Hematopoietic Stem Cells Are Uniquely Selective in Their Migratory Response to Chemokines

Hematopoietic stem cells (HSC)^* are rare, pluripotent, self-renewing cells that give rise to all hematopoietic lineages (1) and migrate among various hematopoietic tissues. The site of hematopoiesis changes several times during embryonic and fetal development in a process thought to be dependent on HSC migration (for review see reference 2). In normal adult animals, HSC continuously migrate from the bone marrow (BM) to the blood and back to the BM, in such a way that circulating HSC are always available in peripheral blood to fill open BM HSC niches (3–5). Large-scale HSC mobilization from BM to blood is induced in adult animals by treatment with cytokines (e.g., G-CSF) and/or cytotoxic drugs (e.g., cyclophosphamide [Cy]) (1). Drug-induced HSC mobilization from BM to blood and the rehoming of transplanted HSC from blood to BM are critical for modern therapeutic transplants that include stem cells. Finally, HSC have recently been shown to be capable of giving rise to hepatocytes (6), and BM cells enriched for HSC have been shown to be capable of giving rise to epithelial cells (7) and cardiac muscle cells (8). It has been proposed that HSC migration into sites of injury may be a mechanism by which damaged tissues are repaired (6). These new discoveries lend urgency to efforts to understand the mechanisms of HSC migration.

Adhesion molecules, such as α4β1 integrin, have been implicated in both HSC mobilization as well as in the rehoming of transplanted HSC placed in the circulation (9, 10). Roles for chemotactic cytokines (chemokines) in hematopoietic stem and progenitor cell biology have also been proposed. HSC are rapidly mobilized from BM to blood after the administration of the chemokine IL-8 to mice (11) or to monkeys (12), and serum IL-8 levels spike just before mobilization after the induction with G-CSF (13). However, the effect of IL-8 in inducing HSC mobilization appears to be indirect (14, 15). The possibility of a direct role for the chemokine stromal cell–derived factor-1α (SDF-1α) in progenitor mobilization was raised by a study of human BM and mobilized peripheral blood (MPB) CD34⁺ cells, in which MPB CD34⁺ cells did not respond as well to SDF-1α as to CD34⁺ cells isolated from BM (16). The authors proposed that the alteration in SDF-1α responsiveness might be part of the mechanism of progenitor mobilization, although only a fraction of human CD34⁺ cells are HSC (17, 18). Finally, SDF-1α has been implicated in the strikingly specific homing of transplanted hematopoietic progenitors to BM. SDF-1α is constitutively expressed on human BM endothelium and appears to cooperate with α4β1 integrin (very late antigen-4) and αLβ2 integrin (LFA-1) in inducing the arrest of circulating progenitors on vascular endothelium (19, 20). As of this writing, however, large panel studies on the ability of chemokines to induce chemotaxis of normal and mobilized HSC have not been published. Therefore, we sought to examine chemotaxis of HSC in a mouse model using a panel of chemokines that bind most known chemokine receptors.

We have previously shown that virtually all of the reconstituting activity of HSC isolated from the BM of untreated C57BL/Ka-Thy-1.1 mice and of mobilized HSC isolated from Cy/G-CSF–treated C57BL/Ka-Thy-1.1 mice is contained in two phenotypically defined populations: Thy-1.1^loSca-1⁺Lin⁻c-Kit⁺Mac-1⁻ cells that are mostly long-term repopulating HSC (LT-HSC), and Thy-1.1^loSca-1⁺Lin⁻c-Kit⁺Mac-1^lo cells that are mostly short-term repopulating HSC (ST-HSC) (21–23). Here, the in vitro chemotaxis of LT-HSC and ST-HSC, obtained from the BM of untreated mice, in response to a panel of chemokines binding most currently described chemokine receptors from the CC and CXC families, was investigated. Chemotaxis of HSC obtained from mice treated with a mobilizing regimen of Cy/G-CSF in response to SDF-1α was also studied. In parallel, the expression of mRNA of chemokine receptors on both LT-HSC and ST-HSC derived from the BM of untreated mice was examined. These studies provide a large-scale survey of HSC chemotactic responses and provide evidence that mouse HSC possess mRNA for several chemokine receptors, yet only migrate in vitro in response to one chemokine, SDF-1α.

