p53 Loss Increases the Osteogenic Differentiation of BMSCs

Abstract

The tumor suppressor, p53, plays a critical role in suppressing osteosarcoma. Bone marrow stromal cells (BMSCs, also known as bone marrow-derived mesenchymal stem cells) have been suggested to give rise to osteosarcomas. However, the role of p53 in BMSCs has not been extensively explored. Here, we report that p53 regulates the lineage choice of mouse BMSCs (mBMSCs). Compared to mBMSCs with wild type p53, mBMSCs deficient in p53 have enhanced osteogenic differentiation, but with similar adipogenic and chondrogenic differentiation. The role of p53 in inhibiting osteogenic lineage differentiation is mainly through the action of Runx2, a master transcription factor required for the osteogenic differentiation of mBMSCs. We find that p53 indirectly represses the expression of Runx2 by activating the microRNA-34 family, which suppresses the translation of Runx2. Since osteosarcoma may derive from BMSCs, we examined whether p53 has a role in the osteogenic differentiation of osteosarcoma cells and found that osteosarcoma cells with p53 deletion have higher levels of Runx2 and faster osteogenic differentiation than those with wild type p53. A systems biology approach reveals that p53-deficient mBMSCs are more closely related to human osteosarcoma while mBMSCs with wild type p53 are similar to normal human BMSCs. In summary, our results indicate that p53 activity can influence cell fate specification of mBMSCs, and provide molecular and cellular insights into the observation that p53 loss is associated with increased osteosarcoma incidence.

Introduction

The tumor suppressor, p53, plays an important role in suppressing osteosarcoma, which is one of the most common cancer types in patients with Li-Fraumeni syndrome, a rare genetic disorder caused by p53 deficiencies [1]. The osteosarcoma suppressive function of p53 is conserved in mice, as mice heterozygous for p53 deletion have a high incidence of osteosarcoma [2]. Although the cell of origin of osteosarcoma is still debatable, both uncommitted stromal stem/progenitor cells and committed pre-osteoblasts may serve as the precursors of osteosarcoma [3–6]. The ablation of p53 in uncommitted stem/progenitor cells of the limb buds leads to ~60% incidence of osteosarcoma [4, 6], suggesting that the role of p53 in stem/progenitor cells is associated with its osteosarcoma suppression function.

Bone marrow stromal cells (BMSCs, also known as bone marrow-derived mesenchymal stem cells) contain a subset of multipotent stem cells that have the ability to develop into several cell types, such as osteoblasts, adipocytes, and chondrocytes as well as hematopoiesis-supportive stroma. BMSCs are central mediators of bone homeostasis, due to their ability to not only form bone, but also due to their regulation of bone resorption [7]. Similar to most post-natal stem/progenitor cells, BMSCs utilize key transcription regulators to orchestrate their lineages choices. Runx2 governs osteogenesis [8], Sox9 controls chondrogenesis [9], and Pparg (peroxisome proliferator-activated receptor gamma) regulates adipogenesis [10]. Pathologically, BMSCs are relevant to osteosarcomagenesis. Some osteosarcomas and BMSCs share overlapping transcriptional programs and features, suggesting that osteosarcomas either originate from transformed BMSCs or from committed osteoblasts that are reprogrammed or dedifferentiated into BMSC-like cells [11, 12]. Both mouse and human BMSCs can generate osteosarcoma by induced transformation [13]. In addition, murine (but not human) BMSCs spontaneously become tumorigenic after prolonged in vitro culture [14].

p53 has been shown to negatively regulate bone homeostasis in vivo [15, 16]. On the one hand, bone density is higher in p53 knockout mice [17]. On the other hand, abnormally enhanced p53 activity is associated with decreased osteoblast differentiation and bone development [15, 16]. However, the role of p53 in uncommitted primary murine BMSCs has largely been elusive, mainly due to the challenge of isolating and maintaining BMSCs from adult mice, the controversy and lack of specificity of the cell surface markers used to isolate these cells, and the potential complication of using immortalized mesenchymal cell lines.

In this study, we used a recently validated approach [18] to isolate primary mouse BMSCs (mBMSCs) from p53 wild type (WT) and knockout (KO) mice, and studied the roles of p53 in these cells. We uncovered a role for p53 in controlling the lineage specification of primary mBMSCs; i.e., p53 loss had a pro-osteogenic function. Mechanistic studies showed that p53 induced microRNA-34s (miR34s), which suppressed Runx2, the osteogenic master gene. This lineage specification role for p53 appeared to be conserved in human osteosarcoma cells. In addition, we performed genome-wide analyses of these cells and found that genes up-regulated in p53_KO mBMSCs were significantly enriched in genes up-regulated in human osteosarcomas, thereby establishing an association between the role of p53 in mBMSCs and osteosarcoma.

Materials and Methods

Mice strains and isolation of MSCs from bone marrow

p53_LSL_R172H (LSL: Loxp-STOP-Loxp) heterozygous mice were obtained from the NCI Mouse Repository. The p53 gene in p53_LSL_R172H allele was disrupted by a STOP cassette. Mice homozygous for p53_LSL_R172H alleles are functionally p53 null mice [19]. Therefore, throughout the paper, p53_KO refers to p53_LSL_R172H homozygous mice. Mice were maintained under the strict guidelines of the Institutional Animal Care and Use Committee (IACUC)-approved protocols of the National Cancer Institute.

