CXCL12 Mediates CCR7-independent Homing of Central Memory Cells, But Not Naive T Cells, in Peripheral Lymph Nodes

Abstract

Central memory CD8⁺ T cells (T_CM) confer superior protective immunity against infections compared with other T cell subsets. T_CM recirculate mainly through secondary lymphoid organs, including peripheral lymph nodes (PLNs). Here, we report that T_CM, unlike naive T cells, can home to PLNs in both a CCR7-dependent and -independent manner. Homing experiments in paucity of lymph node T cells (plt/plt) mice, which do not express CCR7 ligands in secondary lymphoid organs, revealed that T_CM migrate to PLNs at ∼20% of wild-type (WT) levels, whereas homing of naive T cells was reduced by 95%. Accordingly, a large fraction of endogenous CD8⁺ T cells in plt/plt PLNs displayed a T_CM phenotype. Intravital microscopy of plt/plt subiliac lymph nodes showed that T_CM rolled and firmly adhered (sticking) in high endothelial venules (HEVs), whereas naive T cells were incapable of sticking. Sticking of T_CM in plt/plt HEVs was pertussis toxin sensitive and was blocked by anti-CXCL12 (SDF-1α). Anti-CXCL12 also reduced homing of T_CM to PLNs in WT animals by 20%, indicating a nonredundant role for this chemokine in the presence of physiologic CCR7 agonists. Together, these data distinguish naive T cells from T_CM, whereby only the latter display greater migratory flexibility by virtue of their increased responsiveness to both CCR7 ligands and CXCL12 during homing to PLN.

Introduction

Upon Ag encounter, naive T cells differentiate into effector cells poised to eradicate foreign or tumor Ags. Upon completing this task, most effector T cells die, leaving behind a small population of memory T cells that respond more rapidly to reencountered Ags (1–3). Based on the expression of certain adhesion molecules and chemokine receptors, two distinct subsets of memory T cells have been shown to arise during immune responses as follows: central memory T cells (T_CM) express L-selectin and CCR7 and localize to secondary lymphoid organs, whereas effector memory T cells (T_EM) are L-selectin⁻CCR7⁻ and found mainly in nonlymphoid tissues (4). When isolated from human blood, CD8⁺ T_EM show greater cytotoxicity and produce more effector cytokines than T_CM. However, adoptive transfer of CD8⁺ T_CM into naive recipients is more efficient than transfer of T_EM in conferring protective immunity. This may be due to the higher proliferative potential of T_CM as well as their greater capacity to persist in vivo (5). CD8⁺ T_EM can differentiate into T_CM over several weeks; i.e., they reexpress L-selectin and CCR7 and acquire an enhanced capacity of homeostatic and Ag-driven proliferation (5). In addition, naive T cells can give rise to T_CM without first experiencing a full-fledged effector phase (6).

The localization of T_CM in LNs might further contribute to their key role in immune protection. Professional APCs, in particular DCs, transport Ag from the periphery into the T cell zones of the LN (7). Thus, T_CM are strategically positioned to interact with DCs, which provide optimal stimulatory signals if T_CM detect a recall Ag on their surface.

We have recently described a method that allows for the generation of central memory–like CD8⁺ T cells in vitro (6). Ag-primed CD8⁺ T cells cultured in IL-15 for 5–7 d express high levels of CD44, L-selectin, and CCR7. Upon TCR stimulation, they produce IFN-γ, but are not acutely cytotoxic in vitro. After adoptive transfer, these cells survive for long periods of time, and mount rapid Ag-specific recall responses. We have shown that these cells migrate to all secondary lymphoid organs, including peripheral LNs (PLNs), mesenteric LNs (MLNs), Peyer's patches, and spleen (8). Similar to naive T cells, migration to PLNs occurred via high endothelial venules (HEVs) and depended on L-selectin (8).

