Introduction
Useful immune responses require the regulated trafficking of T lymphocytes to secondary lymphoid organs, where naive T lymphocytes are activated by antigen-bearing dendritic cells (DCs), and subsequently to inflamed nonlymphoid tissues, where mainly effector T lymphocytes mediate their activity1. Many insights into the molecular mechanisms underlying the compartmentalization of the inductive and effector phases of immune responses have been gained over the past decade. Naive T cells selectively migrate from blood to peripheral lymph nodes with a 'three-digit code' in which rolling, sticking and firm adhesion are mediated by the interaction of L-selectin, the chemokine receptor CCR7 and the integrin LFA-1 with their ligands (peripheral node addressin (PNAd), CCL21 and intercellular adhesion molecule 1 (ICAM-1), respectively) on the lumenal surfaces of high endothelial venules (HEVs)2, 3. Once they have entered the lymph node, naive T cells migrate to the paracortical T cell zone by establishing physical contacts with fibers of the fibroblastic reticular network bearing the CCR7 ligand CCL21 (ref. 4). CCR7 is also required for the migration of maturing DCs from the peripheral tissues to T cell areas of the draining lymph nodes5, where these DCs seem to integrate into the fibroblastic reticular cell network, positioning themselves for efficient interaction with T cells4, 6. The migrating naive T cells 'scan' the surfaces of the DCs and establish transient or stable contacts of variable duration depending on the nature of the stimulus7, eventually undergoing activation, proliferation and differentiation into effector cells.
Most of the effector T cells that emerge from a productive immune response have lost expression of the lymph node–homing receptors L-selectin and CCR7 and therefore the capacity to migrate to lymph nodes. At the same time, they have acquired the expression of a different set of adhesion molecules and chemokine receptors that equip them for migration into inflamed peripheral nonlymphoid tissues8. Among these are ligands for E- and P-selectin, CXCR3 (the receptor for the interferon-
(IFN-)–induced chemokines CXCL9, CXCL10 and CXCL11) and CCR5 (the receptor for CCL3, CCL4 and CCL5)9, 10, 11, 12.
The segregation of immune functions by migratory constraints described for naive and effector T cells is maintained in the memory compartment13. Indeed, central memory T cells, which are devoid of immediate effector function but rapidly proliferate in secondary responses, express CCR7 and L-selectin and are found in lymph nodes, whereas effector memory T cells, which are characterized by immediate effector function, lack L-selectin and CCR7 and are found in blood, peripheral tissues and spleen but not in lymph nodes14, 15, 16. The exclusion of effector and effector memory T cells, especially cytolytic CD8+ T cells, from the lymph nodes can be viewed as a mechanism to prevent killing of antigen-presenting DCs and therefore preserve their capacity to trigger sustained primary and secondary immune responses. However, it has been shown that immature and mature DCs can be killed in peripheral tissues by cytolytic T cells or natural killer (NK) cells17, 18, although those studies did not find any evidence of such killing in lymph nodes themselves.
Studies have suggested that lymphocyte traffic in lymph nodes might change in inflammatory conditions. Thermal stress has been reported to increase lumenal expression of CCL21 and ICAM-1, allowing increased adhesion of T cells to HEVs and entry into lymphoid tissues19. It has also been shown that NK cells, which are normally excluded from resting lymph nodes, are rapidly recruited to reactive lymph nodes that develop in response to migrating mature DCs or adjuvants20. In the lymph node, NK cells establish interactions with DCs and T cells and enhance the differentiation of T helper type 1 cells through the production of IFN-20, 21. Recruitment of NK cells in reactive lymph nodes is CCR7 independent and is mediated by CXCR3, suggesting that in inflammatory conditions, CXCR3 ligands might be displayed on HEVs and may mediate lymphocyte extravasation. Indeed, it has been shown that tumor necrosis factor (TNF) controls expression of CXCL9 on HEVs and that pDC precursors enter reactive lymph nodes in a CXCR3-dependent way22.
Given the data outlined above on the changing ability of reactive lymph nodes to support the entry of diverse cell types, we have reinvestigated here the issue of whether effector and effector memory CD8+ T cells can enter lymph nodes. In contrast to the generally held view that circulating effector T cells lacking lymph node–homing receptors are segregated from sites of inductive immune response, and in agreement with evidence that effector responses to various viruses act in lymphoid tissues23, 24, 25, 26, 27, we show here that L-selectin-negative CCR7- effector and memory CD8+ T cells were rapidly recruited to lymph nodes produced in response to migrating mature DCs, adjuvant or TNF.
Results
Effector CD8+ T cells migrate to reactive lymph nodes
To investigate whether and how circulating effector T cells may gain access to lymph nodes, we stimulated naive OT-I transgenic CD8+ T cells in vitro and analyzed the migration of primed T cells to resting and reactive lymph nodes after intravenous adoptive transfer into syngeneic mice. After stimulation with large numbers of mature bone marrow–derived DCs pulsed with the MHC class I–binding ovalbumin (OVA) peptide 'SIINFEKL', most of the primed T cells had an effector phenotype characterized by high expression of CD44, loss of L-selectin and CCR7, upregulation of CXCR3, CD49d and CD11a, and acquisition of the ability to bind P-selectin (Fig. 1a). Primed OT-I cells had effector function, as they expressed granzyme B, produced IFN- and killed antigen-bearing target cells in an in vivo cytotoxicity assay (Fig. 1b). We then isolated L-selectin-negative effector T cells by magnetic-activated cell sorting, labeled the cells with CFSE (5-(and-6)-carboxyfluorescein diacetate succinimidyl ester) and injected them intravenously into syngeneic mice. After an equilibration period of 12 h to allow the exit of injected T cells from the lungs28, we gave the mice a subcutaneous injection of CpG DNA–matured, bone marrow–derived DCs in one footpad to produce a reactive popliteal lymph node. We killed the mice 24 h later and analyzed cell suspensions from reactive draining and contralateral resting lymph nodes (Fig. 1c). We recovered very few T cells from the transferred effector population from resting lymph nodes, consistent with their lack of L-selectin and CCR7. In contrast, we recovered effector T cells at higher frequency (about 10-fold) and in larger numbers (about 40-fold) from DC-draining reactive lymph nodes than from resting lymph nodes (Fig. 1c). Tissue sections showed that migrated effector T cells localized mainly in the paracortical T cell areas (data not shown). Effector T cells in lymph nodes were negative for L-selectin (Fig. 1d), ruling out the possibility of selective migration of a small number of residual L-selectin-positive cells or in vivo reversion of the transferred cells to an L-selectin-positive phenotype. Reactive lymph nodes (popliteal or inguinal) capable of recruiting effector CD8+ T cells were produced not only by injection of mature DCs but also by inoculation of TNF or adjuvant, such Ribi or complete Freund's adjuvant, and by local infection induced by Listeria monocytogenes (Fig. 1e and data not shown). We conclude that effector T cells lacking L-selectin and CCR7 are mostly excluded from resting lymph nodes but can be recruited in substantially greater numbers to reactive lymph nodes that develop in response to migrating mature DCs, TNF, adjuvant or local infection.