We used a validated and published protocol to isolate CD45⁻/CD11b⁻ cells from adherent bone marrow cells of 8-week old mice [18]. Briefly, whole bone marrow cells (around 50 million) from femurs and tibias were allowed to attach in 10 cm culture dishes (Corning) for 48 hours in MSC medium (Stem Cell Technologies). Then, non-adherent cells (more than 95% of cells from bone marrow) were washed away. These freshly isolated bone marrow cells were defined as passage 0 (P0). We grew P0 cells for one passage (P1) before cell sorting. During this propagation, most cells in the lymphoid and myeloid lineages detached and cell numbers reached about 5 to 20 million depending on the genotypes at the end of P1. After cell sorting and plating, CD45⁻/CD11⁻ cells were designated as P2. In most cases, mBMSCs with passage less than 4 (less than two passages after cell sorting) were used for differentiation. For p53_KO mBMSCs clonal strains, cells with passage less than 8 (less than 6 after cell sorting) were used due to the need for generating sufficient numbers of cells (see above for definition of passage numbers). CD45-FITC (a pan-leukocyte marker) and CD11b–PE (a myeloid lineage marker) antibodies were used to sort out a double negative population, which normally contains ~0.1–1 million cells.

Colony forming efficiency assay

The colony forming efficiency assay measures the frequency of Colony Forming Unit-Fibroblasts in the mBMSC population that are able to form colonies in a density-independent fashion. Freshly isolated bone marrow cells were seeded in 6-well plates (0.5 million cells/well or 1 million cells/well) in MSC cell medium and cultured for 13 days without changing the medium unless it became highly acidic. For Giemsa staining, the medium was removed, and the plate was gently washed twice with PBS, and then air-dried for 5 minutes. Methanol was added and incubated at room temperature for 5 minutes, followed by addition of Giemsa staining solution (Sigma) and incubation at room temperature for 5 minutes. The plate was washed with water and dried, and the number of colonies was counted.

Antibodies

CD45-FITC, CD11b–PE, Sca1-PE, CD105-PE, CD34-PE, CD31 (PECAM1), CD44-FITC antibodies were from BD Biosciences. Runx2 and Pparγ antibodies were from MBL International Corporation and Cell Signaling Technologies. Sox9 antibody was from Millipore. The p53 antibody was from Santa Cruz (M-19, Cat: sc-1312, lot: K1907), and β-actin and Tubulin antibodies were from Sigma.

Osteogenic differentiation

50,000 primary mBMSCs were seeded in 6-well plates and grown in BMSC cell medium for 36–48 hours. Medium was changed to osteogenic medium, which contains DMEM with low glucose plus 2% FBS, 50 µM ascorbic acid, 10 nM dexamethasone, and 10 mM β-glycerolphosphate. Cells were stained with 1% Alizarin red (pH 4.6) to detect deposited mineral at different time points.

Adipogenic differentiation

Cells were seeded in 6-well plates as for osteogenic induction. After 36 to 48 hours, medium was changed to adipogenic differentiation medium, which contains DMEM with low glucose plus 2% FBS, 50 µM indomethacin, 1 µM dexamethasone, 500 nM IBMX, and 5 µg/ml insulin. Differentiated adipocytes were stained with a 3:2 mixture of 0.5% oil red O in isopropanol and ddH2O.

Chondrogenic differentiation

10 million mBMSCs were mixed with 900 µl 1.5% (w/v) alginate solution in 150 mM NaCl. The alginate/cell suspension was rapidly dispensed with a 3 mL syringe fitted with a 27 gauge needle into a 100 mM CaCl₂ solution while stirring to form chondrogenic microbeads. The chondrogenic microbeads were grown in 4 wells of a 48-well plate in chondrogenic induction medium, which contains DMEM high glucose plus 6.25 µg/ml insulin, 6.25 µg/ml transferrin, 6.25 ng/ml sodium selenite, 5.33 µg/ml linoleic acid, 1.25 mg/ml bovine serum albumin, 100 nM dexamethasone, 10 ng/ml transforming growth factor-β3 (TGFβ3), and 50 µM ascorbic acid. The medium was changed every 3 days up to 3 weeks. After differentiation, the alginate beads were fixed overnight in 4% paraformaldehyde, 100 mM sodium cacodylate and 50 mM BaCl₂. The fixed alginate beads were paraffin embedded, sectioned and stained with 0.5% alcian blue in 0.1 N HCl for 30 min, which stains sulfated proteoglycans.

In vivo ectopic ossification assay

Briefly, 2 million mBMSCs were suspended in 20 µl culture medium, absorbed into 7 × 5 × 5 mm gelatin sponge cubes (Gelfoam™, Pfizer) and implanted subcutaneously in the back of 8-week-old female immunocompromised NSG mice (NOD.Cg-Prkdc^scidIl2rg^tm1WjI/SzJ; Jackson Laboratories) to form an ectopic ossicle [20]. Mice were maintained under the guidelines of an ACUC-approved protocol at the National Institute of Dental and Craniofacial Research (NIDCR). Each mouse was transplanted with 4 different sponges, loaded with WT_BMSCs, p53_KO BMSCs, DM5 BMSCs (a spontaneously immortalized mBMSC line that forms bone and supports formation of marrow) or culture medium loaded into the sponge, the latter two being positive and negative controls, respectively. BMSCs from three different donors for each genotype were tested in three different experiments and harvested at two time points (n=9 transplants/genotype/type point) Mice were monitored weekly by X-ray for ossification, starting at two weeks after the surgeries. X-ray images were taken at 21 days after transplantation. 24 days and 42 days after surgery, transplants were harvested, fixed in 4% PFA for 24 hours at 4 degree, and scanned and reconstructed with 10-µm isotropic voxels on a micro-CT analysis system (µCT 50, Scanco Medical). Afterwards, transplants were decalcified using 0.3 mM EDTA, paraffin embedded, sectioned and stained with hematoxylin and eosin (H&E).

