Splenic CD8α+ dendritic cells undergo rapid programming by cytosolic bacteria and inflammation to induce protective CD8+ T-cell memory


  • Laura Campisi,

    1. Institut National de la Santé et de la Recherche Médicale Unité 924, Groupe Avenir, Valbonne, France
    2. Université de Nice-Sophia Antipolis, UFR Sciences, Nice, France
    Search for more papers by this author
  • Saidi M'Homa Soudja,

    1. Albert Einstein College of Medicine, Department of Microbiology and Immunology, Bronx, NY, USA
    Search for more papers by this author
  • Julie Cazareth,

    1. Université de Nice-Sophia Antipolis, UFR Sciences, Nice, France
    2. Centre National de la Recherche Scientifique, UMR6090, Valbonne, France
    Search for more papers by this author
  • Delphine Bassand,

    1. Institut National de la Santé et de la Recherche Médicale Unité 924, Groupe Avenir, Valbonne, France
    2. Centre National de la Recherche Scientifique, UMR6090, Valbonne, France
    Search for more papers by this author
  • Anne Lazzari,

    1. Institut National de la Santé et de la Recherche Médicale Unité 924, Groupe Avenir, Valbonne, France
    2. Centre National de la Recherche Scientifique, UMR6090, Valbonne, France
    Search for more papers by this author
  • Frédéric Brau,

    1. Université de Nice-Sophia Antipolis, UFR Sciences, Nice, France
    2. Centre National de la Recherche Scientifique, UMR6090, Valbonne, France
    Search for more papers by this author
  • Emilie Narni-Mancinelli,

    1. Institut National de la Santé et de la Recherche Médicale Unité 924, Groupe Avenir, Valbonne, France
    2. Centre National de la Recherche Scientifique, UMR6090, Valbonne, France
    Search for more papers by this author
  • Nicolas Glaichenhaus,

    1. Institut National de la Santé et de la Recherche Médicale Unité 924, Groupe Avenir, Valbonne, France
    2. Centre National de la Recherche Scientifique, UMR6090, Valbonne, France
    Search for more papers by this author
  • Frédéric Geissmann,

    1. King's College London, Centre for Cellular and Molecular Biology of Inflammation, London, England, UK
    Search for more papers by this author
  • Grégoire Lauvau

    Corresponding author
    1. Institut National de la Santé et de la Recherche Médicale Unité 924, Groupe Avenir, Valbonne, France
    2. Université de Nice-Sophia Antipolis, UFR Sciences, Nice, France
    3. Albert Einstein College of Medicine, Department of Microbiology and Immunology, Bronx, NY, USA
    • Albert Einstein College of Medicine, Department of Microbiology and Immunology, 1301 Morris Park Avenue, Price Center, Room 520, Bronx, NY 10461, USA Fax: +1-718-678-1085
    Search for more papers by this author


Memory CD8+ T lymphocytes are critical effector cells of the adaptive immune system mediating long-lived pathogen-specific protective immunity. Three signals – antigen, costimulation and inflammation – orchestrate optimal CD8+ T-cell priming and differentiation into effector and memory cells and shape T-cell functional fate and ability to protect against challenge infections. While among the conventional spleen DCs (cDCs), the CD8α+ but not the CD8α cDCs most efficiently mediate CD8+ T-cell priming, it is unclear which subset, irrespective of their capacity to process MHC class I-associated antigens, is most efficient at inducing naïve CD8+ T-cell differentiation into pathogen-specific protective memory cells in vivo. Moreover, the origin of the required signals is still unclear. Using mice infected with the intracellular bacterium Listeria monocytogenes, we show that splenic CD8α+ cDCs become endowed with all functional features to optimally prime protective memory CD8+ T cells in vivo within only a few hours post-immunization. Such programming requires both cytosolic signals resulting from bacterial invasion of the host cells and extracellular inflammatory mediators. Thus, these data designate these cells as the best candidates to facilitate the development of cell-based vaccine therapy.


Defining the cells and molecules that control CD8+ T-cell priming and differentiation into effector and memory cells in vivo is still being hotly debated in both basic and vaccine immunology. Three signals – antigen, costimulation and inflammation – are necessary for optimal CD8+ T-cell priming and differentiation into effector and memory cells 1. During priming, CD8+ T cells form stable contacts with APCs such as DCs that present pathogen-derived peptides on their cell-surface MHC class I molecules. Among the mouse DC subsets 2, CD11chigh conventional dendritic cells (cDCs), the major subsets of which include CD8α+ and CD8α DCs, are essential for the priming of primary pathogen-specific CD8+ T-cell responses 3, 4. Upon recognition of pathogen-associated molecular patterns (PAMPs), i.e. danger signals and sensing of the inflammatory cytokine environment, DCs undergo rapid maturation. The extent of their activation depends on the initial triggering stimuli 5 that can directly impact the fate of CD8+ T cells differentiation 1. In mice infected by Listeria monocytogenes (Lm), inadequate cDC activation correlates with impaired development of protective CD8+ T-cell memory 6–8.

