Critical Role for CXCR4 Signaling in Progenitor Localization and T Cell Differentiation in the Postnatal Thymus ¹

Like all cells of hemopoietic origin, T lymphocytes must be generated throughout life to replace losses from cellular senescence, trauma, and, in the case of lymphocytes, Ag-driven clonal expansion. The self-renewing progenitors that initiate this process do not reside in the thymus, but, rather, are recruited by the thymus from marrow-derived progenitors that circulate in the blood. Although progenitors with lymphoid lineage preference have been identified within the bone marrow itself (1), numerous studies suggest that T lineage commitment occurs mainly after thymic entry, since many early intrathymic progenitors display multilineage potential (reviewed in Ref.2). During intrathymic residence, these multilineage progenitors are induced to undergo a series of differentiative and proliferative events that produce mature T cells. Thymic stromal cells play a key role in this process, by providing uncommitted progenitors with the signals that induce progressive phases of lineage commitment and proliferative expansion, as well as other functions (3). Different stromal environments within the thymus perform discrete functions, and consequently, cells at various stages of development are found in different tissue regions. Blood-borne progenitors arrive through large venules deep in the tissue (4, 5, 6, 7, 8, 9, 10, 11), near the junction of cortical and medullary compartments (cortico-medullary junction (CMJ) ³). Early intrathymic progenitor cells (defined as CD4/CD8 lineage double-negative (DN)) then migrate specifically outward, undergoing a series of developmental events as they move across the cortex toward the capsule (4, 11). Residence in the subcapsular region coincides with differentiation to the CD4/CD8 lineage double-positive (DP) stage, and a subsequent reversal in the polarity of travel, back into the cortex toward the medulla (12). However, only cells that express TCR of appropriate specificity are allowed to move into the medulla and become fully mature. Thus, T cell differentiation in the postnatal thymus is intimately linked to migration into and between tissue regions that induce and support various elements of the differentiation process.

A number of requirements are implicit in the directional migration of cells within tissues. These include adhesive interactions between migrating cells and a stable matrix as well as signals that induce directional guidance. We have recently shown that the adhesive requirements include α₄ integrin-mediated adhesion to a matrix consisting of VCAM-1⁺ stromal cells (13). In this manuscript we address the question of the directional signals for cortical migration. Using a variety of biochemical and functional assays in vitro and in vivo, we found a nonredundant and critical role for the CXCR4/CXCL12 axis in signaling progenitors arriving from the blood to migrate specifically in the direction of the cortex and thus to encounter the differentiative and proliferative signals required for proper T cell differentiation in the postnatal thymus.

Single-cell suspensions of thymocytes from 5-wk-old C57BL/6 male mice were depleted of small CD4⁺8⁺ cells by density gradient centrifugation, followed by staining with a cocktail of lineage Abs recognizing CD3 (clone KT3), CD4 (clone GK1.5), CD8 (clone 53-6.7), Mac-1 (clone M1/70), Gr-1 (clone RB6-8C5), and erythroid (clone TER-119). Depleted cells were stained with CD24, CD25, and CD44 Abs and sorted using the CD44/CD25 criteria described in the text, together with lymphoid forward and side scatter gates, a dead cell exclusion gate, a CD24⁺ gate, and a doublet exclusion gate (forward side scatter-width). Postsort purities were generally >99%.

RNA was extracted from purified progenitor populations using RNeasy Mini Kit columns (Qiagen, Valencia, CA). DNase I digestion was performed on the column, using 2 U of amplification grade DNase I (Invitrogen, Carlsbad, CA) for 20 min at 37°C. RNA was eluted in 30 μl of diethylpyrocarbonate-treated water and was used for RT-PCR. All RNA samples were tested for DNA contamination by PCR without RT. RT-PCR was performed using the Superscript One-Step RT-PCR with Platinum Taq (Invitrogen, San Diego, CA). Primer sequences were as follows: CXCR4 (GenBank accession no. NM_009911): forward, gtcagaggccaaggaaactg; reverse, cgaggaaggcatagaggatg; CXCL12 (GenBank accession no. NM_021704): forward, gtggcttcatggcaagattc; reverse, ctgtagcctgacggaccaat; and hypoxanthine guanine phosphoribosyl transferase (HPRT; GenBank accession no. NM_013556): forward, atcagtcaacgggggacata; reverse, ttgcgctcatcttaggcttt.