Methylene blue staining to count cells during osteogenic differentiation

Counting cells during osteogenic differentiation has been challenging since trypsin-based approaches do not effectively detach cells from tissue culture plates. To overcome this challenge, we designed a methylene blue-based method [21]. Briefly, cells were allowed to differentiate in 6-well plates, rinsed gently with 2 ml PBS, stained with 1 ml 1% methylene blue in 20% ethanol solution for 2 minutes at room temperature, and gently washed with cool water 3 times. Afterwards, the stained cells were covered with water and 15 random images were taken under a 10X objective lens using a ZEISS AXIO Imager A2 microscope. Cell numbers of 15 images were averaged. The number of cells in each well of a 6-well plate was calculated as the average cell number times a magnification ratio (1.58×10³), the area of one well of a 6-well plate (9.61 cm²) divided by the area of a view under the 10X magnification (0.61 mm²).

Alkaline phosphatase and MTT assay

For microscopic views, we used the BCIP/NBT kit (Sigma) to detect alkaline phosphatase activity. Cells were washed once with Ca⁺⁺ and Mg⁺⁺-free PBS containing 0.05% Tween 20, then fixed with 10% formalin for 1 minute, and stained with BCIP/NBT substrate in the dark for 5–10 minutes. Afterwards, cells were washed once with the wash buffer, and pictures were taken. Positive staining is a dark blue color. To quantify the relative alkaline phosphatase activity, we used the SIGMAFAST™ p-Nitrophenyl Phosphate kit (Sigma) for AP measurement and the Thiazolyl Blue Tetrazolium Blue (MTT) assay (Sigma) for cell number.

miR detection and miR mimics transfection

To detect miRs, we used Taqman assays and the Taqman Universal PCR Master Mix on a 7500HT real-time PCR machine (Life Technologies). 100 ng of Trizol-extracted total RNA was used in reverse transcription with primers in the kit for each miR. Mmu-miR-34a (Assay ID: MC11030), mmu-miR-34b-5p (MC12558), and mmu-miR-34c-5p mimic (MC11039) were purchased from Life Technologies and transfected into mBMSCs at a concentration of 25 nM using DharmaFECT™ siRNA Transfection Reagents (Fisher Scientific).

Generation of miR34 decoy constructs to sequester miR34s

To sequester miR34s from their targets, the sense and antisense oligonucleotides containing two miR34 binding sites (ACCGTCACAGcccGACCAACA) were designed according to the methods published previously [22]. The sequences of sense and anti-sense (one decoy repeat contains two miR34 binding sites) were:

Sense, 5’-Phos-gtcccACCGTCACAGcccGACCAACAaattACCGTCACAGcccGACCAACAgg-3’

Antisense, 5’-Phos-gacccTGTTGGTCgggCTGTGACGGTaattTGTTGGTCgggCTGTGACGGTgg-3’

Multiple miR34 decoy duplexes in tandem were cloned into SanDI restriction site of the adapted pMSCV vector by inserting a SanDI restriction site containing linker. Forward primer: 5’-TTTATCCAGCCCTCACTCC-3’ and the reverse primer: 5’-TTGTGTAGCGCCAAGTGCC-3’ were used for screening correct clones. We choose a clone containing 4 repeats of decoy sequence (8 miR34 binding sites) for sequencing miR34 in p53_WT mBMSCs. See Supplementary Table 1 for the oligo sequences.

RNA-seq analyses and ChIP-seq analyses

Total RNA was extracted from mBMSCs using Trizol (Life Technologies). 1 µg of total RNA was sent to the facility at the Center for Cancer Research (CCR) to perform RNA-seq. RNA with PolyA tails was purified and subjected to fragmentation and library construction. Libraries were sequenced either with the Genome Analyzer (GIIx) or HiSeq 2000 platform. All of the procedures were carried out at the CCR Next Generation Sequencing Facility according to Illumina’s protocols. Sequencing tags were aligned to the mouse genome (mm9, 2007) using Tophat. We used Cufflinks to calculate the fragment per kilobase per million (FPKM). ChIP-seq was carried out as previously described [23] with modifications: 1) 20 µg of p53 antibody and 30 million mBMSCs were used for each ChIP; 2) mBMSCs were not treated to induce DNA damage since we have observed the effect of p53 on osteogenesis without DNA damage. Sequencing was performed at the CCR Next Generation Sequencing Facility using standard procedures. Peak calling was performed with MACS [24]. Peaks were assigned to transcripts using 50 kb as a cutoff (downloaded on May, 2013 from the UCSC Genome Browser).

shRNA knockdown of Runx2 and p21

shRNAs for knocking down Runx2 and p21 were cloned into pLKO.1-puro vector as previously described [25]. The sequences of the shRNAs for Runx2 were:

shRNA1, 5’-CCGGCGGTTCTCAAAGGAGCACAAACTCGAGTTTGTGCTCCTTTGAGAACCGTTTTT-3’;

shRNA4, 5’-CCGGGCAGAATGGATGAGTCTGTTTCTCGAGAAACAGACTCATCCATTCTGCTTTTT-3’.

