Memory CD8⁺ T cells mediate antibacterial immunity via CCL3 activation of TNF/ROI⁺ phagocytes

In a first set of experiments, we have compared the immune response induced by secA2⁻ and actA⁻ bacteria (Fig. 1). Mice were primary infected with 0.1 × LD₅₀ of mutant or WT bacteria. Control animals were injected with PBS, heat-killed L. monocytogenes (HKLM) (20), or 0.1 × LD₅₀ of llo⁻ (ΔLLO) (21), three protocols that do not induce protective memory. Mice were challenged 1 mo later with 10 × LD₅₀ of WT bacteria and monitored for bacterial titers in the spleen and liver. As expected, control animals immunized with PBS or HKLM rapidly died and exhibited, respectively, 1,000- and 240-fold more bacteria in the spleen and liver than mice immunized with WT bacteria (not depicted) that survived and cleared infection (18). Interestingly, mice immunized with the secA2⁻ mutant or the llo⁻ control mutant failed to control secondary infection and exhibited 662- and 912-fold more bacteria in the spleen and liver, respectively, as compared with those immunized with WT or actA⁻ bacteria. Therefore, while the actA⁻ mutant induced protective memory, the secA2⁻ mutant did not.

To investigate why the secA2⁻ mutant failed to induce protective immunity, we have characterized the CD8⁺ T cell response in groups of mice immunized with secA2⁻, actA⁻, or WT bacteria (primary response) and further challenged the same groups of mice 1 mo later with WT L. monocytogenes (secondary response). Spleen cells were harvested at different times after infection and analyzed by FACS after staining with LLO_91-99/H2-K^d tetramers, anti-CD62L, a cell surface marker that is lost upon activation and acquisition of effector functions (2, 22), and anti-CD127 (IL-7 receptor), a marker of long-living memory T cells (23, 24). Mice from all groups exhibited similar frequencies of CD62L^low LLO_91-99/H2-K^d tetramer⁺ CD8⁺ T cells at the peak of primary and secondary responses (Fig. 2 A) (25). Moreover, LLO_91-99/H2-K^d tetramer⁺ CD8⁺ T cells expanded with the same kinetics upon secondary challenge in mice immunized with the secA2⁻ and actA⁻ mutant (Fig. 2 B). Therefore, the absence of a protective secondary response in mice immunized with the secA2⁻ mutant did not result from a decreased ability of Listeria-specific memory CD8⁺ T cells from these mice to proliferate upon secondary infection. We also found that mice primary immunized with secA2⁻ bacteria were not impaired in the generation of CD127^high CD62L^low effector–memory (T_EM) cells as compared with mice immunized with WT and actA⁻ bacteria (Fig. S1). Thus, the inability of mice primary immunized with secA2⁻ bacteria to control secondary infection did not result from a reduced number of T_EM cells (26).

We next assessed whether Listeria⁻specific CD8⁺ T cells from mice immunized with the secA2⁻ mutant were impaired in effector functions such as cytolytic activity, expression of granzyme B, CD107a (27) or CD95 ligand, and secretion of IFN-γ or TNF-α. To measure cytolytic activity, spleen cells were pulsed or not with LLO_91-99, labeled with high or low concentrations of CFSE, respectively, and injected into mice primarily immunized with secA2⁻, actA⁻, or WT bacteria that were secondary challenged with WT bacteria. Spleen cells were recovered 3 h or 1 or 4 d after secondary infection and analyzed by FACS to determine the specific lysis of peptide-pulsed CFSE^high cells. Lysis of these target cells was similar among all groups (Fig. 3 A and not depicted). In addition, CD8⁺ T cells from mice immunized with the secA2⁻ mutant were not impaired in the expression of granzyme B, CD107a, and CD95 ligand (not depicted). To monitor the presence of IFN-γ- and TNF-α–secreting cells, spleen cells from secondary challenged mice were incubated with LLO_91-99 peptide–pulsed splenocytes and analyzed by intracellular cytokine staining. The total numbers of IFN-γ– and TNF-α–secreting CD8⁺CD3⁺ T cells per spleen, as well as the frequencies of IFN-γ– and TNF-α–secreting cells among CD8⁺CD3⁺ T cells, were similar in mice primarily immunized with secA2⁻, actA⁻, or WT bacteria and challenged with WT bacteria at early and late (2- or 3-d) time points (Fig. 3 B and not depicted). Collectively, although Listeria-specific reactivated memory CD8⁺ T cells from secA2⁻ immunized mice did not confer protective immunity, they exhibited effector functions such as cytolytic activity, expression of granzyme B, CD107a and CD95 ligand, and secretion of IFN-γ and TNF-α. Thus, the inability of mice primary immunized with secA2⁻ bacteria to control secondary infection was neither due to ineffective memory CD8⁺ T cell development nor to impaired known effector activities of these cells.

