After an infection, the immune system generates long-lived memory lymphocytes whose increased frequency and altered state of differentiation enhance host defense against reinfection. Recently, the spatial distribution of memory cells was found to contribute to their protective function. Effector memory CD8+T cells reside in peripheral tissue sites of initial pathogen encounter, in apparent anticipation of reinfection. Here we show that within lymph nodes (LNs), memory CD8+ T cells were concentrated near peripheral entry portals of lymph-borne pathogens, promoting rapid engagement of infected sentinel macrophages. A feed-forward CXCL9-dependent circuit provided additional chemotactic cues that further increase local memory cell density. Memory CD8+ T cells also produced effector responses to local cytokine triggers, but their dynamic behavior differed from that seen after antigen recognition. These data reveal the distinct localization and dynamic behavior of naive versus memory T cells within LNs and how these differences contribute to host defense.
► CM CD8+ T cells reside in the cortical ridge and near the LN capsule ► Naive and CM CD8+ T cells largely engage distinct APCs upon LN infection ► CXCL- and CXCL9-dependent feed-forward circuit recruits local CM CD8+ T cells ► TCR versus cytokine stimulation leads to distinct CM CD8+ T cell dynamics
Host defense against infection operates on multiple spatial and temporal scales. Epithelial and mucosal tissues form physical barriers to pathogen entry and elaborate broadly active antimicrobial substances (Ashida et al., 2012; Kim et al., 2010). Soluble and cellular components of the innate immune system are the next layer of protection, operating within minutes of a barrier breach (Janeway and Medzhitov, 2002) and contributing to subsequent adaptive immunity involving antigen-specific B and T lymphocytes (Pulendran and Ahmed, 2006). These lymphocytes generate effector cells and antibodies over several days to weeks and play crucial roles in the clearance of infections (Boehm, 2011).
Tissue microanatomy and cellular positioning enable the immune system to perform its functions efficiently. Although chemokine guidance of inflammatory cells is well appreciated (Rot and von Andrian, 2004), there is a renewed interest in how cells are localized in tissues between infectious episodes so as to enhance responses when pathogens invade. We have shown how such positioning operates in terms of innate lymphoid elements in lymph nodes (LNs). γδ, NKT, NK, and a subset of innate-like CD8+ T cells reside near the sites of LN pathogen entry (subcapsular sinus [SCS], interfollicular area [IFA], and medullary sinus [MS]) (Kastenmüller et al., 2012). They respond within an hour of skin infection to locally released cytokines from sentinel SCS macrophages that first contact the invading organism. The resulting IFN-γ promotes an antimicrobial state in the macrophages and limits systemic pathogen spread. Others have described locally resident innate lymphoid cells in lung and elsewhere that likewise contribute to rapid antipathogen responses (Nanno et al., 2007; Shi et al., 2011; Spits and Cupedo, 2012).
CD8+ T cells play particularly important roles in adaptive immune host defense against intracellular pathogens (Harty et al., 2000), producing effector cytokines such as IFN-γ or TNF-α (Harty et al., 2000;Zhang and Bevan, 2011) or directly killing infected cells via perforin or granzymes (Cullen and Martin, 2008). Static immunohistochemistry and dynamic intravital imaging have revealed that naive CD8+ T cells reside within the central paracortical region of LNs (Lämmermann and Sixt, 2008) where they scan for antigen-bearing dendritic cells by migrating in contact with fibroblastic reticular cells along which dendritic cells are aligned (Bajénoff et al., 2006). Upon contact with DCs bearing cognate antigen, CD8+ T cells arrest and interact with the presenting cell, resulting in activation, initiation of proliferation, and acquisition of effector capacity (Bousso and Robey, 2003; Mempel et al., 2004; Stoll et al., 2002). Among the progeny cells, some become short-lived effector cells (SLECs), attacking infected cells either within the LN or at the peripheral site of invasion. Others become central memory (CM) cells that circulate among LNs awaiting signs of reinfection (Cui and Kaech, 2010). Highly localized spatial positioning contributes to the protective activity of other memory CD8+ T cells. Effector memory T cells maintain residence in peripheral tissues at the site of a cleared infection (Gebhardt et al., 2009; Jiang et al., 2012; Wakim et al., 2010), where they rapidly respond to the same pathogen re-entering through the same portal or reaching the same organ site.
The strategic positioning of effector memory T cells raises the question of whether CM CD8+ T cells might also show preferential localization within LN to augment their capacity to fight a secondary infection. Although the location and early postactivation dynamic behavior of naive T cells have been well studied (Henrickson et al., 2008a), less is known about where CM CD8+ T cells reside in LN, their motility, and their behavior upon reinfection (Chtanova et al., 2009). To acquire insight into these issues, we have employed ex vivo cell analysis, immunohistochemistry, and dynamic intravital 2-photon (2P) imaging. Surprisingly, we found that most CM CD8+ T cells did not reside in the deep paracortex like naive CD8+ T cells but were predominantly prepositioned in the IFA and beneath B cell follicles close to high endothelial venules (HEVs), strategically prepositioned to rapidly encounter cells infected by pathogens that spread via lymphatics. A feed-forward circuit involving CXCL9-CXCR3 interaction played a crucial and nonredundant role in concentrating these cells at the sites of infection within the LN, where they arrested and mediated antigen-specific effector function. This prepositioning appeared to enhance the efficiency of antigen-specific as well as noncognate, cytokine-driven host defense. Taken together with related findings published while this study was under review (Sung et al., 2012), these data reinforce the emerging theme of tissue anatomy as a central player in effective immune responses and provide specific insight into the distinct locations and dynamic behaviors of naive and memory CD8+ T cells within LNs.