To address the role of CCR7 in PLN homing, T_CM were injected into paucity of lymph node T cells (plt/plt) mice in which the lymphoid organ–expressed CCR7 ligands CCL19 (ELC) and CCL21-Ser (SLC) are deleted (9–12). The few T cells present in plt/plt PLNs are enriched for memory cells (10), shown here to be predominantly T_CM, indicating that this subset can home to LNs in the absence of CCR7 ligands. Indeed, although naive T cell homing was reduced to 5%, T_CM migration to plt/plt PLNs was reduced to only 20% of WT levels (8). This suggests involvement of one or more CCR7-independent pathways that enable at least some T_CM to home to PLNs.

In this paper, we show that T_CM rolled and adhered firmly in plt/plt PLN HEVs. T_CM sticking and migration to plt/plt PLNs were pertussis toxin (PTX) sensitive and CXCL12 dependent. Homing of T_CM to WT PLNs was also partially dependent on CXCL12. In contrast, CXCL12 had no detectable effect on naive T cell trafficking to PLNs. Thus, T_CM use at least two different chemokine pathways, CCR7–CCL19/21 and CXCR4–CXCL12, to enter PLNs under steady state conditions.

Materials and Methods

Mice

DDD/1-plt/plt, BALB/c-plt/plt, and DDD1-mtv/mtv control mice were provided by A. Matsuzawa (University of Tokyo, Tokyo, Japan). Transgenic T–green fluorescent protein (GFP) mice were generated in our laboratory (13). CXCR3^−/− mice were provided by C. Gerard (Harvard University, Boston, MA). C57BL/6 and BALB/c mice were purchased from The Jackson Laboratory. Mice were housed and bred in a specific pathogen-free and viral antibody-free animal facility. Experiments were in accordance with National Institutes of Health guidelines and approved by the Committees on Animals of Harvard Medical School and The CBR Institute for Biomedical Research.

Reagents

Fluorochrome-labeled mAbs were obtained from BD Biosciences as follows: CD3ɛ, CD4, CD8α, CD44, CD25, L-selectin, CD122, CD69, and TCRβ. To detect CCR7 and P-selectin ligand expression, recombinant human CCL19-Ig and P-selectin–Ig chimeras were used (6). Antimurine CXCL12, recombinant murine CXCL12, CCL2, CCL5, CCL19, and recombinant human IL-15 were obtained from R&D Systems. PTX was obtained from Calbiochem.

In Vitro Differentiation of T_CM

T_CM were generated as described previously (6). In brief, splenocytes were incubated with 1 μg/ml anti-CD3ɛ and 2 d later with media containing 20 ng/ml IL-15 (for 5–8 d). Before each experiment, activation marker and homing molecule expression were assessed by flow cytometry (8).

ELISA and Immunofluorescence

PLNs were removed from WT and plt/plt mice, homogenized in lysis buffer (radioimmunoprecipitation assay buffer with 1 mM PMSF, 10 μg/ml aprotinin, and 10 μg/ml leupeptin), and centrifuged (14k g at 4°C for 10 min). The supernatant was assayed for CXCL12 immunoreactivity by ELISA (R&D Systems). Immunostaining of frozen sections was performed as described previously (14).

Homing Assays

Homing assays were performed as described previously (8). In brief, tetramethylrhodamine-5-isothiocyanate (TRITC)-labeled T_CM were mixed with LN cells from T-GFP mice and injected i.v. into recipients. After 1 or 24 h, 1 ml PBLs, spleen, PLNs, and MLNs were harvested, immunostained, and analyzed by flow cytometry. The homing index in organs was calculated as the ratio between the number of CD8⁺TRITC⁺ (T_CM) and CD8⁺GFP⁺ (naive) T cells divided by the ratio of CD8⁺TRITC⁺ and CD8⁺GFP⁺ cells in the input (8). In some experiments, T_CM were pretreated with 100 ng/ml PTX for 2 h before adoptive transfer. For blocking experiments, 100 μg/mouse anti-CXCL12 or control mAb were injected i.v. 15 min before adoptive transfer of T cells.