6–12-wk-old C57BL/Ka-Thy-1.1 mice were bred and maintained at the Stanford University Laboratory Animal Facility, and given acidified water (pH 2.5) and rodent chow ad libitum.

Cy and G-CSF were administered as previously described (23). HSC were analyzed on day 4 of the regimen, due to the expansion of HSC in the BM and the large numbers of HSC in the blood and spleen on that day (23).

Single cell suspensions of BM and spleen cells were prepared as previously described (23). Blood was collected from animals after heparin injection and induction of anesthesia with methoxyflurane (Mallinckrodt Veterinary, Inc.). After the incision of the right atrium, blood was removed from the thoracic cavity with a Pasteur pipette. The left ventricle was then perfused via a 25-gauge needle with 10 mM EDTA in HBSS (GIBCO BRL) and blood/EDTA from the thoracic cavity was removed. Blood was subjected to dextran sedimentation and ammonium chloride erythrocyte lysis as previously described (23). Splenic erythrocytes were lysed in all experiments and BM erythrocytes were lysed in most experiments.

mAbs used in immunofluorescence staining were prepared from hybridomas and included 19XE5 (anti–Thy-1.1), 2B8 (anti–c-Kit), and E13 (anti–Sca-1, Ly6A/E). Lineage marker mAbs included KT31.1 (rat anti–mouse CD3), GK1.5 (rat anti–mouse CD4), 53-7.3 (rat anti–mouse CD5), 53-6.7 (rat anti–mouse CD8), Ter119 (rat anti–mouse erythrocyte-specific antigen), 6B2 (rat anti–mouse B220), 8C5 (rat anti–mouse Gr-1), and M1/70 (rat anti–mouse Mac-1). Fc receptors were blocked with mAb 2.4G2 (anti-FcR) at appropriate points in staining procedures.

Cells were stained in HBSS/2% calf serum with the lineage marker antibodies listed above at 10⁸ cells/ml of previously determined optimal concentrations of antibody solution on ice for 25 min, rinsed twice in staining medium (SM), and enriched by incubation on a rocking platform at 4°C for 20 min with SM-washed sheep anti–rat IgG magnetic beads (Dynabeads; Dynal) at a 1:1 ratio of beads/cells in SM. After the removal of beads and attached cells, the remaining cells were incubated on a rocking platform at 4°C for 30 min with fresh washed beads at a 4:1 ratio of beads/cells, and a second round of bead/cell removal was accomplished. Typical lineage staining profiles of whole BM and lineage-depleted BM are shown in Fig. 1 A.

Lineage-depleted tissues were suspended in RPMI medium (GIBCO BRL) with 10% serum and incubated in polystyrene tissue culture flasks at 37°C for 1 h to remove adherent cells and allow time for resensitization of potentially desensitized chemokine responses (24). Cells were then placed into a transwell chemotaxis assay as previously described (25) at 5 × 10⁵ − 2 × 10⁶ cells per top well. The following chemokines were added to the bottom well, and cells were allowed to migrate for 2 h at 37°C: mouse JE, mouse eotaxin, human thymus- and activation-regulated chemokine, mouse regulated upon activation, normal T expressed and secreted (RANTES), mouse macrophage inflammatory protein (MIP)-1α, human MIP-3α, human I-309, mouse KC, and human SDF-1α (PeproTech); mouse MIP-1β, human MIP-3β, mouse thymus–expressed chemokine (TECK), mouse monokine induced by IFN-γ (MIG), and mouse B lymphocyte chemokine (BLC)/BCA-1 (R&D Systems); and human IL-8 (a gift from K. Matsushima, University of Tokyo, School of Medicine, Tokyo, Japan). Optimal chemokine concentrations were determined in preliminary chemotaxis assays using BM cells or splenocytes. Two chemokines (I-309 and eotaxin) did not elicit responses from any BM or splenic subset, and therefore were used at published concentrations (24). Biological activity of I-309 and eotaxin was confirmed using cell lines expressing CCR8 and CCR3, respectively. Chemokines were used at final concentrations of 100 nM, with the following exceptions: 1 nM JE; 3 nM MIP-1β; 5 nM G-CSF; 50 nM SDF-1α; 300 nM KC and TECK; and 500 nM BLC.