CCL3 was the only effector activity that was known to be required for primary activated CD8⁺ T cells to protect naive recipients (14). We have therefore investigated whether Listeria-specific secondary effector CD8⁺ T cells from mice primary immunized with the secA2⁻ mutant were impaired in the secretion of CCL3 during secondary infection. We observed that the total number of CCL3-secreting CD8⁺CD3⁺ T cells per spleen was twofold lower in mice immunized with the secA2⁻ mutant or PBS as compared with those immunized with actA⁻ or WT bacteria (Fig. 4 A). Furthermore, CD8⁺CD3⁺ T cells from mice immunized with the secA2⁻ mutant exhibited a lower mean fluorescence intensity upon staining with anti-CCL3 serum than those from mice immunized with actA⁻ or WT bacteria (27.8 for secA2⁻ vs. 44 and 62 for actA⁻ and WT, respectively) (Fig. 4 A). Likewise, CD8⁺CD3⁺ T cells from mice immunized with the secA2⁻ mutant or PBS secreted less CCL3 when restimulated with LLO_91-99 than those from the two other groups (Fig. 4 B). Therefore, the secretion of CCL3 was the only known effector function of secondary effector CD8⁺ T cells that was significantly altered in mice immunized with the secA2⁻mutant as compared with those immunized with the actA⁻ mutant.

To determine whether CCL3 was required during the effector phase of secondary protective response against L. monocytogenes, mice were immunized with PBS or WT bacteria, injected 30 d later with anti-CCL3 or control goat serum, and challenged with WT bacteria. PBS-injected mice exhibited the same number of bacteria when treated with control or anti-CCL3 serum (Fig. 4 C and Fig. S2 A). In contrast, mice immunized with WT bacteria exhibited 100- and 12-fold more bacteria in the spleen and liver, respectively, when treated with anti-CCL3 serum as compared with control serum. Although mice injected with control serum survived to secondary infection, those treated with anti-CCL3 serum rapidly died (Fig. S3). Therefore, in vivo neutralization of CCL3 during secondary infection abrogated protective immunity.

To investigate if the CCL3 secreted by secondary effector CD8⁺ T cells was responsible for protection, WT CD8⁺ T cells from secondary challenged mice were transferred into CCL3^−/− recipient animals that received anti-CCL3 or control serum and were subsequently infected. In this experiment, 2 × 10⁷ CD8⁺ T cells, accounting for ∼40,000 LLO_91-99/H-2K^d–specific CD8⁺ T cells (Fig. 3 B), were purified from WT mice and transferred into naive CCL3^−/− or control WT recipients that were treated or not with anti-CCL3 serum and infected with 3 × 10⁵ WT bacteria (Fig. 4 D and Figs. S2 B and S4). Protection was similar whether recipient mice were WT or CCL3^−/− with 4.4- (spleen) and 5.7- (liver) fold less bacteria in mice infused with CD8⁺ T from secondary infected animals as compared with mice transferred with CD8⁺ T cells from PBS-injected mice (Fig. 4 D and Figs. S2 B and S4). Although the level of protection achieved in these latter experiments was relatively low, it should be noticed that we have transferred 10–15-fold less epitope-specific T cells than other authors (6, 14, 28). Furthermore, the level of protection is also known to be highly dependent on the infecting dose of bacteria. Indeed, when mice transferred with CD8⁺ T cells from secondary infected mice were challenged with 5 × 10⁴ bacteria (instead of 3 × 10⁵), 15–27 less bacteria were enumerated in their organs as compared with those infused with CD8⁺ T cells from PBS-injected animals (Fig. S5).

Most importantly, protection transferred to CCL3^−/− mice by WT CD8⁺ T cells from mice immunized with WT bacteria was completely abrogated upon treatment with anti-CCL3 serum, further showing that CCL3 derived from WT memory CD8⁺ T cells was required for secondary protection. We also found that CCL3 from memory CD8⁺ T cells accounted for the majority of the CCL3-mediated protective effect in the liver of CCL3^−/− recipient mice transferred with WT CD8⁺ T cells from secondary infected WT animals (Figs. S2 B and S4).

To independently demonstrate that CCL3 from secondary effector CD8⁺ T cells was responsible for protection, CCL3^−/− or WT control mice were primary and secondary infected with WT bacteria, and CD8⁺ T cells from these mice were transferred into naive WT mice (Fig. 4 D and Figs. S2 B and S4). Although the transfer of WT CD8⁺ T cells resulted in 4.4-fold less bacteria in the spleen, CCL3^−/− CD8⁺ T cells only conferred a marginal protective effect in the spleen and liver, further confirming a critical role of the CCL3 derived from secondary effector CD8⁺ T cells. Collectively, our data demonstrated that CCL3, and not other cytokines/chemokines from reactivated memory CD8⁺ T cells (and not from other cell types), is mediating secondary protective immunity.

Although CCL3 mediates the chemotaxis of monocytes and T and NK cells (29, 30), this chemokine also promotes TNF-α secretion from macrophages in vitro (31, 32). Because TNF-α has been shown to be critical for L. monocytogenes clearance and host survival during secondary infection (9–11), we hypothesized that mice immunized with the secA2⁻ mutant did not clear secondary infection because their CD8⁺ T cells were impaired in their ability to secrete CCL3 and subsequently to induce myeloid cells to release TNF-α. In agreement with previous reports, mice secondary challenged with WT bacteria rapidly died (Fig. S3) and exhibited 10,000- and 300-fold more bacteria in the spleen and liver, respectively (Fig. 5 A and not depicted), when treated with soluble recombinant p75 TNF-α receptor (Enbrel) as compared with control animals. TNF-α was produced by myeloid cells that expressed high levels of CD11b (Fig. 5 B). Further characterization of the TNF-α–secreting cells demonstrated that 90% of these cells were monocytic-derived cells (F4/80⁺CD11c^+/−CD11b^med/highLy-6C^highGr1^low) and 10% were neutrophils (F4/80⁻CD11c⁻CD11b^highLy-6C^med/highGr1^high) (Fig. 5 B and not depicted). These monocytic-derived cells expressed MHC class II molecules, CD80, and CD86 and produced inducible nitric oxide synthase (iNOS) (Fig. S6, A and B). They also contained bacterial antigens (Fig. 5 C) as well as live replicating bacteria (not depicted) and exhibited ultrastructural characteristics of phagocytic mononuclear cells (Fig. S6, C and D). They were therefore designated TP-MPCs for TNF-α–producing mononuclear phagocytic cells.