Intravital Microscopy and Image Analysis

Subiliac LN Preparation.

Surgical preparation of LNs was performed as described previously (15). T_CM were labeled with calcein (Molecular Probes), and small boli of cells were injected intraarterially. T cell–endothelial cell interactions in subiliac LNs downstream from the injection site were recorded. For experiments testing the role of CXCL12, cell behavior was analyzed in the same vessels before and after mAb injection. To assess the role of Gα_i-protein coupled receptors (GPCRs), PTX-treated and untreated T_CM were compared.

Cremaster Muscle Preparation.

Cremaster muscles of C57BL/6 mice were prepared as described previously (16). Calcein-labeled T_CM were injected, and several postcapillary and small collecting venules were recorded during a 15-min control period to determine baseline rolling and sticking. Subsequently, superfusion buffer was replaced with buffer containing 100 nM CXCL12, and T_CM behavior was recorded in the same vessels for an additional 15 min.

Offline frame-by-frame video analysis was performed as described previously (17). Rolling fraction was determined as the percentage of cells interacting with HEVs in the total number of cells passing through a vessel during the observation period. Sticking fraction was defined as the percentage of rolling cells that adhered in HEVs for ≥30 s.

Statistical Analysis.

All data are presented as mean ± SEM. Homing indices were compared using the unpaired Student's t test. Rolling and sticking fractions were compared using the paired Student's t test.

Results and Discussion

Preferential Recruitment of T_CM to plt/plt PLNs.

T cells can enter LNs from the blood or via afferent lymphatic vessels draining tissues such as the skin (18). From the blood, naive T cells must adhere to HEVs, a process requiring L-selectin, CCR7, LFA-1, and their respective ligands (14, 18, 19). Similar to naive T cells, T_CM home to PLNs via HEVs in an L-selectin– and partially CCR7-dependent manner (8, 14, 19). Although naive T cells do not arrest in HEVs and home very poorly to plt/plt PLNs (14), T_CM are capable of entering these organs via HEVs, albeit considerably less efficiently than in WT mice (8). Competitive homing of naive T cells and T_CM in DDD/1-plt/plt mice revealed that the total number of homed naive T cells was reduced 19.2-fold compared with WT mice, whereas T_CM were only reduced by 4.9-fold (8). Recent observations suggest the defect in T cell migration to plt/plt LNs can vary with the genetic background (20). Thus, we repeated competitive homing assays in plt/plt mice on both BALB/c and DDD/1 backgrounds. In accordance with previous results (8), homing indices in BALB/c and DDD/1-mtv/mtv mice were 0.6 ± 0.1 and 0.4 ± 0.01, respectively; i.e., naive T cells accumulated approximately twofold more efficiently in WT PLNs than T_CM (Fig. 1, A and B). In contrast, homing indices were 1.4 ± 0.1 and 1.4 ± 0.2 in BALB/c-plt/plt and DDD/1-plt/plt PLN, respectively, indicating the CCR7 pathway may not be absolutely required for T_CM homing to PLNs. Both T_CM and naive T cells were equally affected in homing to MLNs in both DDD/1-plt/plt mice (by 89 and 88%, respectively) and BALB/c-plt/plt mice (by 92 and 94%, respectively). Thus, CCR7 is indispensable for T_CM trafficking to MLNs, whereas PLNs can recruit at least some T_CM in the absence of CCL19 and CCL21.

Figure 1.

Accumulation of in vitro–cultured and endogenous T_CM in plt/plt PLNs. TRITC-labeled T_CM together with T-GFP LN cells were injected i.v. into BALB/c-plt/plt and BALB/c (A), and DDD/1-plt/plt and DDD/1-mtv/mtv mice (B). After 24 h, PBLs and secondary lymphoid organs were harvested, stained for CD8α, and analyzed by flow cytometry for the presence of TRITC⁺ T_CM and naive GFP⁺ T cells. Homing indices were calculated as described in Materials and Methods (n = 6 mice). n.s., not significant. *, P < 0.05; ***, P < 0.001. (C) Total PLN cells from plt/plt and WT mice were analyzed by flow cytometry. Representative dot plots from one experiment out of four are shown. Gates were set on TCRβ⁺CD8α⁺ cells. Numbers represent percentage of gated cells.