Evidence accumulated over the past years suggested that CD8α+ cDCs play a unique role in priming CD8+ T cells, in particular because of intrinsic features of their MHC class I processing machinery 9. CD8α+ cDCs have also been shown to be endowed with optimized functional characteristics to induce pathogen- and tumor-specific CD8+ T cells to differentiate into primary effector cells 10–13. However, whether these cells or even CD8α cDCs, independently of their respective capacity to process MHC class I-associated antigens, are capable of integrating all pathogen-derived signals and conveying them to naïve CD8+ T cells to become long-lasting pathogen-specific protective memory cells in vivo is not known. While both cytosolic and/or extracellular-derived signals likely contribute to such cDC licensing, the relative impact of these signals has not been extensively investigated. Lack of such knowledge is mostly due to technical limitations. In fact, adoptive transfer of DC subsets from immunized animals has been difficult to interpret since these cells contain virulent pathogens that can directly infect recipient hosts and activate long-term immunity. Selective in vivo depletion of APC subsets also suffered from the specificity of the depletion 4, 14. To circumvent these issues, we designed an experimental system in which APC subsets could be purified from mice immunized with the intracellular bacterium Lm lacking the SecA2 auxiliary secretion system (secA2 or ΔSecA2 Lm) 15, 16 which induce protective immunity only upon infection with high numbers of bacteria (107). SecA2Lm also exhibit impaired spreading from cell to cell and do not efficiently infect APCs from recipient mice. Thus, taking advantage of this experimental set-up, we could ask whether a subset of cDC is indeed more efficient at inducing protective CD8+ T-cell memory in vivo.


Immunization of mice with increased numbers of secA2Lm restores protective CD8+ T-cell memory

We previously demonstrated that mice immunized with low numbers (106) of secA2Lm develop memory CD8+ T cells that do not protect against a secondary infection with wt bacteria 16, 17. Since SecA2 partially controls the secretion of a subset of bacterial proteins, we hypothesized that induction of protective memory CD8+ T cells may require the secretion of a sufficient amount of at least one SecA2 substrate protein inside the cytosol of infected host cells to generate the appropriate priming environment. Therefore, we reasoned that the cytosolic signaling defect should be restored by immunizing mice with an increased dose of secA2Lm. To test this hypothesis, mice were either immunized with the usual numbers (106) or tenfold more (107) secA2Lm, with wt Lm (3000) or injected with PBS. Three weeks later, all groups were challenged with high numbers of wt Lm (3×105) and viable bacteria inside the spleen and the liver were enumerated 48 h later (Fig. 1A). As expected, PBS-injected animals exhibited 36 000- and 1500-fold more bacteria in spleen and liver respectively than protected mice, i.e. primarily immunized with wt Lm. Mice inoculated with 106secA2Lm also failed to control the wt Lm challenge infection with 3400- and 140-fold more bacteria in their organs than protected animals. Interestingly, mice injected with the higher dose of secA2Lm (107) exhibited few viable bacteria in their organs, and were similarly protected as the wt Lm-immunized group. Comparable results were obtained using wt BALB/c or C57BL/6 mice, suggesting no or minimal impact of the genetic background in this phenomenon (not shown). Also, even though a tenfold range of secA2Lm were injected, the kinetics of bacterial clearance from infected organs was comparable (not shown), likely ruling out a much longer presentation of the bacterial antigens in protected animals. As expected 18, protection in these mice was abolished upon CD8+ T-cell depletion (not shown), demonstrating that protective immunity also required memory CD8+ T cells. Therefore, increasing the immunizing dose of secA2Lm restores the development of CD8+ T-cell-mediated long-term protection.

Figure 1.

Increasing the immunizing dose of secA2Lm restores memory CD8+ T-cell-mediated protective immunity. Mice (BALB/c) were injected with PBS or immunized with 3000 wt, 106 or 107secA2Lm. (A) Three wk later, animals were challenged with 3×105 wt Lm and 48 h later, the number of live Lm (mean+SE) in the spleen (left panel) and the liver (right) enumerated. (B) 6 and 48 h after secondary challenge, spleen cells were stained with H2-Kd/LLO91–99 tetramers (Tet+) and anti-CD8 mAb and the number of Tet+cells enumerated. (C) 6 h after the recall infection, splenocytes were restimulated with the LLO91–99 peptide for 4 h and CD8+ T cells analyzed by FACS for intracellular CCL3 staining. Numbers of CCL3-producing CD8+ T cells and frequencies among tet+ CD8+ T cells (% of activation) are shown out of a pool of 3 (n=11 mice) (A) and two experiments (n=8 mice) (B and C). p-Values were calculated between groups immunized with 106 and 107secA2Lm or with 106secA2 and wt Lm.