We were only able to obtain one shRNA that significantly knocked down p21 in mBMSCs. The sequence of the shRNA for p21 was:

5’-CCGGACAGGAGCAAAGTGTGCCGTTCTCGAGAACGGCACACTTTGCTCCTGTTTTTT-3’.

Lentiviruses expressing the shRNAs were prepared as previously described [25]. Concentrated lentiviruses were used to infect mBMSCs.

Cloning of 3’ UTR of the Runx2 mRNA and the expression vector of miR34s

Through miRTarBase, we found that Runx2 has been shown to be one of the targets of miR34c (http://mirtarbase.mbc.nctu.edu.tw/php/detail.php?mirtid=MIRT001958). The 3’ UTR of Runx2 was cloned into the pSiCheck2 vector (Promega) using Xho I and Not I sites. Site-directed mutagenesis was done using the QuikChange XL Site-Directed Mutagenesis Kit (Agilent) in accordance to the protocol in the kit. To generate the expression vector for miR34, we used PCR to clone the genomic fragment corresponding to miR34a, miR34b, or miR34c and cloned each fragment into pMXs-miR-GFP_puro vector (Cell Biolabs). To prevent the interference of luciferase activity on the pSiCheck2 vector, we removed the GFP cassette from pMXs-miR-GFP_puro to generate a vector called pMXs-miR-puro. During the luciferase assay, pSiCheck2 vector and pMXs-miR-puro vector were co-transfected. All the cloning primers are provided in Supplementary Table 1.

Cycloheximide treatment

p53_WT mBMSCs and p53_KO mBMSCs were seeded into a 6-well plate one day before being treated with ethanol or 100 µg/ml Cycloheximide in ethanol for various hours. Cells were collected for standard Western blot analyses using Runx2 and β-actin antibodies. Band intensity was calculated with GeneTools software (SYNGENE).

Human osteosarcoma cell lines

Saos2 (HTB-85), U2OS (HTB-96), and TE85 cells were purchased from ATCC and grown in DMEM+10% FBS+antibiotics. Hu09-M112 is a subclone (generous gift from Dr. Jun Yokota, Biology Division, National Cancer Center Research Institute, Japan) from Hu09 cells and grown in RPMI1640+10% FBS+antibiotics [26]. Osteogenic differentiation was performed using an osteogenic differentiation kit (PCS-500-052) from ATCC.

Gene set enrichment analysis and Fisher’s Exact test

From mouse p53_WT and p53_KO mBMSCs, we derived two sets of genes that were up and down regulated in the mouse p53_WT, respectively. To test whether the up-regulated and down-regulated genes in mouse are relevant to human osteosarcoma and human MSCs, we generated two human up/down-regulated gene lists by mouse-human homolog mapping. We then used the GEO dataset GSE42352 to rank all genes by comparing the osteosarcoma and human BMSCs, and applied the Gene Set Enrichment Analysis (GSEA) method to analyze the human up/down-regulated genes. We also used Fisher’s exact test to analyze if there is an association between the two classification methods. In this case, a gene can belong to the down-regulated list or not, and it can also belong to the gene set in the left leading edge (from the leading edge analysis in GSEA) or not. A small p-value from the Fisher’s exact test of such a 2 by 2 table indicates association; i.e., enrichment of the down-regulated genes in the left leading edge (from the leading edge analysis in GSEA).

Results

Primary mBMSCs are capable of tri-lineage differentiation

To study the roles of p53 in mBMSCs, we used a published procedure based on attachment followed by flow cytometry to isolate primary CD45⁻/CD11b⁻ cells from the bone marrow of p53 wild type (p53_WT) and p53 knockout (p53_KO) mice (Figure 1A, Methods) [18]. We consistently obtained two-fold more CD45⁻/CD11b⁻ cells from p53_KO mice than from p53_WT mice immediately after cell sorting (Figure 1A). We also performed the colony forming efficiency (CFE) assay, which is based on the ability of a single cell to form a colony when plated at clonal density, a feature of BMSCs (See Methods) [7, 27]. The CFE assay showed that the frequency of colonies was 9.5+1.23 per million cells for p53_WT and 37.5+4 per million cells for p53_KO (Figure 1B), suggesting that p53 loss either increases the frequency of mBMSCs in vivo or allows mBMSCs to have a survival advantage in low density (clonal) conditions.

Figure 1. Primary mBMSCs have tri-lineage differentiation capacity in the absence and presence of p53.

A, Flow cytometry to sort CD45⁻/CD11b⁻ population (mBMSCs) from P1 cells (see Materials and Methods for definition of passage). Right panel is the percentage of mBMSCs (CD45⁻/CD11b⁻) within the previous gated cells, error bars are ± SEM; n=9, **, p<0.01. After sorting and plating, CD45⁻/CD11b⁻ cells are designated as P2. B, Colony forming efficiency assay of fresh bone marrow cells stained with Giemsa (low magnification, upper panels; high magnification, lower panels). Right panel is the frequency of colonies per million freshly isolated bone marrow cells, n=12, error bar is ± SEM; **, p<0.01. C, RNA-seq showing that both p53_WT and p53_KO mBMSCs express master regulators for osteogenesis (Runx2), chondrogenesis (Sox9), and adipogenesis (Pparg). D, Realtime PCR to measure mRNA levels of Runx2, Sox9, and Pparγ in three single cell-derived clones (#4, #8, #12) of p53_KO mBMSCs. Error bars are ± SEM, n=4. RT, reverse transcription; nonRT, no reverse transcriptase control.