Figure 1: In vitro–generated L-selectin-negative CCR7- effector CD8+ T cells are excluded from resting lymph nodes but migrate to reactive lymph nodes.
![Figure 1 :
In vitro|[ndash]|generated L-selectin-negative CCR7|[minus]| effector CD8+ T cells are excluded from resting lymph nodes but migrate to reactive lymph nodes.](/ni/journal/v8/n7/thumbs/ni1469-F1.jpg)
Analysis of naive CD45.1 OT-I cells and of effector CD45.1 OT-I cells generated by in vitro culture of naive cells with SIINFEKL-pulsed mature DCs. (a) Phenotypes of naive OT-I cells and effector OT-I cells after 7 d of culture. (b) Expression of granzyme B, production of IFN- and in vivo cytotoxic activity of effector OT-I cells. Granzyme B and IFN- were measured by intracellular staining after 4 h of stimulation with phorbol 12-myristate 13-acetate and ionomycin. Cytotoxicity was measured by flow cytometry 16 h after mice were injected with CFSEhi SIINFEKL-pulsed and CFSElo unpulsed splenocytes at a ratio of 1:1 in the absence (Control) or presence of effector T cells (4 
106 cells). Numbers above bracketed lines indicate percent positive cells. (c) Selective recruitment of effector OT-I cells in reactive but not resting lymph nodes (LN). In vitro–generated effector OT-I cell samples at day 7 depleted of contaminating L-selectin-positive cells were labeled with CFSE and were injected intravenously (4 106 cells per mouse) into naive C57BL/6 mice; 6–12 h after T cell transfer, a reactive popliteal lymph node was induced by subcutaneous injection of 3 106 mature DCs in the footpad. Data are presented as percent and absolute numbers of CFSE+ CD45.1+ cells recovered after 24 h (mean 
s.d. of three mice). (d) Expression of L-selectin in CD8+ CD45.1+ effector OT-I cells recovered from reactive lymph nodes and spleens after 24 h. Numbers in quadrants indicate percent cells in each. (e) Recruitment of effector OT-I cells to reactive lymph nodes produced in response to mature DCs, TNF, adjuvant or localized infection. Mice previously injected with effector OT-I cells (2.5 106 to 3 106 cells) were then injected with CpG-matured DCs (3 106; CpG DC), TNF (200 ng), Ribi (25 
g) or L. monocytogenes (LM; 0.6 106 colony-forming units). Control (open bars), PBS injection or untreated lymph nodes. Data are the absolute number of effector OT-I cells recovered in the reactive lymph nodes after 24 h (mean s.d. of three to four lymph nodes). Data are representative of at least three (a,b), five (c,d) or two (e) separate experiments.
In vivo effector and memory T cells migrate to lymph nodes
To investigate whether in vivo–generated effector and effector memory CD8+ T cells would also be capable of migrating to reactive lymph nodes, we adoptively transferred small numbers of naive CD45.1 OT-I cells into CD45.2 congenic wild-type mice and immunized the mice with L. monocytogenes expressing OVA. At 7 d after immunization, most of the OT-I cells recovered from the spleen had an L-selectin-negative CXCR3+ phenotype, similar to that of in vitro–primed OT-I cells (Fig. 2a). At a later time point (day 30), a substantial fraction of OT-I cells had an effector memory (L-selectin-negative) CXCR3+ or CXCR3- phenotype in the spleen (Fig. 2a). We obtained similar results with effector and memory T cells induced by immunization with SIINFEKL-pulsed mature DCs (not shown). At this time point, lymph nodes contained, as expected, L-selectin-positive central memory T cells (data not shown), but in mice given an injection of mature DCs or adjuvant, we readily detected L-selectin-negative effector memory T cells (Fig. 2b). When transferred intravenously into CD45.2 congenic mice, the in vivo–generated L-selectin-negative effector and effector memory CD8+ T cells efficiently migrated to reactive lymph nodes (produced by subcutaneous injection of mature DCs) but not to the contralateral resting lymph node (Fig. 2c). As in vivo–generated effector T cells equilibrate more quickly than in vitro–generated cells, we were able to measure early migration of T cells to lymph nodes that had been conditioned 8 h before by injection of mature DCs. We detected effector T cells in the reactive lymph node as early as 35 min after intravenous injection (Fig. 2d). Notably, nearly all of the migrating effector T cells recovered at early as well as late time points expressed CXCR3 (Fig. 2d and data not shown).
Figure 2: Effector and effector memory CD8+ T cells generated by in vivo priming rapidly migrate to reactive lymph nodes.