PLN in plt/plt mice are smaller than in WT animals, and contain a larger fraction of Ag-experienced lymphocytes (9, 10). However, the exact composition of memory T cells in these organs has not been determined. Because in vitro–generated T_CM retained, in part, the capacity to home to plt/plt PLNs, we examined the phenotype of resident CD8⁺ memory T cells by staining BALB/c-plt/plt or BALB/c PLN single cell suspensions for CD122 (21). Expression of this memory marker correlated well with that of CD44 (Fig. 1 C). There was an approximately fivefold increase in the percentage of CD8⁺CD122⁺ T cells in plt/plt PLNs as compared with WT PLNs. Almost all of them expressed L-selectin and bound CCL19-Ig chimera, but little or no P-selectin–Ig, thus displaying a central memory phenotype (4). These observations strongly suggest that, in contrast to naive T cells, which depend almost entirely on CCR7 ligands, there may be a second homing pathway in PLNs that allows T_CM recruitment in the absence of CCL21 and CCL19.

T_CM Roll and Stick in plt/plt HEVs.

There are two nonexclusive explanations for preferential T_CM traffic to plt/plt PLNs as follows: T_CM may have the unique ability to interact with HEVs by responding to an integrin-activating signal distinct from CCR7 ligands and/or they may enter via afferent lymphatic vessels. The latter mechanism was proposed by Mori et al., who observed delayed, but enhanced and prolonged T cell–mediated immune responses to contact sensitizers in plt/plt PLNs (22). Due to CCL21-leu expression in lymphatic vessels in plt/plt mice, both T cells and DCs can enter these vessels to migrate to PLNs (11, 22, 23). However, we show here that several T_CM accumulated in plt/plt PLNs as early as 1 h after i.v. injection. Given this brief interval, it seems unlikely that the homed cells could migrate to peripheral tissues and find their way into afferent lymphatics and from there into the cortex of draining LNs. Thus, our findings strongly suggest that T_CM can enter plt/plt PLN from the blood.

To test this hypothesis, we used intravital microscopy (IVM) to analyze T_CM behavior in the subiliac LN microcirculation of plt/plt mice. The venular tree of normal subiliac LNs consists of five successive venular branching orders, distinguishable by IVM (15). Order V represents postcapillary venules in the cortex and order I is the large collecting vessel at the LN hilus. In WT mice, naive T cells and T_CM roll and stick mainly in order III–V venules. These venules express the most peripheral node addressin, which mediates the rolling of L-selectin⁺ T cells (8, 18, 19, 24). Firm adhesion requires activation of the integrin LFA-1 on rolling lymphocytes, which is usually induced by CCR7 ligands displayed in HEVs (14).

plt/plt PLNs are smaller than WT PLNs and their microarchitecture is disorganized, but the venular tree is readily discernible by IVM in DDD/1-plt/plt PLNs, even though order V venules are typically absent (14). By contrast, BALB/c-plt/plt PLNs are not amenable to IVM because the rarefied venular tree in this strain is obscured by overlying B cell follicles (14). Calcein-labeled T_CM were recorded in 43 HEVs (7 order I, 12 order II, 16 order III, and 8 order IV) in 6 DDD1-plt/plt mice (Fig. 2). Mean rolling fractions of T_CM were 17 ± 5% (order I), 23 ± 3% (order II), 40 ± 5% (order III), and 33 ± 7% (order IV). This is comparable to rolling fractions of T_CM in WT PLNs (8). For comparison, naive GFP⁺ T cells (14) were analyzed in 29 venules (2, order I; 8, order II; 8, order III; and 11, order IV) in seven mice. Their mean rolling fractions were 7 ± 2% (order I), 35 ± 9% (order II), 58 ± 4% (order III), and 59 ± 6% (order IV). Naive T cells rolled at similar frequencies in PLN HEVs of DDD/1-plt/plt and DDD/1-mtv/mtv mice, but completely failed to arrest in the former, indicating that CCR7 ligands are absolutely required (Fig. 2 and reference 14). However, T_CM were unmistakably capable of sticking in DDD/1-plt/plt PLN HEVs (Fig. 2 B, mean sticking fractions: order I, 2.5 ± 1.8%; order II, 11.5 ± 3.3%; order III, 21.0 ± 4.4%; and order IV, 31.5 ± 1.8%). We conclude that integrins are efficiently activated on rolling T_CM in plt/plt PLNs, indicating that an alternative chemoattractant pathway can be used by these cells. Moreover, HEVs are the likely port of PLN entry for T_CM in plt/plt mice.

