Hematopoietic stem cells (HSCs) primarily reside in the bone marrow where signals generated by stromal cells regulate their self-renewal, proliferation, and trafficking. Endosteal osteoblasts 1,2 and perivascular stromal cells including endothelial cells 3 , CXCL12-abundant reticular (CAR) cells 4,5 , leptin-receptor positive stromal cells 6 , and nestin-GFP positive mesenchymal progenitors 7 have all been implicated in HSC maintenance. However, it is unclear if specific hematopoietic progenitor cell (HPC) subsets reside in distinct niches defined by the surrounding stromal cells and the regulatory molecules they produce. CXCL12 (stromal-derived factor-1, SDF-1) regulates both HSCs and lymphoid progenitors and is expressed by all of these stromal cell populations 7–11 . Here, we selectively deleted Cxcl12 from candidate niche stromal cell populations and characterized the effect on HPCs. Deletion of Cxcl12 from mineralizing osteoblasts has no effect on HSCs or lymphoid progenitors. Deletion of Cxcl12 from osterix-expressing stromal cells, which includes CAR cells and osteoblasts, results in constitutive HPC mobilization and a loss of B lymphoid progenitors, but HSC function is normal. Cxcl12 deletion in endothelial cells results in a modest loss of long-term repopulating activity. Strikingly, deletion of Cxcl12 in nestin-negative mesenchymal progenitors using Prx1-Cre is associated with a marked loss of HSCs, long-term repopulating activity, HSC quiescence, and common lymphoid progenitors. These data suggest that osterix-expressing stromal cells comprise a distinct niche that supports B lymphoid progenitors and retains HPC in the bone marrow, while expression of CXCL12 from stromal cells in the perivascular region, including endothelial cells and mesenchymal progenitors, support HSCs. CXCL12 plays a crucial role in maintaining HSC function, including retention in the bone marrow 8,12–14 , quiescence 15,16 , and repopulating activity 16 . To test the hypothesis that CXCL12 production by different stromal cell populations has distinct effects on HSCs and lineage-committed HPC, we generated a floxed allele of Cxcl12 (Cxcl12fl ) to conditionally delete Cxcl12 from candidate niche cells in the bone marrow (Suppl. Fig. 2). Deletion of Cxcl12 in endothelial cells and mature osteoblasts was mediated by the Tie2-Cre recombinase (Cre) and osteocalcin (Oc)-Cre transgenes, respectively. To target Cxcl12 deletion in osteoprogenitors, we used the osterix (Osx)-Cre transgene, which mediates efficient recombination in mature osteoblasts and osteoblast progenitors 17 . It also targets CAR cells, a perivascular stromal cell population implicated in HSC and B lymphoid progenitor maintenance 5,11 . Finally, we used the Prx1-Cre transgene to target multipotent mesenchymal progenitors in the appendicular skeleton. Prx1 is a transcription factor expressed early during limb bud mesoderm development, and Prx1-Cre targets all cells derived from limb bud mesoderm 18 . Lineage mapping studies were performed using a Cxcl12gfp knock-in mouse to define CAR cells 11 . These studies showed that both the Osx- and Prx1-Cre transgenes efficiently targeted recombination in mature osteoblasts, osteocytes, and CAR cells in long bones (Fig. 1a–d and Suppl. Fig 3). Triple transgenic mice were generated containing one floxed Cxcl12 allele, one null allele (Cxcl12fl/ −), and a Cre-recombinase transgene. Total CXCL12 mRNA expression in the femoral bone marrow of Oc- and Tie2-Cre-targeted mice was similar to that observed in control mice (Fig. 1e). In contrast, CXCL12 mRNA expression was reduced by 70% in Osx-Cre-targeted mice and nearly undetectable