Analysis of effector and memory OT-I cells after immunization of C57BL/6 mice given adoptive transfer of low numbers of naive CD45.1 OT-I cells and primed with L. monocytogenes expressing OVA. (a) Expression of L-selectin and CXCR3 on in vivo-primed effector (day 7) and memory (day 30) OT-I cells (gated on CD8+CD45.1+V
2+) isolated from the spleen and on effector OT-I cells primed in vitro (as described in Fig. 1). (b) Analysis of effector memory OT-I cells that migrated into reactive lymph nodes of mice containing memory OT-I cells; 3 106 mature DCs were injected into a footpad 34 d after immunization to produce reactive lymph nodes. Data represent absolute numbers (mean s.d. of three mice) of CD45.1+ CD8+ L-selectin-negative effector memory OT-I cells recovered after 24 h from resting or reactive lymph nodes. (c) Analysis of intravenously transferred effector and memory OT-I cells that migrated into reactive lymph nodes. At 6 d (effector) or 40 d (memory) after immunization, OT-I cells were isolated from the spleen, labeled with CFSE and injected intravenously (effector, 10 106 cells; memory, 2 106 cells) into C57BL/6 mice in which reactive lymph nodes were then produced by injection of 3 106 mature DCs. Data are presented as absolute numbers (mean s.d. of four (effector) or two (memory) mice) of CD45.1+ CFSE+ L-selectin-negative OT-I cells recovered in the resting and reactive lymph nodes 24 h after DC injection. (d) Analysis of the expression of L-selectin and CXCR3 on intravenously transferred effector OT-I cells that migrated early to reactive lymph nodes. At 6 d after immunization, OT-I cells were isolated from the spleen, were labeled with CFSE and were injected intravenously (3 106 cells per mouse) into C57BL/6 mice containing a reactive lymph node produced by injection of mature DCs into a footpad 8 h before T cell injection; organs (spleen and resting and reactive lymph nodes) were analyzed 35 min after T cell injection. Data are gated on CD8+CD45.1+V2+ cells. (e) Analysis of intravenously transferred polyclonal L-selectin-negative effector CD8+ T cells that migrated into reactive lymph nodes. C57BL/6 mice were immunized intravenously with L. monocytogenes; on day 6 after immunization, CD8+ T cells were enriched by magnetic-activated cell sorting from the spleen or perfused liver, were labeled with CFSE and were injected intravenously (2 106 CD8+ L-selectin-negative cells per mouse) into C57BL/6 mice, and reactive lymph nodes were induced by injection of 3 106 mature DCs. Data are absolute numbers (mean s.d. of three mice) of CFSE+ CD8+ CD44+ L-selectin-negative cells recovered in resting or reactive lymph nodes 24 h after DC injection. Numbers in quadrants (a,d) indicate percent cells in each. Data are representative of three (a,d) or two (b,c,e) separate experiments.
In a final series of experiments, we isolated polyclonal CD8+ T cell populations from spleens and perfused livers of mice challenged 6 days earlier with L. monocytogenes and adoptively transferred the cells into mice that received injection of mature DCs in the footpad. CD44hi L-selectin-negative effector T cells were present in greater numbers in reactive than in resting lymph nodes (Fig. 2e). These results indicate that the ability to migrate to reactive lymph nodes is a general property of effector CD8+ T cells and that effector T cells present in tissues, once back in the circulation, also have the ability to migrate to reactive lymph nodes.
Effector T cells roll on HEVs of reactive lymph nodes
The rapid kinetics of the recruitment of effector T cells to reactive lymph nodes would be consistent with direct migration from the blood through HEVs. To visualize this migration step, we labeled effector OT-I cells with the fluorescent dye BCECF-AM and, as a control, labeled naive CD8+ T cells with the fluorescent dye TRITC, injected the cells intravenously into syngeneic mice and monitored by intravital video microscopy their interactions with the HEVs of resting and reactive lymph nodes (Fig. 3). Individual vessels analyzed in resting lymph nodes showed efficient rolling of naive T cells but no or limited rolling of effector T cells (Fig. 3a), consistent with the low expression of L-selectin of the latter cell population. In contrast, vessels of reactive lymph nodes supported efficient rolling and adhesion of effector T cells to an extent similar to that of naive T cells in resting lymph nodes (Fig. 3a,b). Furthermore, more effector T cells stuck to HEVs in reactive lymph nodes than in resting lymph nodes (Fig. 3b). These results collectively demonstrate selective and efficient rolling and sticking of effector T cells to HEVs of reactive lymph nodes.
Figure 3: Effector T cells roll and stick on HEVs of reactive lymph nodes.
Intravital microscopy of resting and reactive lymph nodes from mice given effector OT-I cell samples generated in vitro as described in Figure 1, depleted of contaminating L-selectin-positive cells to a purity of over 99% by magnetic beads on day 7, labeled with BCECF and then transferred (1 107 cells per mouse). (a) Rolling fraction, determined as the percent rolling cells among all cells passing through a given venule (containing at least ten events) as a function of vessel diameter. Left, naive CD8+ T cells (from C57BL/6 mice) on HEVs of resting lymph nodes. Middle, effector OT-I cells on HEVs of resting lymph nodes. Right, effector OT-I cells on HEVs of reactive lymph nodes induced by injection of 3 106 mature DCs 16 h before. (b) Pooled rolling fractions and number of sticking cells per mm2 in pooled HEVs of resting or reactive lymph nodes (mean s.d.). Data (a,b) are representative of four separate experiments.
CXCR3 in effector T cell migration to lymph nodes
To investigate the requirements for chemokine receptors in the process of the migration of effector CD8+ T cell to reactive lymph nodes, we backcrossed OT-I transgenic mice onto different chemokine receptor–deficient backgrounds to produce mice whose T cells lacked CCR7, CXCR3 or both. We mixed chemokine receptor–deficient naive CD45.2 OT-I cells with an equal number of chemokine receptor–sufficient naive CD45.1 OT-I cells and stimulated the cells in vitro to generate mixed populations of effector T cells in the same activation conditions. Wild-type and chemokine receptor–deficient T cells demonstrated a similar ability to expand clonally and differentiate into effector cells lacking L-selectin and CCR7 (Fig. 4a). We purified L-selectin-negative effector T cells from these mixed cultures by magnetic-activated cell sorting (Fig. 4b), labeled the cells with CFSE, injected them intravenously into mice and measured the proportion of chemokine receptor–deficient T cells relative to wild-type T cells in reactive lymph nodes (Fig. 4c). Migration into reactive lymph nodes was only modestly impaired for the CCR7-deficient T cells, was severely but variably impaired for the CXCR3-deficient T cells and was nearly completely abolished for the double-deficient T cells. The variability of the results obtained with CXCR3-deficient T cells would be consistent with residual functional CCR7 expression undetectable with the specific monoclonal antibody. That possibility was supported by the finding that some of these cells migrated in transwell chamber assays in response to the CCR7 ligand CCL19 (data not shown). The lack of migration of cells doubly deficient in CXCR3 and CCR7 and the efficient migration of CCR7-deficient cells are consistent with a chief function for CXCR3 in the process of the migration of effector CD8+ T cells to reactive lymph nodes.