After the addition of a fixed number of 15 μm polystyrene beads (Polysciences) for normalization of cell numbers among wells as previously described (25), migrated cells were carefully collected from bottom wells, centrifuged, restained with lineage marker antibodies, and then stained with PE-conjugated goat anti–rat polyclonal antibody (Jackson ImmunoResearch Laboratories) to visualize lineage markers. Cells were then washed twice and stained with a cocktail containing FITC-conjugated (Molecular Probes) anti–Thy-1.1 (19XE5), allophycocyanin-conjugated (APC; Cyanotech) anti–c-Kit (2B8), and TxR-conjugated (Molecular Probes) anti–Sca-1 (E13). In some experiments, PE-conjugated lineage antibodies were also included with the anti–Thy-1.1, anti–c-Kit, and anti–Sca-1 antibodies. Stained cells were resuspended in SM containing propidium iodide in preparation for flow cytometry.

Cells were analyzed and sorted by multiparameter flow cytometry on either a modified 2-laser FACS Vantage^® (Becton Dickinson), or a modified 3-laser cytometer (Cytomation, Inc. and Becton Dickinson), made available through the flow cytometry shared user group at Stanford University. Flow cytometry data were analyzed using FloJo^® software (Treestar, Inc.). Representative gating for LT-HSC and ST-HSC is shown in Fig. 1 A.

HSC were enriched by positive selection for the c-Kit antigen with magnetic beads (Miltenyi Biotec) as previously described (26), and stained for flow cytometry with PE-conjugated lineage mAbs as previously described, FITC-conjugated mAb against Thy1.1, TxR-conjugated mAb against Sca-1, and APC-conjugated mAb against c-Kit. Lin^−/loThy-1.1^loSca-1⁺c-Kit⁺ HSC were double sorted for RT-PCR analysis.

RNA from FACS^®-sorted HSC (∼20,000 cells) was extracted with Trizol (GIBCO BRL) according to the manufacturer's instructions. Isolated RNA was resuspended in diethylpyrocarbonate (DEPC)-treated dH₂O and incubated with DNaseI (RNase-free; Boehringer) for 20 min at 37°C to remove contaminating genomic DNA. Before RT, DNaseI was inactivated and RNA was denatured by incubation at 70°C for 10 min.

After chilling on ice, RNA samples were split equally into two 0.5-ml Eppendorf tubes (3.5 μl per tube), and used for either the RT reaction (+RT), or for a control reaction in which the RT was omitted (−RT control). 16.5 μl of the appropriate (+ or −RT) RT master mix was added to each tube containing 3.5 μl RNA, for a final volume of 20 μl.

4 μl 5 × RT buffer; 2 μl 0.1 M dithiolthreitol; 1 μl random hexamer (137 pmoles/μl); 8 μl dNTPs (2.5 mM each dNTP); 0.5 μl RNAsin (Promega); plus either 1 μl Superscript II RT (+RT; GIBCO BRL) or 1 μl DEPC-dH₂O (−RT control). RT reactions were performed by incubation at 42°C for 1 h.

Specific cDNA were amplified from the RT reactions by hot-start PCR using 2 μl per tube of RT reaction as template and a total volume of 50 μl per reaction. PCR master mix per tube: 5 μl 10× PCR buffer; 5 μl 25 mM MgCl₂; 4 μl dNTPs (2.5 mM each); 1.5 μl primer 1 (16.7 μM stock concentration); 1.5 μl primer 2 (16.7 μM stock concentration); 0.25 μl Taq polymerase; 30.75 μl dH₂O; and 2 μl template (RT reaction). PCR parameters were optimized for each primer set. PCR amplification was performed for 40 cycles. B. Fife (Northwestern University, Chicago, IL) provided some of the chemokine receptor PCR primer sequences.

Oligonucleotide sequences specific for chemokine receptors are listed in Table I.

The RxC test of independence (using the G test) with the Williams' correction was used to analyze the data in Table III. The Student's t test was used to analyze data in Figs. 1 and 4 A. P value of 0.05 was chosen as the limit of statistical significance.

