Osteogenic Factor Runx2 Marks a Subset of Leptin Receptor-Positive Cells that Sit Atop the Bone Marrow Stromal Cell Hierarchy

Bone marrow mesenchymal stem and progenitor cells (BM-MSPCs) maintain homeostasis of bone tissue by providing osteoblasts. Although several markers have been identified for labeling of MSPCs, these labeled cells still contain non-BM-MSPC populations. Studies have suggested that MSPCs are observed as leptin receptor (LepR)-positive cells, whereas osteoblasts can be classified as positive for Runx2, a master regulator for osteoblastogenesis. Here, we demonstrate, using Runx2-GFP reporter mice, that the LepR-labeled population contains Runx2-GFPlow sub-population, which possesses higher fibroblastic colony-forming units (CFUs) and mesensphere capacity, criteria for assessing stem cell activity, than the Runx2-GFP− population. In response to parathyroid hormone (PTH), a bone anabolic hormone, LepR+Runx2-GFPlow cells increase Runx2 expression and form multilayered structures near the bone surface. Subsequently, the multilayered cells express Osterix and Type I collagen α, resulting in generation of mature osteoblasts. Therefore, our results indicate that Runx2 is weakly expressed in the LepR+ population without osteoblastic commitment, and the LepR+Runx2-GFPlow stromal cells sit atop the BM stromal hierarchy.

potential to form CFU-F colonies or clonal mesenspheres 6,7,11 . These results suggest that the fractions are impure and still contain non-BM-MSPC populations.
Runt-related transcription factor 2 (Runx2) is a master regulator for osteoblast differentiation [12][13][14] . Osteoblastogenesis is fully suppressed by the global knockout of Runx2 13,14 . Exon 8 of Runx2 gene conditional deletion in mature osteoblasts, which express Cre recombinase under the control of a 2.3-kb fragment of the type I collagen α[(Col1 (2.3)] promoter, exhibit low bone mass phenotype 15 . In contrast, conditional knockout mice lacking exon 4 of Runx2 gene in mature osteoblasts have no effect on osteoblastic activity 16 . These studies indicate that the necessity of Runx2 in osteoblastic activity is still controversial. On the other hand, in vivo lineage tracing studies have demonstrated that Runx2 is essential for osteoblast lineage commitment 17 . Interestingly, Runx2 overexpression approaches revealed that the late stage of osteoblastogenesis is negatively regulated by Runx2, whose levels were found to decrease with osteoblast maturation 18,19 . Overall, these findings suggest that Runx2 is required for osteoblast commitment from immature mesenchymal stromal cells. These results raise the intriguing possibility that Runx2 may be expressed in a portion of LepR + stromal cells, which have osteogenic-committed sub-populations.
Osteoblastogenesis is completely diminished in knockout mice lacking Osterix (Osx), a transcription factor that acts downstream of Runx2 20 . Furthermore, bone formation is inhibited by conditionally deleting Osx in mature osteoblasts 21 . These results suggest that Osx is necessary not only for osteoblast differentiation, but also for their functions. On the other hand, during endochondral bone ossification, BM-MSPCs are generated from part of the developing chondrogenic cell populations 17 . The expression levels of Osx are increased throughout the development of chondrogenic cell populations that subsequently differentiate into BM-MSPCs 5,17,22 . Although Osx protein expression in BM-MSPCs is completely lost in the adult stage, mRNA expression is maintained 5,23 . However, the Osx expression pattern during osteoblastogenesis from BM-MSPCs has yet to be elucidated.
Teriparatide, a biologically active amino acid 1-34 fragment of human PTH [hPTH ], is clinically used in treatment of osteoporosis patients 24 . Several studies have demonstrated that intermittent PTH treatment induces remedial action against osteoporosis due to anabolic effects on bone tissue [25][26][27][28] . Researchers have found that osteoblast precursors are increased along the bone surfaces in response to PTH treatment [27][28][29][30] . These results suggest that the anabolic effects of PTH on bone tissue are exerted by the acceleration of osteoblastogenesis from immature BM mesenchymal precursors. However, it still remains unclear which BM stromal cells give rise to osteoblasts in response to PTH treatments, thereby mediating the therapeutic response in osteoporosis.
Here we demonstrate, using Runx2-GFP reporter mice, that the LepR + cell population contains Runx2-GFP low cells, and unexpectedly, that stem cell capacity is enriched in the Runx2-GFP low sub-population. In addition, our studies have shown that the LepR + Runx2-GFP low cells differentiate into mature osteoblasts via multilayered cell formation adjacent to bone surfaces in response to PTH-induced bone anabolic effects. These results provide evidence that LepR + Runx2-GFP low cells sit atop the BM mesenchymal stromal cell hierarchy.