Purified progenitor cells and RNA were isolated as described above. cDNA synthesis and labeling, hybridization to the Affymetrix U74A gene chip, imaging, and data analysis were performed by the Genomic Core Facility at Memorial Sloan-Kettering Cancer Center. Briefly, double-stranded cDNA was synthesized from a total of 5 μg of pooled RNA template from each progenitor population. Linear amplification with T7-RNA polymerase and biotin labeling were performed during an in vitro transcription step. The resulting biotin-labeled cRNA was fragmented and hybridized to the array for 16 h at 45°C. Following hybridization to the U74A array, an automated washing and staining protocol was performed on a dedicated fluidics station. The array was then immediately scanned on a GeneArray Scanner (Hewlett-Packard, Palo Alto, CA). Gene expression results were calculated using MicroArray Suite 5.0 and Wilcoxon rank analysis of differences for each of the eight matched and eight mismatched oligonucleotide probe sets in the array. For comparison of expression levels between progenitor populations (i.e., between arrays), the mean expression level for all genes characterized as present (i.e., where Wilcoxon rank differences between matched and mismatched probe sets had a null hypothesis significance of p < 0.04) was taken. This global mean was then scaled up or down, as appropriate, to an arbitrary value of 500, and all individual gene expression values from that array were adjusted proportionally.

Transverse sections of 10-μm thickness were prepared from whole cryopreserved thymus. Sense and antisense probes for CXCL12 were cloned from PCR products amplified using the primers described above, after cloning into pCRII-TOPO using the TA Cloning Dual Promoter kit (Invitrogen) according to the manufacturer’s instructions. Orientation of the insert was confirmed by restriction mapping and sequencing. Digoxigenin-labeled probes were synthesized by in vitro transcription from linearized plasmid, using DIG RNA Labeling Mix (Roche, Indianapolis, IN) and the appropriate enzyme. For antisense probe, template was linearized with SpeI, and cRNA was synthesized using T7 polymerase (Roche). For sense probe, template was linearized with EcoRV, and cRNA was synthesized using SP6 polymerase (Promega, Madison, WI). All subsequent steps were performed at room temperature unless noted. Sections were fixed for 20 min in 4% formaldehyde/PBS, washed, and treated for 8–10 min with proteinase K at 7.5 μg/ml. Tissue sections were fixed again in formaldehyde and treated with 0.25% acetic anhydride in 0.1 M triethanolamine for 10 min. Sections were then incubated in hybridization buffer (50% formamide, 5× SSC, 5× Denhardt’s reagent, and 250 μg/ml yeast tRNA) for 2 h and hybridized overnight at 55°C in fresh buffer containing a digoxygenin-labeled CXCL12 probe (0.8 μg/ml) and denatured herring sperm DNA (5 μg/ml). After sequential washes in 5× and 0.2× SSC at 60°C, bound probe was detected using peroxidase-conjugated, anti-digoxin Ab (Jackson ImmunoResearch Laboratories, West Grove, PA) and tyramide signal amplification (PerkinElmer, Boston, MA) as recommended by the manufacturer. For immunohistochemistry, detection was performed using 3,3′-diaminobenzidine (Vector Laboratories, Burlingame, CA). For two-color detection, the RNA probe was imaged using Alexa594 streptavidin conjugate (Molecular Probes, Eugene, OR), followed by standard immunofluorescent staining for cytokeratin using FITC-conjugated clone C11 (Sigma-Aldrich, St. Louis, MO).

Transverse sections of 10-μm thickness were prepared from whole cryopreserved thymus. Sections were fixed in ice-cold 75% ethanol, stained for 20 s in Mayer’s hematoxylin, and dehydrated through 75, 95, and 100% ethanol. Dehydrated slides were stored desiccated at 4°C until used. At least 30 min before dissection, slides were brought to room temperature under desiccated conditions. Dissection was performed using the P.A.L.M. system (Carl Zeiss, Thornwood, NY) as recommended by the manufacturer. Briefly, computerized images were acquired, and regions of interest (see Fig. 2) were outlined on the screen. The margins of the regions of interest were ablated by exposure to a high power laser, and the central regions were catapulted into microfuge caps containing mineral oil. RNA extraction and RT-PCR were performed as described above.