We assayed the migration of LT-HSC and ST-HSC to a panel of chemokines that bind the receptors listed in Table II. In addition, G-CSF, a nonchemokine cytokine widely used for HSC mobilization in experimental animals and in humans, was tested. BM enriched for HSC was placed in the top well of a transwell chamber, and chemokine-containing medium was placed in the bottom well. Responding cells were collected from the bottom well and stained with antibodies to identify and enumerate HSC by flow cytometry (Fig. 1 A). Migration data were expressed as the percentage of input HSC contained in the population placed in the top well that migrated to the bottom well. Strikingly, both LT-HSC (Fig. 1 B) and ST-HSC (Fig. 1 C) migrated only in response to SDF-1α, and did not respond significantly to any of the other chemokines or to G-CSF. Although the magnitude of SDF-1α–induced migration varied among experiments (5.8–49% for LT-HSC and 7.3–57% for ST-HSC), there was no significant difference in the degree of migration to SDF-1α of LT-HSC and ST-HSC. Furthermore, the pattern of nonresponsiveness to other chemokines did not vary between LT-HSC and ST-HSC.

To further characterize SDF-1α–induced HSC migration, we tested the requirement for an SDF-1α gradient to induce HSC migration, or for non-HSC BM cells to stimulate HSC migration to SDF-1α in vitro. HSC migration to SDF-1α was chemotactic (polarized migration toward a chemokine source) not chemokinetic (randomly induced migration), because HSC migration only occurred in a gradient of SDF-1α (bottom well only) and not when SDF-1α was presented in a uniform manner (top and bottom wells) (Fig. 2 A). In addition, HSC chemotaxis did not require the presence of non-HSC BM cells. When HSC alone were sorted directly into the top well of the transwell chamber, these individual sorted HSC migrated in response to SDF-1α, with a migration index in the same range as that determined using a heterogeneous mixture of lineage-depleted BM cells (Fig. 2 B).

We addressed the possibility that the functional reconstituting ability of phenotypically identified migrated HSC was altered by exposure to SDF-1α over the time course of our assay. We tested whether ex vivo incubation or exposure to SDF-1α altered the engraftment potential at near limit dilution doses of LT-HSC. As shown in Table III, neither ex vivo incubation nor exposure to SDF-1α significantly altered the engraftment potential of LT-HSC. 31 wk after the injection of 40 LT-HSC that had been incubated alone for 1 h, followed by incubation with SDF-1α for 2 h, all recipient mice were engrafted by donor cells and four out of eight mice displayed long-term multilineage reconstitution. In a second experiment, 17 wk after the injection of 30 LT-HSC that had been incubated alone for 1 h, followed by incubation with SDF-1α for 2 h, four out of seven mice were reconstituted with donor cells, and three out of seven mice displayed long-term multilineage reconstitution.

In parallel with the chemotaxis studies, we tested for expression of mRNAs encoding the chemokine receptors for the chemokines addressed in this study, plus the chemokine receptors XCR1 (lymphotactin [XCL1] receptor) and CX3CR1 (fractalkine [CX3CL1] receptor). Chemokine receptor expression was determined by RT-PCR from samples representing 1,000 double-sorted HSC (this sorted population contained both LT-HSC and ST-HSC) or 1,000 unfractionated whole bone marrow (WBM) cells as a positive control. CXCR4 (SDF-1α receptor) mRNA was easily detected in HSC at the 1,000 cell level. CCR3 (eotaxin and RANTES receptor) and CCR9 (TECK receptor) messages were also detected (Fig. 3 A and Table IV), and XCR1, CXCR2 (KC/IL-8 receptor), and CXCR5 (BLC receptor) mRNA were observed inconsistently. Other chemokine receptors were not detectably expressed in HSC, although they were clearly present in RNA isolated from an equivalent number of WBM cells. Because it is conceivable that the in vivo migratory behavior of LT-HSC and ST-HSC differ, and that these differences could be reflected in chemokine receptor expression, we further examined CCR3, CCR9, and CXCR4 mRNA expression by LT-HSC and ST-HSC individually. CCR3, CCR9, and CXCR4 mRNA expression was detected in both LT-HSC and ST-HSC (Fig. 3 B).

It has been suggested that decreased responsiveness to SDF-1α may play a role in the mobilization of human hematopoietic progenitors from BM to blood (16). We tested this hypothesis in the mouse by examining the chemotaxis of HSC derived from BM, blood, and spleens of Cy/G-CSF–treated animals. When compared with HSC from the BM of untreated mice, PB-, BM-, and spleen-derived HSC from Cy/G-CSF–treated animals showed indistinguishable responsiveness to SDF-1α (Fig. 4). In addition, both MPB and mobilized spleen HSC expressed easily detectable CXCR4 mRNA by RT-PCR (unpublished data).