Naive CD8+ T Cells Are Primed in the Cortical Ridge of the LN upon Viral Infection
To examine the localization and dynamic behavior of naive versus CM CD8+ T cells before and after infection, we first re-explored the temporal and spatial events that occur during the priming of a naive CD8+T cell response (Hickman et al., 2008). We used a replication-deficient vaccinia virus (VV), modified vaccinia virus Ankara (MVA), as a model for acute viral infection. MVA NP-S-GFP (that targets GFP to the nucleus of infected cells) was injected into the footpad of mice, and the primary targets of virus in the dLN (draining lymph node) were analyzed by intravital 2P microscopy. GFP-expressing cells were seen as early as 1 hr after infection (Figure 1A; Movie S1A available online), with the majority situated directly beneath the capsule. Through s.c. injection of labeled anti-CD169 to mark CD169-expressing cells in situ, we identified the most infected (GFP+) cells as being positive for that marker (Figures 1B and S1A; Movie S1A). After infection with MVA encoding cytosolic GFP, fine processes extending from the infected CD169+ cells and their movements were readily visualized (Figure 1C; Movie S1B). Previously, both CD169+CD11c− and CD169+CD11c+ cells were regarded as macrophages based on sensitivity to clodronate killing (Asano et al., 2011). However, based on the recent identification of a DC-specific transcription factor, CD169+CD11c+ cells have been reclassified as DCs (Meredith et al., 2012; Satpathy et al., 2012). Because of this heterogeneity of CD169+ cells, we refer to such cells as SCS myeloid cells.
Figure 1. Priming of Naive CD8+ T Cells Predominates in the Cortical Ridge(A–E) Maximum projections of 2P images (z stack 90–120 μm) of pLNs acquired in situ.(A) Time course after MVA NP-S-GFP infection; second-harmonic generation (SHG, capsule); time is shown as hr:min:s.(B) Image from the SCS region 4 hr after infection with MVA NP-S-GFP (nuclear) and s.c. injection with anti-CD169.(C) Image from the SCS region 6 hr after infection with MVA GFP (cytosolic).(D and E) Images from the SCS region at different time points after MVA OVA infection showing OT-I T cells and capsule (D) and at 4 hr postinfection showing HEV (blood tracer) in relationship to OT-I T cells (E).(F and G) Track length and average speed of OT-I T cells and polyclonal control CD8+ T cells 3 hr p.i. ∗∗p ≤ 0.01.(H) Number of OT-I T cells present in the field of imaging in the course of MVA OVA infection.(I) Positioning of OT-I T cell clusters (white circles) in relation to virus-infected cells, B cells, and collagen IV, 4 hr after infection with MVA OVA GFP. Inserts show GFP signal within OT-I clusters.The data are representative of three (A–C) or five (D–F) similar experiments; bars show mean ± SEM. See also Figure S1 and Movies S1 and S2.
We then examined the dynamics of naive T cells after infection and identified where they encounter antigen that leads to migration arrest and clustering. Naive (CD44lo) OT-I T cells (specific for ovalbumin) and (CD44lo) polyclonal CD8+ T cells were sorted, stained with distinct cell tracker dyes, and transferred into recipient animals 1–4 days prior to infection and imaging. Mice were then infected with MVA OVA in the footpad and intravital 2P analysis of the draining popliteal LN was started 90 min later. Initially, both antigen-specific and polyclonal CD8+ T cells moved randomly in the interfollicular area (IFA), as reported (Movie S2A and data not shown; Henrickson et al., 2008b). By 2 hr after infection, we could observe migration arrest of some antigen-specific OT-I T cells (Movie S2A). 3 hr p.i. we observed increasingly large clusters of arrested OT-I T cells, and by about 4 hr p.i. the majority of OT-I but not polyclonal T cells had stopped migrating (Figure 1D; Movie S2A).
To better understand the relationship of the larger clusters to key histologic features of the LN, fluorescent wheat-germ agglutinin was injected i.v. to visualize the blood vessels. This revealed that the clusters were close to HEVs (high endothelial venules) (Figure 1E; Movie S2A), consistent with previous findings (Bajénoff et al., 2003). Additionally, antigen-specific T cells were found near the capsule in the peripheral IFA. At the 3–4 hr time point, the average speed of antigen-specific OT-I T cells was significantly reduced (4 μm/min) as compared to nonspecific polyclonal CD8+ T cell (10 μm/min), as was the track length (Figures 1F and 1G;Movie S2B). Despite this migration arrest behavior, the total number of OT-I T cells in the imaging field remained largely stable over time, arguing against directed recruitment of naive cells (Figure 1H).
These data were obtained in the superficial 200 μm of the LN accessible to 2P imaging. To examine where T cell antigen recognition and clustering took place in the whole LN, OT-I T cells were labeled and transferred, and recipient mice were infected with MVA OVA GFP. 8 hr later the LNs were fixed, sectioned, and stained for confocal analysis. This analysis confirmed clustering in the IFA but additionally revealed clusters in a subfollicular region previously named the cortical ridge (Katakai et al., 2004). Additionally, T cell clusters were found in the peripheral paracortical region adjacent to the medullary zone, which, in analogy toKatakai et al. (2004), we term the medullary ridge. High-power examination showed that there was always a GFP+ (infected) DC in the middle of a cluster, indicating direct interaction with infected cells at this early time point (Figures 1I and S1B). These dynamic and static imaging analyses thus revealed a distributed but mostly peripheral rather than central paracortical localization of early sites of CD8+ T cell interaction with directly infected cells.