Mice immunized with 107secA2Lm exhibit increased numbers of CCL3+ memory CD8+ T cells

We next analyzed the primary and secondary CD8+ T-cell responses as well as memory CD8+ T cells in all groups of mice. Mice primarily immunized with 107secA2Lm exhibited increased numbers of primary effector CD8+ T cells (day 8, Supporting Information Fig. 1A–C) as compared with those infected with wt Lm. Interestingly, the number of memory cells 30 days later, and 6 and 48 h after the secondary infection (Fig. 1B, C and Table 1 and the Supporting Information Fig. 2A) also increased. In all groups, primary and secondary activated as well as memory (day 30) CD8+ T cells specific for distinct Lm-presented antigenic peptides exhibited comparable surface expression of CD62L, CD44, CD127, KLRG-1, expressed granzyme B, and secreted IFN-γ and TNF-α to comparable extent (Fig. 1 and the Supporting Information Figs. 1 and 2). Because we had previously shown that early (6 h) secretion of the chemokine CCL3/MIP1α by memory CD8+ T cells is required for protective response against secondary listeriosis and is lacking in mice immunized with the low (106) dose of secA2Lm17, we monitored CCL3 production in all groups of non-challenged and challenged animals (Fig. 1B, C, Table 1 and the Supporting Information Fig. 2B). As expected, the number of CCL3+ memory CD8+ T cells in animals immunized with 106secA2Lm was lower than in mice that received wt Lm. Importantly, mice primarily injected with 107secA2Lm exhibited increased numbers of CCL3-secreting memory CD8+ T cells than those that received 106secA2Lm. The frequency of cells producing CCL3 among tetramer+ CD8+ T cells was also twice as high and equivalent to that of mice immunized with wt Lm (Fig. 1C) suggesting that increasing the immunizing dose of secA2Lm restored the ability of reactivated memory CD8+ T cells to secrete CCL3 in vivo. Of note, this analysis gave comparable results on two distinct mouse genetic backgrounds and over three distinct naturally presented Lm-derived H2-Kd-restricted epitopes 19 and the H2-Kb-restricted SIINFEKL OVA-derived model epitope (Table 1). Therefore, protective immunity in mice immunized with wt and 107secA2Lm correlates with CCL3 expression and higher numbers of effector memory CD8+ T cells. Thus, we established an original experimental system using different doses of the same mutant bacteria that do or do not prime protective immunological memory, and in which the signals integrated by the priming APCs are likely distinct.

Table 1. Summary of Lm-specific memory CD8+ T cells cell-surface and functional characteristics (day 30)
Cell surface phenotypeNumber of tet+ cellsCD62LCD44CD127KLRG-1
Primary Lm immunizationwtΔSecA2wtΔSecA2wtΔSecA2wtΔSecA2wtΔSecA2
  106107 106107 106107 106107 106107
  1. a

    This table summarizes the data presented in the Supporting Information Fig. S1 on the phenotype of memory CD8+ T cell (day 30, steady state) under the different conditions of immunization and for distinct antigen-specificities on BALB/c and C57BL/6 mouse genetic backgrounds. For clarity purposes, the relative frequencies or numbers between the different experimental groups are represented with ‘+’ or ‘+/−’; identical numbers of these symbols are comparable; lo, low; hi, high; n.d., not determinable.

Functional features (peptide restimulation)% IFN-γ+ cells among CD8+ T cells% CCL3+ cells among IFN-γ+ T cells% IFN-γ+ cells among IFN-γ+ T cells      
Primary Lm immunizationwtΔSecA2wtΔSecA2wtΔSecA2      
  106107 106107 106107      
P60449–457/Kb+++/−+ n.d.  n.d.       

CD8α+ cDCs are permissive to Lm growth in vivo

Efficient induction of long-term protective immunity requires the escape and the growth of Lm in the cytosol of infected cells 16, 20. We therefore looked for the cell subsets that sustain active Lm growth inside their cytoplasm in vivo. To define such cells, mice were immunized i.v. with 106 or 107secA2Lm-expressing GFP that is only expressed by viable Lm as GFP expression is rapidly lost upon bacterial death 16. 2.5, 5 and 10 h later, spleens were harvested and stained with cell surface markers allowing the discrimination of the different myeloid-derived cell subsets containing live bacteria (Supporting Information Figs. 2 and 3). At both early time points analyzed (2.5 and 5 h), CD8α+ cDCs were the main subset of cells expressing GFP (75.2 and 64.4 % respectively), and containing viable bacteria (Supporting Information Fig. 3A), 16), as also reported for wt Lm21. Interestingly, intracellular staining of spleen cells using serum against Lm antigens, which detects both live and dead Lm as well as secreted bacterial antigens, showed that innate phagocytes, i.e. neutrophils, inflammatory monocytes and macrophages, represented 69 and 62% of the positive spleen cells 2 and 5 h after the immunization respectively (Fig. 2 and data not shown), a result supporting their role in the uptake and the killing of Lm22. Therefore, while CD8α+ cDCs represent 20–30% of the Listeriapos cells, they are the major cell type exhibiting live Lm (65–75%), likely providing the most ‘hospitable’ intracellular environment for Lm growth in vivo. Since CD8α+ cDCs are permissive to Lm growth, it makes them likely to integrate and convey signals from cytoplasmic bacteria early after immunization.

Figure 2.