Both the number (Fig. 5 D) and the frequency (Fig. S7 A) of TP-MPCs in the spleen increased more slowly and to lower levels in mice immunized with the secA2⁻ mutant or PBS and secondary challenged as compared with those immunized with actA⁻ or WT bacteria. In contrast, the frequency of TNF-α–secreting neutrophils in the spleen remained identical in mice immunized with secA2⁻ and actA⁻ bacteria (not depicted). Moreover, the frequency of MPCs that secreted TNF-α among total MPCs was twofold lower at 6 and 10 h after challenge in mice immunized with the secA2⁻ mutant compared with those immunized with actA⁻ or WT bacteria (Fig. S7 B). Therefore, the kinetics of TNF-α secretion by MPCs, but not by neutrophils, was strongly altered in mice that received a primary immunization with the secA2⁻ mutant.

To assess whether CCL3 promotes TNF-α secretion by MPCs, we measured the number of TP-MPCs in primary and secondary infected mice treated or not with anti-CCL3 serum. In mice primary immunized with WT bacteria, neutralization of CCL3 before secondary infection resulted in a lower number of TP-MPCs in the spleen (Fig. 6 A, left). In contrast, the number of TNF-α–secreting neutrophils remained the same (not depicted). A similar reduction in the number of TP-MPCs, but not neutrophils, was observed in mice depleted of CD8⁺ T cells before the secondary infection (Fig. 6 A, right, and not depicted). Furthermore, recombinant CCL3 induced MPCs to secrete TNF-α in vitro as efficiently as HKLM (Fig. 6 B). As expected, anti-CCL3 prevented TNF-α release, demonstrating that this effect was dependent on CCL3. CCR1 and CCR5 are the two major receptors of CCL3 (33). Although we could not detect the expression of CCR5 at the surface of TP-MPCs, 50% of these cells expressed CCR1, a result in agreement with their responsiveness to CCL3 (Fig. 6 C). Collectively, these data support the hypothesis that CCL3 secreted by CD8⁺ T cells during a secondary infection induces MPCs to rapidly release TNF-α.

TNF-α is an in vitro inducer of ROIs that directly allows infected myeloid cells to kill bacteria (34, 35). Therefore, the rapid release of TNF-α by MPCs may be critical for the production of high levels of ROI and subsequent clearance of bacteria. To test this hypothesis, mice received a primary immunization with PBS, WT, or mutant bacteria and were further challenged with WT L. monocytogenes. Spleen cells were analyzed for ROI production upon staining with anti-CD11b and anti–Ly-6C mAbs. In both PBS- and bacteria-immunized mice, challenge with WT bacteria induced ROI production (Fig. 7 A). At each time, the frequency of ROI-producing MPCs was 1.3- to 1.5-fold lower in mice primarily immunized with the secA2⁻ mutant or PBS as compared with those immunized with actA⁻ or WT bacteria. In contrast, although neutrophils accounted for the majority (∼80%) of ROI-producing CD11b⁺ cells (Fig. 7, A and B, Fig. S8), their numbers did not differ between mice primarily immunized with secA2⁻, actA⁻, or WT bacteria. In PBS-injected animals, however, the number of ROI-producing neutrophils was significantly lower compared with mice primarily immunized with WT bacteria (Fig. S8). Both the neutralization of CCL3 or TNF-α and the depletion of CD8⁺ T cells before the secondary challenge in mice primarily immunized with WT bacteria reduced the number of ROI-producing MPCs and neutrophils to similar numbers as those measured in the PBS-immunized control group (Fig. 7 C). Collectively, these data reveal a causal relationship between the secretion of CCL3 and the production of TNF-α by MPCs, and between the secretion of TNF-α and the production of ROIs by both MPCs and/or neutrophils.

To determine whether ROIs were required for bacterial clearance upon secondary infection, mice that were unable to generate ROIs as the result of a targeted deletion of the p47^phox gene (36) were immunized and challenged or not with WT bacteria. It has been reported that p47^phox−/− mice are able to clear primary L. monocytogenes infection (37). In contrast, upon immunization and subsequent acute challenge with WT bacteria, p47^phox−/− mice rapidly died and exhibited 3,000- and 9,000-fold more bacteria in the spleen and liver, respectively, as compared with WT animals (Fig. 8 and Figs. S3 and S9).

To further confirm that the lack of protection in p47^phox−/− mice actually resulted from the inability of innate cells to generate ROIs and not from a defective memory response, we compared primary and secondary CD8⁺ T cell responses in WT and p47^phox−/− mice. This was done using HKLM to restimulate CD8⁺ T cells because p47^phox−/− mice were on the 129 genetic background in which Listeria epitopes have not been characterized. Upon primary infection, the spleens of p47^phox−/− mice contained similar numbers of granzyme B–expressing and IFN-γ–secreting CD8⁺ T cells reacting against HKLM as those from WT mice. Upon secondary infection, these numbers were three- to fivefold increased as expected for a recall response (Fig. S10). Most importantly, this memory response was similar in p47^phox−/− and WT mice, suggesting that the lack of protective immunity in secondary infected p47^phox−/− mice did not result from defective CD8 T cell memory.