Assays were performed using the ChemoTx 96 microplate system (NeuroProbe, Gaithersburg, MD). Briefly, plates containing 3.2-mm diameter wells and Transwell membranes with 5-μm pores were precoated with purified mouse fibronectin (Invitrogen) at 10 μg/ml for 1 h at 37°C, followed by air-drying. Murine CXCL12 (PeproTech, Rocky Hill, NJ) was added to the lower chamber in medium (RPMI 1640 containing 0.5% BSA and 5 mM HEPES); the optimal concentration, determined from a number of preliminary studies, was 8 nM. The fibronectin-coated membrane was placed carefully on top, and 25 μl of cell suspension containing 1–1.5 × 10⁴ cells was added over the membrane. Controls included medium only in the bottom well or 25 μl of cell suspension plated directly in the bottom well. Plates were incubated for 90 min at 37°C, following which cells remaining above the membrane were carefully removed by wiping and rinsing with PBS. Cells remaining trapped within the membrane were loosened by treating each cell well with 25 μl of 2 mM EDTA in PBS for 5 min, followed by centrifugation. The contents of the wells were photographed, and the number of cells in a central field were counted. Cell counts were also confirmed by pooling the contents of replicate wells and counting on a hemocytometer. The percentage of cells migrating was calculated based on the controls where cells were added directly to the bottom chamber.

Generation of CXCR4^loxP mice (D. Littman, Skirball Institute, New York University School of Medicine, New York, NY) will be described elsewhere. Briefly, loxP sites were introduced into the flanking regions surrounding exon 2 of CXCR4. Mice displaying germline transmission were bred to mice expressing Cre recombinase under the lck proximal promoter (14). The lck[Cre]-transgenic offspring of this cross (CXCR4^loxP/+) were backcrossed to generate lck[Cre]/CXCR4^loxp/loxP mice, which served as marrow donors. Mice expressing lck[Cre] only were used as controls.

Donor bone marrow was prepared by flushing marrow from tibias and fibulas of lck[Cre]/CXCR4^loxP/loxP or control lck[Cre] mice, followed by hypotonic lysis of RBC. Recipients for marrow transplantation were sex-matched CD45.1 congenic mice (B6.SJL-Ptprc^a Pep3^b/BoyJ; The Jackson Laboratory, Bar Harbor, ME). Recipient mice received 6 Gy of gamma irradiation ∼20 h before transplantation. Irradiated recipients received ∼3 × 10⁷ donor cells in total, which were a mixture of mutant or control donor cells together with syngeneic (CD45.1) marrow. After 5–7 wk, chimeric mice were sacrificed, and hemopoietic tissues were harvested (thymus, bone marrow, blood, spleen). For analysis of CD4/CD8 phenotype in donor thymocytes, single-cell suspensions were stained with CD45.1-Alexa633, CD45.2-Alexa680, CD4-Alexa488, CD8-PE, and CD90-biotin/PE-Texas Red streptavidin. For analysis of progenitor thymocyte stages, suspensions were first stained with the cocktail of lineage Abs described above, followed by PE-Texas Red-conjugated anti-rat IgG, and then nonspecific rat IgG to block excess binding sites. Following this, cells were stained with the CD45 Abs described above as well as CD44-Alexa488 and CD25-PE. In all cases, 4′,6-diamidino-2-phenylindole dihydrochloride was added at 0.1 μg/ml to discriminate dead from live cells. Analysis was performed on an LSR Cytometer (BD Biosciences, San Jose, CA) with modifications as previously described (15). For in situ localization of donor cells by immunofluorescent microscopy, 5-μm transverse sections of cryopreserved thymus were fixed in ice-cold acetone, followed by staining in PBS/5% FBS using the following Abs: CD45.2-Alexa594, FITC-conjugated pan cytokeratin (clone C-11; Sigma-Aldrich), CD25-Alexa488. 4′,6-Diamidino-2-phenylindole dihydrochloride (0.25 μg/ml) was used as a counterstain, followed by fluorescent imaging using a mercury light source.