Individual Canonical Chemokine Receptors Are Not Required for Formation of Naive T Cell-APC Clusters
A role for chemokine guidance in enhancing naive T cell interaction with DCs under certain conditions has been previously reported (Castellino et al., 2006; Semmling et al., 2010), as has a role for CCR5 in the recruitment of naive CD8+ T cells to LN sites of VV infection (Hickman et al., 2011). The data in Figure 1showing that for several hours, CD8+ T cell density did not increase in the IFA despite its heavy concentration of infected cells, raised the question of whether any canonical type 1 chemokine receptor (CCR) such as CCR2, CCR5, CXCR3, or CXCR6 promoted efficient clustering of naive CD8+ T cells around infected APCs. To examine this issue, OT-I T cells deficient in each of these CCR were transferred together with WT OT-I T cells and the mice infected with MVA OVA. 8 hr later, the infected LNs were harvested, fixed, sectioned, and stained to evaluate the ratio of WT to CCR-deficient T cells in clusters. Under these conditions we could not detect any difference in antigen-induced clustering between WT and CCR-deficient T cells (Figure S1C). There were also no significant differences between WT and any CCR-deficient T cells tested (including Cxcr3 and Ccr5) in terms of accumulation in the peripheral IFA close to the SCS myeloid cell layer when using highly purified naive (CD44lo) CD8+ T cells (Figure S1D). These data reveal that canonical CCR do not play a major role in guiding CD8+ T cells to initially infected cells in LN, in contrast to their role in promoting efficient interaction among certain lymphoid and myeloid cell populations.
Memory CD8+ T Cells Are Prepositioned in the Cortical and Medullary Ridge and IFA of the LN
We next examined how memory CD8+ T cells behaved under such conditions. To generate memory CD8+T cells, we employed two strategies (see Experimental Procedures), generating CM CD8+ T cells in LN based on their high CD62L/CD127 and low KLRG-1 expression (Figure S2A). 60 days after priming we additionally transferred a new cohort of naive dye-labeled OT-I T cells into these memory mice. Comparison of the distribution of the memory and naive OT-I T cells within the LN prior to viral challenge revealed a striking and unexpected difference. Most of the memory CD8+ T cells were found in a subfollicular (cortical ridge) location as well as close to the medulla (medullary ridge) with some residing even more peripherally in the IFA (Figure 2A; MD4 model not shown). In contrast, naive CD8+ T cells were predominantly in the deep paracortex, as previously shown. This qualitative visual impression was confirmed quantitatively by calculating the average minimal distance of memory and naive CD8+ T cells from a virtual LN center (Figures 2B and 2C) or from the LN capsule (Figure S2B).
Figure 2. CM CD8+ T Cells Are Positioned in LN Periphery in the Steady State(A) IF images showing pLNs harboring naive and CM OT-I T cells.(B and C) Relative distances from a virtual LN center of naive and CM CD8+ T cells displayed as a dot plot (B) or as a frequency distribution (C).(D and E) Average speed and track length comparing naive and CM OT-I T cells in the steady state.(F and G) Average speed and track length of memory OT-I T cells near to the LN capsule in the steady state.(H) Combined tracks of CM CD8+ T cells (near capsule) over 40 min. White lines indicate border to B cell follicles.The data are representative of five similar experiments; bars show mean ± SEM (∗∗∗p ≤ 0.001; ns, nonsignificant). See also Figure S2 and Movie S3.
2P imaging was utilized to further examine this differential localization and to investigate the dynamic behavior of CD8+ memory T cells in the popliteal lymph node (pLN). There was a much higher density of memory OT-I T cells in the region close to the capsule as compared to deeper areas in the paracortex (Figure S2C; Movie S3A). In contrast, naive CD8+ T cells were rather sparse in the LN periphery and at much greater density in the paracortex (Figure S2C; Movie S3A). Staining for CD44 on LN sections also showed a strong signal in the LN periphery, representing both memory and innate lymphocyte populations (Figure S2D;Kastenmüller et al., 2012). Transfer of sorted polyclonal CD44hi versus CD44lo CD8+ T cells from LCMV-infected animals 12 months postinfection showed a similar peripheral positioning of memory versus naive CD8+ T cells (Figure S2E). Despite the difference in location, most naive and memory OT-I T cells had similar speeds (10 μm/min) and similar track lengths (Figures 2D and 2E; Movie S3B). Some memory CD8+T cells migrated close to the LN capsule, scanned the SCS area between B cell follicles in the IFA, and moved at a lower speed (∼7 μm/min) (Figures 2F–2H; Movie S3B), in line with previous findings (Stoll et al., 2002; Worbs et al., 2007). These imaging studies thus revealed an unexpectedly distinct intramodel distribution of naive versus memory CD8+ T cells in the steady state.