To independently confirm that ROIs were critical for secondary protection against L. monocytogenes, we used a different experimental approach in which WT BALB/c mice primarily immunized with WT bacteria were treated with a superoxyde scavenger (38) before secondary challenge to neutralize the oxidative burst in vivo (Fig. S11). As compared with untreated animals, mice treated with a superoxyde scavenger exhibited 1,073- and 230-fold more bacteria in the spleen and liver, respectively. Therefore, collectively, these results on two distinct genetic backgrounds demonstrate that the generation of ROIs is required for both bacterial clearance and survival of secondary infected mice.

To further clarify the role of MPCs and neutrophils in ROI production and bacteria elimination, 4 × 10⁵ MPCs or 4 × 10⁶ neutrophils from WT or p47^phox−/− naive mice were transferred into p47^phox−/− mice primary immunized with WT bacteria or PBS and secondary challenged with WT bacteria (Fig. 8 and Fig. S9). Mice reconstituted with MPCs from WT mice exhibited 170- and 10-fold less bacteria in the spleen and liver, respectively, than those reconstituted with MPCs from p47^phox−/− mice. Similarly, mice reconstituted with neutrophils from WT mice exhibited 1,000- and 80-fold less bacteria in the spleen and liver, respectively, than those reconstituted with neutrophils from p47^phox−/− mice. Also, PBS-immunized p47^phox−/− recipient mice failed to clear infection with WT bacteria even when reconstituted with MPCs or neutrophils from WT mice. Collectively, these results show that both MPCs and neutrophils from WT mice, but not those from p47^phox−/− mice, were able to restore a protective secondary response when transferred into p47^phox−/− mice. Because protection was only observed in p47^phox−/− mice primary immunized with WT bacteria, these data strongly suggested that the presence of memory cells was required to activate ROI production and bacteria killing.

We next verified that the restoration of protective immunity by MPCs or neutrophils in secondary challenged p47^phox−/− mice was dependent on CCL3 and CD8⁺ T cells. p47^phox−/− mice were primary immunized with WT bacteria, reconstituted with MPCs or neutrophils from WT mice, treated or not with anti-CCL3 or with anti-CD8, and secondary challenged with WT bacteria. Neutralization of CCL3 or CD8⁺ T cell depletion prevented bacterial clearance both in mice reconstituted with MPCs or neutrophils (Fig. 8 and Fig. S9). Therefore, secretion of CCL3 by reactivated memory CD8⁺ T cells during secondary infection induces the accelerated release of TNF-α by MPCs that in turn promotes the production of ROIs by MPCs and neutrophils, leading to subsequent clearance of the bacteria (Fig. 9).

One important prediction of our findings is that CCL3-secreting memory CD8⁺ T cells should promote an antigen-independent antimicrobial effector phase by innate immune cells upon antigen-specific restimulation. Therefore, we have postulated that any ROI-sensitive pathogen would be killed upon restimulation of memory cells in a bystander fashion. To test this hypothesis, BALB/c mice were immunized with WT L. monocytogenes and infected 3 wk later with the intracellular ROI-sensitive L. major promastigotes (39). Parasite-containing mice were challenged 2 wk later either with L. monocytogenes or LPS-matured DCs pulsed or not with the dominant Listeria-derived LLO_91-99 peptide (Fig. 10 and not depicted). Animals challenged with L. monocytogenes or peptide-pulsed DCs exhibited 10,000- and 1,000-fold less L. major parasites in their spleen than those injected with PBS and control LPS-matured DCs, respectively. This dramatic decrease in parasite burden was completely abrogated upon injection of a CCL3-neutralizing, but not of a control, serum (not depicted). Therefore antigen-driven restimulation of L. monocytogenes–specific memory CD8⁺ T cells can induce antigen-independent CCL3-mediated bystander killing of an unrelated intracellular parasite.

Several studies have demonstrated that neither cytolytic activity nor secretion of IFN-γ or TNF-α by CD8⁺ T cells was absolutely required for resistance to L. monocytogenes in mice (4, 6, 13). In contrast, CCL3-deficient CD8⁺ T cells isolated from primary infected mice did not confer protective immunity upon transfer into recipient mice (14). Although this suggested that secretion of CCL3 by primary effector CD8⁺ T cells is required for protection, it was not known whether CCL3 derived from secondary effector CD8⁺ T cells could confer protection to primary infected animals. In this study, we show that protection conferred by reactivated memory CD8⁺ T cells depends on their ability to secrete CCL3. This was independently shown by experiments in which CD8⁺ T cells from WT or CCL3^−/− mice were injected into CCL3^−/− mice, as well as in neutralization experiments using anti-CCL3 mAb. In agreement with these data, we found that secondary effector CD8⁺ T cells from mice immunized with secA2⁻ bacteria did not secrete CCL3 as compared with those from mice immunized with actA⁻ or WT bacteria. Although these differences were modest, they were significant and reproducible. How could relatively small differences account for such a dramatic effect in protection? At least two explanations can account for these observations. First, the CCL3-secreting cells that were present in mice immunized with actA⁻ or WT bacteria may be localized in strategic areas of the spleen that are critical for the development of protective immunity. A small difference in the number of CCL3-secreting cells could therefore have a major impact on the outcome of the recall infection. Second, several cytokines such as TNF-α (40), IFN-α/β (41), or IL-12 (42) have been shown to be biologically active and/or to exert a distinct effect above a specific threshold. Likewise, small differences in the amount of secreted CCL3 could lead to a distinct biological response, i.e., protective versus nonprotective.