Figure 2.

T_CM roll and stick in HEVs of plt/plt subiliac LNs. IVM analysis of calcein-labeled T_CM or naive GFP⁺ T cells in DDD/1-plt/plt subiliac LNs. Interactions between fluorescent T cells and HEVs were visualized by epifluorescence IVM. Rolling (A) and sticking (B) fractions were determined for each venular order (n = 6 mice). n.s., not significant. *, P < 0.05; ***, P < 0.001.

Homing of T_CM to plt/plt PLN Is PTX Sensitive.

Chemokine receptors are GPCRs whose function is efficiently blocked by PTX (25, 26). To determine whether CCR7-independent integrin activation signals on T_CM were perceived through such GPCRs, T_CM were preincubated with PTX before in vivo testing. PTX did not affect cell viability, but inhibited in vitro chemotaxis of T_CM toward CCL21 and CXCL12 by 80% (unpublished data). PTX treatment greatly reduced the ability of T_CM to home to BALB/c-plt/plt PLNs and MLNs, but had no effect on the number of T_CM in PBLs or spleen (Fig. 3, A and B). These results were further substantiated by IVM of DDD/1-plt/plt PLNs. There was no difference in the rolling fractions of PTX-treated and untreated T_CM (Fig. 3 C, mean: 22 ± 2% and 28 ± 3%, respectively; P > 0.05). This is not surprising because L-selectin–mediated rolling is insensitive to PTX (19). In contrast, the sticking fraction of control T_CM was fivefold higher than that of PTX-treated cells (Fig. 3 D, mean: 15.1 ± 3.4% and 2.8 ± 1.6%, respectively; P < 0.01). This indicates that one or more GPCRs, other than CCR7, mediates rapid integrin activation on rolling T_CM in PLN HEVs of plt/plt mice.

Figure 3.

Gα_i-coupled signaling mediates CCR7-independent homing of T_CM to PLNs and is required for T_CM sticking in PLN HEVs of plt/plt mice. (A and B) TRITC-labeled control T_CM and CFSE-labeled PTX-treated T_CM were adoptively transferred into BALB/c-plt/plt recipients. After 24 h, PBLs, spleen, PLNs, and MLNs were harvested, and homed cells were analyzed by flow cytometry. Data are shown as the number of homed cells per 10³ or 10⁶ injected cells (n = 3 mice). (C and D) IVM analysis of rolling (C) and sticking fractions (D) of calcein-labeled control or PTX-treated T_CM in order III and IV HEVs in DDD/1-plt/plt PLNs (n = 3 mice). n.s., not significant. *, P < 0.05; **, P < 0.01.

Homing of T_CM to DDD/1-plt/plt PLNs Is Partially Mediated by CXCL12.