Results
Runx2 is heterogeneously expressed in the LepR + BM stromal cell population. To detect Runx2 expressing cells in bone tissue, we analyzed Runx2-GFP reporter mice, in which GFP is driven by a bacterial artificial chromosome (BAC) of Runx2 locus 31 . FITC-conjugated anti-GFP antibody was used to amplify the intensity of the GFP signal when imaging bone tissue sections. Consistent with a previous study 31 , Runx2-GFP + expression in bone tissues was detected in osteoblasts, osteocytes and chondrocytes ( Fig. 1A-D). Interestingly, the Runx2-GFP signal was observed not only in bone tissues, but also in the BM cavity (Fig. 1A, right panel). Some hematopoietic cells were observed as GFP-positive cells due to nonspecific binding of the anti-GFP antibody (compare wild-type and Runx2-GFP mice in Fig. 1E). On the other hand, VE-cadherin (VE-Cad) and CD31positive endothelial cells, and Perilipin-positive adipocytes did not express Runx2-GFP (Fig. 1F,G). It is interesting to note that leptin receptor-positive (LepR + ) BM stromal cells, which are considered to have characteristics of BM-MSPCs, also weakly expressed Runx2-GFP throughout the bone marrow cavity (Fig. 1H). These results indicate that the osteoblastic master regulator Runx2 may already be expressed in the pre-or early osteoblastic lineage-committed LepR + sub-population.
To further analyze Runx2-GFP expression pattern in the BM stromal population, we performed flow cytometric analyses of BM cells in the Runx2-GFP mice. Runx2-GFP + cells were observed in the BM stromal cell population without enhancing the GFP signal by FITC-conjugated anti-GFP antibody ( Fig. 2A). We found two distinct types of Runx2-GFP + cells on the basis of their GFP expression levels and cellular morphology (designated as Runx2-GFP low and -GFP high cells) ( Fig. 2A). Most of the Runx2-GFP low cells (82.0 ± 1.4%), but not Runx2-GFP high cells (3.3 ± 1.8%), were positive for LepR ( Fig. 2A). Previous reports suggested that mature osteoblasts are negative for LepR 5, 6 . These results indicate that the mature osteoblasts are contained in the Runx2-GFP high population. Next, we analyzed the Runx2-GFP expression pattern in LepR + cells. Interestingly, most of the LepR + cells were positive for Runx2-GFP (64.6 ± 2.0%) (designated as LepR + Runx2-GFP low cells) (Fig. 2B, right panel). The frequency and absolute number of LepR + Runx2-GFP low cells per femur were approximately 3-times higher than those of LepR + Runx2-GFP − cells (Fig. 2C,D). These results indicate that LepR + cells consist of two populations: LepR + Runx2-GFP low and LepR + Runx2-GFP − sub-populations.
Stem cell activity is enriched in the Runx2-GFP low sub-population of LepR + cells. Upregulation of Runx2 expression is thought to indicate osteoblastic commitment of multipotent BM stromal progenitors 32 . We next generated LepR-Cre/ROSA26-loxP-stop-loxP-tdTomato/Runx2-GFP (LepR/Tomato/Runx2-GFP) mice, and analyzed whether stem cell activity in LepR + cells increases in inverse proportion to Runx2 expression levels. Previous reports demonstrated that the differentiation of LepR + cells into mature osteoblasts is mainly observed in adult mice, and only rarely seen in young mice 5,6 . In order to avoid contamination of the LepR-Cre-labeled population with LepR + cell-derived osteoblasts, we analyzed young mice (5-7 weeks old) in this experiment. Histological analysis demonstrated that bone-lining osteoblasts were strongly positive for Runx2-GFP, but negative for LepR/Tomato (Fig. 3A, asterisks). These results indicate that LepR-Cre-labeled Tomato + (LepR/ Tomato + ) cells do not contain the osteoblastic population in young mice. On the other hand, the LepR/Tomato + cells in the BM cavity were positive for Runx2-GFP (Fig. 3A, arrows). Immunofluorescence staining showed that LepR/Tomato + cells express Runx2 protein (Fig. 3B, arrows). Flow cytometric analysis also demonstrated that the majority of LepR/Tomato + cells in BM expressed Runx2-GFP (60.9 ± 2.5%) (Fig. 3C). We then assessed the stem cell activity of both Runx2-GFP low and Runx2-GFP − sub-populations in LepR/Tomato + cells by performing CFU-F assays of cell sorted BM stromal fractions (Fig. 3D). Consistent with previous reports 6,7 , there were no CFU-F capable cells in the LepR/Tomato − BM stromal population (Fig. 3D). Contrary to our expectations, the CFU-F capacity of the LepR/Tomato + /Runx2-GFP low BM stromal population was high (Fig. 3D). CFU-F colonies derived from the LepR/Tomato + /Runx2-GFP low BM stromal population were positive for LepR/Tomato and weakly Runx2-GFP-positive (Fig. 3E,F). Furthermore, when BM stromal fractions were plated at clonal densities