Chemokines are heavily implicated in the directional migration of lymphocytes, such as that which occurs during steady state differentiation in the thymus (4). To elucidate the role of chemokine signals in this process, we first performed semiquantitative RT-PCR screening for all known murine chemokine receptors, using RNA template extracted from defined progenitor stages: these were DN1 (CD3⁻4⁻8⁻25⁻44^high), DN2 (CD3⁻4⁻8⁻25⁺44^high), DN3 (CD3⁻4⁻8⁻25⁺44^low), and pre-DP (CD3^low4^low8^low 25⁻44^low). A large variety of chemokine receptors was found to be expressed on one or more of these stages (data not shown). Among these, CXCR4 emerged as the most abundant and most ubiquitously expressed chemokine receptor message (Fig. 1). Message levels by RT-PCR were highest in the CD25⁺ stages (DN2 and DN3) that are actively migrating across the cortex (4), but were found in all progenitor cells at appreciable levels. These RT-PCR findings were confirmed by hybridization to the Affymetrix U74A microarray (Fig. 1 b). Global expression analysis of microarray results (see Materials and Methods) indicated that the lowest levels of CXCR4 expression, occurring in DN1 and pre-DP cells, were equivalent to the average expression levels for all genes expressed, while expression in DN2 and DN3 cells was higher. Together, these data indicate that all early intrathymic progenitors are potentially synthesizing CXCR4 and therefore have the potential to respond to its ligand.

To further evaluate the involvement of CXCR4 in facilitating the migration of progenitors into the thymic cortex, expression of its ligand, CXCL12, was evaluated. RNA in situ hybridization showed that CXCL12 was expressed on scattered cells throughout the cortex, but not the medulla (Fig. 2,a), consistent with the results of others (16). The morphology of cells expressing CXCL12 was reticular, consistent with that of cortical stromal cells (17). Differential expression of CXCL12 in thymic tissue regions was further evaluated by RT-PCR analysis (Fig. 2, b and c), using regionally dissected thymic tissue. Semiquantitative comparison to a ubiquitously expressed gene (HPRT; see Fig. 2) indicated that CXCL12 levels were highest in the outer regions of the cortex, but were detectable throughout, and were virtually undetectable in the medulla. To further characterize the nature of cells producing CXCL12 in the cortex, costaining of CXCL12 mRNA and cytokeratin protein was performed (Fig. 2 d). Cells expressing CXCL12 were uniformly cytokeratin⁺, although not all cytokeratin⁺ cells expressed CXCL12. These data indicate that the ligand for CXCR4 is expressed in the thymus in a manner consistent with the ability to guide progenitors into the cortex and away from the medulla.

The data in Figs. 1 and 2 indicate the potential of early intrathymic progenitors to respond to CXCL12 via CXCR4. Functional analysis of these correlative RNA expression results was next sought. In the first instance, Transwell migration assays were performed using CXCL12 as a chemoattractant and purified progenitors as target cells (Fig. 3). A variety of CXCL12 concentrations and incubation times were tested in initial experiments. All progenitor populations responded proportionally to changes in the CXCL12 concentration; preliminary experiments showed that optimal transmigration occurred at a concentration of 8 nM. Ninety minutes provided a maximal signal-to-noise ratio in this system, i.e., at longer times, response to chemokine did not increase proportionally to random migration (no CXCL12 added). Consistent with the activities predicted by receptor expression levels (Fig. 1), transition to the CD25⁺ stages of development correlated with the most robust migration response, but all progenitor populations demonstrated a significant response to CXCL12. These data confirm the prediction of mRNA screening, by showing that early intrathymic progenitors are indeed capable of responding to CXCL12-mediated signals via CXCR4.