Figure 4: CXCR3 is required for the migration of effector CD8+ T cells to reactive lymph nodes.
OT-I mice were crossed with mice deficient in CCR7, CXCR3 or both CCR7 and CXCR3; Ccr7-/-, Cxcr3-/- and Ccr7-/-Cxcr3-/- naive OT-I cells were each mixed with an equal number of congenic CD45.1 naive OT-I cells and were stimulated in vitro as described in Figure 1 to generate effector cells. (a) Expression of surface markers on congenic CD45.1+ chemokine receptor–sufficient OT-I cells (dashed lines) or CD45.1- chemokine-receptor–deficient OT-I cells (solid lines) at 7 d after priming. (b) Percent congenic CD45.1+ chemokine receptor–sufficient OT-I cells (WT) and CD45.1- chemokine receptor–deficient OT-I cells at day 7 after priming after depletion of contaminating L-selectin-positive cells and before injection. Numbers in quadrants indicate percent cells in each. (c) Ratio of chemokine receptor–deficient OT-I cells (KO) to congenic CD45.1 OT-I cells (WT) in reactive lymph nodes at 24 h after intravenous injection into mice (1 107 cells per mouse) in which a reactive lymph node was produced by subcutaneous injection of 3 106 mature DCs into the footpad. Data are normalized to the ratio in the spleen of the same mouse, which was similar to the ratio measured before injection (b), and are the mean s.d. of five to six mice. Dots represent individual mice. Data (a–c) are representative of two separate experiments.
Transient CXCL9 expression on the lumenal side of HEVs
The results presented above raised the issue of whether a CXCR3 ligand is present on HEVs of reactive lymph nodes. To address that possibility, we intravenously injected antibodies to the three CXCR3 ligands and visualized the antibodies bound on PNAd+ HEVs by immunofluorescence microscopy of lymph node sections. We detected CXCL9 on PNAd+ HEVs of reactive lymph nodes conditioned by injection of mature DCs, but it was absent from the HEVs of resting lymph nodes (Fig. 5a). CXCL9 was detectable as early as 6 h after DC injection and remained detectable for 1–3 d, then decreased substantially and disappeared at later time points. The kinetics of CXCL9 expression on HEVs correlated with the kinetics of effector T cell recruitment (Fig. 5b), which was maximal at 12–24 h after DC injection and was substantially lower by day 4. In addition to detecting CXCL9, we also detected CXCL10 but not CXCL11 on HEVs of reactive lymph nodes at early time points after DC injection (data not shown). These results indicate that CXCR3 ligands are expressed on the lumenal surfaces of HEVs in a time- and stimulus-dependent way and, along with the data on the chemokine receptor–deficient T cells described above, suggest that CXCR3-CXCR3 ligand pairs have a chief function in the recruitment of effector and memory CD8+ T cells in reactive lymph nodes.
Figure 5: CXCL9 is transiently displayed on the lumenal surfaces of HEVs in reactive lymph nodes.
(a) Immunofluorescence microscopy of reactive lymph nodes induced by subcutaneous injection of mature DCs. At various times after DC injection, mice were injected intravenously with Alexa Flour 594–labeled anti-CXCL9; after 30 min, lymph nodes were collected, fixed and counterstained with rat IgM anti-PNAd followed by Alexa Fluor 488–labeled anti–rat IgM for identification of HEVs. Original magnification, 200. (b) Temporal window of the migration of effector OT-I cells to reactive lymph nodes induced in mice (n = 16) by subcutaneous injection of 3 106 mature DCs; 2 106 L-selectin-negative effector T cells were injected intravenously at various time intervals after DC injection. Data are the absolute numbers of L-selectin-negative effector OT-I cells recovered in reactive lymph nodes at 24 h after T cell injection (mean s.d. of four lymph nodes of two mice per time point). 0, T cell migration into resting lymph node. Data are representative of three separate experiments (a,b).
Effector T cells kill antigen-presenting DCs in lymph nodes
The finding that effector and effector memory CD8+ T cells migrated from the blood into the paracortical regions of reactive lymph nodes suggested the possibility that these cells might interact with antigen-presenting DCs in the lymph nodes. To investigate that possibility, we labeled SIINFEKL-pulsed and unpulsed DCs with different dyes and injected the cells into the footpads of mice that had previously received effector OT-I cells intravenously (Fig. 6a). In the absence of T cell transfer, we recovered both pulsed and unpulsed DCs in the draining lymph node in similar proportions. In contrast, in the presence of effector T cells, we recovered very few pulsed DCs from the lymph node, whereas unpulsed DCs were present in the expected numbers, suggesting that DCs were killed in a cognate way by effector CD8+ T cells. We obtained similar results when we labeled pulsed and unpulsed DCs with the reciprocal combination of dyes (data not shown). The disappearance of SIINFEKL-pulsed DCs was proportional to the number of injected effector OT-I cells, reaching 90% after injection of 2 106 cells (Fig. 6b), which resulted in an effector/target ratio of approximately 4:1 in the draining lymph node (calculated assuming an equal number of pulsed and unpulsed DCs and counting the number of effector T cells in the lymph node). Notably, the loss of SIINFEKL-pulsed DCs was also mediated by in vivo–generated effector and memory CD8+ T cells (Fig. 6c).
Figure 6: Effector and memory CD8+ T cells kill antigen-presenting DCs in reactive lymph nodes.