under nonadherent culture conditions, the LepR/Tomato + /Runx2-GFP low population formed spheres (mesenspheres) at greater frequency than other fractions (Fig. 3G,H). The LepR/Tomato + /Runx2-GFP low BM stromal cells exhibited tri-lineage differentiation potential ( Fig. 3I-K). However, it is not clear whether the LepR/Tomato + / Runx2-GFP low BM stromal cells generate other types of BM stromal cells. Real-time PCR analyses revealed that the LepR/Tomato − sub-population did not express the MSPC markers PDGFRa 7, 33 LepR 5, 6 or CXCL12 23 ( Fig. 3L-N). Interestingly, the expression levels of all three of these MSPC markers in the LepR-Cre/Tomato + Runx2-GFP low sub-population were significantly higher than in the LepR-Cre/Tomato + Runx2-GFP − sub-population ( Fig. 3L-N). These results also indicated that the LepR-Cre/Tomato + Runx2-GFP low sub-population overlaps with CXCL12 abundant reticular (CAR) cells 23 , which are generated from part of the developing chondrogenic cell populations 17 . The expression levels of Runx2 mRNA in the Runx2-GFP low sub-population were significantly higher than those in the Runx2-GFP − sub-population from LepR-Cre/Tomato + stromal cells (Fig. 3O). The Runx2 expression was hardly detected in the LepR-Cre/Tomato − stromal cell population at the mRNA level (Fig. 3O). Consistently, the LepR-Cre/Tomato − stromal population contained almost no Runx2-GFP + cells (0.7 ± 0.2%) (Suppl. Fig. 1). Taken together, these results indicate that stromal stem cell activity in BM is high in LepR + Runx2-GFP low stromal cell populations. Therefore, our findings provide evidence that LepR + Runx2-GFP low cells sit atop the BM stromal cell hierarchy, and the osteoblastic master transcription factor Runx2 is weakly expressed in BM-MSPC populations without osteoblastic lineage commitment. LepR + Runx2-GFP low cells differentiate into osteoblasts through multilayered cell formation in response to PTH anabolic effects. In vivo genetic lineage tracing analysis demonstrated that LepR + cells differentiate into osteoblasts 5, 6 . As we found that stem cell activity is enriched in LepR + Runx2-GFP low BM stromal cell populations, we next examined whether LepR + Runx2-GFP low cells differentiate into mature osteoblasts in vivo by lineage tracing. Because intermittent treatment of parathyroid hormone (PTH) (1-34) increased bone volume by inducing bone formation (Suppl. Fig. 2) [24][25][26]28 , we injected PTH into LepR/Tomato/Runx2-GFP mice. Bone-lining osteoblasts were detected as Runx2-GFP single-positive cells in the control group (Fig. 4A, asterisks). In contrast, LepR/Tomato and Runx2-GFP double-positive mature osteoblasts were significantly increased on the endosteal surface in the PTH-treated bone tissue (Fig. 4B, asterisks and Suppl. Fig. 3A). These results suggest that osteoblastogenesis from LepR + cells is accelerated by PTH treatment. It is noteworthy that  while LepR + Runx2-GFP low cells were observed in the control BM cavity (Fig. 4A and Suppl. Fig. 4A, arrows), LepR + Runx2-GFP + multilayered cells (designated as ML-cells) were observed in the vicinity of bone tissue by PTH treatment (Fig. 4B and Suppl. Fig. 4B, arrows). Quantification of vertical-and cross-section images revealed that ML-cells were significantly increased by PTH treatment (Suppl. Figs 3B and 4C-E). The expression level of Runx2-GFP in ML-cells was higher than in LepR + Runx2-GFP low cells (Fig. 4A,B, right panels, arrows). In contrast, the expression level of Runx2-GFP in ML-cells was lower than in cuboidal-shaped osteoblasts (Fig. 4B, right  panel, arrows and asterisks). These results suggest that the LepR + Runx2-GFP low cells differentiate into mature osteoblasts through ML-cell formation with increasing levels of Runx2 expression. LepR + Runx2-GFP low cell-derived multilayered cells differentiate into mature osteoblasts with increasing expression of Osterix and type I collagen α. We next analyzed the expression pattern of Osterix (Osx), a transcription factor downstream of Runx2, in ML-cells using Osx-Cre ERT2 (iOsx)/Tomato/ Runx2-GFP mice, administering tamoxifen for labeling of Osx + cells 34 . Histological analyses demonstrated that mature osteoblasts express both iOsx/Tomato and Runx2-GFP, but Runx2-GFP low cells located away from bone surfaces were negative for iOsx/Tomato (Fig. 4C, asterisks and arrowheads). ML-cells located far from osteoblasts were iOsx/Tomato negative, but those near bone-lining osteoblasts were positive for Tomato (Fig. 4D, arrowheads and arrows). Both the iOsx/Tomato-negative and -positive ML-cells (Runx2-GFP + cells) were significantly increased by the PTH treatment (Suppl. Fig. 3C and D).
