Cutting Edge: CXCR4 Is Critical for CD8⁺ Memory T Cell Homeostatic Self-Renewal but Not Rechallenge Self-Renewal

A key feature of adaptive immunity is the capacity to develop long-lived memory T cells that control recurrent or persistent infection. Memory CD8⁺ T cells are heterogeneous and can be divided into two main subsets: central memory (CM) cells and effector memory (EM) cells. CM (CD44^hi CD62L^hi) cells, which preferentially home to secondary lymphoid organs, have longer life spans and greater capacity for homeostatic proliferation than do EM (CD44^hi CD62L^lo) cells (1). In the absence of rechallenge, CM CD8⁺ T cells slowly replenish themselves to maintain a stable population. Upon rechallenge, CM T cells produce differentiated effector cells while renewing the CM cell fate through asymmetric cell division (2), thereby avoiding the depletion of cells needed to respond to subsequent or persistent challenge (3).

CM T cells preferentially accumulate and undergo homeostatic proliferation in the bone marrow (BM) (4–6). However, the functional consequences of BM homing on homeostatic self-renewal or rechallenge self-renewal have not been evaluated directly. CXCR4 binds to CXCL12 and has an essential role in homing of hematopoietic stem cells (HSCs) to the BM (7). In this study, we analyzed the impact of the lack of CXCR4 on CD8⁺ T cell responses to lymphocytic choriomeningitis virus (LCMV) infection. CM CXCR4-deficient T cells exhibit defective homeostatic self-renewal, which correlates with impaired homing to the BM. Upon rechallenge, however, CM CXCR4-deficient T cells can proliferate efficiently and differentiate while self-renewing. Thus, homeostatic self-renewal and rechallenge self-renewal are mechanistically separable regenerative properties of memory T cells.

All animal work was performed in accordance with Columbia University Institutional Animal Care and Use Committee guidelines. CXCR4^F/F mice (8), expressing granzyme (Gzm)B–Cre or not (9), were infected as intact animals. For adoptive-transfer experiments, naive CD8⁺ P14 T cells were sorted from wild-type (WT) Thy1.1/1.1 and CXCR4^F/F–CD4–Cre⁺ Thy1.1/1.2 mice. A total of 10³ cells from each genotype was transferred i.v. into WT Thy1.2/1.2 recipients 1 d prior to infection. Mice were infected with 2 × 10⁵ PFU LCMV Armstrong strain by i.p. injection. For rechallenges, 5 × 10⁵ CFU Listeria monocytogenes expressing GP33–41 (gp33) were injected i.v. Results depict the percentage of CXCR4-deficient P14 T cells among transferred cells at the indicated time postinfection when normalized to the proportion of CXCR4-deficient P14 T cells among transferred cells at the time of transfer. To identify sinusoidal lymphocytes, mice were injected i.v. with 1 μg anti-CD45.2 mAb coupled to PE (BD Biosciences) and sacrificed 2 min after mAb injection, as previously described (10). To assess proliferation, mice were treated with 2 mg BrdU (Sigma-Aldrich) i.p. every 2 d for 15 d prior to tissue harvest and analysis.

Single-cell suspensions of spleen, BM, and lymph nodes (LNs; pool of mesenteric and s.c.) were stained with a LIVE/DEAD Fixable Dead Cell Stain Kit (Invitrogen) prior to staining with the indicated Abs. H-2D^b gp33 tetramer was used to identify LCMV-specific CD8⁺ T cells. The following mAbs from BD Biosciences, BioLegend, or eBioscience were used: anti-CD4 (RMA4-5), anti-CD8a (53-6.7), anti-CD19 (1D3), anti-CD44 (IM7), anti-CD62L (MEL14), anti-CD127 (A7R34), anti-KLRG1 (2F1), anti-Thy1.1 (Ox-7), and anti-Thy1.2 (53-21). BrdU incorporation was assessed using a BrdU Flow Kit (BD Biosciences), according to the manufacturer’s instructions. Cells were analyzed on an LSR II (BD Biosciences), and data were analyzed with FlowJo software (TreeStar).

Statistical analyses were performed using two-tailed t tests run on Prism Version 5 (GraphPad) software.