Rapid Memory CD8+ T Cell Movement to the Site of LN Infection
To investigate the spatiotemporal behavior of CM CD8+ T cells after infection, naive dye-labeled OT-I T cells were transferred into mice harboring memory OT-I T cells, the animals were infected with MVA OVA, and intravital 2P microscopy was performed. By using high numbers of transferred naive T cells (5 × 106), we were able to visualize similar numbers of naive and memory CD8+ T cells in the IFA despite the differential localization of the two cell populations, although the CM CD8+ T cells still showed a more peripheral localization within this region (Figure 3A). By 90 min after infection, some memory CD8+ T cells and a few naive CD8+ T cells showed migration arrest (Movie S4A). The initial T cell clusters rapidly expanded with more and more memory CD8+ T cells accumulating in the imaged area. Indeed, the number of CD8+ memory T cells in the imaging volume doubled within 1 hr (from 100 to 160 min) after infection (Figure 3B), in contrast to the stable number of naive T cells in a similar imaging volume prior to and during virus-induced clustering in the absence of antigen-specific CM T cells (Figure 1H). Unexpectedly, in this potentially competitive situation, the numbers of naive T cells actually declined over time in the area showing strong CM CD8+ T cell accumulation (Figure 3B). This observation implied both a selective active recruitment of memory CD8+T cells but not naive CD8+ T cells to account for the numerical increase in the former and a limitation of naive T cell access to account for the decline in the latter. In line with these data, the average speed of memory CD8+ T cells was lower than that of naive CD8+ T cells 3 hr after infection, reflecting a larger fraction of cells showing migrational arrest after antigen-specific interaction with infected APCs (Figure 3C).
Figure 3. CM CD8+ T Cells Are Rapidly Recruited to the Site of Infection(A) Maximum projections of 2P images of pLNs acquired in situ showing various time points after MVA OVA infection (z stack of 120 μm, starting beneath the capsule/SHG).(B) Number of naive and CM OT-I T cells present in the imaging field (from A) at the indicated times after MVA OVA infection.(C) Average speed of naive and CM OT-I T cells 3 hr p.i. ∗∗p ≤ 0.01.(D) Positioning of naive and CM OT-I T cell clusters, B cells, and collagen IV 4 hr after infection with MVA OVA. Inserts show magnification of naive and CM OT-I T cell clusters from different locations; yellow numbers indicate distance of cluster from LN center (yellow circle).(E) Maximum projections of 2P images from the pLN acquired in situ (z stack of 120 μm, starting beneath the capsule). White arrows indicate T cell-macrophage interaction before killing and disruption of macrophages.The data are representative of five similar experiments; bars show mean ± SEM. See also Figure S3 and Movie S4.
To more broadly analyze the distribution of memory and naive CD8+ T cells after infection, we stained and imaged frozen LN sections. In comparison to the preinfection state, many CM CD8+ T cells were found even closer to the LN periphery near the infected myeloid cell layer in the medulla and the IFA (Figure 3D). In contrast, naive CD8+ T cells were very sparse in those areas, remaining predominantly paracortical. As seen in animals lacking specific CM CD8+ T cells, clusters of naive CD8+ T cells were found in the cortical ridge as in Figure 1I. Many of these deeper clusters were exclusively formed by naive CD8+ T cells whereas more peripheral clusters were dominated by memory CD8+ T cells (Figures 3D and S3A), indicating a striking distinction in the APCs with which the two T cell populations interacted at these time points. When CM CD8+T cells reached the peripheral infected macrophages, they arrested and killed the infected cells (Figure 3E;Movie S4A). Because the CM CD8+ T cells were prepositioned in the IFA, they first reached the myeloid cell layer covering that area and eliminated the infected cells (Figure S3B). From there they moved beneath the capsule to reach infected myeloid cells that covered the B cell follicles (Figure S3C; Movies S4A and S4B). This finding is of particular interest because this cohort of cells has been recently shown to allow for enhanced viral replication, forming a potentially dangerous niche for pathogen replication (Honke et al., 2012;Moseman et al., 2012). Our analysis thus revealed the repositioning of memory and not naive CD8+ T cells to even more peripheral sites in LN after infection and the distinct sets of presenting cells with which these two subsets of lymphocytes interacted in their zones of localization.
CXCR3 Mediates Memory CD8+ T Cell Recruitment to the SCS after Infection
Given the increase in CM CD8+ T cells in the IFA and their selective postinfection translocation to the SCS (Figures 3B and 3D), we speculated that chemokine signals may recruit CM CD8+ T cells to the infected myeloid cells at the SCS and the medulla. To address this hypothesis, MD4 mice harboring CM OT-I T cells (and other virus-specific memory T cells, unlabeled) were infected with MVA WT to determine whether CM OT-I T cells are recruited to the infected cell layer in the absence of cognate antigen. Indeed, 3–5 hr after infection, we observed a substantial movement of CM OT-I T cells to the SCS (Figure 4A; Movie S5A), where they swarmed in localized regions, suggestive of chemokine-mediated attraction (Figure 4A, yellow insert).
Figure 4. CXCR3 Is Required for CM T Cell SCS Recruitment and Local Swarming(A) Maximum projections of 2P images of pLN acquired in situ showing various time points after MVA WT infection (z stack of 120 μm, starting near the capsule/SHG). Red square indicates magnified area, yellow square insert shows the combined tracks of cells from the blue squares over 40 min.(B) pLN of CD11cYFP mouse 4 hr after infection with MVA WT harboring WT and CXCR3-deficient (Cxcr3−/−) CM OT-I T cells.(C) Relative distance of WT and Cxcr3−/− memory OT-I T cells from the LN capsule 4 hr after infection with MVA WT. ∗∗∗p ≤ 0.001.(D) Combined tracks displaying the movement of CD11cYFP cells, WT, and Cxcr3−/− CM OT-I T cells after MVA WT infection (140–180 min); clusters of WT OT-I T cells are encircled.(E) pLN 4 hr after infection with MVA WT showing WT and Cxcr3−/− CM OT-I T cells, B cells, and collagen IV.The data are representative of five similar experiments; bars show mean ± SEM. See also Figure S4 and Movie S5.