CD8α+ cDCs contain mostly live bacteria at early time points. Mice were injected with PBS or immunized with 1×107secA2-GFP+Lm. About 2.5 and 5 h later, splenocytes were positively enriched for CD11c+ or CD11b+ cells and further stained with mAbs against CD11c, CD8α, CD11b, Ly-6C and intracellular Lm antigens (Listeria+). Pie charts show the proportions of different cell subsets containing whole Lm antigens (upper) or live Lm (lower) among the total number of Listeria+ or GFP+ cells respectively and are representative of two replicate experiments (n=6 mice).

Transferred CD8α+ cDCs are more efficient than CD8α cDCs in priming Lm-protective CD8+ T-cell memory in vivo

Previous reports showed that CD8α+ and CD8α cDCs prime naïve Lm-specific CD8+ T cells with equivalent efficiency when loaded with exogenous peptide ex vivo 11. Since CD8α+ cDCs are the main infected cells, we wondered whether they were functionally better to induce long-term protective immunity in vivo than CD8α cDCs, independently of their antigenic load and ability to process the antigen. For this, the two splenic populations of cDCs were purified from mice immunized with a protective number (107) of secA2Lm early after injection (5 h) and adoptively transferred to naïve recipient animals (Fig. 3A). To minimize live bacteria transfer, cells were incubated in vitro with ampicillin (less than 100 viable secA2Lm were enumerated after such treatment, data not shown). To rule out the effect of epitope density, cells were pulsed with an excess of the ovalbumin (OVA)-derived SIINFEKL MHC class I epitope, an exogenous model antigen that is not naturally expressed by wt Lm. Of note, the cell surface expression level of MHC class I molecules was comparable between the different subsets of DCs and under the distinct immunization procedures (Supporting Information Fig. 4). Thus, with this experimental protocol, bacterial immunization was used as an adjuvant to induce cDC maturation, allowing the assessment of the impact of Lm infection on the DCs. Three wk later, recipient mice were challenged with a high dose of Lm-expressing OVA (Lm-OVA) or not (control), and their ability to clear the infection was monitored by determining splenic bacterial titers after 3 days (Fig. 3B). As shown, after challenge with Lm-OVA, mice transferred with CD8α+ and CD8α cDCs exhibited respectively 70- and 3-fold less viable bacteria than non-transferred animals. Moreover, CD8α+ cDCs were more than 20-fold more efficient at inducing protective immunity than CD8α cDCs from the same animals (Fig. 3B). Of note, when challenged with wt Lm that does not express OVA, mice did not efficiently clear the infection, demonstrating that OVA peptide-pulsed DCs transfer only primed OVA-protective responses (Fig. 3B). Therefore, as early as 5 h following primary infection, CD8α+ cDCs have acquired all the functional features necessary to induce protective immunity.

Figure 3.

Infected CD8α+ cDCs control protective memory CD8+ T cells development. (A) Schematic outline of the experimental protocol. Mice were injected with anti-CD8β mAb (H35) prior to the immunization to prevent the concomitant transfer of primed CD8+ T cells with purified/pulsed cDC. (B, C) 2.5×105 OVA-derived SIINFEKL peptide-pulsed CD8α+ or CD8α cDCs were i.v. transferred into naïve mice. (D, E) Splenic CD8α+ cDCs were sorted into GFP+ and GFP CD8α+ cDCs and 5×102 or 2.5×105 cells were respectively transferred into recipients. In (C) and (E) mice received 5×104 GFP-expressing OT-I cells 48 h before cDC transfer. Three wk later, all recipient mice were challenged with 2×104 wt Lm or 105 wt Lm-OVA. Data show splenic bacteria titers (B and D) and absolute numbers of GFP+ OT-I cells (C and E) 3 and 5 days after the challenge infection respectively (mean+SE). In (B) bar graphs show a pool of two replicate experiments (n=12 mice). In (D) is shown the number of bacteria for each mice pooled from two replicate experiments (n=5–13 mice). In (E) is shown the average of n=5–15 mice pooled from two replicate experiments. p-Values were calculated between groups immunized with CD8α+ and CD8α cDCs or with GFP+ and GFP CD8α+ cDCs from mice infected with 107secA2Lm.

We then monitored the memory CD8+ T-cell response in mice transferred with the two distinct subsets of cDCs (Fig. 3C). To best track memory cells, we took advantage of an adoptive transfer system in which recipient mice were injected with 5×104 GFP-expressing naïve OT-I CD8+ T cells. GFP+ OT-I cells were purified from OT-I×ubiquitin–GFP 23 mice and because these cells constitutively expressed the GFP, we could easily follow their fate inside Lm-OVA immunized hosts as we previously described 24. Following the same experimental scheme as in Fig. 3A, mice were challenged with Lm-OVA and the number of secondary activated OT-I cells was enumerated after 5 days. While ∼3×105 primary expanded OT-I cells were recovered from control mice that did not receive immunizing cDCs, 2×106 OT-I cells were found in animals transferred with CD8α+ cDCs purified from mice infected with 107secA2Lm (Fig. 3C). OT-I memory cells accounted for the eight-fold better expansion observed in the latter group of mice. Interestingly, mice that received CD8α cDCs exhibited 6×105 OT-I cells, a number only twice increased in comparison with control animals. These differences are directly correlated to the lower proliferation of primary activated Lm-specific CD8+ T cells in mice immunized with 106 but not 107secA2 or wt Lm (Supporting Information Fig. 1A). Collectively our results suggest that CD8α+ cDCs most efficiently induce bacteria-specific memory CD8+ T cells that can mediate protective immunity against a recall infection in vivo.