Although progenitor thymocytes were capable of responding to optimal concentrations of CXCL12 in vitro, there was no way of accurately determining that these concentrations of CXCL12 recapitulated in vivo conditions. Therefore, confirmation of the role of CXCR4 signaling in the directional movement of early intrathymic progenitors in vivo was sought. To preclude potential complications related to a role for CXCR4 in the homing of circulating progenitors to the thymus, inactivation of CXCR4 was targeted specifically to thymocytes by constructing a mouse with lox-P recombination sites flanking the CXCR4 gene (CXCR4^loxP), and crossing homozygous targeted mice offspring to mice expressing Cre recombinase under control of the lck proximal promoter (14). The lck promoter first becomes active in immature thymocytes (18), apparently during the DN1 stage (19), and can thus be used to target deletion very early during T cell differentiation. To evaluate such mutant cells in the context of a normal thymic microenvironment, stable bone marrow chimeras were constructed using donor marrow from lck[Cre]/CXCR4^loxP/loxP mice or control lck[Cre]-only mice transplanted into sublethally irradiated, wild-type, CD45.1-congenic recipients. After return to the steady state (5–7 wk), thymuses from chimeric animals were removed, and the two lobes were separated. One lobe was used immediately for phenotypic analysis by flow cytometry, and the other was frozen for subsequent histology; in this way, developmental stage could be directly correlated with localization in a single chimeric organ.

Fig. 4 shows an example of results obtained from four lck[Cre]/CXCR4^loxP/loxP chimeras, and three control chimeras (lck[Cre] only). In all four of the former, the proportion of thymocytes derived from lck[Cre]/CXCR4^loxP/loxP donors was very low (0.2 ± 0.1%), especially when compared with non-T lineage cells in other tissues (such as granulocytes in bone marrow; Fig. 4,a). When control (lck[Cre] only) marrow was used, DN, DP, and mature cells of donor origin were present in normal proportions (Fig. 4 b) and, other than the CD45 congenic marker, were essentially indistinguishable from cells of recipient origin. In all four lck[Cre]/CXCR4^loxP/loxP chimeras, however, all donor cells were DN, while DP and mature SP cells were completely absent (recipient thymocytes were present in normal numbers and proportions; data not shown). Further characterization of the DN progeny of transplanted lck[Cre]/CXCR4^loxP/loxP donors showed that in three of these four chimeras, virtually all DN cells present were arrested at DN1 (93 ± 5%), with a few DN2 (3 ± 2%) and DN3 (2 ± 2%) cells. In the fourth case there were substantially more DN2 (24%) and DN3 (62%) cells; nonetheless, as stated above, no DP or SP cells were found. The basis for the differences in this chimera are not clear, but since thymic compartmentalization is somewhat amorphous, it is possible that discrimination between regulatory signaling environments is somewhat leaky. Nonetheless, the absence of DP and more mature cells is consistent in all lck[Cre]/CXCR4^loxP/loxP donor progeny, indicating that signaling through CXCR4 is linked to the proper differentiation of early intrathymic progenitors. The nature of this linkage is further evaluated in the next section.

The fundamental mechanism for arrest at the DN stage in CXCR4-deficient thymocytes was further revealed by immunofluorescent localization of targeted donor cells in chimeric thymus lobes (Fig. 5). The progeny of control (lck[Cre] only) marrow donors were found throughout the thymus, consistent with the presence of all developmental stages (Fig. 4,b). In contrast, thymocytes derived from lck[Cre]/CXCR4^loxP/loxP donor cells were found only at the CMJ or, more rarely, in the deep cortex. Some cells with nonlymphoid morphology were also found in the medulla; these represent CD11c⁺ dendritic cells that normally develop in or home to the thymus (36). The location of CXCR4-targeted lymphoid progeny in the thymus correlates with the sites where thymus-homing progenitors first enter the organ (4, 5, 6, 7, 8, 9, 10, 11). These results indicate that in the absence of CXCR4 signaling, progenitors recruited from the blood fail to move efficiently into the cortex, where they would otherwise differentiate further. This is further emphasized by comparing the localization of mutant donor thymocytes to CD25⁺ progenitors of recipient origin (Fig. 5 b); mutant donor cells fail to move into the regions where recipient CD25⁺ cells are found, and consequently fail to differentiate past the DN1 stage. Together, our findings show that CXCR4-mediated signaling in response to stromally derived CXCL12 is crucial for mediating cortical localization of progenitors homing to the thymus from the blood and consequently for the success of steady state T cell differentiation.