One chemoattractant candidate that could mediate T_CM homing to PLNs in the absence of CCR7 ligands is CXCL12. mRNA for CXCL12 is expressed by stromal cells in close vicinity to PLN HEVs, and CXCL12 protein was found in the lumen of some cortical HEVs (20). Moreover, CXCL12 and its receptor CXCR4 partly mediate homing of B cells to LN via HEV (20). In the case of T cell homing to plt/plt PLNs, the role of CXCR4–CXCL12 is more complex. Okada et al. found that CXCR4^−/− T cells have no discernible defect in homing to WT PLNs, but CXCR4^−/− splenic T cells were more severely compromised than WT T cells in their ability to migrate to plt/plt PLNs (20). Notably, the homing experiments with CXCR4^−/− T cells were performed with total splenocytes in which the T cell population typically contains ∼30% memory cells. Conceivably, CXCL12-responsive T cells recovered from plt/plt PLNs by Okada et al. may have contained a disproportionate fraction of T_CM, rather than naive T cells, which predominate in WT PLNs.

Normal expression of CXCL12 in plt/plt PLNs makes this a good candidate for CCR7-independent T_CM migration to PLNs. Thus, we performed IVM of the subiliac LNs in DDD/1-plt/plt mice before and after anti-CXCL12 treatment. A total of 22 HEVs of orders III and IV in three DDD/1-plt/plt recipients were analyzed. Rolling fractions did not change after injection of anti-CXCL12 (Fig. 4 A, before: 25.0 ± 3.7%; after: 30.7 ± 5.0%; P > 0.05). In contrast, anti-CXCL12 markedly reduced the sticking fraction of T_CM from 11.9 ± 2.8% to 2.6 ± 1.2% (Fig. 4 B, P < 0.01).

Figure 4.

CXCL12 mediates CCR7-independent homing of T_CM to PLNs. (A and B) IVM analysis of rolling (A) and sticking fractions (B) of calcein-labeled T_CM in DDD/1-plt/plt PLN HEVs. Interactions of T_CM and HEVs in the same venules were recorded before and after injection of anti-CXCL12 (n = 4 mice). (C and D) TRITC-labeled T_CM were injected into BALB/c-plt/plt mice (C) or DDD/1-plt/plt mice (D) pretreated with control or anti-CXCL12 mAb. 1 h later, PBLs, spleen, PLNs, and MLNs were harvested and analyzed by flow cytometry. Data are expressed as percentage of homed cells in the total number of leukocytes (n = 6 mice). (E) TRITC-labeled T_CM were adoptively transferred into C57BL/6 mice injected previously with control or anti-CXCL12 mAb. 1 h later, PLNs were harvested and analyzed by flow cytometry. Data are expressed as percentage of homed cells relative to control mice (n = 6 mice). n.s., not significant. *, P < 0.05; **, P < 0.01.

To determine whether CXCL12 mediates T_CM homing to plt/plt PLNs, anti-CXCL12 mAb was injected i.v. into DDD/1-plt/plt mice 15 min before injection of fluorescently tagged T_CM. Anti-CXCL12 had no effect on the accumulation of T_CM in PBLs, spleen, or MLNs after 1 h (Fig. 4, C and D). In contrast, the number of T_CM that homed to PLNs was ∼30% lower than in control mice (Fig. 4 D, P < 0.05). We also performed competitive homing experiments using naive GFP⁺ T cells and TRITC-labeled T_CM in DDD/1-plt/plt mice after anti-CXCL12 or control mAb injection (n = 2). Anti-CXCL12 did not alter the low number of naive T cells that homed to PLNs (unpublished data). However, consistent with the aforementioned data, the homing index decreased from 1.4 ± 0.1 to 1.0 ± 0.2, indicating that T_CM, but not naive T cells, home to PLNs in a partially CXCL12-dependent fashion. In contrast with DDD/1-plt/plt mice, anti-CXCL12 had no significant effect on T_CM homing to BALB/c-plt/plt PLNs (Fig. 4 C). As a possible explanation for this apparent discrepancy, we asked whether PLN HEVs in BALB/c-plt/plt mice present less CXCL12 on their luminal surface than their counterparts in DDD/1-plt/plt. However, ELISA of PLN lysates and immunostaining of frozen PLN sections from both strains did not reveal detectable differences in CXCL12 protein content and expression pattern, at least at the light microscopic level (unpublished data). Therefore, it is more likely that the lack of CXCL12 contribution to T_CM homing in BALB/c-plt/plt mice occurred because the venular tree in PLNs of this strain is very poorly developed compared with that in DDD/1-plt/plt (14). The available HEV surface area in the former may simply be too small to permit effective T_CM recruitment, despite CXCL12 expression. This explanation is consistent with the finding that the absolute number of T_CM that homed to BALB/c-plt/plt PLNs was three times lower than that recovered from DDD/1-plt/plt PLNs (85 ± 20 homed T_CM/10⁶ injected cells versus 267 ± 116/10⁶ injected cells, respectively; P = 0.05; Fig. 4, C and D, and not depicted).