Both HSCs and CM T cells face similar demands of long-term homeostatic renewal and the capacity to produce differentiated progeny while self-renewing the less differentiated fate. Therefore, we tested whether CM T cells, like HSCs (11), require BM homing to ensure durability during homeostasis and differentiation. We used mice with a conditional allele of CXCR4 (8) to assess the role of BM homing in CD8⁺ T cell responses to LCMV infection. CXCR4^F/F mice were bred to CD4-Cre mice to avoid defects in double-negative thymocytes that were observed in CXCR4^F/F–Lck–Cre⁺ mice (12). Compared with CXCR4-proficient animals, the BM of CXCR4^F/F–CD4–Cre⁺ mice contained ∼8-fold fewer naive and 14-fold fewer CM CD8⁺ T cells (Fig. 1A). The abundance of CXCR4-deficient EM CD8⁺ T cells in the BM was normal, consistent with prior studies showing impaired migration of CM, but not EM, phenotype cells toward a CXCL12 gradient (4).

CXCL12-abundant reticular cells are an important BM stromal source of homeostatic cytokines and a putative niche that could trans-present IL-15 to T and NK cells in a cognate fashion (5, 6, 13–19). Therefore, we asked whether CXCR4-deficient T cells are even capable of accessing BM parenchyma or instead are simply circulating in BM capillary sinusoids. Intrasinusoidal cells were identified by injection of CD45.2 Ab (10). Most CXCR4-proficient cells were in the BM parenchyma proper, whereas the majority of naive CXCR4-deficient CD8⁺ T cells and a substantial fraction of CXCR4-deficient memory CD8⁺ T cells were intrasinusoidal (Fig. 1B). Thus, CXCR4 may play a critical role in homing of naive and memory CD8⁺ T cells into the BM parenchyma.

To assess the antiviral response, we generated peripheral chimeras by transferring WT Thy1.1/1.1 and CXCR4^F/F–CD4–Cre⁺ (knockout [KO]) Thy1.1/1.2 P14 TCR–transgenic T cells into WT Thy1.2/1.2 recipients prior to infection of recipients with LCMV. Postinfection, we evaluated recovery of KO P14 T cells normalized to the proportion of KO P14 T cells present on the day of adoptive transfer. At day 8 postinfection, there was a specific defect in CXCR4-deficient P14 T cells populating the BM, but there was no significant defect in blood, spleen, or LNs (Fig. 2A, 2B).

CD4-Cre–expressing P14 TCR–transgenic T cells have largely completed gene deletion prior to transfer (20), whereas GzmB-Cre–expressing T cells will undergo most deletion rapidly following Ag activation (21). Therefore, we analyzed endogenous T cell responses to virus in mice with deletion of CXCR4 following infection. Tetramer staining of gp33-specific T cells in CXCR4^F/F–GzmB–Cre⁺ mice and CXCR4^F/F–GzmB–Cre⁻ mice at day 8 revealed little difference in relative numbers of tetramer⁺ CD8⁺ T cells in the blood (Fig. 2C) or in the absolute numbers of gp33 tetramer⁺ CD8⁺ T cells in other organs (Fig. 2D). Together, these results suggest that CXCR4-deficient T cells can undergo normal effector cell clonal expansion whether CXCR4 is deleted prior to or following Ag challenge. It is possible that the prechallenge loss of CXCR4 in CD4-Cre–expressing cells compared with the acute loss of CXCR4 in GzmB-Cre–expressing cells explains the difference in onset of the BM-homing phenotype.

Serial analyses of mice that received WT P14 T cells and CXCR4^F/F–CD4–Cre⁺ P14 T cells revealed that the ratio of WT/KO P14 T cells was stable in the blood until day 30, after which CXCR4-deficient cellularity began to decline. By day 140, KO P14 T cells decreased to 40% of their original representation (Fig. 2A). To rule out the possibility that recipient mice preferentially rejected CXCR4-deficient P14 cells, we followed the LCMV response of CXCR4^F/F–GzmB–Cre⁺ versus CXCR4^F/F–GzmB–Cre⁻ littermate controls using gp33 tetramers. Similar to the analysis of P14-transgenic T cells, we found that CXCR4-deficient tetramer⁺ cells were present in the blood stably until day 30 (Fig. 2C). However, by day 140, CXCR4^F/F–GzmB–Cre⁺ mice had 3-fold fewer gp33 tetramer⁺ CD8⁺ T cells in the blood relative to CXCR4^F/F–GzmB–Cre⁻ mice.