A role for chemokines in homing to or migration within the thymus has been indicated by several earlier studies (reviewed in Ref.20). In contrast to the present study, most of these have focused on migration of progenitors into the fetal primordium (21, 22), migration from the outer cortex toward the medulla (16, 23, 24, 25, 26), or migration out of the organ (27, 28). The role of chemokines in enabling progenitor migration outward across the cortex during steady state differentiation has not been characterized, although it is clearly implicated. Similar to the results presented here, studies of progenitor cells from fetal thymus indicated that they can respond to CXCL12 in Transwell migration assays (23), although it should be noted that the migratory requirements for fetal and postnatal progenitors cannot be assumed to be the same, since fetal progenitors do not migrate outward across the cortex during differentiation, nor do they enter the thymus through blood vessels at the CMJ, as the fetal organ is neither structured nor vascularized at the time of progenitor seeding (29). Numerous other chemokines have been shown to be expressed in the fetal thymic primordium (21, 22) and to induce chemotaxis of fetal thymic progenitors (21). The multiplicity of chemokines that can attract fetal progenitors (21) may explain the results obtained from CXCR4 germline knockouts, where fetal organs transplanted to wild-type mice developed normally (30), since other chemokines/chemokine receptors can compensate for the absence of CXCR4 signaling in the embryonic rudiment. In contrast, our studies show an absolute requirement for CXCR4 signaling in the postnatal thymus, a function that cannot be compensated by other signals despite the expression of numerous other chemokine receptors (data not shown). The inability of cells lacking CXCR4 to differentiate past the DN1 stage also clearly illustrates that migration into the cortex and interaction with signals found only in the cortical microenvironment are obligatory processes for continuous production of T lymphocytes during postnatal life. This does not preclude additional roles for CXCR4 in thymic T cell differentiation, such as recruitment of new progenitors into the thymus or movement of more mature thymocytes inward toward the medulla, although these associations remain to be proven.

Recently, two works from the same group have shown that germline mutation of the CXCR4 gene results in a reduction of T cells produced by the thymus (31, 32). One important fact is that the effects of CXCR4 deficiency are much more severe when deficient cells are placed in competition with wild-type cells using hemopoietic chimerism, as shown by our present work and that of Ara et al. (32). The effects of CXCR4 deficiency in our system are even more severe than those found in the germline mutant model; this could reflect the use of fetal progenitors (in germline mutant studies) vs adult marrow progenitors (in the present study) and could also be affected by the use of lethal (32) vs sublethal (present study) irradiation. Nonetheless, these studies and others (33) reveal an important role for CXCR4 signaling in the T cell development process. Our results reveal the mechanism for this effect by showing that CXCR4 signaling is required to facilitate entry of thymic homing progenitors into the thymic cortex. Nonetheless, it is important to note that the distribution of CXCL12-producing cells in the thymic cortex appears to be fairly homogenous, and thus it is not intuitive that CXCR4 should polarize migration all the way across the cortex to the capsule. However, several other mechanisms may participate in this process. First, although the producer cells themselves may be evenly distributed, the CXCL12 protein may not be and could, in fact, be present at higher levels in the outer cortical or subcapsular regions. However, it has been shown by others that once the direction of migration is polarized, it is not necessary to maintain a gradient of CXCL12, although CXCR4 signaling must be maintained (34). Thus, initial polarization of newly arrived progenitors in the direction of the cortex and away from the medulla together with continuous presence of CXCL12 throughout the cortex may be sufficient to facilitate the directional migration of early progenitors across the cortex to the capsule. It is also worth pointing out that our comprehensive screen of thymocyte progenitors (see Results) revealed the presence of numerous chemokine receptors, some of which were expressed in a stage-specific fashion (data not shown). Thus, it is possible that chemokines other than CXCL12 may ultimately induce the direction of migration, while CXCL12 serves to activate integrin-mediated adhesion to substrates for migration, consistent with its known functions (35). Experiments are now underway to determine the relative roles of other chemokines in the transcortical migration process. Nonetheless, it is clear that the CXCL12/CXCR4 axis is critical for the initial events in this process and thus for the successful differentiation of T lymphocytes in the fully formed, steady state thymus.

We thank D. Littman (New York University, New York, NY) for bone marrow from lck[Cre] and CXCR4^loxP/loxP mice, L. Berg (University of Massachusetts, Worcester, MA) for advice regarding RNA in situ hybridization, K. Gordon (Memorial Sloan-Kettering Cancer Center, New York, NY) for assistance with cell sorting, and A. Viale (Memorial Sloan-Kettering Cancer Center) for microarray analysis.