Because our competitive homing experiments indicate that some T_CM home to BALB/c-plt/plt PLNs in a CCR7-independent fashion (Fig. 1 A), it is likely that one or more other chemoattractants may be expressed in PLN HEVs of BALB/c-plt/plt mice that mediate T_CM homing. The existence of additional GPCR-dependent recruitment signals appears likely, even in DDD/1 mice because PTX treatment blocked T_CM homing to PLNs in both plt/plt strains more completely than anti-CXCL12 (Fig. 3 B). Because in vitro–generated T_CM respond to CCL2 (monocyte chemoattractant protein–1) and CCL5 (regulated on activation, normal T cell expressed, and secreted [RANTES]; reference 8), we tested whether these chemokines contribute to T_CM homing to BALB/c-plt/plt PLNs. Furthermore, very low levels of CCL21-leu mRNA have been detected in plt/plt mice (27). Therefore, we desensitized T_CM with a 40-min incubation in 1 μM CCL2, CCL5, or CCL19 before adoptive transfer. The desensitized cells homed to BALB/c-plt/plt PLNs as efficiently as control cells (unpublished data), indicating that these chemokines do not play a role here. Other possible candidates include CXCR3 and its ligands as well as lipid chemoattractants, such as LTB4, which contribute to effector T cell homing to sites of inflammation (28). However, homing of CXCR3^−/− T_CM was identical to that of WT T_CM (unpublished data), and T_CM do not chemotax toward LTB4 gradients (28). Thus, the nature of other chemoattractants involved in CCR7-independent T_CM homing to PLNs remains to be identified.

Together, these results strongly suggest that CXCL12–CXCR4 triggers integrin activation on rolling T_CM, but not on naive T cells, in PLN HEVs of DDD/1-plt/plt mice. This pathway accounts for a significant fraction of T_CM that home to PLNs. In addition, there is probably at least one other PTX-sensitive recruitment signal whose identity remains to be uncovered.

A Physiological Role for CXCL12 in T_CM Homing to PLNs.

Given the severely disturbed architecture of secondary lymphoid organs in plt/plt mice (9, 10), we asked whether the observed contribution by CXCL12 to T_CM migration reflects a truly physiological event. Therefore, we examined the effect of CXCL12 inhibition on T_CM homing to WT PLNs. Indeed, homing was moderately by ∼20%, but significantly, reduced (Fig. 4 E, P < 0.05). This reduction in homing occurred in both C57BL/6 and BALB/c mice, indicating that the role of CXCL12 in WT mice may not depend on genetic modifiers.

We conclude that the CCR7 pathway plays a dominant role in T_CM homing to PLNs, but CXCL12 also contributes to this process. Our finding that this nonredundant effect of CXCL12 is specific for T_CM is also supported by earlier conclusions that CXCR4 deficiency does not reduce the ability of naive T cells to home to PLNs that express normal CCR7 ligands (20).

CXCL12 Induces Firm Adherence of T_CM, But Not Naive T Cells, in Cremaster Muscle Venules.