CD45⁻/CD11b⁻ cells from both p53_WT and p53_KO mice had similar fibroblast-like morphology (Figure S1A), and they were positive for cell surface markers, such as stem cell antigen 1 (Sca1⁺), CD105 (CD105⁺) but negative for CD34 (CD34⁻), a cell surface marker on hematopoietic stem/progenitor cells (HSCs), and CD31 (CD31⁻), a marker for endothelial cells (Figure S1B) [28]. These cell surface markers were selected according to the guidance from the International Society for Cellular Therapy [29], although it is noted that these markers are not specific [7]. Of note, we did not detect the expression of CD44 in our freshly isolated CD45⁻/CD11b⁻ cells. This observation agrees with a recent report that CD44 is acquired de novo during prolonged in vitro culture of mBMSCs [30]. To characterize these cells at a molecular level, we performed RNA-seq analyses (Figure 1C). One of the prominent observations was that CD45⁻/CD11b⁻ cells from both p53_WT and p53_KO mice simultaneously expressed the master regulators of osteogenesis (Runx2), chondrogenesis (Sox9), and adipogenesis (Pparγ), in congruence with their capacity for tri-lineage differentiation (Figure 1C and S1C). Moreover, the mRNA levels of these master regulators in p53_WT and p53_KO cells were comparable, while those of two well established p53 direct target genes, Cdkn1a and Mdm2, were much higher in p53_WT cells than in p53_KO cells (Figure 1C and S1C).

The expression of master regulators of the three lineages could result from either a fraction of multipotent cells or a mixture of committed progenitors for the three lineages. To distinguish these two possibilities, we isolated single clones from p53_KO mBMSCs immediately after sorting and randomly selected three clones (#4, #8, #12) for realtime PCR analysis. All of the p53_KO single clones expressed markers for the three lineages, although in variable levels from clone to clone (Figure 1D). Thus, p53_KO mBMSC populations contain multipotent cells that express markers for the three lineages. For p53_WT mBMSCs, we were not able to propagate single clones because they either senesced or lost tri-lineage differentiation capacity after isolation. In summary, p53 loss allowed mBMSCs to grow for more passages and still be able to express multipotent markers.

Taken together, CD45⁻/CD11b⁻ cells within the adherent bone marrow population from p53_WT and p53_KO mice are BMSCs, and have tri-lineage differentiation ability (Figure 2A and S2). Of note, we did not detect any Oct4, Sox2, and Nanog expression in these cells based on RNA-seq (data not shown), suggesting that they are not like the newly reported multipotent R1 cells isolated from somatic tissues [31].

Figure 2. p53_KO mBMSCs have a higher tendency to differentiate into osteogenic lineage than p53_WT mBMSCs.

A, Alizarin red staining of p53_WT and p53_KO mBMSCs at different time points (0–27 days) during in vitro osteogenesis. B, Alkaline phosphatase (Alpl) staining of p53_WT and p53_KO mBMSCs at different time points during in vitro osteogenesis. C, Realtime PCR to measure the mRNA levels of Alpl during osteogenesis. D, High magnification of Alpl staining of p53_WT and p53_KO mBMSCs before differentiation. E, Quantification of Alpl activity by normalizing Alpl activity to cell number measured by MTT assay (Methods). The normalized Alpl activity of p53_WT mBMSC was set to 1; error bar are ± SEM; n=3; **, p<0.01. F, Macroscopic view of transplants generated p53_WT BMSCs and p53_KO mBMSCs in collagen sponges in NSG mice and corresponding µCT images of bone formed after 24 days. G, Bone volumes at 24 and 42 days; error bars are ± SEM; n=9 mice; *, p<0.05; **, p<0.01. H, Representative H&E staining of transplants (low power, upper right; boxes indicate high power views) showing the histology of bone formed 24 days after transplantation; CB – cortical bone, TB – trabecular bone; BM – bone marrow. I, Representative H&E staining of transplants (low power, upper right; boxes indicate high power views) showing the histology of bone formed 42 days after transplantation; CB – cortical bone, TB – trabecular bone; BM – bone marrow, A – adipocytes.

Loss of p53 predisposes mBMSCs to the osteogenic lineage

We did not observe obvious differences in in vitro adipogenesis and chondrogenesis between p53_WT and p53_KO mBMSCs (Figure S2). However, we found that p53_KO mBMSCs differentiated into mature osteoblasts much faster than p53_WT mBMSCs because alizarin red staining was positive within 6–9 days for differentiating p53_KO mBMSCs versus 21 days for differentiating p53_WT mBMSCs (Figure 2A). Alizarin red is a dye binding to calcium deposited in the matrix by mature osteoblasts, and therefore its positive staining represents a relatively late marker for osteogenesis [32]. To assess the osteogenic potential of mBMSCs at an earlier stage, we performed alkaline phosphatase (Alpl) staining. Alpl is a marker of osteogenic progenitors and mediates the maturation of osteoblasts by regulating the ratio of phosphate and pyrophosphate [33, 34]. In p53_WT mBMSCs, Alpl signal was detected as early as 6 days after differentiation (Figure 2B). Strikingly, we detected much higher Alpl activity in p53_KO mBMSCs than in p53_WT mBMSCs, suggesting that p53_KO mBMSCs have a higher tendency to develop into osteoblasts than p53_WT mBMSCs (Figure 2B). Using realtime PCR, we observed that the Alpl mRNA level gradually increased in p53_WT mBMSCs during osteoblast differentiation while its level in p53_KO mBMSCs was already high before differentiation, peaked at 6 days upon differentiation, dropped to the lowest point at day 15, and then increased again to a similar level as in p53_WT mBMSCs (Figure 2C). The bimodal pattern of Alpl levels (Figure 2B and 2C) in p53_KO mBMSCs may relate to the unique proliferation feature of p53_KO mBMSCs since Alpl levels in bone cells are associated with cell cycle [35].