in Prx1-Cre-targeted mice. A similar decrease in CXCL12 protein levels was observed (Fig. 1f). To confirm Cxcl12 deletion in CAR cells, mice containing Cxcl12fl/gfp and either the Osx- or Prx1-Cre transgenes, were generated (the Cxcl12gfp allele is a null allele). Indeed, CXCL12 mRNA was nearly undetectable in CXCL12-GFPbright CAR cells that were sorted from these mice (Fig. 1g). As expected, CXCL12 mRNA was nearly undetectable in endothelial cells sorted from Tie2-Cre-targeted mice (Fig. 1h). Together these data suggest that, under basal conditions, the majority of CXCL12 is produced by CAR cells, while mature osteoblasts and endothelial cells are only minor contributors. All conditional knockout mice exhibited normal peripheral blood counts and the same relative percentage of granulocytes, monocytes, B cells, and T cells (Suppl. Table 1). However, bone marrow cellularity in femurs was reduced by approximately 50% in both the Osx-Cre- and Prx1-Cre-targeted mice, which was due, in part, to a loss of B cells. HPC subsets in the bone marrow were quantified by flow cytometry (Fig. 2a). The number of c-Kit+ Sca+ Lineage− (KSL) cells, short-term HSCs, multipotent progenitors, and myeloid-committed progenitors was similar in all mice with the exception of a two-fold decrease in common myeloid progenitors in Prx1-Cre-targeted mice (Suppl. Fig. 4). Loss of CXCL12 expression in endothelial cells or mature osteoblasts had no effect on the number of phenotypic HSCs (Fig. 2 b–d). The frequency of phenotypic HSCs in the bone marrow of Osx-Cre-targeted mice was comparable to control mice (data not shown); however, since bone marrow cellularity was reduced, a modest decrease in the absolute number of HSCs was observed. In contrast, a significant decrease in both the frequency and absolute number of phenotypic HSCs in Prx1-Cre-targeted mice was observed, with nearly undetectable dormant HSCs (Flk2− CD34− CD150+ CD48− KSL cells). Consistent with these findings, competitive repopulation assays showed a significant multi-lineage long-term repopulating defect using bone marrow from Prx1-Cre- but not Osx-Cre-targeted mice (Fig. 3a–b). Despite the normal number of phenotypic HSCs, a small, but significant, decrease in long-term repopulating activity also was observed using Tie2-Cre-targeted bone marrow. Serial transplantation of bone marrow from Prx1-Cre- or Tie2-Cre-targeted mice showed no further decrease in repopulating activity in secondary recipients, suggesting that self-renewal capacity may be restored when HSCs are exposed to a normal stromal microenvironment (Suppl. Fig. 5). Quiescence is a fundamental property of HSCs, which is closely related to long-term repopulating activity 19 . Increased cycling of HSCs was observed in Prx1-Cre- but not Osx-Cre- or Tie2-Cre-targeted cells. In contrast, increased cycling of more mature KSL progenitors was observed in both Prx1-Cre- and Osx-Cre-targeted cells (Fig. 3c–d). Collectively, these data show that CXCL12 production from Prx1-Cre-targeted stromal cells and, to a lesser extent, endothelial cells is required for maintenance of HSC repopulating activity and quiescence. Consistent with results from the companion paper by Ding et al., our data suggest that CXCL12 production from mature osteoblasts and osteoblast precursors is dispensable for HSC maintenance. Since CXCL12 has been shown to play an important role in the retention of HPC within the marrow 5,20 , we next quantified HPCs in the blood and spleen. In Osx-Cre-targeted mice, the number of colony-forming cells and KSL cells was increased in the blood and spleen, demonstrating constitutive HPC mobilization (Fig. 3e–f and