_CM population following transfer into Rag1^−/− mice suggested that these cells were byproducts of cell division. However, because the lymphopenic Rag1^−/− environment results in robust homeostatic proliferation that exceeds normal memory turnover in a WT host, we next tested whether functionally impaired CD62L^Lo/CD27^Lo memory CD8 T cells could be generated from T_CM following transfer into a WT environment. Indeed, at 35 d following transfer of T_CM into a WT host, a substantial percentage of the recovered cells now exhibited the CD62L^Lo/CD27^Lo phenotype (Fig. 5A). In addition, these newly generated cells that were found in both the spleen and lymph node were unable to produce IFN-γ in response to antigenic stimulation (Fig. 5B). Finally, the CD62L^Lo/CD27^Lo cells generated from T_CM in WT hosts also exhibited characteristics of apoptosis, including reduced size, activation of caspase-3 and caspase-7, and expression of surface phosphatidylserine (Fig. 5C). Therefore, these data demonstrate that nonfunctional apoptotic CD62L^Lo/CD27^Lo memory CD8 T cells are generated from T_CM during memory turnover in WT hosts.

Fig. 5.

Functionally impaired CD62L^Lo/CD27^Lo cells are generated following transfer of T_CM CD8 T cells into WT mice. (A) CD62L^Hi/CD27^Hi T_CM P14 CD8 T cells were sorted (>99% purity) as shown in Fig. 4D and transferred into WT B6 mice. Thirty-five days after transfer, the adoptively transferred population was analyzed for expression of CD62L and CD27. (B) Same as in A, with the exception that production of IFN-γ was analyzed on the adoptively transferred population following stimulation with GP_33–41 peptide from cells isolated from either the spleen or inguinal lymph node. (C) Same as in A, with the exception that cell size by forward scatter (FSC), activation of caspase-3 and caspase-7, and annexin V staining were analyzed on the adoptively transferred population. Data are representative of two independent experiments.

Discussion

Persistence of stable numbers of memory CD8 T-cell populations has long been appreciated as a hallmark characteristic of cellular immunity. In addition, the ability to “self-renew” through steady-state proliferation has been recognized as being critical for maintaining a constant number of memory CD8 T cells for long periods of time. In fact, memory CD8 T-cell populations numerically decline at a slow rate when steady-state proliferation is diminished (7–12). On the contrary, because memory CD8 T-cell populations remain quantitatively stable, this proliferation must be balanced by a tightly regulated rate of cell death; otherwise, memory CD8 T-cell populations would gradually inflate over time. However, until now, models to explain how memory CD8 T-cell numbers are maintained in the face of memory turnover have not been experimentally addressed.

Herein, we demonstrate that a nonfunctional apoptotic subset of cells is a general feature of memory CD8 T-cell populations that form following infection. Mechanistically, we provide direct evidence that as T_CM CD8 T cells turn over by proliferation, they generate this population of cells that lack functionality and exhibit signatures of apoptosis. Thus, we would argue that this newly identified memory CD8 T-cell subset constitutes a previously unidentified “death-intermediate memory” (T_DIM) population, generated during memory turnover, that will ultimately die, and thus contribute a critical regulatory event in maintaining the numerical stability in memory CD8 T-cell populations. Furthermore, we show that generation of the T_DIM population can be linked to cell division of T_CM during memory turnover. Therefore, our data support an important role for a division-linked model wherein memory turnover produces two different daughter cells: a “renewed” T_CM and a T_DIM (Fig. 1A, Left). Whether this division-linked mechanism is the sole pathway for regulating CD8 T-cell numbers during all phases of the memory immune response remains to be determined. However, identification of the T_DIM population provides a potential “missing link” that will permit future studies aimed to determine their precise cellular origins.

Additionally, our results open a unique avenue for future studies aimed at understanding the specific molecular and cellular mechanisms that regulate the process of T_DIM formation. One possibility is that the T_DIM population is simply unable to compete adequately for survival cytokines, which then causes it to become apoptotic. However, our data suggest this is not the case, because this population still forms during memory CD8 T-cell proliferation in the noncompetitive Rag1^−/− environment. Therefore, T_DIM generation is likely the result of changes in either gene regulation or the activation of signaling pathways that occur during or following cell division. In fact, several recent studies have suggested that asymmetrical cell division is a general feature of CD8 T cells that occurs following stimulation of the TCR (27–29). However, whether asymmetrical cell division occurs in proliferating memory CD8 T cells in a TCR-independent fashion is currently unknown. Overall, a complete understanding of the intrinsic and/or extrinsic signals that regulate generation of T_DIM populations might ultimately lead to therapies that could increase the quantity of memory CD8 T-cell populations.

Materials and Methods

Mice and Pathogens.

C57BL/6 and BALB/c mice were obtained form the National Cancer Institute and used for experiments at 6–10 wk of age. P14 (30), OT-I (31), and CS₂₈₀-tg (32) mice have previously been described and maintained by sibling × sibling mating. Rag1^−/− mice (33) were obtained from the Jackson Laboratories. LCMV–Armstrong was propagated according to standard protocols and injected (2 × 10⁵ plaque-forming units) i.p. as indicated. Attenuated Listeria monocytogenes expressing OVA or CS protein from P. yoelii have been previously described (34, 35). Attenuated L. monocytogenes were injected i.v. at a dose of 1 × 10⁷ cfu. All animal experiments followed approved Institutional Animal Care and Use Committee protocols.

Additional methods can be found in SI Materials and Methods.