The inhibitory effect of p53 on osteogenesis is not due to changes in cell cycle arrest, apoptosis, and cell confluence

After observing that p53 has a negative effect on osteogenic differentiation, we wanted to further study the underlying mechanism(s). We first tested the possibility that p53_KO mBMSCs differentiate into osteoblasts faster because they proliferate faster than p53_WT mBMSCs (Figure S3A). For this, we used short hairpin RNA (shRNA) to knockdown Cdkn1a (also called p21 or Waf1) in p53_WT mBMSCs (Figure S3B) because Cdkn1a is a major mediator of cell cycle arrest downstream of p53 [36, 37]. Knockdown of Cdkn1a made p53_WT mBMSCs become immortal and proliferate even faster than p53_KO mBMSCs (compare Figure S3C to S3A). Surprisingly, knockdown of Cdkn1a in p53_WT BMSCs led to lower levels of Alpl and slower maturation of osteogenic cells (Figure S3D and S3E). Therefore, although cell cycle arrest is generally associated with differentiation, our results indicate that p53 does not inhibit osteogenic differentiation through cell cycle arrest.

We noticed that osteogenesis in vitro generally started at the edge of tissue culture plates where the local cell-cell contact was higher than in other areas (Figure 2A). When we started with 50,000 p53_WT mBMSCs and p53_KO mBMSCs, we ended with a higher cell number for p53_KO mBMSCs than p53_WT mBMSCs because p53_KO mBMSCs proliferated faster. This raises the possibility that the higher cell-cell contact of p53_KO mBMSCs, at one point during differentiation, is associated with faster osteoblast maturation of these cells. To test this possibility, we assayed the osteogenic differentiation of different numbers of plated p53_WT mBMSCs (starting with 100,000 and 200,000 cells) and p53_KO mBMSCs (starting with 10,000 and 50,000 cells) and counted the cell numbers at each time point during differentiation using methylene blue staining (Figure S4A). The cell numbers under all these conditions were dynamic (Figure S4A and S4B). However, cell confluence cannot completely explain the faster osteogenic differentiation of p53_KO mBMSCs versus p53_WT mBMSCs because p53_KO mBMSCs starting at 10,000 have fewer cells at every time points compared with p53_WT mBMSCs starting at 200,000 (Figure S4B), but yet they differentiated faster (Figure S4A).

We also measured the apoptotic population in p53_WT and p53_KO mBMSCs and did not observe any difference of percentage of apoptotic cells between p53_WT and p53_KO mBMSCs (Figure S4C). Together, these studies reveal that p53 regulates the osteogenic differentiation of mBMSCs through a mechanism that is beyond its canonical cellular roles, such as cell cycle arrest, apoptosis, and cell confluence, and that this mechanism is likely a novel function of p53 in mBMSCs.

Single clones of p53_KO mBMSCs have the potential for tri-lineage differentiation

We were intrigued by the observation that p53_KO mBMSCs contained a higher percentage of Alpl-positive cells than p53_WT mBMSCs even before induced differentiation (Figure 2D and 2E). The high percentage of Alpl-positive cells in p53_KO mBMSCs suggests that either p53 deficiency increases the Alpl activity of mBMSCs or a higher percentage of Alpl-positive non-mBMSC cells were co-purified with p53_KO mBMSCs. To test these two possibilities, we picked and propagated single clones from p53_KO mBMSCs and then performed Alpl staining. Interestingly, we found that all the single cell-derived clones of p53_KO mBMSCs displayed different percentages of Alpl positive-staining cells (data not shown). Single clones from p53_KO mBMSCs had about 40% Alpl-positive cells while low-passage (less than 8) p53_WT mBMSCs contained less than 5% Alpl-positive cells (Figure 2D), suggesting that p53 either directly or indirectly controls the Alpl activity of mBMSCs.

We performed tri-lineage differentiation with three clones (#4, #8, #12) of p53_KO mBMSCs and found that they all have the ability to differentiate into three lineages (Figure S5A–C). Of note, one clone (#8) was able to differentiate into adipocytes with 100% efficiency (Figure S5A), indicating that clones of p53_KO mBMSCs were not fully committed to the osteogenic lineage and capable of differentiation into other lineages upon induction.

In summary, our results showed that the high percentage of Alpl-positive cells in freshly isolated primary p53_KO mBMSCs is not likely caused by contamination of Alpl-positive non-mBMSCs during purification. Instead, p53 loss encourages the spontaneous conversion of mBMSCs towards pre-osteoblasts. This effect of mBMSCs upon p53 loss may contribute to the rapid osteogenic differentiation.