Suppl. Fig. 6). Interestingly, though CXCL12 expression in the bone marrow is significantly lower, Prx1-Cre-targeted mice displayed a similar magnitude of HPC mobilization. Thus, our data suggest that, although CXCL12 production from Osx-Cre-targeted stromal cells is largely dispensable for HSC maintenance, it is required for the efficient retention of HPCs in the bone marrow. CXCL12 is required for normal B and T cell development 19,21 . Pre-pro-B cells are found in close association with CAR cells 11 , and ablation of CAR cells is associated with a loss of CLPs and pro-B cells 4 . Here, we show that deletion of Cxcl12 in mineralizing osteoblasts or endothelial cells has no effect on CLPs, B lymphoid progenitors (BLPs), or pre-pro-B cells (Fig. 2e–h, Suppl. Figure 7). In contrast, Cxcl12 deletion in CAR cells using Osx-Cre results in a marked loss of pre-pro B cells, and a trend towards a loss of BLP. However, CLPs and earliest thymic progenitors (ETPs) in the thymus are normal. Deletion of Cxcl12 in Prx1-Cre-targeted stromal cells results in a similar phenotype but also results in a marked loss of CLPs. In the companion paper, Ding et al show that deletion of Cxcl12 in osteoblasts using Col2.3-Cre also results in a modest decrease in CLP and lymphoid-primed multipotential progenitors (LMPP). Together, these data suggest that CXCL12 production from CAR cells or osteoblast precursors, but not mineralizing osteoblasts or endothelial cells, is required for the maintenance of B-lymphoid committed progenitors, while CLP maintenance is supported by CXCL12 production from both endosteal osteoblasts and a Prx1-Cre-targeted perivascular stromal cell population. The normal CLP in Osx-Cre-targeted mice may be secondary to compensatory changes related to the severe loss of pre-pro-B cells. We next performed studies to define the stromal cell population(s) differentially targeted by Prx1-Cre and Osx-Cre. We first considered the possibility that Prx1-Cre may target endothelial cells in the bone marrow. However, we detected no tdTomato expression in bone marrow endothelial cells from Prx1-Cre reporter mice (Fig. 4a–b). Moreover, expression of CXCL12 mRNA from sorted CD31+ endothelial cells from Prx1-Cre-targeted mice was comparable to control mice (Suppl. Fig. 8). Thus, loss of CXCL12 from bone marrow endothelial cells does not account for the loss of HSCs in Prx1-Cre-targeted mice. We extended the lineage mapping studies to the CD45− lineage− PDGFRα+ Sca+ (PαS) cell population, which is enriched for mesenchymal stem cells 22 . Whereas Osx-Cre did not target this cell population, approximately 50% of cells were targeted by Prx1-Cre, including a subpopulation that expressed intermediate levels of CXCL12 (Fig. 4c–e). To evaluate the mesenchymal progenitor activity of the Prx1-Cre-targeted cells, we sorted Prx1-Cre-targeted (tdTomato+) and non-targeted PαS cells and assessed their clonogenic capacity. All of the colony-forming unit-fibroblast (CFU-F) activity was contained with the Prx1-targeted PαS cell population, with greater than 10% of these cells having CFU-F activity (Fig 4f–g). This compares to a CFU-F frequency of approximately 4% in unselected PαS cells 22 and less than 1% in nestin-GFP+ stromal cells 7 . The Prx1-targeted PαS cells have osteogenic and adipogenic differentiation potential in vitro, consistent with a mesenchymal stem cell phenotype (Fig 4h–i). RNA expression profiling of Prx1-targeted PαS cells is notable for the lack of nestin 7 , CD146 9 , or leptin receptor 6 , all of which have been used to mark stromal cells contributing to HSC maintenance (Suppl. Table 2). Interestingly, other than moderate