(a–c) Analysis of cells after SIINFEKL-pulsed mature DCs (OVA-DC) and unpulsed mature DCs (U-DC) were labeled with CFSE or CMTMR, respectively, and were injected at a ratio of 1:1 (3 106 DCs per mouse) into the footpads of C57BL/6 mice given L-selectin-negative effector OT-I cells generated in vitro (6 106 cells (a) or the numbers along horizontal axis in b) 6 h before, or mice containing naive, effector (day 7) and memory (day 36) OT-I cells primed in vivo with L. monocytogenes expressing OVA (as described in Fig. 2). (a) Recovery of CFSE+ OVA-pulsed DCs and CMTMR+ unpulsed DCs in reactive lymph nodes 30 h after DC transfer. Numbers below outlined gates indicate percent cells in each. (b) Recovery of OVA-pulsed DCs relative to unpulsed DCs in reactive lymph nodes of mice given 'graded' numbers of in vitro–generated L-selectin-negative effector OT-I cells. Data are the mean s.d. of four mice. (c) Recovery of OVA-pulsed DCs relative to unpulsed DCs from reactive lymph nodes 30 h after subcutaneous injection of OVA-pulsed and unpulsed DCs into mice containing naive or in vivo–generated effector or memory OT-I cells. (d) Analysis of interactions between antigen-pulsed DCs and effector OT-I cells in reactive lymph nodes, measured by intravital two-photon lymph node imaging of effector OT-I cells with DCs pulsed with SIINFEKL peptide (OVA) or control SIYRYYGL peptide (CONT) at various times after adoptive transfer of T cells (186 individual T cell–DC contacts). Short horizontal bars indicate median contact duration value; each symbol represents an individual contact. (e) Time-lapse imaging sequences of cell body fragmentation (arrows) of SIINFEKL-pulsed DCs (green) after prolonged contact with effector OT-I cells (red). DCs pulsed with control peptide (purple) are either 'ignored by' or show transient interaction with T cells. Clock, h:min:s. Scale bars, 20 m. More images available in Supplementary Videos 1 and 2. Data are representative of four (a), two (b,c) or three (d) independent experiments or are two examples of eight videos of three independent experiments (e).
The finding of many fewer antigen-pulsed DCs in mice with effector and memory CD8+ T cells recognizing antigen presented by DCs could be the result of interaction with the DCs in peripheral sites, where the T cells might physically retard DC migration or kill the presenting cells while they are still in these tissues. Evidence for such peripheral killing of antigen-bearing DCs by CD8+ effector T cells has been reported18. However, our evidence that effector and effector memory CD8+ T cells gain access to the T cell zone of reactive lymph nodes raised the possibility that killing of the DCs might also occur in this site. To directly investigate that possibility, we used intravital two-photon microscopy to visualize in reactive lymph nodes any interaction between effector OT-I cells and DCs pulsed with SIINFEKL or a control peptide (Fig. 6d and Supplementary Videos 1 and 2 online). Effector T cells interacted transiently with control DCs but established stable interactions with SIINFEKL-pulsed DCs. Concurrent with the establishment of stable interactions, several pulsed DCs showed prominent blebbing and fragmentation consistent with the induction of apoptotic cell death (Fig. 6e and Supplementary Videos 1 and 2) beginning 6–18 h after the introduction of effector T cells into the reactive lymph nodes. These results provide direct evidence that effector CD8+ T cells can recognize and kill antigen-bearing DCs in reactive lymph nodes, a process that may contribute to the small number of antigen-bearing DCs that can be recovered from the lymph nodes of mice with effector T cells present.
Effector and memory CD8+ T cells inhibit T cell priming
The finding that effector and memory CD8+ T cells migrated from the blood to reactive lymph nodes where they seemed to kill antigen presenting DCs indicated that these T cells might limit the activation of naive T cells entering the lymph node or promote the termination of an ongoing response by reducing effective antigen presentation. To examine that issue, we measured the proliferative responses of CFSE-labeled naive OT-II cells that we adoptively transferred into mice containing naive OT-I or in vivo–primed effector or memory OT-I cells. We then challenged the mice with DCs pulsed with major histocompatibility complex (MHC) class I and class II OVA peptides and measured the proliferation of naive OT-II cells by CFSE dilution. OT-II cells that had divided several times were readily detected in the presence of naive OT-I cells, but such divided cells were nearly absent when effector or memory OT-I cells were present (Fig. 7a,b). In different experiments this led to a considerable reduction in the number of CD4+ OT-II cell blasts accumulating in the recipients (2%–5% of control). There was also strong inhibition of the OT-II cell response when the challenge occurred 30 d after immunization with L. monocytogenes expressing OVA, indicating that memory OT-I cells generated in these conditions can also suppress naive T cell clonal expansion (Fig. 7c), as well as when L-selective-negative effector T cells (purity, over 99%) generated in vitro were transferred into mice (data not shown). As a control, we challenged mice with a mixture of DCs pulsed with the MHC class I and class II peptides on separate DCs. In this case, the extent of the OT-II cell response in mice containing effector OT-I cells was similar to that of naive mice (Supplementary Fig. 1 online). These results are consistent with the hypothesis that the lower response is caused by direct killing of antigen-presenting DCs.
Figure 7: Priming of naive CD4+ and CD8+ T cells is inhibited in mice containing effector or memory CD8+ T cells.
(a–c) Proliferation of naive heterozygous CD45.1 and CD45.2 OT-II cells in mice given adoptive transfer of naive CD45.1 OT-I cells and then left untreated (Naive) or immunized (Effector or Memory) with SIINFEKL-pulsed DCs (a,b) or infected with L. monocytogenes expressing OVA (c); recipient mice containing naive or effector OT-I cells (day 7; a) or naive and memory OT-I cells (day 35; b,c) then received a second transfer of CFSE-labeled naive heterozygous CD45.1 and CD45.2 OT-II (1 106 cells per mouse) and were challenged by subcutaneous injection into the footpad of 1 106 DCs pulsed with both MHC class I and class II OVA peptides. (a,b) CFSE dilution and absolute numbers of proliferating heterozygous CD45.1 and CD45.2 OT-II cells in reactive lymph nodes (day 5) of mice containing naive or effector (a) or naive or memory (b) CD45.1+ OT-I cells. (c) Response of heterozygous CD45.1 and CD45.2 OT-II cells in lymph nodes (day 4) of mice in which memory OT-I cells were obtained after infection with L. monocytogenes expressing OVA. (d–f) Proliferation of naive heterozygous CD45.1 and CD45.2 OT-I cells in mice given adoptive transfer of naive CD45.1 OT-I cells and then left untreated or immunized with SIINFEKL-pulsed-DCs (d,e) or infected with L. monocytogenes expressing OVA (f); recipient mice containing naive or effector OT-I cells (day 7; d) or naive and memory OT-I cells (day 35; e,f) then received a second transfer of CFSE-labeled naive heterozygous CD45.1 and CD45.2 OT-I cells (0.5 106 cells per mouse) and were challenged as described in a–c. (d,e) CFSE dilution and absolute numbers of proliferating heterozygous CD45.1 and CD45.2 OT-I T cells in reactive lymph nodes (day 5) of mice containing naive or effector (d) or naive or memory (e) CD45.1+ OT-I cells. (f) Response of heterozygous CD45.1 and CD45.2 OT-I cells in lymph nodes (day 4) of mice in which memory OT-I cells were obtained after infection with L. monocytogenes expressing OVA. All data are from CD45.1+ CD45.2+ CD4+ or CD8+ gated cells. Data (mean s.d. of three mice) are representative of two to four separate experiments.