In both P14 chimeras and CXCR4^F/F–GzmB–Cre⁺ mice, we observed a pronounced impairment of memory CD8⁺ T cells in the BM after resolution of infection. In addition to the BM defect, there was a defect in representation of CXCR4-deficient memory CD8⁺ T cells in the spleen and LNs (Fig. 2B, 2D). The results, so far, suggest a critical role for CXCR4 and BM homing in the long-term maintenance of the memory CD8⁺ T cell population, but a limited role for homeostatic BM conditioning on the effector expansion of a naive T cell upon Ag challenge.

After viral clearance, CM CD8⁺ T cells gradually and steadily express increasing amounts of the transcription factor eomesodermin and its target gene, CD122, a key receptor of IL-15 responsiveness (22–24). The phenotype of eomesodermin-deficient memory CD8⁺ T cells, which also have reduced expression of CXCR4, is marked by progressively poorer occupancy of BM and a selective loss of CM CD8⁺ T cells (13).

At day 140 following LCMV infection, we found that the proportion of CXCR4-deficient CM cells was most obviously deficient in the BM (Fig. 3A), although the total number of CXCR4-deficient CM cells was reduced in both BM and LNs, another characteristic location of CM cells (Fig. 3B). As might be predicted from their similarity to cells deficient in eomesodermin or IL-15 signaling (13, 14, 16, 19), we found that CXCR4-deficient CM T cells undergo reduced homeostatic proliferation, as assessed by BrdU incorporation, but without obvious evidence of increased apoptosis (Fig. 3C). The defect in homeostatic proliferation was more pronounced for CXCR4-deficient CM cells than for EM CD8⁺ T cells (Fig. 3C, 3D). Although impaired homeostasis of CXCR4-deficient memory cells may be due to defective BM homing and reduced homeostatic division of CM cells, it ultimately seems to impact the entire CD8⁺ memory T cell pool because total EM cell numbers are also significantly reduced in BM and spleen (Fig. 3B).

CM CD8⁺ T cells function much like adult stem cells, homeostatically maintaining themselves in the absence of recurrent infection and asymmetrically producing differentiated progeny while renewing themselves during rechallenge (2). Despite the critical role of CXCR4 and BM homing in maintaining homeostatic proliferation of CM CD8⁺ T cells, we found that CXCR4-deficient memory CD8⁺ T cells underwent normal and possibly heightened re-expansion upon rechallenge with L. monocytogenes expressing gp33 (Fig. 4A, 4B). Importantly, CXCR4-deficient memory CD8⁺ T cells retained the capacity to maintain a population of CD44^hi CD62L^hi CM cells while generating CD44^hi CD62L^lo secondary effector cells (Fig. 4C).

The results presented in this article suggest that homeostatic and rechallenge self-renewal are differentially dependent on CXCR4 and BM homing, probably owing to differences in the signaling requirements for each type of self-renewal. Homeostatic proliferation, which seems to be CXCR4 dependent, is thought to be primarily cytokine driven and Ag independent (25), whereas rechallenge differentiation, which seems to be CXCR4 independent, is induced by Ag (26), occurs mainly in the spleen and LNs, and achieves functional self-renewal by asymmetric production of CM and secondary effector cells (2). In chronic infection, exhausted T cells no longer undergo homeostatic self-renewal (27), but progenitors stimulated by Ag can divide and reproduce themselves while simultaneously producing more differentiated cells that suppress viral burdens (3). The importance of rechallenge self-renewal in maintaining equilibrium with the virus is finally unveiled when continual Ag-induced division ultimately results in the collapse of both forms of self-renewal (3, 27). Recently, there has been interest in identifying stem cell–like CD8⁺ T cells with durability and protective capacity (28, 29). Our results suggest that the long-term maintenance and protective capacity of stem cell–like T cells might not be regulated by the same signals, which will be an important consideration in the design of vaccines.

We thank D. Littman and Y-R. Zou for providing the CXCR4 conditional allele and B. Barnet, M. Ciocca, S. Gordon, Y. Grinberg-Bleyer, J. Johnnidis, K. Mansfield, and M. Paley for advice and assistance.

The authors have no financial conflicts of interest.