Infected, rather than uninfected, CD8α+ cDCs exhibit a superior ability to prime CD8+ T-cell memory

To test whether Lm growth inside the cytosol of CD8α+ cDCs is licensing these cells to optimally prime memory CD8+ T cells, we performed the same experiment as above (Fig. 3A) by transferring either purified GFP (2.5×105 cells) or GFP+ CD8α+ cDCs (∼500 among 2.5×105 DCs, which is equivalent to that of the transferred CD8α+ cDCs in the previous experiments, Fig. 3B and C) from animals immunized with the protective dose of GFP+secA2Lm. These cells contained live bacteria at the time of purification, thus had received signals from cytosolic Lm. As shown in Fig. 3D, the majority of mice (9 out of 13) transferred with GFP+ CD8α+ cDCs exhibited a substantial protection (1.5–3 and more logs) in contrast to those that received the non-infected DCs.

We next monitored the memory CD8+ T-cell response in transferred animals (Fig. 3E). As before, recipient mice were injected with GFP-expressing OT-I CD8+ T cells before cDC immunization, challenged with Lm-OVA after 3 wk and the number of OT-I cells enumerated 5 days later. As shown, the number of OT-I cells recovered from animals immunized with GFP CD8α+ DCs was similar to non-transferred mice (Fig. 3E). Interestingly, the small number of transferred GFP+ CD8α+ DCs induced at least five-fold more memory CD8+ T cells than control groups. Thus, in the presence of OT-I, the few transferred DCs consistently promoted the differentiation of higher numbers of memory CD8+ T cells. Of note, we observed much less variability in this assay than in the protection assay (Fig. 3D), likely because we transferred OT-I cells which increased the probability of encounter of the few transferred DC with their cognate T cells inside the secondary lymphoid organs. Collectively, our results suggest that cytosolic signals delivered by replicating bacteria are required for CD8α+ cDCs to become functionally capable of inducing protective bacteria-specific memory CD8+ T cells.

Cytosolic and extracellular signals synergize to ‘license’ CD8α+ DCs to activate protective memory

We next investigated whether the cytosolic signals delivered inside CD8α+ cDCs from mice immunized with the protective dose of secA2Lm was the result of increased numbers of replicating bacteria inside their cytosol. We quantified the number of viable bacteria per infected GFP+ CD8α+ cDC 2.5, 5 and 10 h after immunization with the protective (107) and the non-protective (106) doses of secA2 Lm (Fig. 4A). Surprisingly, at all time points and in both conditions, CD8α+ cDCs contained the same number of bacteria per cell. To further analyze if infected CD8α+ cDCs from protected and non-protected animals show distinct functional features, we next directly purified and transferred the same numbers of infected (GFP+) OVA-pulsed-CD8α+ cDCs from mice immunized with the two doses of secA2Lm into recipient mice that were subsequently challenged with a high dose of Lm-OVA 3 wk later (Fig. 4B). Mice immunized with GFP+ CD8α+ cDCs from non-protected mice had equivalent bacterial titers as non-transferred animals upon challenge infection. In fact, only GFP+ CD8α+ cDCs from mice immunized with the protective dose of secA2Lm were able to induce substantial levels of immunity. Since the number of bacteria per infected cell is the same between the two conditions of immunization, it suggested that other signals distinct from those given by cytosolic bacteria are allowing CD8α+ cDCs from protected animals to be optimally conditioned to induce CD8+ T-cell protective memory.

Figure 4.

Only infected CD8α+ cDCs from protected mice prime potent long-term protective immunity. Mice were immunized with 106 or 107secA2-GFP+Lm. (A) 2.5, 5 and 10 h after infection, splenic CD8α+ cDCs were positively purified, counted, lysed and plated to determine the number of viable Lm (mean+SD). Data represent a pool of three replicate experiments (n=6 mice). (B) Spleen cells were positively enriched, sorted and 500 SIINFEKL-pulsed GFP+ CD8α+ cDCs from mice infected with 107 or 106secA2Lm were transferred into naïve animals. Three wk later recipient mice were challenged with 105 wt Lm-OVA and splenic bacteria titers determined after 3 days. Data represent a pool of two replicate experiments (n=6 mice).