CCL3 is an inflammatory molecule that activates macrophages to secrete TNF-α (31, 32) and exhibits chemotactic activities on several cell types (29, 30). Neutralization of CCL3 in protected animals reduces both the total number of MPCs in the spleen and the number of these cells that secrete TNF-α. It remains unclear whether CCL3 induces activation or recruitment of MPCs or both. We have found (a) that recombinant CCL3 induces MPCs to secrete TNF-α in vitro, and (b) that the total number of MPCs in the spleen is identical in mice immunized with secA2⁻ bacteria as compared with those immunized with actA⁻ bacteria, whereas the frequency of TP-MPCs among MPCs is twofold lower. Thus, although it remains to be determined whether CCL3 also promotes MPC recruitment, it does activate MPCs to secrete TNF-α.

We and others have demonstrated that TNF-α blockade using mAbs or soluble receptor prevents protective immunity after secondary infection, therefore demonstrating a critical role of this cytokine in this process (9, 10). These results are in apparent contrast to those obtained in TNF-α^−/− mice or in mice exhibiting TNF-α–deficient macrophages and neutrophils that are able to develop long-term protective immunity (8). How can these data be reconciled? Regarding TNF-α^−/− mice, it is well established that these animals do not regulate properly the development and the organization of splenic follicular architecture and exhibit defects in their lymphoid organs and germinal centers (43, 44). Furthermore, these mice are unable to regulate the homeostasis and the extent and duration of inflammatory processes. Therefore, it is conceivable that the development of protective immunity in TNF-α^−/− mice results from compensatory mechanisms that do not reflect what is happening in WT mice.

In a recent study, Grivennikov et al. (8) have taken advantage of the lysozyme M (lys-M) promoter to generate mice in which macrophages and neutrophils did not express TNF-α. These mice were protected against secondary infection, further suggesting that lys-M–expressing cells were not required for protection. We have found that at least 50% of the MPCs that are the main source of TNF-α during secondary infection do not express lys-M in the lys-M–GFP^+/− knock-in mouse (Fig. S12) (45). Therefore, this cell-specific TNF-α knockout strategy did not completely abrogate the production of TNF-α from all CD11b-expressing cells, further providing an explanation for the survival of these mice after a secondary challenge.

Our results also highlight that the fine-tuning of TNF-α production is crucial for efficient control of secondary bacterial growth. Although TNF-α is produced by several cell types including CD8⁺ T cells, this is the TNF-α derived from non–CD8⁺ T cells that is required for secondary protection (12). As mentioned before, 80–90% of TNF-α–secreting cells during secondary infection are MPCs, and depletion of CD8⁺ T cells or neutralization of CCL3 abrogated both the release of TNF-α by MPCs and mice resistance. This result, together with the lys-M experiments described above, strongly suggests that TNF-α produced by MPCs, but not by other cells, is critical for secondary protective immunity.

A recent study has identified an inflammatory subset of DCs that produce large amounts of TNF-α and are necessary for mice survival upon primary infection. These cells were designated Tip-DCs for TNF-α/iNOS-producing DCs (46). Tip-DCs and MPCs express the same DC-specific surface markers, and both produce iNOS that generates reactive nitric intermediates (RNIs). However, in contrast to Tip-DCs, MPCs express the macrophage marker F4/80, contain live and dead bacteria, and produce ROIs. It remains to be determined whether Tip-DCs and MPCs are related or not.

A previous study has suggested that RNIs are not necessary for bacterial clearance during a secondary response (47). Our results further show that ROI generation is the final bacteria killing effector mechanism required for secondary protection. In contrast, RNIs, but not ROIs, are necessary for bacterial clearance and mice survival during primary infection, (48). RNI depends on iNOS synthesis that is regulated at the transcriptional level (49). The generation of ROIs is dependent on phosphorylation of the cytosolic regulatory protein p47^phox. This event leads to activation of the NADPH oxidase, a TNF-α–induced activity that is far more rapid than the regulation at the transcriptional level (50). All the mechanisms involved in this regulation loop are rapid, including secretion of CCL3 by memory CD8⁺ T cells, secretion of TNF-α by MPCs, phosphorylation of p47^phox upon binding of TNF-α to its surface receptor, activation of NADPH oxidase, and ROI production. The ability of TNF-α to induce both MPCs and neutrophils to produce ROIs results in an amplification loop that allows rapid clearance of high doses of bacteria, a mechanism that does not occur in naive mice because of the lack of memory CD8⁺ T cells.

The aim of this study was to elucidate the molecular and cellular events that orchestrate the control of bacterial growth in secondary infected mice. However, the molecular signals that are required for the priming of fully functional memory CD8⁺ T cells remain to be identified. Although both secA2⁻ and actA⁻ bacteria escape from phagocytic vacuoles, multiply in cytosol, and exhibit the same LD₅₀, these mutants nevertheless exhibit distinct defects. ActA⁻ bacteria cannot mobilize actin (17, 19). SecA2⁻ bacteria lack an ATPase that is critical for an auxiliary protein secretion system important for the secretion of at least 17 bacterial proteins (16). It is conceivable that egress in the cytosol of one or more of the SecA2-dependent proteins is required for the differentiation of bacteria-specific naive CD8⁺ T cells into fully functional memory cells.

