The data presented in this study identified an important role for DCs in generating efficient memory recall responses to multiple pathogens. Thus far, a potential role for DCs in mounting a secondary CD8 T cell response to infection has not been described. Indeed, only limited data are available with regard to the role of endogenous DCs in initiating immune responses in vivo. Although the transfer of antigen-loaded, in vitro-generated DCs has been used to drive responses in vivo, only recently has a system become available to study DC involvement in immune response initiation in situ (Jung et al., 2002
). This system was employed to demonstrate a DC requirement for driving a primary CD8 T cell response in the spleen to Lm infection. We have now utilized this model to analyze the secondary CD8 T cell response to various infectious agents. Considering that many memory cells are located in nonlymphoid tissues and are closely apposed to potentially infected MHC class I-bearing parenchymal cells (Masopust et al., 2001
), we reasoned that DCs might be dispensable for secondary immune response initiation. Moreover, the lower activation threshold of memory CD8 T cells suggests that costimulation may not be required, and thus, nonprofessional APCs could mediate reactivation as has been previously suggested (Bachmann et al., 1999b
). However, our results indicated an important role for DCs in activation of CD8 memory T cells after secondary infection.
Our findings demonstrated that the recall response to Lm infection was decreased >90% in all organs analyzed when DCs were depleted at the time of infection. Other effects of the absence of DCs in our system have potential to influence the CD8 T cell recall response. For example, DC removal could affect lymphoid architecture and thus the location of memory T cells. Distinguishing such effects from the role of DCs themselves in memory cell reactivation, however, will be difficult if both occur simultaneously. In addition, DC ablation may also prevent the DCs from acting as a conduit between the helper CD4 T cell and the CD8 memory cell. Nevertheless, the inhibition of the Lm-specific CD8 T cell response could not be solely attributed to a loss of CD4 T cell help, because depletion of CD4 T cells generally reduces the CD8 T cell recall response by ~50% (Marzo et al., 2004
). Thus, our data indicated that other cell types such as macrophages, which can be directly infected by Lm (Unanue, 1997
), were unable to replace the role provided by DCs after Lm rechallenge. Unlike the recall response to Lm, the secondary CD8 T cell response to VSV is not CD4 T cell dependent (Marzo et al., 2004
; Sun et al., 2004
), and VSV is thought to infect a wide range of cell types. Hence, many cell types are potentially capable of presenting viral antigens via MHC class I. Nevertheless, in the absence of DCs, the recall response to VSV infection was reduced by ~81%–90% in all organs analyzed. Although this is a dramatic reduction compared to control mice, the response obtained is nevertheless the result of a significant increase in the numbers of antigen-specific cells that were originally transferred, and this was the case for Lm and influenza virus recall responses as well. This finding suggested that either not all DCs were depleted by DT treatment or that other APCs or parenchymal cells were involved in antigen presentation and activation of memory cells. Importantly, the reduced recall responses were not due to any unintended effects of DT treatment, as the VSV N peptide recall response could be restored in our model by the transfer of N peptide-loaded DCs. Likewise, it seems unlikely that the small population of GFP+
B cells deleted by DT treatment contributed substantially to the decreased memory recall responses observed, although this has not been formally excluded.