Extracellular inflammation likely contributes to optimal ‘licensing’ of CD8α+ DCs

Protected mice were immunized with ten-fold more bacteria than non-protected animals, likely leading to a stronger inflammatory environment at the time of DC maturation. To provide support for this hypothesis, we measured the early inflammatory environment (5 h) under the two conditions of immunization (Fig. 5). As proposed, we readily detected a strong inflammatory response that included cytokines and chemokines involved in DC maturation in mice that received 107secA2Lm. Animals injected with the lower numbers of bacteria were comparable to non-immunized control groups and exhibited low levels of inflammation. We next sought to determine whether this finding held true for animals immunized with other well-established protective Lm immunizations, e.g. wt Lm or the attenuated mutant actALm25 (Supporting Information Fig. 5) and monitored several inflammatory mediators (IL-1β, CCL2, IL-12p70 and TNF-α) over a 48 h kinetics. In all groups that received protective immunization (e.g. 107secA2 Lm, 106actALm and 3000 wt Lm), inflammation reached levels that were never measured in mice immunized with the non-protective dose of secA2Lm. In the case of wt Lm immunization, however, such levels of inflammation were only observed at later time points (24–48 h), a result in agreement with former studies 26, which also correlates with the low initial inocula and the growth kinetics of wt Lm in vivo 16. Therefore, collectively these data favor the idea that during a protective immunization, CD8α+ cDCs receive stronger extracellular inflammatory signals than during non-protective immunization, which likely contribute to their optimal maturation in vivo.

Figure 5.

Infection with 107 but not 106secA2L. monocytogenes induces rapid secretion of inflammatory mediators in the spleen. Mice were infected with 106 or 107secA2Lm or injected with PBS. 5 h later, the levels of the indicated chemokines/cytokines were measured in whole spleen homogenates from individual mice (mean+SE). Data represent a pool of two replicate experiments with n=9 mice. p-Values were calculated between groups of mice immunized with 106 and 107secA2Lm.

Infected CD8α+ cDCs from protected mice exhibit the best maturation profile

To further support to our interpretation that both cytosolically delivered and extracellular signals are conditioning CD8α+ cDC optimal programming, we compared the maturation profiles of infected and non-infected CD8α+ cDCs from mice immunized with the two doses of secA2Lm. For this, we monitored the cell surface expression of several maturation markers (CD80, CD86, CD40, PDL1, 41BBL and CCR7) on (i) DCs that underwent maturation only due to extracellular stimuli and (ii) on those that received intracellular signals from live bacteria or bacterial antigens by gating on cells that contained or not detectable Lm antigens (Listeriapos or Listerianeg, respectively) (Fig. 6). As shown, Listerianeg CD8α+ DCs up- (or down-) regulated the distinct maturation markers with a 1.5- to a 2.5-fold difference between mice that received a protective and a non-protective dose of secA2Lm. In agreement with our hypothesis, Listeriapos CD8α+ DCs purified from protected animals also exhibited a stronger modulation of their maturation markers (∼two-fold) than those from non-protected mice. In correlation with this result, cell-surface expression levels of CD86, CD80 and CD40 costimulatory molecules on infected GFP+ CD8α+ cDC only but not on the non-infected cDC (CD8α+ or CD8α) was 2–3 times stronger (Supporting Information Fig. 6). Therefore, in addition to receiving signals from bacteria replicating inside their cytosol, CD8α+ DCs from protected mice integrated additional signals – likely from the stronger inflammatory environment – which accounted for the observed difference of maturation with non-protected mice.

Figure 6.

Infected CD8α+ cDCs from protected mice undergo the best maturation. Mice were injected with PBS or immunized with 106 or 107secA2Lm. 12 h later, splenocytes were positively enriched using anti-CD11c beads, stained for CD11c, CD8α, the indicated extracellular markers and Listeria antigens, using a Lm-specific rabbit antiserum (Supporting Information Table S1). Data show the MFI of the selected cell surface molecules gated on Listeriapos or Listerianeg CD8α+ cDCs and are representative of two replicate experiments on a pool of spleen cells from three individual mice in each experiment.


We have investigated the ability of the two splenic cDC subsets to induce antibacterial memory CD8+ T cells that can protect against a recall infection. We found that CD8α+ cDCs from primary immunized hosts are the most efficient cDC subtype for transferring long-term, anti-Lm memory CD8+ T-cell-mediated protection to naïve recipient animals. Since both DCs subsets were loaded with saturating amounts of the same antigenic peptide and expressed equivalent cell-surface levels of MHC class I molecules, such features are independent of their capacity to process MHC class I-associated antigens. Interestingly, CD8α+ cDCs become endowed with these functional features as early as 5 h following the primary immunization and this requires cytosolic signals that are potentiated by extracellular inflammatory signals delivered by bacterial infection of the host.