The measure of cytolytic IFN-γ– and TNF-α–secreting activity of memory CD8⁺ T cells is widely used to predict the ability of an immunized individual to develop a protective immune response. Although these activities may accurately reflect the ability of the host to resist to a secondary challenge in some infectious diseases, this study suggests that it may not always be the case and that another effector activity such as CCL3 secretion could be an important marker to predict vaccine efficacy. Importantly, we have shown that it is also possible to induce, in an antigen-dependent fashion, a CCL3-mediated bystander killing of an unrelated pathogen, a mechanism that could potentially be used for therapeutic immunologically based protocols. In fact, some microbial infections cannot be treated with conventional medications anymore because the pathogens have become resistant to multiple drugs. For instance, this is the case for some strains of Mycobacterium tuberculosis or Staphylococcus aureus. If such patients were previously vaccinated with attenuated L. monocytogenes bacteria, it should be possible to specifically reactivate the CCL3-secreting memory CD8⁺ T cells to mediate the bystander elimination of multidrug-resistant pathogens without the need of any other medication. We believe that taking advantage of this mechanism may have important therapeutic implications for curative medicine.

6–8-wk-old BALB/cByJ female mice (Janvier Laboratories) were used in all experiments unless otherwise indicated. p47^phox−/− mice (36) were obtained from European Mouse Mutant Archive, and CCL3^−/− (51) were obtained from The Jackson Laboratory. Aged-matched 129S2/SvPasCrl control mice were purchased from Charles River Laboratories. lys-M GFP-expressing knock-in mice (lys-M-GFP^+/−) (45) were provided by T. Graf (Albert Einstein College of Medicine, New York, NY). L9.6 TCR transgenic mice have been described (2). p47^phox−/−, CCL3^−/−, lys-M–GFP^+/−, and L9.6 mice were housed and bred in our specific pathogen-free animal facility. All mice were used in accordance with Animal Care and Use regional Côte d'Azur Committee procedures.

The L. monocytogenes 10403s WT strain was used in all experiments. The actA⁻, llo⁻, and secA2⁻ mutant strains were on the 10403s WT strain background and have been described. The WT, actA⁻, llo⁻, and secA2⁻ strains exhibit an LD₅₀ of 3 × 10⁴, 5 × 10⁹, 10⁷, and 10⁷, respectively, in BALB/c mice (16, 19, 21). HKLM was prepared as described previously (2). All bacteria were prepared from clones grown from organs of infected mice. Stocks of bacteria were kept frozen at −80°C. DsRed-expressing L. major promastigotes (World Health Organization strain WHOM/IR/-/173) (52) were grown in M199 medium containing 20% FCS.

For L. monocytogenes infections, bacteria were grown to a logarithmic phase (OD₆₀₀ = 0.05–0.15) in broth heart infusion (BHI) medium (Sigma-Aldrich) diluted in PBS and injected i.v. into the lateral tail vein. In all experiments, mice were primary immunized with a 0.1 × LD₅₀ of bacteria (3 × 10³). Secondary infections were performed 1 mo later with 10 × LD₅₀ of WT bacteria (3 × 10⁵) except when specified otherwise. To measure bacterial titers in the spleen and liver, organs were harvested and dissociated on metal screens in 10 ml of 0.1% Triton X-100 (Sigma-Aldrich). Serial dilutions were performed in the same buffer, and 100 μl was plated onto BHI media plates. Mice survival was monitored twice a day.

For L. major infections, mice were injected with PBS or infected with 0.1 × LD₅₀ of WT bacteria and infected i.v with 2 × 10⁶ stationary phase promastigotes 3 wk later. Mice were secondary challenged after 2 wk with WT bacteria or LPS-matured DCs pulsed with the Listeria-derived LLO_91-99 peptide and co-treated or not with anti-CCL3. To determine the parasite burden in the spleens, organs were harvested 48 h after challenge and dissociated on nylon screens in 5 ml of M199 medium containing 20% FCS, 100 μg/ml penicillin, and 100 U/ml streptomycin. Serial dilutions were performed in the same buffer, and cells were incubated in flat-bottom 96-well plates at 27°C for 3–6 d. The number of parasite was determined using a FACS Array (Becton Dickinson) to monitor the red fluorescence of the live parasite.

The following mAbs were purchased from BD Biosciences: anti–CD8α (53–6.7)-FITC, -PE, –peridinin chlorophyll protein (PcP), or -allophycocyanin (APC), anti–CD3ε-FITC (145-2C11), anti–Ly-6C–FITC (AL-21), anti–CD107a-FITC (1D4B), anti–CD11c-PE (HL3), anti–CD11b-PE, -PcP, or -APC (M1/70); CD80-PE (16-10A1), anti–CD86-PE (GL1), anti–I-A^d–PE (AMS-32.1), anti–CCR5-PE (C34-3448), anti–CD62L-APC (MEL-14), anti–CD127-APC (A7R34), anti–IFN-γ–APC (XMG1-2), anti–TNF-α–APC (MP6-XT22), and control rat IgG₁ mAb. Anti–NOS-2 (M-19) polyclonal rabbit and anti–CCR-1 (C-20) antibodies were purchased from Santa Cruz Biotechnology, Inc. Difco Listeria O rabbit polyserum was purchased from Fischer. Goat and donkey anti–rabbit Alexa 647 and CFSE were purchased from Invitrogen. Goat polyclonal anti-CCL3 antibody and recombinant mouse CCL3 (rCCL3) was purchased from R&D Systems. Endotoxin content of 1 μg rCCL3 from Gram-negative bacteria was <1.2 pg/ml as determined using the Limulus Amebocyte Lysate QCL-1000 (Cambrex). PE-conjugated LLO_91-99/H2-K^d tetramers were obtained from the National Institutes of Health (NIH) tetramer core facility.