Why are naive T cells apparently incapable of using CXCL12 for homing to PLNs? Naive T cells respond to CXCL12 gradients in chemotaxis assays at least as efficiently as T_CM (unpublished data). Furthermore, CXCL12 induces LFA-1–mediated adhesion of human naive T cells to ICAM-1 in flow chamber assays (29). In contrast, murine naive T cells fail to arrest in cremaster muscle venules upon CXCL12 superfusion, but undergo immediate arrest in muscles superfused with CCL21 (16).

Because our homing data also indicate that CXCL12 fails to attract naive T cells to PLNs, but can recruit T_CM, we performed IVM to determine whether circulating T_CM can respond to CXCL12 in situ. To this end, we asked whether CXCL12 superfusion induces T_CM sticking in cremaster muscle venules (16). Fluorescently labeled T_CM were injected intraarterially and their interactions with cremaster muscle venules were recorded. Without CXCL12 superfusion, the mean rolling and sticking fractions were 18.3 ± 5.0% and 1.3 ± 0.9%, respectively (n = 3 mice/30 vessels). Upon subsequent superfusion with CXCL12, rolling fractions remained unchanged (Fig. 5 A, 23.9 ± 3.0%, P > 0.05). In contrast, the T_CM sticking fraction was significantly increased (Fig. 5 B, 6.9 ± 1.8%, P < 0.05 vs. control). To determine whether this was an effect of CXCL12 on the venular endothelium or on circulating T_CM, we desensitized T_CM to CXCL12 before injection. Desensitized T_CM rolled similarly to untreated cells, but were incapable of sticking (Fig. 5, A and B). Thus, T_CM, unlike naive T cells, are responsive to CXCL12 displayed on venular endothelium.

Figure 5.

T_CM express higher levels of CXCR4 than naive T cells and stick in cremaster muscle venules upon topical exposure to CXCL12. Calcein-labeled untreated or CXCL12-desensitized T_CM were injected into a femoral artery of C57BL/6 mice, and rolling (A) and sticking (B) of fluorescent cells in cremaster muscle venules were recorded. Interactions between T_CM and the endothelium of the same muscle venules were recorded before and after superfusion of the preparation with 100 nM CXCL12 solution (n = 3 mice, 30 venules). *, P < 0.05. (C) Expression of CXCR4 on T_CM (shaded histogram) and CD8⁺ naive T cells (solid line) was analyzed by flow cytometry. Isotype control staining is shown as a dotted line. Histograms are representative for results of n = 2. Gates were set on TCRβ⁺CD8α⁺ cells.

What could account for the distinct intravascular responsiveness of T_CM versus naive T cells to CXCL12? One possibility is that these two subsets express distinct densities of CXCR4. Indeed, flow cytometry revealed that T_CM expressed markedly higher levels of CXCR4 compared with naive CD8⁺ T cells (Fig. 5 C; mean fluorescence intensity: T_CM, 278 and naive T cells, 126). This finding is also in accordance with a recent microarray analysis of T_CM and naive T cells, which showed higher mRNA expression of CXCR4 in the former subset (30).

What might be the functional relevance of CCR7-independent homing of T_CM to PLNs? Wherry et al. have recently shown that CD8⁺ effector CTL can give rise to T_EM, which differentiate into L-selectin⁺ T_CM over the course of several weeks (5). Effector CTL and T_EM do not express L-selectin and CCR7 (4, 6, 8). Although the kinetics of L-selectin and CCR7 induction on T_EM in the course of their transition into T_CM is unknown, it has been shown that some Ag-experienced T cells express L-selectin but not CCR7 (4). For this subset, expression of CXCR4 could provide an alternative pathway to gain CCR7-independent access to PLNs. Moreover, CXCL12 is constitutively expressed in many tissues, especially in the BM, and is up-regulated during inflammation. In addition, human T_CM isolated ex vivo, and murine T_CM generated in vitro express several inflammatory chemokine receptors (4, 8). This broad expression pattern of chemokine receptors provides T_CM with maximum flexibility to survey not only secondary lymphoid organs but also multiple other sites in the body.