To identify the putative CCR required for such a behavior, we generated CM CD8+ T cells with various Ccr-deficient OT-I and WT OT-I T cells. 60 days later, we challenged these animals with MVA WT and looked for differential localization of Ccr−/− versus WT CM CD8+ T cells. Cxcr3−/− but not Ccr2−/−, Ccr5−/−, or Cxcr6−/−OT-I memory CD8+ T cells showed a different localization after infection as compared to WT OT-I memory CD8+ T cells (Figure S4A). Importantly, all tested Ccr−/− CM OT-I T cells, including Cxcr3−/−, showed steady-state peripheral prepositioning similar to that of WT CM OT-I T cells (Figures S4A and S4B or not shown). To further examine the differences seen between WT and Cxcr3−/− CM T cells after infection and to gain insight into cell dynamics, we generated tdTomato transgenic mice and crossed them with WT OT-I. Then we generated CM T cells with Cxcr3−/− GFP and WT tdTomato OT-I cells in CD11cYFP recipients. 60 days later, these animals were infected with MVA WT and the behavior of the fluorescent-protein-expressing cells was imaged dynamically by intravital 2P microscopy and at sequential time points by confocal microscopy of fixed frozen sections. In line with the above data, we found a specific recruitment of WT (tdTomato) but notCxcr3−/− (GFP) memory CD8+ T cells to the SCS and medullary area after infection (Figures 4B and S4C). Quantitative analysis of the minimal average distance to the capsule of WT versus Cxcr3−/− OT-I T cells showed a significant difference (Figure 4C). Intravital imaging revealed that WT but not Cxcr3−/− CM OT-I T cells swarmed around CD11cYFP+ cells close to the SCS (Figure 4D; Movie S5B). Magnified images of the IFA clearly confirmed the differential distribution of WT versus Cxcr3−/− T cells after MVA WT infection (Figure 4E). These data clearly identified CXCR3 as a key contributor to the peripheral repositioning of CM CD8+ T cells early after LN infection.
A Feed-Forward Local Recruitment Circuit Involving CXCR3+ Memory CD8+ T Cells and CXCL9
There are three known ligands for CXCR3: CXCL9, CXCL10, and CXCL11. Cxcl11 is not transcribed in C57BL/6 mice due to a stop codon. CXCL9 and CXCL10 can both be induced by IFN-γ, and CXCL10 can additionally be induced by type I IFN (Groom and Luster, 2011). We first examined where and by which cells in the LN CXCL9 and CXCL10 are produced. CD11cYFP reporter mice were infected with MVA WT and 4 hr later stained for these chemokines and various cellular markers. CXCL10 staining was found colocalized with medullary macrophages (data not shown). CXCL9 was found in the medulla and in the IFA where it partially colocalized with SCS myeloid cells and CD11cYFP+ cells close to the SCS (Figures 5A and S5A). At least some of the remaining CXCL9 staining involved stromal cells (Figure S5B). A few HEVs close to the SCS also showed intense CXCL9 signal (data not shown). Specific depletion of LN-resident myeloid cells by clodronate liposomes reduced CXCL9 staining to background levels 4 hr after infection with MVA (Figure S5C). The specificity of CXCL9 staining on LN sections was validated with MVA-infected Cxcl9−/−animals (Figure S5D).
Figure 5. CM CD8+ T Cells Amplify CXCL9 Production(A) Confocal IF image showing CXCL9 staining in the IFA of a pLN from a CD11cYFP mouse 4 hr p.i. MVA WT.(B) Absence of WT OT-I CM T cell recruitment to the SCS in a pLN from a Cxcl9−/− mouse 4 hr after MVA WT infection.(C) Quantitative ELISA for CXCL9 in LN homogenates 4 hr after infection of the indicated mouse strains with MVA OVA; insert shows statistical significance (ns, not significant; ∗p ≤ 0.05,∗∗p ≤ 0.01, ∗∗∗p ≤ 0.001).(D) IF images of a pLN 4 hr p.i. MVA OVA Tomato showing CM OT-I and IFN-γ staining.(E) IF images of pLN 4 hr p.i. MVA OVA of CD11cYFP mice. 7 days before infection mice received an s.c. injection with clodronate liposomes to deplete LN-resident myeloid cells.Data are representative of three independent experiments (A, B, D, and E) or show pooled data from two independent experiments (C); bars show mean values. See also Figure S5.
To assess the role of CXCL9 in CM CD8+ T cell recruitment to the SCS upon viral infection, we transferred CM OT-I T cells into Cxcl9−/− animals and analyzed their position after MVA WT infection. In this circumstance, CM CD8+ T cells failed to accumulate in the SCS upon viral infection (Figure 5B). There were no differences in localization of WT and Cxcr3−/− OT-I T cells in infected Cxcl9−/− recipient animals, with both populations failing to show movement toward the SCS (Figure S5E). In contrast, recruitment of WT CM OT-I but not Cxcr3−/− CM OT-I to the SCS was intact in Cxcl10−/− mice (Figure S5F). These data imply a nonredundant role of CXCL9 with respect to CM CD8+ T cell recruitment to the IFA and medulla.