Organs were cut in small pieces and incubated at 37°C for 20 min in HBSS medium (Invitrogen) containing 4,000 U/ml collagenase I (Invitrogen) and 0.1 mg/ml DNase I (Roche). Red blood cells were lysed for 2–3 min in 170 mM NH₄Cl, 17 mM Tris HCl, pH 7.4. All steps of intracellular CCL3 staining as well as purification of CD8⁺ T cells for transfer experiments were performed in the presence of 2 μg/ml Golgi plug (BD Biosciences).

Cells were stained with the specified antibodies in 100 μl of PBS containing 0.5% BSA (FACS buffer). For intracellular staining, splenocytes were incubated at 37°C and 5% CO₂ for 3–4 h in RPMI 1640 (Invitrogen) containing 5% FCS and 2 μg/ml Golgi Plug (BD Biosciences) with or without 5 × 10⁸ HKLM/ml, 1 nM LLO_91-99 (Mimotopes), or 1 μg/ml of recombinant mouse CCL3 (R&D Systems). Cells were incubated for 20 min on ice with the indicated cell surface marker mAbs, fixed in 1% paraformaldehyde FACS buffer for 20 min on ice, and permeabilized for 30 min in 1XPerm/Wash (BD Biosciences). For intracellular staining of TNF-α or IFN-γ, cells were incubated for 20 min on ice in FACS buffer containing anti–IFN-γ, anti–TNF-α, or control rat IgG₁. For intracellular staining of iNOS or Listeria antigens, cells were incubated for 20 min on ice in FACS buffer containing anti-iNOS rabbit polyclonal, anti-Listeria Difco Listeria O rabbit polyserum, or control normal goat IgG, and staining was revealed using goat anti–rabbit Alexa 647 mAb. For intracellular staining of CCL3, cells were incubated for 20 min on ice in FACS buffer containing anti-CCL3 goat polyclonal antibodies, and staining was revealed using donkey anti–goat Alexa 647 mAbs. In all cases, cells were washed, fixed for 30 min in 1% paraformaldehyde FACS buffer, and analyzed on a FACSCalibur cytofluorometer (Becton Dickinson). When indicated, cells were sorted on a FACSVantage SE cell sorter (Becton Dickinson).

Supernatants were analyzed for CCL3 content by ELISA using the DuoSet ELISA kit (R&D Systems). The sensitivity of the assay was 2 pg/ml.

Spleen cells were pulsed or not with 1 nM LLO_91-99 for 1 h at 37°C. Pulsed and unpulsed cells were incubated for 10 min at 37°C in PBS containing 5 or 0.5 μM CFSE, respectively. Cells were washed in PBS, and 10⁷ of peptide-pulsed and 10⁷ of unpulsed cells were co-injected i.v. into recipient mice. Mice were killed 15 h later, and spleen cells were analyzed by FACS. Cytolytic activity against peptide-pulsed target cells was determined by measuring the frequency of CFSE^high and CFSE^low cells and by calculating the percentage of specific lysis using the following formula: 100 − ([(% peptide pulsed in infected/% unpulsed in infected) / (% peptide pulsed in uninfected/% unpulsed in uninfected)] × 100) (53).

For neutralization of TNF-α, mice were injected with 1.5 mg of soluble p75 TNF-α receptor (Enbrel) i.p. and infected 1 d later. For superoxyde neutralization (MnTBAP; A.G. Scientific, Inc.), mice were injected with 10 μg/g of body weight (38) and infected 1 d later. For neutralization of CCL3, mice were treated with 75 μg anti-CCL3 or goat IgG control i.v at the time of the secondary infection and 24 h later. For depletion of CD8⁺ T cells, mice were injected three times daily with 100 μg of the anti-CD8β H-35 mAb or its control isotype i.p., and then infected 1 d after.

WT or CCL3^−/− mice were PBS injected or primary immunized with WT bacteria and secondary challenged 1 mo later with 3 × 10⁵ WT bacteria. Mice were killed 6 h later, and splenic CD8⁺ T cells were negatively enriched using 5 μg/ml of homemade anti–MHC-II (MKD6), anti-CD4 (GK1.5), anti-B220 (RA3), and anti-CD19 (1D3) mAbs in 5% FCS RPMI 1640 containing 5% FCS and 2 μg/ml Golgi Plug. CD8⁺ T cells were further flow cell sorted, and 20 × 10⁶ cells (∼40,000 LLO_91-99/H2-K^d–specific CD8⁺ T cells) with a purity of >90% were transferred per naive WT or CCL3^−/− recipients. Mice were subsequently treated with anti-CCL3 or control goat serum and infected with 3 × 10⁵ or 5 × 10⁴ WT bacteria.