Lastly, we examined the hierarchical relationship between Osx and Type-1 collagen α (Col1), a marker for mature osteoblasts, in osteoblastogenesis from ML-cells using iOsx/Tomato/Col1(2.3)-GFP mice. Tamoxifen-induced iOsx/Tomato + cells were observed as mature osteoblasts with Col1(2.3)-GFP expression in control bone tissue (Fig. 4E, asterisks). However, iOsx/Tomato + cells were observed as ML-cells after PTH-treatment (Fig. 4F). Only the population of these cells localized in the vicinity of the bone surface overlapped with Col1(2.3)-GFP expression (Fig. 4F, compare arrowheads and arrows). Both the Col1(2.3)-GFP-negative and -positive ML-cells (iOsx/Tomato + cells) were significantly increased by the PTH treatment (Suppl. Fig. 3E and F). These results thus suggest that LepR + Runx2-GFP low cells differentiate into ML-cells adjacent to the bone surface, and that PTH treatment enhances Runx2 expression, which subsequently induces Osx expression, resulting in differentiation into Col1 + mature osteoblasts (Fig. 5).

Discussion
To clarify in vivo osteoblastogenesis from LepR + cells, we analyzed the expression pattern of the early osteoblastic transcription factor, Runx2, in LepR + cells using Runx2-GFP transgenic mice 31 . Here we report that BM-MSPCs in adult BM are confined to the weak Runx2-GFP-expressing LepR + stromal cell population, which differentiates into Col1 + mature osteoblasts in response to PTH anabolic effects. In this process, the LepR + Runx2-GFP low cells form multilayered structures along the bone surface, subsequently increasing expression levels of Runx2 and Osx.
Our results demonstrating that the stem cell capacity is enriched in Runx2-GFP low populations are consistent with recently published data in which the most primitive stromal population in the calvaria, gated as Prx1 + Scal + cells, expresses Runx2 at low levels 35 . On the other hand, a single-cell assay demonstrated that CXCL12 abundant reticular (CAR) cells, which largely overlap with the LepR + cell population, express not only Runx2 and Osx, but also peroxisome proliferator-activated receptor γ (PPARγ), an essential transcription factor for adipogenesis 36 , at the mRNA level 23 . Interestingly, osteoblastogenesis is enhanced by decreased transcriptional activity and haploinsufficiency of PPARγ in BM stromal progenitors 37,38 . In contrast, adipogenesis is accelerated due to the stromal deletion of Wnt/β-catenin-signaling, an essential signaling pathway for osteoblastogenesis 39 . These reports suggest that the undifferentiated state of LepR + Runx2-GFP low cells is sustained due to reciprocal inhibition between osteogenic and adipogenic factors. Several studies have reported that the lineage differentiation of BM-MSPCs is skewed toward osteoblasts by intercellular expression of vascular endothelial growth factor A (VEGF-A) 40 . In contrast, the adipocyte lineage commitment from BM-MSPCs is increased by up-regulation of MicroRNA-188 41 , deletion of transcription factor Foxc1 9 and peripheral Leptin/Leptin receptor signaling 10 . Further studies will clarify the mechanistic details of the cell fate decision of BM-MSPCs. Our data demonstrate that CFU-F and mesensphere forming capacities are rarely observed in the Runx2-GFP − stromal sub-population of LepR + cells. These results indicate that the LepR + Runx2-GFP − stromal sub-population contains some kind of committed cells other than stem cells. Previous studies provide in vivo evidence demonstrating that LepR + cells differentiate into not only osteoblasts but also adipocytes with aging and during tissue regeneration processes after injury 5,6 . In the process of adipogenesis, the expression levels of Runx2 decrease, as opposed to the increase of PPARγ 40 . Furthermore, our immunofluorescence data demonstrated that adipocytes are negative for Runx2 (Fig. 1G). These data suggest that the LepR + Runx2-GFP − stromal sub-population may be adipocyte-committed precursors. Further analysis of this fraction will provide information about the process of adipocytogenesis from BM-MSPCs in vivo.