In the case of influenza virus, infection is known to occur primarily in the epithelial lining of the lungs (Collins et al., 2004
). Hence, the initiation of a primary immune response is thought to occur when antigen-loaded DCs migrate to the draining mediastinal LN (Banchereau and Steinman, 1998
). However, a recent study demonstrates that in mice lacking the spleen and LN, the primary CD8 T cell response to influenza virus infection can be initiated, with delayed kinetics, in bronchus associated lymphoid tissue (BALT), whose formation is induced by the infection (Lund et al., 2002
; Moyron-Quiroz et al., 2004
). In our experiments, although many influenza virus-specific memory cells are located in the lung parenchyma, the recall response to infection was reduced by ~60%–90% in the various organs analyzed. Like in the studies by Lund et al. (2002)
, a delayed increase in total antigen-specific cells in the lung parenchyma, BAL, and spleen was observed. Thus, the number of antigen-specific cells in the parenchyma doubled in the DT-treated mice between days 8 and 11 after infection, whereas increases of 54% and 85% were observed in the BAL and spleen, respectively. In contrast, the response in control mice decreased during this time in the lung parenchyma, presumably due to clearance of the infection. Both the BAL and spleens of the control mice exhibited slight increases in antigen-specific cells though not of the same magnitudes as that seen in the DTR mice. Despite an increase in antigen-specific cells in the BAL, the response was still considerably lower when DCs were removed, although there was an equivalent total cellularity compared to control BAL fluid. The large cell number in the DT-treated animals was a reflection of the nonspecific recruitment of cells into the lung airways due to inflammation (Ely et al., 2003
). In any case, the results suggested that as compared to secondary responses to VSV and Lm infection, the recall response to influenza virus infection was less dependent on DCs. In accordance with this finding, a previous report examining epitope dominance in the CD8 T cell response to influenza virus infection suggested that non-DCs are involved in initiation of at least a portion of the secondary response (Crowe et al., 2003
). Despite this result, protection against infection was clearly DC dependent (Figure S1
Whether DC-dependent recall responses are being initiated in the BALT, LN, or nonlymphoid tissue is unknown, but our results suggested that at least some measure of memory cell activation may occur in the parenchymal tissues, and further, that effector memory cells in these tissues were responding to secondary challenge. A recent study suggests that although effector and central memory T cells both exhibit rapid effector function, central memory T cells are more proficient at mounting a proliferative recall response (Wherry et al., 2003
). In contrast, our results showed that splenic CD62Lhigh
memory cells both mounted robust recall responses in response to VSV infection (Figure S2
). Moreover, effector memory cells in the lung can mount a vigorous proliferative recall response to respiratory virus infection, perhaps more effectively than central memory cells (Roberts and Woodland, 2004
). Interestingly, CD62Llow
memory cells were less dependent on DCs for reactivation than were CD62Lhigh
cells. This may be due to a greater DC requirement for activation of lymphoid versus nonlymphoid memory cells, though additional analysis is required to address this possibility. Whether DCs, non-DC APCs, or parenchymal tissues are capable of inducing effector functions in either memory T cell population remains to be seen and is a subject of further investigation.
Our findings also shed light on the duration of memory cell-DC interaction required for optimal reactivation after VSV infection. Not surprisingly, DCs were essential in the first 24 hr of the recall response. However, substantial DC dependence was noted when DCs were removed 24 and 48 hr after infection. Whether this effect was the result of disruption of preexisting T cell-DC conjugates or due to inhibition of formation of new interactions will require analysis with in situ visualization techniques. These results were perhaps unexpected given the current data for naive CD8 T cell activation, which suggests that a short (<24 hr) encounter with antigen is sufficient to drive an optimal response (Badovinac et al., 2002
; Kaech and Ahmed, 2001
; Mercado et al., 2000
; Van Stipdonk et al., 2001
). Because memory cells respond more rapidly than do naive T cells, one would not expect memory cells to require a longer period of antigen stimulation to reach a full response. The differences observed may be due to the use of in vitro systems in some cases and the incomplete removal of antigen/APCs in the in vivo studies. For example, in a study of naive CD8 T cell activation, removal of infectious Lm by antibiotic treatment 24 hr after infection did not affect the overall CD8 T cell response (Mercado et al., 2000
). In our model, the infectious agent was not removed, but the means to present antigen via DCs was eliminated at various time points after infection. Thus, as previously noted (Mercado et al., 2000
), removal of the pathogen in vivo does not remove DCs that are primed and loaded with antigen, and our data indicated that, at least for memory T cells, antigen presentation by DCs is required for more than 24 hr in order to drive the maximal proliferative response of memory CD8 T cells.
By relying on DCs to initiate a recall response, the immune system adds a layer of protection against auto-reactivity in the same manner as it does for naive T cells. The rapidity and strength of the recall response, which are the advantages gained in having memory T cells, occur primarily by increased sensitivity at the level of the T cell without sacrificing the need for antigen presentation by DCs. The relative contribution of other potential APCs, professional or otherwise, in mounting and sustaining the recall response remains the focus of further investigation.