In the same set of experiments, we measured the response of CFSE-labeled naive OT-I cells transferred into mice along with unlabeled naive, effector or memory OT-I cells (Fig. 7d–f). Mice with only naive OT-I cells had a strong proliferative response. In contrast, in mice with effector or memory OT-I cells, the frequency of naive T cells showing multiple divisions was much lower. In particular, some T cells entered cell division but failed to accumulate, a phenomenon that has been noted in conditions of low or costimulation-deficient antigenic stimulation29, 30. The results presented above collectively indicate that effector and memory CD8+ T cells entering reactive lymph nodes can inhibit naive CD4+ and CD8+ T cell proliferation and that at least part of this effect may result from elimination of DCs presenting the cognate antigen in secondary lymphoid tissues.
Discussion
Here we have shown that circulating effector and effector memory CD8+ T cells lacking L-selectin and CCR7 were excluded from resting lymph nodes but were rapidly recruited to reactive lymph nodes conditioned by migrating mature DCs, TNF, adjuvant or local infection. Effector CD8+ T cells were recruited to reactive lymph nodes in a CXCR3-dependent way through HEVs and seemed to directly kill antigen-bearing DCs in lymph nodes. Such activity contributed to the suppression of naive CD4+ and CD8+ T cell proliferation in the same lymph nodes.
Published studies have shown that effector and memory T cells that extravasate into peripheral tissues can reach the draining lymph nodes through afferent lymphatics31. The idea of the existence of a direct pathway for the extravasation of effector CD8+ T cells into reactive lymph nodes that we have described here is supported by the rapid kinetics of T cell recruitment (30–40 min), by the coincidence of the recruitment of CXCR3+ effector T cells with the presence of the CXCR3 ligand CXCL9 on HEVs and, most importantly, by the direct visualization of efficient rolling and sticking of effector T cells on HEVs of reactive lymph nodes. In addition, the similar recovery of CCR7-sufficient and CCR7-deficient T cells in reactive lymph nodes is inconsistent with migration from tissues, as recovery of the latter cells has been shown to depend on CCR7 expression32, 33. We also have preliminary evidence that CD4+ effector T cells can enter reactive lymph nodes through HEVs. However, the molecular requirements for migration and the consequences for the immune response are very different (A.M.-F. et al., unpublished data).
Our results have demonstrated a nonredundant function for CXCR3 in the recruitment of effector CD8+ T cells to reactive lymph nodes. Within a few hours of subcutaneous injection of mature DCs, CXCL9 became detectable on the lumenal side of HEVs in the draining lymph nodes and it remained associated with HEVs for 3–4 d. It will be important to determine whether CXCL9 is produced directly by HEVs or by tissue cells and/or by migrating activated DCs34. Indeed, chemokines produced in inflamed peripheral tissues or injected subcutaneously can be transported through lymph and conduits to the 'peri-HEV' region, where they are moved by transcytosis by endothelial cells of HEVs for lumenal display35, 36. It is also unclear at present which adhesion molecules render HEVs in reactive lymph nodes permissive for interactions with effector or memory T cells. L-selectin seems to be dispensable, as highly purified L-selectin-negative T cells migrate efficiently to reactive lymph nodes and blocking antibodies to L-selectin do not interfere with effector T cell migration (G.G., unpublished observations). In addition, it has been shown that low expression of L-selectin (less than 1 104 molecules per cell) does not allow for efficient rolling of T cells on native HEVs37, 38. It is possible that expression of endothelial selectins may be higher in reactive lymph nodes. Such selectins can mediate rolling through appropriately glycosylated P-selectin glycoprotein ligand 1 present on effector and memory CD8+ T cells. The latter cells also have high expression of LFA-1 and 
4 integrins that may promote efficient adhesion of rolling cells, perhaps even in the absence of chemokine-induced integrin activation.
The elimination of antigen-presenting DCs has been always considered an important mechanism for the regulation of T cell responses39. Several in vitro studies have shown that DCs can be killed by NK cells and by cytotoxic CD8+ T cells, although the susceptibility of killing decreases as DCs mature, possibly because of upregulation of the granzyme B inhibitor Spi-6 (refs. 17,40,41). However, in contrast to those data on NK cell–mediated killing and the postulated protective function of Spi-6 in mature DCs, several in vivo studies have provided evidence for the killing of both immature and mature DCs18, 25, 26, 27, 42. In those studies, it was suggested that DC killing occurs either in peripheral tissues, where it is mediated by extravasating effector T cells, or in infected lymph nodes, where it is mediated by effector T cells generated in situ in response to viruses. Our study has shown that blood-borne effector CD8+ T cells and even effector memory CD8+ T cells can re-enter lymph nodes where the immune response is taking place. Our intravital imaging analysis also directly visualized the killing of mature DCs by effector T cells in reactive lymph nodes. In contrast to peripheral tissues, where T cells are more dispersed and target cells are more heterogeneous, the lymph node environment provides an ideal condition for this mechanism to be particularly effective because of the high cell density and motility favoring DC–T cell encounters.
Killing of antigen-presenting DCs by effector T cells in lymph nodes can be viewed as a classical mechanism of feedback regulation in which the end product removes the initiating stimulus. At the peak of the immune response, effector T cells re-entering lymph nodes will be particularly effective in killing antigen-presenting DCs at these sites, thus terminating ongoing responses and preventing excessive T cell stimulation. That idea is consistent with the observation that perforin deficiency is associated with greater accumulation of antigen-specific CD8+ T cells after infection or DC immunization18, 43, 44. In the memory phase, recruitment of effector memory T cells to lymph nodes may serve to provide immediate protection against pathogens that actively replicate in lymph nodes. However, killing of DCs presenting viral antigens may also limit the population expansion of virus-specific central memory T cells and particularly of virus-specific naive T cells, which require sustained contact with antigen-presenting cells for activation45. The final outcome would depend on the size of the effector memory T cell pool and on the efficiency and kinetics of effector T cell recruitment in lymph node.