To quantitatively examine the basis for CXCL9 production in the SCS region, we measured the amount of CXCL9 produced in the absence of type I or type II IFN (Groom and Luster, 2011) and in the absence or presence of antigen-specific CM CD8+ T cells (a likely source of IFN-γ). We infected WT, Ifng−/−, Ifnar−/−, or WT mice harboring CM OT-I T cells with MVA OVA and harvested the draining LN 4 hr later. By quantitative ELISA we detected ∼1.7 ng of CXCL9 per LN 4 hr after infection (Figure 5C). This amount was significantly reduced in LN from Ifnar−/− (0.69 ng/LN) and Ifng−/− (0.85 ng/LN) mice, though these values were still significantly above those of LN from mock-infected WT animals. Because CXCL9 is not known to be directly induced by type I IFN, the latter may act indirectly by inducing IFN-γ production, e.g., by NK cells in the LN (Kastenmüller et al., 2012). Additionally, the induction of CXCL9 independently of IFN-γ argues for an IFN-independent mechanism (Huang et al., 2012; Xia et al., 2009). In the presence of antigen-specific CM OT-I T cells, the amount of CXCL9 induced after infection was dramatically increased by >5-fold to 10.5 ng/LN. A likely explanation was CXCL9 induction by IFN-γ produced by antigen-activated CM CD8+ T cells upon encounter with infected cells. In accord with this idea, mice bearing CM OT-I T cells and infected with MVA OVA tdTomato showed a strong IFN-γ signal at the synapse between CM CD8+ T cells and infected APCs (Figure 5D).
To further examine the role of CM OT-I T cells as part of a feed-forward loop involving IFN-γ and CXCL9, we made use of the near absence of CXCL9 staining in myeloid cell-depleted LNs after MVA infection (Figure S5C). We predicted that, given their prepositioning, memory OT-I T cells should efficiently find the few infected DCs in the cortical ridge in clodronate-treated LNs. IFN-γ derived from OT-I T cells upon antigen encounter should lead to local CXCL9 production that should be detectable by immunofluorescence and this was found to be the case (Figure 5E), strengthening the concept that the CM T cells contribute to their own recruitment to the SCS region upon viral infection (Figure S6).
Noncognate Memory T Cell Activation Leads to Extensive IFN-γ Production without T Cell Arrest
These findings revealed the unanticipated peripheral LN localization of CM CD8+ T cells in the steady state and the CXCL9-guided repositioning of these lymphocytes during reinfection, a process amplified by a feed-forward mechanism that depended on cognate antigen recognition. We next compared these observations with the dynamics of CM CD8+ T cell upon cytokine-mediated, noncognate activation during infection. As we (Kastenmüller et al., 2012) and others (Kupz et al., 2012; Soudja et al., 2012) have shown, inflammasome-driven release of IL-18 from pathogen-sensing myeloid cells in concert with additional cytokines leads to IFN-γ production by various innate and memory T cells in LN. We considered that such IFN-γ might not only provide enhanced antimicrobial resistance by acting on the myeloid cell population, but also induce CXCL9 production by these cells that could enhance recruitment of additional antigen-specific and nonspecific CM to the infected site.
We first compared to what extent CM OT-I T cells produce IFN-γ upon cognate activation using MVA OVA versus upon noncognate, IL-18-dependent activation induced by Pseudomonas aeruginosa (PA) infection (Kastenmüller et al., 2012). We used mice harboring CM OT-I T cells and infected one footpad with MVA OVA and the other with PA. 4 hr after infection with MVA OVA, about 60% of memory OT-I T cells produced IFN-γ as analyzed by direct ex vivo intracellular staining (Figure 6A). Surprisingly, 4 hr after PA infection as many as 80% of all memory OT-I T cells, significantly more than after MVA OVA infection, produced IFN-γ. After PA infection, the memory OT-I T cells did not arrest in any area of the LN analyzed (SCS, IFA, medulla) (Figures 6B–6D; Movie S6), in striking contrast to the situation after MVA OVA infection (Figure 3C). The average speed of these T cells was slightly below the steady-state speed (7 μm/min versus 10 μm/min;Figures 2D and 6B), but the origin of this effect remains unclear. Confocal analysis of frozen sections of LNs infected with PA showed that CM CD8+ T cells remained in the cortical ridge and did not translocate further to the SCS, unlike after MVA OVA infection (Figures 3D, 6E, and 6F). In line with this we found strong CXCL9 staining in proximity to the cytokine-activated CM CD8+ T cells in the cortical ridge and within HEV, but not among cells near the SCS as seen with MVA OVA infection (Figure 6G). These data are consistent with IFN-γ from the cytokine-activated lymphocytes being responsible for local induction of CXCL9 in the area to which these cells are constrained.
Figure 6. Noncognate Activation of CM CD8+ T Cells Leads to Extensive Nonarrested IFN-γ Production(A) Intracellular IFN-γ staining of CM OT-I T cells in the pLN 4 hr after s.c. infection with MVA OVA or PA.(B and C) Average speed (B) and track length (C) of CM OT-I T cells 3 hr after PA infection.(D) Combined tracks displaying the movement of CM OT-I T cells 3 hr after PA infection.(E and F) IF images of CM OT-I T cells in the pLN 4 hr after PA infection.(G) IF images of CM OT-I T cells after PA infection in relation to CXCL9.The data are representative of three similar experiments; (A) p values and ±SEM are shown; (B) bars represent mean ± SEM. See also Movie S6.