HSC migrate in several circumstances: (a) HSC are thought to migrate through the blood to seed new sites of hematopoiesis during embryonic and fetal development (for review see reference 2); (b) HSC migrate continuously between BM and blood in normal adult animals (3–5); (c) HSC migrate in large numbers from BM to blood after the treatment of animals with cytotoxic agents such as Cy and/or cytokines such as G-CSF (23, 27); and (d) injected HSC migrate efficiently to hematopoietic tissues after transplantation in experimental animals, or after clinical transplantation in humans (28). Because HSC have now been shown to also have the capacity to give rise to nonhematopoietic tissues such as liver, migrating HSC may represent a source of pluripotent cells that are constantly available for the repair of damaged organs. Here, we report the first comprehensive study to address the chemotactic responsiveness of HSC to chemokines.

The role of chemokines in hematopoiesis includes guiding progenitor cells to microanatomical sites in BM or thymus for proper maturation (29). However, the roles of chemokines in the migrations of HSC during embryonic and fetal development, the maintenance of adult HSC niches, and the physiological flux of HSC between BM and blood in adults are unclear. Here we showed that in normal adult animals, BM-derived LT-HSC and ST-HSC displayed a sharply restricted responsiveness pattern to chemokines; they only migrated appreciably in response to SDF-1α. Consistent with this, HSC express mRNA for CXCR4, the only known receptor for SDF-1α. The SDF-1α–CXCR4 interaction is important in hematopoietic development, and may have a role in engraftment of BM by HSC. Among other defects, mice lacking either SDF-1α or CXCR4 lack lymphomyeloid hematopoiesis in fetal BM (30, 31). The antibody blockade of CXCR4 prevented engraftment of adult SCID mouse BM by CD34⁺-enriched human cord blood cells (32).

Our data suggest that HSC, in contrast to other leukocyte subsets, only respond to SDF-1α. However, the possibility exists that HSC may respond to other chemokines or chemoattractants that we have not tested in this study. Although HSC express mRNA for CCR3 and CCR9 (lack of antibodies against murine chemokine receptors precluded determination of whether CCR3 or CCR9 proteins were present on the cell surface of HSC), they did not migrate in response to eotaxin or RANTES, ligands for CCR3, or in response to TECK, the ligand for CCR9. These data indicate that receptor mRNA expression alone does not adequately predict chemotactic responsiveness, as has been observed in other studies (24, 33, 34). Knockout mice that do not express functional CCR3 have been prepared and no hematopoietic defects have been observed (Gerard, C., personal communication). Similarly, preliminary analyses of CCR9 knockout mice do not reveal notable defects in the development of most lymphoid compartments (Wurbel, M.A., and Malissen, B., personal communication [35]). Lack of migration by HSC to eotaxin, RANTES, and TECK is also consistent with the absence of essential roles for CCR3 and CCR9 in hematopoiesis.

Current models of leukocyte trafficking hold that chemotactic responsiveness plays a critical role in the homing of cells to particular microenvironments and in positioning cells within these microenvironments (36). Most leukocyte subsets express multiple chemokine and chemoattractant receptors, and migrate in response to several chemokines (37, 38). The possession of multiple chemokine and chemoattractant receptors has been shown to allow cells to maneuver in a stepwise fashion through spatial arrays of chemokine and chemoattractant gradients (39). This raises the question of the significance of the extremely selective responsiveness pattern to chemokines exhibited by HSC.

BM has the challenging task of promoting the growth of several cell lineages. The presumption that various lineages have unique requirements for maturation has led to the notion that BM is partitioned functionally into specialized microenvironments, or “niches.” Although HSC niches have not been characterized, evidence of their existence comes from the observation that cloned stromal lines are heterogeneous in their ability to maintain hematopoiesis in vitro (40–43). A second observation supporting the niche hypothesis is that transplanted HSC, when injected intravenously as discrete boluses, do not engraft BM of syngeneic recipient animals unless the dose of HSC is extremely large (44–48), or the recipient's BM is injured by radiation/cytotoxic drugs.