p53_KO mBMSCs have enhanced osteogenic differentiation ability in vivo

To examine the differentiation ability of p53_KO BMSCs compared with p53_WT mBMSCs in vivo, cells were transplanted in conjunction with a gelatin sponge into immunocompromised mice to form an ectopic ossicle [38]. Each mouse received four independent transplants, loaded with p53_WT mBMSCs, p53_KO BMSCs, DM5 BMSCs (an immortalized line) as a positive control, and culture medium in a sponge as a negative control (Figure S6A). Compared to p53_WT transplants, p53_KO transplants were larger as shown by gross morphology and µCT (Figure 2F and Figure S6B), and had a larger bone volume (BV, Figure 2G). Histological examination at 24 days (Figure 2H), revealed that p53_WT transplants were composed of trabecular bone, while p53_KO transplants had already established cortical bone as well as trabecular bone (Figure 2G). At 42 days, p53_KO transplants were still significantly larger (Figure S6B) and displayed mature cortical bone, but were also noted to contain a dramatic increase in adipocytes of a larger size than in p53_WT transplants. By histomorphometric analysis, differences in bone density (BV/TV, Figure S6C), trabecular number (Tb.N, Figure S6D), trabecular spacing (Tb.Sp, Figure S6E) and connectivity density (Conn.D, Figure S6F) were also noted in transplants of different genotypes and at different time points. Taken together, these results indicate that p53_KO BMSCs displayed accelerated bone formation and establishment of an ectopic, albeit with a slightly different composition and architectural layout compared with transplants generated by p53_WT BMSCs.

High levels of Runx2 in p53_KO mBMSCs are mainly responsible for the enhanced osteogenic differentiation

Our previous study of p53 in ES cells found that p53 induced the differentiation of ES cells upon stress by changing the expression of ES cell master regulators [23]. This study led us to hypothesize that p53 affects osteogenic differentiation in mBMSCs by regulating Runx2, a master regulator of osteogenesis [8]. To test this hypothesis, we first performed immunoblotting of p53_WT mBMSCs and p53_KO mBMSCs using antibodies recognizing Runx2, Sox9, and Pparγ (Figure 3A). The result showed that the protein levels of Runx2 were higher in p53_KO mBMSCs than in p53_WT mBMSCs, while the levels of Sox9 and Pparγ were similar in both genotypes (Figure 3A). In addition, we measured the mRNA levels of two Runx2 targets, Sp7 (also called Osterix) and Bglap (also called Osteocalcin) and found that the levels of Sp7 were higher in p53_KO mBMSCs than in p53_WT mBMSCs (Figure S7A). However, we did not detect a difference in Bglap mRNA levels between p53_KO mBMSCs and p53_WT mBMSCs, presumably because osteocalcin expression also depends on other factors (Figure S7A). To investigate whether Runx2 was responsible for the enhanced osteogenic differentiation in p53_KO mBMSCs, we used retroviruses expressing Runx2 shRNAs to reduce Runx2 to a level similar to that in p53_WT mBMSCs (Figure 3B). The down regulation of Runx2 in p53_KO mBMSCs delayed the osteoblast maturation and Alpl staining of these mBMSCs, indicating that the effect of p53 loss on the osteogenic differentiation of mBMSCs is mainly due to the up-regulation of Runx2 (Figure 3C, 3D, and 3E).

Figure 3. Runx2 mediates the enhanced osteogenic differentiation of p53_KO mBMSCs.

A, Western blot analyses of p53_WT and p53_KO mBMSCs with Runx2, Sox9, Pparγ, p53, and β-actin antibodies. B, Western blot analyses of Runx2 in lentivirus-based knockdown of Runx2 in p53_KO mBMSCs. p53_WT mBMSCs are used as a control. C, Alizarin red staining of p53_KO mBMSCs with or without (shLuc) Runx2 knockdown. D, Alpl staining of p53_KO mBMSCs with or without Runx2 knockdown. shRNA1 and shRNA4 were used. E, Quantification of normalized Alpl activity. The normalized Alpl activity of p53_KO mBMSC (shLuc) was set to 1, n=3, error bar is ± SEM; *, p<0.05; **, p<0.01.

To further test whether the up-regulation of Runx2 causes the enhanced osteogenic differentiation of p53_WT mBMSCs and p53_KO mBMSCs, we over-expressed Runx2 in p53_WT and p53_KO mBMSCs (Figure 4A). Over-expression of Runx2 significantly increased the proliferation of p53_WT mBMSCs but not of p53_KO mBMSCs (Figure 4B). However, p53_WT mBMSCs over-expressing Runx2 still grew more slowly than p53_KO mBMSCs, suggesting that Runx2 is not the only factor that causes the faster proliferation of p53_KO mBMSCs. Over-expression of Runx2 also led to an increase of Alpl staining signal (Figure 4C and 4D) and faster osteoblast maturation of both p53_WT and p53_KO mBMSCs (Figure 4E). Therefore, both loss-of-function and gain-of-function experiments demonstrate that p53 regulates osteogenic differentiation of mBMSCs through Runx2.

Figure 4. Over-expression of Runx2 leads to faster osteogenic differentiation of p53_WT mBMSCs and p53_KO mBMSCs.

A, Western blot analyses of p53_WT and p53_KO mBMSCs with or without over-expression of Runx2. B, Proliferation assay for p53_WT mBMSCs and p53_KO mBMSCs with or without Runx2 over-expression. C, Alpl staining of p53_WT and p53_KO mBMSCs with or without over-expression of Runx2. D, Quantification of normalized Alpl activity. The normalized AP activity of p53_WT mBMSC (Luc) was set to 1, n=3, error bar is ± SEM. *, p<0.05; **, p<0.01. E, Alizarin red staining of p53_WT and p53_KO mBMSCs with or without over-expression of Runx2.