Here we show that CM CD8+ T cells are prepositioned in the cortical ridge and IFA of the LN, close to the entry portals of invading pathogens trafficking in the lymphatics. This is in contrast to the deep paracortical localization of naive CD8+ T cells, with implications for how the two CD8+ T cell subsets perceive antigens derived from invading pathogens and their respective contributions to local host defense. Scout CM CD8+T cells residing near to initially infected cells in the LN periphery undergo rapid arrest on the infected cells and secrete IFN-γ that induces CXCL9 expression from local myeloid and stromal cells. This chemokine acts in a feed-forward manner to amplify the response by recruiting more CXCR3+ CM CD8+ T cells to the infected region, where they rapidly kill cells harboring virus. This local prepositioning of CM T cells near sites of pathogen entry can be analogized to evidence that effector memory CD8+ T cells form a stable tissue-resident population at the site of original pathogen encounter, enhancing protection upon secondary challenge (Gebhardt et al., 2009; Jiang et al., 2012; Wakim et al., 2010). It is intriguing that a similar scheme seems applicable to secondary lymphoid organs, with the periphery of the LN where the pathogens first enter representing the recalled site of infection.
Naive CD8+ T cells appear to be excluded from the peripheral region in which antigen-dependent and chemokine-driven CM responses occur, interacting instead with a distinct set of antigen-presenting cells in the paracortical area. This spatial division enables the memory cells to concentrate their effector function on peripherally infected cells, without interfering with the recruitment of naive T cells into the response by sparing more centrally positioned infected DC (Guarda et al., 2007; Yang et al., 2006). This differs from a model in which distinct DC subtypes preferentially interact with memory versus naive CD8+ T cells (Belz et al., 2007); rather, microanatomic localization and chemokine action leads to association of the two T cell populations with different APCs. Our findings also differ from those obtained by tetramer staining to visualize the endogenous CD8+ T cell response upon infection, which suggested that CM CD8+ T cells reside within splenic B cell follicles rather than the T zone (Khanna et al., 2007). We find CM CD8+ T cells only in the latter, although such cells are closer to B cell follicles than their naive counterparts.
Our data show the predominant areas of naive T cell priming to be the cortical and medullary ridges and the IFA, consistent with some (Bousso and Robey, 2003; Mempel et al., 2004; Miller et al., 2002; Qi et al., 2006) but not other (Hickman et al., 2008) studies. We were unable to confirm a critical role for CCR5 in regulating initial antigen encounter involving naive CD8+ T cells (Hickman et al., 2011), a result that may arise from analysis of distinct viruses (MVA versus VV). However, our data suggest that the number of cells, especially CD44hi cells, in the initial transfer can influence the result.
Although CM CD8+ T cells are prepositioned near pathogen entry sites, a feed-forward chemokine-driven guidance mechanism ensures efficient additional local recruitment of these lymphocytes directly to sites rich in infected cells. This recruitment depends on CXCR3 expression by the CM T cells and results in swarming behavior around DCs in an antigen-independent fashion, promoting contact with infected targets. CXCR3 also shapes the differentiation of naive T cells during viral infections by altering the migration of both CD4+and CD8+ effector T cells (Groom et al., 2012; Hu et al., 2011; Kohlmeier et al., 2011; Kurachi et al., 2011). In our study, CXCL9 rather than CXCL10 was crucially involved in CM CD8+ T cell positioning upon viral infection. The amount of CXCL9 produced after infection was reduced in both Ifng−/− and Ifnar−/− mice, although IFN I signaling is not known to directly induce CXCL9 production. However, IL-18 plus IFN-α is a potent stimulator of IFN-γ production by innate-like lymphocytes, which are positioned in the area where we detected production of CXCL9 (Kastenmüller et al., 2012). Therefore, type I IFN may act with IL-18 to promote CXCL9 expression indirectly via induction of IFN-γ.
An important role in the recruited response also appears to be played by the small number of antigen-specific CM CD8+ T cells that roam near the SCS. Upon antigen encounter, these “scout” CM CD8+ T cells make IFN-γ rapidly and strongly increase CXCL9 precisely at the sites of initial infection, quickly recruiting additional CM T cells positioned nearby. This results in a very high local effector T cell concentration that mediates pathogen containment by effective target cell killing (Budhu et al., 2010). An obvious question is why don’t all CM CD8+ T cells reside in close proximity to the SCS myeloid cells like these scout CM T cells or their innate counterparts. One possible answer is that there may be space or trafficking restrictions, given our finding that the influx of CM CD8+ T cells upon MVA infection reduces naive CD8+ T cell density in the region of CM T cell accumulation (Figure 3B).
During the revision of this report, Sung et al. (2012) reported very similar findings on the chemokine guidance of CM CD8+ T cells within the LN after viral infection. With LCMV as a model system, the authors identified CXCL9 as a crucial chemokine to recruit CM to the SCS and characterized a feed-forward mechanism that involves IFN-γ from activated CM and CXCL9. The authors also identified both macrophages as well as stromal elements as sources for CXCL9 in the LN. Our observations are congruent with and support these previously published results but differ on a few points. One is related to the role of CXCL10. Sung et al. (2012) mainly investigated short-term transfer of in vitro generated CM CD8+ T cells. Interestingly, those CM cells were found to populate the deep paracortex like their naive counterparts and we have confirmed these data with in vitro generated CM OT-I T cells (data not shown). In this prior study CXCL10 was required to relocate these centrally disposed CM CD8+ T cells after infection to the peripheral paracortex, the area in which we find our in vivo generated CM CD8+ T cells to be prepositioned in the uninfected state. Interestingly, we found that P14 CM T cells primed in the context of LCMV infection do occupy a more central position than the in vivo generated OT-I memory cells or bulk CD8+ T memory cells we have analyzed in detail here. We are currently investigating the basis of these differences, which could be infection specific, due to residual antigen, or T cell intrinsic. Although the issue of whether CM CD8+ T cells reside close to the sites of pathogen entry in the steady state is an important one, it should be emphasized that the two studies agree in large measure with respect to the events involved in rapid chemokine-orchestrated relocalization of CM CD8+ T cells to the sites of active viral infection in LNs.