Given the unique requirement of HSC niches in supporting both the maintenance of multipotency and differentiation, it is tempting to speculate that the restricted chemotactic responsiveness of HSC we observed might be important in localizing HSC within their niches, at least when they are actively participating in hematopoiesis. Possession by HSC of multiple functional chemokine receptors in a complex environment such as BM would be undesirable as it could lead to the inappropriate migration of HSC out of the HSC niche. Ma et al. (49) recently proposed that CXCR4 is required for the retention of B lineage and granulocytic precursors in the fetal liver and BM. The possibility that CXCR4 might similarly promote the retention of HSC in their BM niches remains to be fully tested, but is supported by the observation that the administration to mice of an SDF analogue that downmodulates CXCR4 resulted in a greater than 30-fold increase in the number of circulating Thy-1.1^loLin⁻Sca-1⁺c-Kit⁺ cells (50).

In addition to promoting the maintenance of HSC in specific microanatomical sites within BM, CXCR4 might allow efficient homing back to BM by those HSC that are released into the bloodstream. Using a cross-circulation model, we have found that HSC continuously flux between BM and blood under physiological conditions. HSC released from the BM into the circulation recolonize BM in other locations (5). Therefore, the circulating pool of HSC must have a mechanism for the efficient rehoming to BM. The SDF-1α–CXCR4 interaction appears to play a key role in the homing of transplanted human hematopoietic progenitors to BM of NOD/SCID mice (32), and SDF-1α promotes integrin-mediated arrest on vascular endothelium of circulating CD34⁺ cells (19). The SDF-1α–CXCR4 interaction may have a similar role in the rehoming to BM of murine HSC released physiologically into the circulation or injected intravenously for transplantation.

Aiuti et al. (16) raised the possibility of a direct role for SDF-1α in hematopoietic progenitor mobilization. In a study of human BM and MPB CD34⁺ cells, fewer MPB CD34⁺ cells responded to SDF-1α than CD34⁺ cells isolated from BM, and it was suggested that the reduction in SDF-1α responsiveness might be part of the mechanism of progenitor mobilization. But only a fraction of CD34⁺ cells are HSC (17, 18). We found that mouse HSC from MPB, or from BM or spleens of Cy/G-CSF–treated animals, were equally responsive to SDF-1α as HSC from BM of untreated control animals. These data do not support a role for changes in SDF-1α responsiveness in cytokine-induced HSC mobilization. The disparity between the results of the current study and those of Aiuti et al. (16) might be due to the differences between progenitors and HSC, or it may reflect species- or mobilization protocol–specific effects. In preliminary experiments, the migration of mobilized HSC to the following chemokine panel was also tested: JE, eotaxin, thymus- and activation-regulated chemokine, RANTES, MIP-1β, MIP-3α, MIP-3β, I-309, KC, IL-8, MIG, and SDF-1α. Similar to untreated BM, mobilized HSC migrated only to SDF-1α and did not respond to the other chemokines or to G-CSF (unpublished data).

In conclusion, we report that HSC migrate to SDF-1α but not to chemokines signaling through other known chemokine receptors. To our knowledge, this is the first report of a leukocyte subset that responds to a single chemokine, which makes HSC highly specialized in this regard. It remains to be determined whether SDF-1α–CXCR4 interactions play a role in HSC localization within BM, or in the rehoming to BM by HSC that are released physiologically into the circulation. SDF-1α responsiveness is preserved in HSC isolated from BM, blood, and spleens of Cy/G-CSF–mobilized mice.

The authors thank Libuse Jerabek for superb laboratory management, Veronica Braunstein for antibody production, Lou Hidalgo and Bert Lavarro (deceased) for animal care, and Jim Campbell for very helpful discussions.

This work was supported by National Institutes of Health grant 5R01 HL-58770 to I.L. Weissman, and by NIH grants GM37734 and AI47822 to E.C. Butcher. D.E. Wright was supported by National Institute of Allergy and Infectious Diseases Training grant 5T32 AI-07290. E.P. Bowman was supported by U.S. Public Health Service grant 5T32CA09151 and was a recipient of an Arthritis Foundation Postdoctoral Fellowship. A.J. Wagers was supported by American Cancer Society grant PF-00-017-01-LBC.