CXCL12 expression, these cells do not express genes classically associated with HSC maintenance, including kit ligand 6 and angiopoietin-1 19 , though high expression of several matrix proteins (e.g., proteoglycan 4 23 and osteonectin 24 ) implicated in HPC regulation is present. Collectively, these data suggest that distinct stromal cell niches in the bone marrow exist that regulate specific HPC populations (Suppl. Fig. 1). Osterix-expressing stromal cells comprise a niche that supports B lymphoid progenitors and retains HPC in the bone marrow, while CXCL12 production from nestin− leptin receptor− mesenchymal progenitors is required for HSC and CLP maintenance. METHODS SUMMARY Targeting of the Cxcl12 locus was accomplished by conventional techniques. Conditional knockouts were accomplished by interbreeding with Osx-Cre 25 , Prx1-Cre 18 , Oc-Cre mice 26 , Tie2-Cre 27 , and Cxcl12 +/− mice 21 . Lineage mapping was accomplished using Ai9 28 and Cxcl12gfp mice 12 . All mice with the exception of Cxcl12gfp mice were maintained on the C56Bl6/J background. Unless indicated otherwise, data are presented as mean ± SEM and were analyzed with the Student's t-test, one-way ANOVA, or two-way ANOVA. METHODS Mice With the exception of Cxcl12gfp mice, all transgenic strains had been backcrossed at least 10 generations onto a C57BL/6 background. Osx-Cre 1 , Prx1-Cre 2 , Tie2-Cre 3 , and Ai9 (B6.Cg-Gt(ROSA)26Sortm9(CAG-tdTomato)Hze/J) 4 mice were obtained from The Jackson Laboratory. EIIa-Cre mice 5 were a gift from Monica Bessler (University of Pennsylvania, Pennsylvania). Oc-Cre mice 6 were a gift from Thomas Clemens (Johns Hopkins University, Maryland), and Cxcl12gfp mice 7 were a gift from Takashi Nagasawa (Kyoto University, Japan). Cxcl12+/− mice 8 were obtained through the RIKEN BioResource Center (Ibaraki, Japan). Mice were maintained under standard pathogen free conditions according to methods approved by the Washington University Animal Studies Committee. Generation of Cxcl12fl mice A floxed allele of Cxcl12 was generated containing LoxP sites flanking exon 2 of Cxcl12; a third LoxP was inserted 3’ of the neomycin selection cassette (Suppl. Fig. 2). Generation of targeted embryonic stem cells and blastocyst injections were performed as previously described 9 . Excision of the neomycin cassette was accomplished through partial recombination by intercrossing mice with mice expressing EIIa-Cre. Mice were genotyped using PCR primers: Cxcl12flox forward, 5’-CTACACCTCCTCTAGGTAAACCAGTCAGCC-3’; Cxcl12flox reverse 5’-GGACACCAGAACCTTGAAACTGACA-3’. Bone marrow transplantation Bone marrow from WT Ly5.1/5.2-expressing mice was mixed at a 1:1 ratio with marrow from experimental or control mice expressing the Ly5.2 locus. A total of 2 × 106 cells injected retro-orbitally into lethally irradiated (1,000 cGy) WT Ly5.1-expressing mice. Since Prx1 is expressed primarily in limb-bud derived long bones, only tibias and femurs were used for transplant and other analyses. Blood, bone marrow, spleen, and thymus analysis Blood, bone marrow, and spleen cells and thymocytes were harvested using standard techniques and quantified using a Hemavet automated cell counter (CDC Technologies). Flow cytometry Cells were stained by standard protocols with the following antibodies (eBiosciences unless otherwise noted): Lineage analysis and chimerism was assessed using peridinin chlorophyll protein complex (PerCP)-Cy5.5–conjugated Ly5.1 (A20, CD45.1), allophycocyanin (APC)-conjugated Ly5.2 (104, CD45.2), and one or more of the following lineage markers: APC-conjugated CD115 (AFS98, monocytes), fluorescein isothiocyanate (FITC)-conjugated Ly6C/G (RB6-8C5, Gr-1, myeloid), phycoerythrin (PE)-conjugated CD3e (145-2C11, T lymphocytes), and APC-eFluor780–conjugated CD45R (RA3-6B2, B220, B lymphocytes). For HSPC analysis, cells were stained with a cocktail of biotin-conjugated B220, TER-119, CD3e, Gr1, and CD41 (MWReg30), PE-conjugated CD150 (TC15-12F12.2, Biolegend), PE-Cy7-conjugated CD48 (HM48-1, BD Biosciences), PerCP-Cy5.5-conjugated Sca1 (D7), APC eFluor780-conjugated c-kit (2B-8), FITC-conjugated CD34 (RAM34), APC-conjugated Flk2 (A2F10), eFluor450-conjugated CD16/32 (93), and eFluor605NC-conjugated streptavidin. For HSC cell cycle staining, cells were stained with the biotin-conjugated lineage panel, PE-conjugated CD150, PE-Cy7-conjugated CD48, PerCP-Cy5.5-conjugated Sca, APC-conjugated c-kit, and APC-eFluor780-conjugated streptavidin. Cells were then fixed using the Cytofix/Cytoperm kit (BD Biosciences), stained with FITC-conjugated Ki-67 (B56, BD Biosciences), and resuspended in 1 mg/mL of 4',6-diamidino-2-phenylindole (DAPI). Doublets were gated out using FSC vs. FSC-W. Data were collected on a Gallios 10-color, 3-laser flow cytometer (Beckman Coulter). Data were analyzed with FlowJo (Treestar). For CLP/BLP analysis, bone marrow cells were stained with a cocktail of PE-Cy7-conjugated B220, TER-119, CD3e, and Gr-1, APC-conjugated CD27 (LG.7F9), biotinylated IL-7Ra (gift of Deepta Bhattacharya, Washington University), PE-conjugated Flk2, FITC-conjugated Ly6D (49-H4, BD Biosciences), and eFluor450-conjugated streptavidin. CLP were defined as B220- TER-119- CD3e- Gr-1- CD27+ IL-7Ra+ Flk2+ Ly6D- cells, and BLP were defined as B220- TER-119- CD3e- Gr-1- CD27+ IL-7Ra+ Flk2+ Ly6D+ cells. For Pre-pro B cell analysis, bone marrow cells were stained with APC-eFluor780-conjugated B220, a cocktail of PerCP-Cy5.5-conjugated CD3e, CD11c (N418), and NK1.1 (PK136), APC-conjugated IgM (II/4), eFluor450-conjugated IgD (11–26c), PE-Cy7-conjugated CD19 (eBio1D3), PE-conjugated CD43 (S7, BD Biosciences), and FITC-conjugated Ly6D (BD Biosciences). Pre-Pro B cells were defined as B220+ CD3e− CD11c− NK1.1− IgM− IgD− CD19− CD43+ Ly6D+ cells. For ETP analysis, thymocytes were stained using a cocktail of FITC-conjugated CD4 (RM4-5), CD8 (53-6.7), and CD11b (M1/70), a cocktail of PE-Cy7-conjugated B220, TER-119, CD3e, and Gr-1, PE-conjugated CD44 (IM7), eFluor450-conjugated CD25 (PC61.5), and APC-eFluor780-conjugated c-kit. ETP were defined as CD4− CD8− CD11b− B220− TER-119− CD3e− Gr−1− CD44+ CD25− c-kit+ thymocytes. Stromal cell analysis and sorting To extract bone marrow stromal cells, intact bones were crushed in PBS by mortar and pestle. Crushed fractions in PBS were collected and stored on ice. The bone chips were digested using subjected to enzymatic digestion by collagenase type II (3mg/mL, Worthington Biochemical) and dispase (4mg/mL, Roche) at 37°C for 45 minutes at 37°C in a shaking water bath. Both crushed and digested fractions were pooled. Following RBC lysis, endothelial cells were stained with APC-conjugated CD45 (30-F11), FITC-conjugated lineage cocktail (CD3e, Gr-1, B220, and TER-119), and biotinylated anti-mouse CD31 (PECAM-1) followed by streptavidin PE. Dead cells were excluded using 7-AAD (EMD Biosciences). Perivascular mesenchymal progenitor cells were stained with APC-eFluor780-conjugated CD45 and APC-eFluor780-conjugated lineage cocktail, APC-conjugated Sca-1, biotinylated CD140a (PDGRFα), and streptavidin Brilliant Violet 421 (Biolegend). FACS analyses were performed using FACScan (BD Biosciences), LSRII (BD Biosciences), or Gallios (Beckman Coulter) flow cytometers. Cell sorting was done on Reflection (iCyt) or Aria (BD Biosciences) flow cytometers. RNA