Several seminal studies established that cDCs are key players to prime naïve antigen-specific CD8+ T cells in vivo 3, 4. While these reports support a critical function for splenic CD8α cDCs in initiating primary CD8+ T-cell responses in vivo 3, 4, 9, 11, 12, 14, 27–30, none of them had addressed the question of their ability to set memory development, i.e. whether – and to which extent – they exhibit the functional capacity to induce antibacterial protective CD8 memory. By showing that splenic CD8α cDCs become rapidly conditioned to induce anti-Lm protective memory CD8+ T cells and are best to provide such effect in vivo, we highlight a novel feature of these cells. In addition, we uncouple this functional property of CD8α+ cDCs from their ability to process the antigens from the bacteria. Of note, in our transfer experiments it is unlikely that protection resulted from secA2Lm spreading and growth into recipient's spleen cells for several reasons: (i) transferred CD8α+ cDCs contained less than 100 live secA2Lm after ampicillin treatment and 107secA2Lm are required to induce protective CD8 memory, (ii) secA2Lm exhibit a strong cell to cell spreading defect 15, 16, and (iii) a strong and only OVA-specific protective CD8+ T-cell response was observed ruling out any effect that may result from a different load of Lm antigenic peptide on these cells purified from animals immunized with the distinct doses of secA2Lm. Our observation that the CD8α DCs were mostly inefficient to induce protective CD8+ T-cell memory may indeed result from an intrinsically low ability to activate naïve CD8+ T cells and/or to efficiently reach the T-cell area of the spleens after the transfer. An alternative explanation may be that only very few CD8α cDCs are infected in vivo, which prevent them from efficiently inducing CD8+ T-cell memory. In that latter scenario CD8α cDCs would still intrinsically be able to prime protective CD8+ T-cell memory, although this mechanism would only be of minor contribution. Whatever the true explanation is, our report supports a crucial role of CD8α+ cDCs cells for most potent induction of CD8+ T-cell memory.

Recent studies have shown a role of CD11c+ cells, and in particular CD8α+ cDCs, in the transport of live Lm from the marginal zones to the splenic white pulps, suggesting that the primary function of these cells may be to uptake pathogens to the organs of infected animals, even before the priming of T cells 8, 21, 31. However, others 22, 32 suggested that marginal zone macrophages, but not CD8α+ cDCs, are taking up particulate antigens as well as dead bacteria (Lm, E. coli and S. aureus) from the blood. Here and in agreement with a previous study 33, we reconcile these discrepancies by showing that (i) the great majority of spleen cells staining positive for Lm antigens (i.e. containing live, dead Lm or soluble Lm antigens) are phagocytes (macrophages, neutrophils and monocytes) that also express antimicrobial effector functions and (ii) CD8α+ cDCs, which are specialized APCs, represent the main subset of live bacteria-containing cells. Even though our experiments used the secA2 mutant of Lm, our results are in line with those from other laboratories that used wt Lm. We had also previously shown that the early distribution of live (GFP+) secA2Lm matched that of wt and actALm16, collectively suggesting that this experimental system may help us unravel the mechanisms of protective immunization. Therefore, our results support the idea that phagocytes rapidly capture and kill the majority of blood-injected bacteria whereas CD8α+ cDC provide a replicative niche, thus representing the most actively infected cell type in vivo. In such context, it is tempting to speculate that only direct priming and not cross-priming is inducing fully competent and protective memory CD8+ T cells, a still ongoing controversy in the field 34–36. However, because of our experimental design in which DCs subsets were loaded with an excess of exogenous antigenic peptide before transfer, we cannot draw any definitive conclusions as to which pathway of antigen processing and presentation is most efficient for inducing protective CD8+ T-cell memory in vivo.

How do splenic CD8α+ cDCs become able to imprint the functional characteristics of memory cells? DCs can sense the environment by expressing intra- and extracellular PRRs 5. During Lm infection, bacterial escape to host cell cytosol and SecA2-dependent cytosolic signaling are both necessary to induce memory CD8+ T-cell-mediated protective immunity 16–18, 20. Here, we further suggest that these signals likely converge to a specific subset of spleen cDCs, the CD8α+ cDCs, that then is sufficient to deliver all information to naïve CD8+ T cells. We also show that direct microbial-derived signals from inside their cytosol are required for this phenomenon. This is in contrast to the LCMV infection model that involves cross-priming by CD8α+ DCs as direct infection of DCs prevents their capacity to initiate the cytotoxic T-cell response 37. Thus, splenic CD8α+ DCs licensing by an intracellular bacteria and a non-cytolytic virus arose from distinct mechanisms.

Since the number of live Lm per infected CD8α+ cDCs is identical in protected and non-protected animals, cytosolically delivered signals are likely similar on a per cell basis. However, immunizing recipient mice with the exact same numbers of infected CD8α+ cDCs purified from both conditions of immunization demonstrated that only cells from protected mice induced protective memory, suggesting that CD8α+ cDCs from protected mice receive distinct extracellular signals that likely play a critical role in optimizing their functional features, independently of the level and duration of presented antigenic peptides (DC were pulsed with exogenous peptide before transfer). In fact, we observed a better maturation profile of CD8α+ cDCs and a much stronger inflammatory environment in the spleen of mice immunized with the protective dose of secA2Lm. Since most Listeria+ spleen cells are phagocytes, they may be the cells that provide such extracellular signals to infected CD8α+ cDCs 38, 39. Of note, the chemokines/cytokines detected within this early splenic inflammatory environment of protected animals are also involved in DCs maturation 39–41.

Previous reports showed that CD4+ T cells optimally differentiate into Th1 effector and memory cells only when primed by DCs that have received direct microbial-derived danger signals 38, 39, 42. Indirect release of inflammatory mediators only or lack of inflammation on PAMP-activated DCs failed to support such differentiation. Here we found that two levels of bacterial signals (i) from inside the cytosol and (ii) from the extracellular microbial-derived inflammation need to be delivered to the priming APC to promote pathogen-specific memory CD8+ T-cell differentiation. This model reconciles former observations demonstrating that DCs maturation can be induced after stimulation with pro-inflammatory cytokines in vitro, but also that direct stimulation by PAMPs is indispensable to generate immunogenic DCs, i.e. able to induce full T-cell differentiation 27, 38, 39.