The results discussed above may be also relevant in the context of the phenomenon of T cell 'original antigenic sin'46, whereby an existing effector memory T cell population, by eliminating infected DCs in lymph nodes, prevents the stimulation of an immune response to new antigens expressed by the same virus. Killing of DCs in lymph nodes by effector memory T cells may also limit the efficacy of 'booster' immunization. Indeed, experimental evidence indicates that CXCR3-deficient effector and memory T cells undergo more population expansion after 'booster' immunization than do wild-type T cells (G.G., unpublished observations). It can be envisaged that blockade of effector T cell recruitment to lymph nodes may enhance secondary responses and the efficacy of multiple immunizations, a hypothesis that needs to be further confirmed experimentally.
Methods
Media and reagents.
RPMI 1640 medium supplemented with 10% (vol/vol) FCS, 1% (vol/vol) Glutamax, 1% (vol/vol) nonessential amino acids, 1% (vol/vol) pyruvate, 50 U/ml of penicillin, 50 g/ml of streptomycin (all from Invitrogen) and 50 g/ml of 2-mercaptoethanol (Merck) was used throughout. Human recombinant IL-2 was produced with the myeloma-based expression system47. CpG oligodeoxynucleotide 1826 (5'-CCATGACGTTCCTGACGTT-3') was from Microsynth. Mouse recombinant TNF was from R&D Systems. Complete Freund's adjuvant, lipopolysaccharide (LPS) and Ribi adjuvant (MPL + TDM Adjuvant system) were from Sigma. OVA, the MHC class I-binding peptide SIINFEKL (OVA amino acids 257–264) and the MHC class II–binding peptide ISQAVHAAHAEINEAGR (OVA amino acids 323–339) were from Invitrogen. The control MHC class I–binding peptide SIYRYYGL was obtained through the Research Technologies Branch of the National Institute of Allergy and Infectious Disease. Fluorescent dyes used for cell labeling were CFSE, CellTracker Orange (5- and 6-(4-chloromethyl) benzoyl amino tetramethylrhodamine (CMTMR)), TRITC (tetramethylrhodamine-5-(and-6)-isothiocyanate) and BCECF-AM (2',7'-bis-(2-carboxyethyl)-5-(and 6)-carboxyfluorescein, acetoxymethyl ester; Molecular Probes).
Mice.
OT-I TCR-transgenic mice48 were bred onto backgrounds of different CD45 alleles in the animal facility of the Institute for Research in Biomedicine. OT-II TCR-transgenic mice49 were also bred onto backgrounds of different CD45 alleles. Ccr7-/- mice5 and Cxcr3-/- mice50 were crossed with OT-I TCR-transgenic mice on the backgrounds of different CD45 alleles. C57BL/6 mice 6–9 weeks old were from Harlan. Mice were treated in accordance with animal regulations of the Swiss Federal Veterinary Office guidelines or the American Association for Laboratory Animal Care and National Institutes of Health.
DC cultures.
Bone marrow–derived DCs were generated in cultures supplemented with 20 ng/ml of recombinant granulocyte-macrophage colony-stimulating factor (R&D Systems). After 7–8 d, DC maturation was induced by overnight incubation with 0.5 g/ml of LPS or 2.5 g/ml of CpG 1826; the resultant cells are called 'mature DCs' here. DCs were pulsed for 90 min at 37 °C with 1–5 M SIINFEKL peptide. In some experiments, mature DCs were labeled with 1.7 M CFSE or 10 M CMTMR. DCs were washed extensively and were injected into syngeneic mice subcutaneously in the footpad or intravenously. Similar results were obtained with DCs matured with LPS or CpG.
Flow cytometry analysis and sorting.
For phenotypic analysis, the following monoclonal antibodies were used: anti-L-selectin (MEL14), anti-CD44 (IM7), anti-CCR7 (4B12), anti-CD4 (L3T4), anti-CD8 (Ly-2), anti-CD45.1 (A20), anti-CD45.2 (104), anti-IFN- (XMG1.2), anti-V2 TCR (B20.1), anti-CD3 (145-2C11), anti-CD11a (2D7; all from eBioscience), anti-CD49d (R1-2; BD Biosciences), anti-CXCR3 (220803; R&D Systems) and anti-human granzyme B (GB12; Caltag). CD62P ligand on T cells was detected by staining with a fusion protein of CD62P and immunoglobulin G (IgG; 555294; BD Biosciences). Six-color staining of the cell surface was done with the appropriate combinations of antibodies conjugated to fluorescein isothiocyanate, phycoerythrin, peridinine chlorophyll protein complex, phycoerythrin-indotricarbocyanine, allophycocyanin, allophycocyanin-indotricarbocyanine or biotin, and with streptavidin labeled with peridinine chlorophyll protein complex or phycoerythrin (BD Biosciences). Samples were acquired on a FACSCanto (BD Bioscience) and were analyzed with FlowJo software (TreeStar). Naive OT-I and OT-II cells were sorted from lymph nodes and spleen by cell sorting (FACSAria; BD Biosciences) as CD8+CD44lo L-selectin-high and CD4+CD8-CD44lo L-selectin-high, respectively, with the following specific antibodies: fluorescein isothiocyanate–labeled antibody to CD8 (anti-CD8), phycoerythrin-labeled anti-CD44, fluorescein isothiocyanate–labeled anti-CD4 and allophycocyanin-labeled anti-L-selectin (all from eBioscience), and peridinine chlorophyll protein complex–labeled anti-CD8 (from BD Biosciences). Before sorting, samples were enriched for CD4+ T cells with anti-CD4 magnetic beads (Miltenyi). Polyclonal T cells were also isolated from spleens and livers. To obtain T cells from liver, the mice were first perfused by intracardiac injection with 8 ml PBS-heparin (75 U/ml; Roche Pharma). Livers were 'smashed' in Hank's balanced-salt solution, 10 mM HEPES (Invitrogen) and 2% (vol/vol) FCS and were filtered. Percoll (Amersham Bioscience) gradient centrifugation was used to separate lymphocytes from hepatocytes.
In vitro generation of effector OT-I cells.