of sorted cells was extracted using NucleoSpin RNA XS kit (Macherey-Nagel) per manufacture recommendations. Quantitative RT-PCR For total bone marrow RNA, femurs were flushed with 1 mL of Trizol (Invitrogen). RNA was prepared according to manufacturer’s specification. One-step qRT-PCR was performed using the TaqMan Universal PCR Master Mix (Applied Biosystems) using no template and no RT controls. Data was collected on a 7300 Real-Time PCR System (Applied Biosystems). Primers were: CXCL12 forward, 5’-GAGCCAACGTCAAGCATCTG-3’; CXCL12 reverse, 5’-CGGGTCAATGCACACTTGTC-3’; CXCL12 dT-FAM/TAMRA probe, 5’-TCCAAACTGTGCCCTTCAGATTGTTGC-3’; β-actin forward, 5’-ACCAACTGGGACGATATGGAGAAGA-3’; β-actin primer; and β-actin dT-VIC/TAMRA probe, 5′-AGCCATGTACGTAGCCATCCAGGCTG-3′. Immunofluorescence Mice were lethally sedated and perfused with PBS followed by 4% paraformaldehyde. Hind limbs were removed and post-fixed in 4% paraformaldehyde overnight at 4°C. Bones were washed in water, decalcified in 14% EDTA pH 7.4, and cryoprotected in 30% sucrose in PBS. Bones were then snap frozen in OCT using liquid nitrogen-cooled 2-methylbutane, and blocks were sectioned at 7 µM. For lineage mapping, sections were washed in PBS and mounted with Prolong Gold Antifade Reagent with DAPI (Invitrogen). Slides were imaged using an ApoTome fluorescent microscope (Zeiss). Colony-forming unit cell (CFU-C) assay 25,000 bone marrow cells or 50,000 spleen cells were plated in 2.75 ml methylcellulose media (MethoCult 3434; Stemcell Technologies). 20 µL of whole peripheral blood was RBC lysed and plated in methylcellulose. Duplicate cultures were incubated at 37°C for 7 days, after which colonies containing at least 100 cells were counted in a blinded fashion. Colony-forming unit-fibroblast (CFU-F) assay PDGFRα+ Sca+ cells were sorted by flow cytometry and directly plated on tissue culture plates containing alpha MEM and 10% fetal bovine serum (Atlanta Biologicals). Media exchanges were performed every 3–4 days for a total of 14 days, after which colonies containing at least 50 cells were counted. On day 14, cells were harvested from the cultures and replated in osteogenic or adipogenic media and cultured for an additional 14 days. Osteogenic media: alpha MEM with 10% fetal bovine serum, 50µM ascorbic acid (Sigma) and 10µM of β-glycerophosphate (Sigma). Adipogenic media: alpha MEM with 10% fetal bovine serum, 5µg/mL insulin, 100µM indomethacin, and 100nM dexamethasone. Osteoblast differentiation was assessed using the Leukocyte Alkaline Phosphatase Kit (Sigma), per manufacture’s recommendations. Adipocyte differentiation was assessed by staining with Oil Red O, as reported previously 10 . RNA expression profiling PDGFRα+ Sca+ cells or CAR cells, pooled from 2–6 mice, were sorted directly into lysis buffer and RNA was prepared using the RNA XS column kit (Macherey-Nagel, Bethlehem, PA) according to the manufacturer’s directions. RNA was amplified using the NuGen Ovation system (NuGen, San Carlos, CA), and hybridized to the Affymetrix MoGene 1.0 ST array. Data normalization was performed using the Robust Multichip Average (RMA) algorithm. Submission of this RNA expression data to Gene Expression Omnibus is in progress. Statistics Significance was determined using Prism software (GraphPad). Unless otherwise stated, statistical significance of differences was calculated using 1- or 2-way ANOVA. P-values indicate the result of Bonferroni post-testing relative to Cxcl12f l/− control mice unless other comparisons are explicitly shown. P-values less than 0.05 were considered significant. All data are presented as mean ± SEM. Supplementary Material 1 2