We have not yet determined the chemokines and/or adhesion molecules that regulate the peripheral positioning of most CM CD8+ T cells. We have tested several of the obvious candidates via individual receptor-deficient animals (CCR2, CCR5, CXCR3, CXCR6), antibody blockade (CD44), or chemical inhibition (EBI2), but to date none of these individual molecular manipulations has shown a detectable effect on CM T cell localization. This implies either redundancy in function or the operation of positioning systems that we have not yet examined.
Both naive and CM T cells constantly recirculate between blood and SLO in a search for antigen. This search is primarily considered to give rise to antigen-dependent activation and the generation of effector cells that mediate their effects once they have left the SLO and entered a site of peripheral infection. Here we provide a very different view of CM CD8+ T cell effector function by revealing the prepositioning of these T cells in a distinct region of the LN from naive cells and the local, chemokine-driven enhancement of CM CD8+ T cell numbers immediately around infected cells at the site of pathogen entry into the LN. Taken together, these findings reinforce the growing awareness of how tissue microanatomy subserves function in orchestrating efficient host defense.
Mice were purchased from Jackson Laboratory, kindly provided by other investigators, or obtained from Taconic Laboratories through a special NIAID contract. All mice were maintained in SPF conditions at an Association for Assessment and Accreditation of Laboratory Animal Care-accredited animal facility. All procedures were approved by the NIAID Animal Care and Use Committee (National Institutes of Health, Bethesda, MD). For details see Supplemental Information.
Bacterial and Viral Infections
MVA OVA (pH 5) GFP and MVA OVA (pH 5) dtTomato were constructed as previously described using pLW-73 (Kastenmuller et al., 2007; Wyatt et al., 2009). 108 IU vaccinia virus (MVA) or 107 CFU PA GFP were diluted in PBS and injected in the footpad (30 μl). For IFN-γ blocking in vivo, mice received 1 mg anti-IFN-γ (XMG 1.2; BioXcell) or isotype control (HPRN; BioXcell) i.p. 5 hr before challenge.
Adoptive T Cell Transfer
OT-I T cells or polyclonal control CD8+ T cells were sorted with a MACS CD8-negative selection kit (Miltenyi) combined with biotinylated anti-CD44 (IM7, BD Biosciences). Cells were stained and transferred (2.5 × 106) 1–5 days prior to imaging. For generation of CM CD8+ T cells, 5 × 103 cells were transferred before i.p. immunization with either MVA OVA or OVA protein (500 μg) in combination with poly(I:C) (50 μg) and CD40 antibody (50 μg) (FGK 4.5, BioXcell) (Ahonen et al., 2004).
In Vivo Depletion of LN Macrophages
For in vivo depletion of LN macrophages, mice were injected in the footpad with 20 μl of clodronate containing liposomes or empty liposomes as control (Encapsula) 7 days before infection (Van Rooijen and Sanders, 1994).
For analysis of intracellular cytokine production, preparation of cell suspensions from dLNs and subsequent staining was done in the presence of brefeldin A (1 μg/ml) (Sigma). Cells were surface stained with anti-CD8 (5H10; Caltag Laboratories) and anti-CD44 (IM7, BD Biosciences). Intracellular cytokine staining was performed with anti-IFN-γ (XMG1.2; BD) with the Cytofix/Cytoperm kit (BD Biosciences). Flow cytometric data was collected on an LSR II (BD Biosciences) and analyzed with FlowJo software (TreeStar).
LNs and spleens were harvested and fixed with PLP buffer (0.05 M phosphate buffer containing 0.1 M L-lysine [pH 7.4], 2 mg/ml NaIO4, and 10 mg/ml paraformaldehyde) for 12 hr, then dehydrated in 30% sucrose prior to embedding in OCT freezing media (Sakura Finetek). 30 μm sections were cut on a CM3050S cryostat (Leica), adhered to Superfrost Plus slides (VWR), stained, mounted with Fluormount G (Southern Biotech), and acquired on a 710 confocal microscope (Carl Zeiss Microimaging). For details on antibodies seeSupplemental Information.
Intravital Two-Photon Imaging
Mice were anesthetized, popliteal LNs were exposed, and intravital microscopy was performed by a protocol modified from a previous report (Bajénoff et al., 2006). Raw imaging data were processed and analyzed with Imaris (Bitplane). For details see Supplemental Information.
Student’s t test (two-tailed) and Mann-Whitney test were used for the statistical analysis of differences between two groups with normal or nonnormal distribution, respectively.
We would like to thank J. Farber and A. Luster for providing critical reagents, M. Gerner for help with imaging analysis, and G. Gasteiger, T. Lämmermann, and N. van Panhuys for critically reading the manuscript. We are especially grateful to A. Luster and U. von Andrian for generously sharing unpublished data while this study was in progress or in revision. MVA NP-S-GFP was kindly provided by I. Drexler. This research was supported by the Intramural Research Program, NIAID, NIH, by DFG, KA 3091/1-1 to W.K., and HE 6068/1-1 to J.H.