Materials and methods


BALB/c ByJ and OT-I TCR-transgenic (Charles Rivers), C57BL/6J (Janvier), and ubiquitin–GFP-expressing mice 23 (Jackson) were housed and bred in our SPF animal facility. Unless otherwise specified in the legend of the figures, wt C57BL/6 mice were used in the experiments. This study was carried out in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the Commitee of Animal Care and Use of the Regional Cote d'Azur. The protocol was approved by the Committee on the Ethics of Animal Experiments of the Institut de Pharmacologie Moléculaire et Cellulaire (Permit Number: B-06-152-5, delivered by the Veterinary Services of the Alpes-Maritimes Prefecture) and by the animal use committees at the Albert Einstein College of Medicine. All efforts were made to minimize suffering and provide humane treatment to the animals included in the study.


We used the L. monocytogenes 10403s background strain in all experiments, either wt or deleted in the secA2 gene, expressing or not GFP 16. Wt Lm-OVA was a kind gift from Hao Shen (University of Pennsylvania, PA, USA).

Infection of mice, measure of protective immunity and mice survival

For infections, Lm were grown to log phase (OD600∼0.05–0.15) in broth heart infusion (BHI) medium (Sigma-Aldrich), diluted in PBS and injected in the lateral tail vein. For Lm titers, organs were dissociated on metal screens (water 0.1% Triton X-100), and serial dilutions plated onto broth heart infusion plates.

Cell suspensions

Spleens were digested 20 min at 37°C in HBSS (Invitrogen) containing 4000 U/mL collagenase I (Invitrogen) and 0.1 mg/mL DNase I (Roche). Red blood cells were lysed for 5 min in 170 mM NH4Cl, 17 M Tris-HCl and pH 7.4.

Antibodies and reagents

All fluorochrome-labeled mAbs are listed in the Supporting Information Table S1. PE-conjugated LLO91-99/H2-Kd tetramers were obtained from the NIH tetramer core facility.

Antibody staining and flow cytometry

Splenocytes were stained with the specified antibodies in PBS containing 0.5% BSA (FACS buffer). For surface staining, cells were incubated for 20 min on ice. For intracellular staining, splenocytes were incubated for 4 h at 37°C, 5% CO2 in RPMI1640 (Invitrogen) 5% FBS, 2 μg/mL Golgi Plug (BD) with or without 100 nM LLO91–99 peptide (Mimotopes), fixed in 1% paraformaldehyde/FACS buffer 10 min, incubated 20 min in 1× Perm/Wash (BD). Cells were analyzed on a FACSCalibur cytofluorometer (BD). When indicated, cells were sorted on a FACSVantage SE cell sorter (BD).

Assay for cytokines secretion

Organs were homogenized in PBS containing a complete protease inhibitor cocktail (Roche), centrifuged 10 min 12 000×g. The supernatants were incubated with the BD Cytometric Bead Assay Flex Sets and analyzed using a FACS Array (BD).

Cell enrichment and transfer experiments

For CD8+ T-cell transfers, splenocytes from OT-I GFP+ mice were enriched using anti-CD8α MACS beads (Miltenyi Biotec, purity>98%) and the indicated numbers of cells were infused i.v. into recipient mice. For DC transfers, 5 h after the immunization, spleens were harvested, collagenase/Dnase digested and cells were centrifuged in dense BSA (35%) to obtain a cell fraction with a low buoyant density 43. CD8α+ cDCs were positively selected using anti-CD8α-specific MACS beads and flow-sorted on CD8α and CD11c expression (purity ∼98% of CD8αhighCD11chighLy6Cneg cells). CD8α cDCs were positively enriched using anti-CD11c-specific MACS beads and flow-sorted as above (purity ∼98% of CD8αnegCD11chigh). Before i.v. transfer into recipient mice, cDCs were pulsed with 1 μM OVA SIINFEKL peptide in RPMI1640 1% FBS and 2 mg/mL ampicillin for 1 h, 37°C.

Statistical analysis

In all experiments, statistical significance was calculated using an unpaired Mann–Whitney test and Instat software. All p-values of 0.05 or less were considered significant and referred to as such in the text.


We thank T. Dilorenzo (AECOM, USA) and M. Dalod (CIML, France) for critical reading of the manuscript, F. Larbret (C3M, France) for cell-sorting and the AECOM Cytofluorometry Facility. Work was supported by grants from INSERM (Avenir), Human Frontier Science Program (CDA), Agence Nationale de la Recherche (ANRs: IRAP-2005, MIE EMICIF-2008) and Fondation pour la Recherche Médicale (Nouvelles Approches en Immunothérapie 2008). L. C. and E. N. M. received MENRT and FRM fellowships.

Conflict of interest: The authors declare no financial or commercial conflict of interest.