In this study, we demonstrated that the PTH treatment accelerated osteoblastogenesis from LepR + Runx2-GFP low cells. It has been shown that mature osteoblasts are continuously replaced by immature precursors in adult bone tissues 42 . Because the LepR + cells differentiate into osteoblasts in adult bone tissues 5, 6 , the LepR + Runx2-GFP low cells contribute to bone remodeling and maintain bone homeostasis in the adult phase. On the other hand, it has been reported that PTH anabolic effects are exerted by activation of quiescent bone lining osteoblasts 43 , suggesting that the PTH has multiple targets for bone anabolism. Our results also demonstrate that ML-cells appear along the bone surface by PTH treatment. Consistent with our results, it is reported that clustered Osx-positive cells are observed as pre-osteoblasts in the vicinity of trabecular bone surfaces from PTH-treated rats 27 . Others, by employing proliferating cell labeling experiments, have also reported that the PTH-induced thick layered cells proliferate with expression of osteoblastic markers such as alkaline phosphatase (ALP), Runx2, osteocalcin and osteonectin 29,30 . These results indicate that the PTH-induced osteoblastic differentiation is associated with cell cycle progression. Previous studies reported that the LepR + cells are quiescent in adult BM 5,6 . However, the LepR + cells markedly proliferate in response to self-depletion, and lineage differentiation into both osteoblasts and adipocytes is accelerated concomitantly in this situation 6 . In contrast, depletion of the transcription factors Snail and Slug in the skeletal stem cells decreases not only proliferative activity but also lineage commitment potential, coincidently 44 . These results suggest that cell cycle quiescence may be critical for maintaining the undifferentiated state of BM-MSPCs. Further analysis of the mechanistic relationship between cell cycle regulation and lineage commitment of LepR + Runx2-GFP low cells will provide a potential therapeutic target for osteoporotic patients. PTH treatment and induction of Cre-mediated recombination. Human PTH (1-34) was kindly provided from Asahi Kasei Pharma Co. Ltd. Four-six-week-old mice were intraperitoneally injected with PTH (80 μg/kg/12 hours) for 10 days. Forty-eight hours after the final PTH injection, mice were sacrificed and used for analyses. For induction of Cre-mediated recombination in Osx-Cre ERT2 mice, CRF-1 chow diet (Oriental Yeast) containing tamoxifen (Sigma-Aldrich) at 400 mg/kg was given from 5 days before the first PTH injection until the end of the experiment.
Cell sorting and flow cytometry. Cell sorting experiments were performed using an Aria III Cell Sorter (BD Biosciences). Flow cytometric analyses were carried out using a Cytomics FC 500 flow cytometer equipped with CXP software (all Beckman Coulter Life Sciences). Dead cells and debris were excluded by FSC, SSC, DAPI (Dojindo) and Fixable Viability Dye eFluor 780 (eBioscience) staining profiles. Data were analysed with FlowJo (Tree Star) software.
In vitro cell differentiation. Sorted LepR-Cre/Tomato + Runx2-GFP low stromal cells were expanded using a MesenCult proliferation Kit with MesenPure (StemCell Technologies) containing 100 U/ml and 100 μg/ml penicillin-streptomycin. Osteogenic, adipogenic and chondrogenic differentiation were induced using a Mouse Mesenchymal Stem Cell Function Identification Kit (R&D Systems). Cells were maintained with 5% CO 2 in a water-jacketed incubator at 37 °C for 2-5 weeks. Mineralized osteogenic cells were identified by Alizarin Red S (Sigma-Aldrich) staining. Adipocytes were identified by characteristic production of lipid droplets and staining with an anti-FABP antibody (R&D Systems). Chondrocytic cells were identified using an Alcian Blue 8GX solution (Sigma-Aldrich).

Statistics.
The results were expressed as mean ± SD. Data were evaluated by unpaired Student′s t-tests.
Experiments were performed three times and similar results were obtained. Statistical analyses were performed with Graph Pad Prism 6. P < 0.05 was considered significant.