Naive OT-I cells from normal or chemokine receptor–deficient mice were stimulated in vitro in 96-well plates with matured SIINFEKL-pulsed DCs at high DC/T cell ratio (6:1). On day 4, cultures were supplemented with 10 U/ml of human recombinant IL-2 and effector cells were collected on day 7 or 8. To obtain a pure population of effector T cells, samples were depleted of contaminating L-selectin-positive T cells with anti-L-selectin magnetic beads (Miltenyi).
IFN- production and granzyme expression by effector OT-I cells were measured by intracellular staining after 4 h of stimulation with phorbol 12-myristate 13-acetate (0.1 M, Sigma) and ionomycin (1 g/ml; Sigma), with the last 2 h of culture in the presence of brefeldin A (10 g/ml; Sigma). Fluorescein-labeled antibodies to IFN- and granzyme B (both from BD Bioscience) were used after cell fixation in 4% (wt/vol) paraformaldehyde and permeabilization with 0.5% (wt/vol) saponin (Sigma-Aldrich). For in vivo cytotoxicity assays, total splenocytes from C57BL/6 mice were labeled with low or high concentrations of CFSE. CFSEhi cells were pulsed with 10 M OVA peptide (amino acids 257–264), were washed extensively, were mixed at a ratio of 1:1 with CFSElo unpulsed splenocytes and were injected intravenously (1.4 107 cells) into mice that received 4 106in vitro–generated effector OT-I cells. Mice were killed 16 h later and splenic cells were analyzed by flow cytometry.
In vivo generation of effector and memory OT-I cells.
CD45.1 OT-I cells (3 104) were injected intravenously into CD45.2 congenic mice. An appropriate combination of CD45 markers was chosen to allow discrimination of injected versus host cells. Mice carrying naive OT-I cells were immunized by intravenous injection of 1 106 matured DCs pulsed with 1 M OVA peptide (amino acids 257–264) or by intravenous injection of 4 103 colony-forming units of L. monocytogenes strain 10403s expressing OVA51. In vivo–generated effector and memory T cells were isolated by positive selection with fluorescein isothiocyanate–labeled anti-CD45.1 (eBioscience) followed by anti–fluorescein isothiocyanate microbeads and magnetic selection (Miltenyi) on days 6–7 or days 30–40 after priming, respectively.
Immunofluorescence.
For analysis of the distribution CXCL9, reactive lymph nodes were induced at different time intervals by subcutaneous injection of mature DCs (3 106 cells). Alexa Fluor 594–labeled goat polyclonal anti-CXCL9 (AF-492-NA; R&D Systems) was then injected intravenously and 30 min later mice were killed. Lymph nodes were then collected, were embedded in optimum cutting temperature compound (Polyfreeze Tissue Freezing Medium; Polysciences) and were 'snap-frozen' in liquid nitrogen. Frozen sections 8 m in thickness were counterstained with anti-PNAd (MECA-79; rat IgM; BD Biosciences) followed by Alexa Fluor 488–labeled goat anti-rat IgM (Molecular Probes) as described34. Air-dried frozen sections were fixed with 2% (wt/vol) paraformaldehyde, were blocked with serum-free protein block (Dako) and were stained with antibodies diluted in antibody diluent (Dako). A Nikon Eclipse E800 microscope was used for immunofluorescence microscopy. Separate images were collected for each fluorochrome and images were overlaid to produce a multicolor image with Openlab 5 software (Improvision). Individual images were juxtaposed with Adobe Photoshop to reconstitute the image of a whole lymph node cross-section.
Intravital microscopy.
Intravital microscopy of resting inguinal peripheral lymph nodes of C57BL/6 mice was done essentially as described52. The adhesive activity of adoptively transferred BCECF- or TRITC-labeled naive or effector OT-I cells in the microvascular network of surgically exposed lymph nodes was visualized with a custom-built microscope (IVM-500; Mikron Instruments) equipped with a Xenon stroboscope lamp and a silicon-intensified target camera (VE-SIT1000; Dage MTI). Adhesive events were recorded (DVCAM; Sony) for later 'offline' analysis of rolling or sticking fractions and adherent cells per area as described53. For intravital microscopy of reactive lymph nodes, DCs were injected into the right footpad 12–20 h before surgical preparation of the draining popliteal lymph node as described7. DC transfer resulted in readily visible swelling of popliteal lymph node, indicative of an inflammatory phenotype (data not shown). Highly pure (over 99%) L-selectin-negative effector T cells were adoptively transferred through a catheter in the contralateral femoral artery and their adhesive activity in the lymph node microcirculation was analyzed as described above.
Two-photon microscopy.
Bone marrow DCs matured with LPS (10 ng/ml) were pulsed for 1 h at 37 °C with 1 M SIINFEKL or SIYRYYGL (control peptide). The two DC populations were then differentially labeled with dye (2.5 M CFSE or 5 M Cell Tracker Blue, respectively) and 0.5 106 cells of each DC population were injected together into the footpads of recipient mice 24 h before T cell injection. At 30 min to 15 h after transfer of T cells labeled with SNARF-1 (benzenedicarboxylic acid, 2- (or 4)-[10-(dimethylamino)-3-oxo-3H-benzo(c)xanthene-7-yl], recipient mice were anesthetized with nebulized isoflurane (2% induction; 1% maintenance) in 30% O2 and 68–69% air. Two-photon microscopy was done on the right popliteal lymph node with a modified surgical technique54 using the Radiance 2100MP system (Bio-Rad Laboratories; Carl Zeiss MicroImaging) equipped with a Nikon 600FN upright microscope, 20 water immersion lens (N.A. 0.95; Olympus) and a 10-Watt Chameleon femtosecond-pulsed laser (Coherent) tuned to 800 nm. Images were collected with typical voxel size of 0.91 0.91 3–5 mm and a volume dimension of 467 467 100 mm. This volume collection was repeated every 20–25 s for up to 5 h to produce four-dimensional datasets, which were then processed with Imaris software (Bitplane) to create sequential two-dimensional maximum-intensity projections and movies.
Statistical analysis.
Rolling fractions and the number of sticking T cells on HEVs (Fig. 3) were analyzed with a Student's t-test (unpaired, two-sided).
Note: Supplementary information is available on the Nature Immunology website.