Naive WT or p47^phox−/− mice were killed, and spleen and bone marrow cells were first positively enriched using anti-CD11b–specific MACS beads (Miltenyi Biotec) according to the manufacturer's standard protocol and further flow cell sorted based on their expression of CD11b and Ly-6C cell surface markers, with monocytic-derived cells being CD11b^med/highLy-6C^high and neutrophils being CD11b^highLy-6C^med/high. PBS-injected or WT-immunized p47^phox−/− mice were reconstituted with 4 × 10⁵ MPCs (purity >87%) or 4 × 10⁶ neutrophils (purity >93%). 1 d later, mice were treated with anti-CCL3 or control goat serum and infected with 3 × 10⁵ WT bacteria.

Spleen cells of naive BALB/c mice were positively enriched using anti-CD11c–specific MACS beads (Miltenyi Biotec) according to the manufacturer's standard protocol. For LPS stimulation, cells were plated at 10⁶ cells/ml in RPMI 1640 medium containing 20% FCS supplemented with 0.1 μg/ml LPS (Sigma-Aldrich). 20 h later, cells were washed and pulsed with 1 μM of the LLO_91-99 peptide for 1 h, and 20 × 10⁶ of these peptide-pulsed mature DCs were transferred per recipient mouse.

Flow-sorted CD11b^med/highLy-6C^high cells were fixed with 1.6% glutaraldehyde in 100 mM phosphate buffer, pH 7.5. Cells were incubated at 4°C in 100 mM cacodylate buffer, pH 7.0, containing 1% osmium tetroxyde. Cellular pellets were washed with distilled water and incubated for 2 h with 0.5% uranyl acetate buffer at room temperature in the dark. Cells were washed in water, and pellets were dehydrated in increasing acetone series and embedded in epoxy resins. Blocks were thin-sectioned using standard procedures and contrasted for the observation on an electron microscope (CM12; Philips).

5-10 × 10⁶ splenocytes were incubated for 3 h at 37°C and 5% CO2 in 5% FCS RPMI 1640 with 5 × 10⁸ HKLM/ml and 160 μM hydroethidine (Polysciences). Hydroethidine is oxidized by ROIs in red fluorescent ethidium bromure, which allows for the detection of ROI-producing cells. Cells were washed in FACS buffer and stained for expression of cell surface markers.

Statistical significance was calculated using an unpaired Mann-Whitney test and Instat software. All p-values of ≤0.05 were considered significant and are referred to as such in the text. In experiments for which we did not have n > 6 subjects to run a Mann-Whitney test, we have applied the rules of SE bars (54) and accordingly provided an estimated p-value for all panels shown in the study.

Fig. S1 shows the frequency of tetramer⁺CD127⁺CD62L⁻ (T_EM) and tetramer⁺CD127⁺CD62L⁺ (T_CM) (± SE) among CD8⁺ T cells. Fig. S2 shows that CCL3 from CD8⁺ T cells is required for protection against secondary infection in the liver. Fig. S3 shows the survival kinetics of WT mice treated or not with anti–TNF-α or anti-CCL3 and of p47^phox−/− mice during secondary infection. Fig. S4 shows that CCL3 from CD8⁺ T cells is required for secondary protective immunity in the spleen and liver (as in Fig. 4 D and Fig. S2), but data are expressed in the percentage of the maximum protective effect achieved by transferring WT CD8⁺ T cells from secondary infected animals. Fig. S5 demonstrates that the protection achieved by transferring WT CD8⁺ T cells increases when recipient animals are challenged with lower numbers of bacteria (5 × 10⁴ instead of 3 × 10⁵). Fig. S6 provides a phenotypic characterization of the TP-MPCs and shows that they are infected using electron microscopy. Fig. S7 shows the frequencies of TP-MPCs in the spleen and among total MPCs during secondary infection in mice primarily immunized with PBS or 0.1 × LD₅₀ of the indicated bacteria. Fig. S8 monitors the numbers of ROI-producing MPCs and neutrophils in primary and secondary infected mice. Fig. S9 shows that the CCL3-producing memory CD8⁺ T cells promote ROI generation by phagocytic cells for secondary bacterial clearance in the liver of infected mice. Fig. S10 shows the number of granzyme B–expressing and IFN-γ–secreting CD8⁺ T cells in WT and p47^phox−/− 129 mice after primary and secondary infection. Fig. S11 demonstrates that ROIs are required for mediating secondary protection against L. monocytogenes infection in WT mice treated with an O₂⁻ scavenger before secondary challenge. Fig. S12 provides an analysis of lys-M expression in TP-MPCs during secondary infection.

We thank C. Schrike and N. Guy for their help with mice, E. Pamer for providing L9.6 TCR transgenic mice, T. Graf for providing the lys-M–GFP knock-in mice, D. Portnoy for the L. monocytogenes–targeted deletion mutants, O. Brocq and A. Bernard for Enbrel, O. Leo for the H35 anti-CD8_β hybridoma, and T. Aebischer for the DsRed-expressing L. major strain. H2-K^d/LLO_91-99 tetramer was obtained through the NIH Tetramer Facility.

This work was supported by grants from INSERM (Avenir), Human Frontier Science Program (CDA), Agence Nationale de la Recherche (ANR-IRAP-2005), and Fondation pour la Recherche Médicale (FRM). E. Narni-Mancinelli and L. Campisi are both recipients of a MENRT fellowship.

The authors have no conflicting financial interests.