p53/miR34 axis suppresses the protein levels of Runx2 in mBMSCs

After establishing that Runx2 mediates the role of p53 in regulating the osteogenic differentiation of mBMSCs, we sought to delineate the mechanisms underlying the regulation of Runx2 by p53. Our previous study of p53 in mouse ES (mES) cells showed that p53 binds to the distal enhancers of some master regulators of ES cells and repress their expression in response to DNA damage [23]. Thus, one of the possibilities is that p53 directly represses Runx2 using the same mechanism in mES cells. However, our data from RNA-seq (Figure 1C and S1C) and realtime PCR (Figure 5A) showed that the RNA levels of Runx2 in p53_WT and p53_KO were comparable, but protein levels were higher in p53_KO (Figure 3A), indicating that p53 does not directly regulate the transcription of Runx2. We also performed p53 ChIP-seq (chromatin immunoprecipitation assay followed by deep sequencing) in both p53_WT mBMSCs and p53_KO mBMSCs. Although p53 readily bound to the loci of Cdkn1a and Mdm2, two well-established p53 targets, we did not detect the binding of p53 in the Runx2 locus, even after we extended the search to 100 kb upstream and downstream of the locus (Figure 5B). Taken together, these results ruled out the possibility that p53 directly represses the transcription of Runx2 in the absence of extrinsic stresses.

Figure 5. p53 indirectly represses Runx2.

A, Realtime PCR to measure the levels of Runx2 and Cdkn1a (p21). Error bars are ± SEM, n=6. **, p<0.01; n.s., not significant. B, ChIP-seq of p53 in p53_WT and p53_KO mBMSCs. Shown are the binding of p53 in the Cdkn1a (positive control), Mdm2 (positive control), and Runx2 loci. C, Western blot analyses (upper panel) and quantification of Runx2 protein (lower panel) in p53_WT mBMSCs and p53_KO mBMSCs treated with 100 µg/ml Cycloheximide for various time points as indicated. Representative Western blot images were selected based on the equal Runx2 intensity at the 0-hour time point for p53_WT mBMSCs and p53_KO mBMSCs. Densitometry was done with the GeneTools Software. Curves are exponential regression. Error bars are ± SEM, n=4, **, p<0.01. D, ChIP-seq of p53 in p53_WT mBMSCs and p53_KO mBMSCs. Shown is the binding of p53 in the miR34a and miR34b/c loci. E, Realtime PCR to measure the levels of mature miR-34a and miR-34b/c in p53_WT and p53_KO mBMSCs. Error bars are ± SEM, n=3, **, p<0.01. F, Western blot analyses of p53_WT mBMSCs, p53_KO mBMSCs transfected with 25 nM of mimics of miR34a, miR34b, or miR34c. G, Western blot analyses of p53_WT mBMSCs transfected with an empty vector control (Ctr) and a vector expressing a decoy containing 8 miR34 binding sites (miR34_sp8).

We then investigated possible post-transcriptional mechanisms. We first determined if the high levels of Runx2 protein (Figure 3A) in p53_KO mBMSCs were caused by slower degradation of Runx2. We treated both p53_WT mBMSCs and p53_KO mBMSCs with 100 ng/ml cycloheximide for various times and then quantified the protein levels of Runx2 (Figure 5C). The half-life of Runx2 is approximately 4.5 hours within both p53_WT mBMSCs and p53_KO mBMSCs, suggesting that decreased degradation of Runx2 is not the cause of the different protein levels of Runx2 in these two types of cells. Because miRNAs (miRs) are one of the major categories of post-transcriptional regulators, we hypothesized that p53 regulates miRs, which in turn suppress the translation of Runx2 mRNA. To identify the miRs regulated by p53 in mBMSCs, we mapped p53 peaks from ChIP-seq datasets to all the annotated miR loci in the Refseq database and found that p53 mainly binds to the loci of miR34a and miR34b/c (Figure 5D). The miR34 family is a major group of miRs regulated by p53 in differentiated cells [39, 40]. Realtime PCR showed that the levels of miR34a and miR34b/c indeed were much higher (>9 folds) in p53_WT mBMSCs than in p53_KO mBMSCs (Figure 5E). Since miR34a, miR34b, and miR34c have the same seed sequences (nucleotides 2–7) [41] (Figure S7B), the miR-34 family could be the major regulator to regulate the levels of Runx2 in mBMSCs. To test this, we transfected p53_KO mBMSCs with miR34a, miR34b, or miR34c mimics, which were chemically synthesized. Transfection of miR34a, miR34b, and miR34c decreased the protein levels of Runx2, comparable to the levels in p53_WT cells (Figure 5F). We also transfected p53_KO mBMSCs with mimics of miR34a, miR34b, and miR34c together and did not detect a synergistic effect of these three miRs in the repression of Runx2, suggesting that they are redundant (Figure 5F). To address whether miR34s regulate Runx2 in a physiological manner, we designed a miR34 decoy construct that sequesters all of members in the miR34 family [22] (see Materials and Methods). This miR34 decoy increased the levels of Runx2 in p53_WT mBMSCs, further supporting the notion that miR34s suppresses Runx2 (Figure 5G).

Using the miRanda algorithm (from the miRTarBase database) [42], we identified three potential target sites within the 3’ UTR (un-translated region) of the Runx2 mRNA (Figure S7C). We then cloned the 3’ UTR of Runx2 into a luciferase reporter vector (Figure 6A) and transfected this reporter into p53_WT mBMSCs and p53_KO mBMSCs. We observed that the luciferase activity in p53_KO mBMSCs was higher than that in p53_WT mBMSCs (Figure 6B), suggesting that higher levels of miR34s in p53_WT mBMSCs suppress the reporter activity of the 3’ UTR of Runx2 mRNA. We then performed site-directed mutagenesis on the three putative targeting sites in the 3’ UTR of the mouse Runx2 mRNA with different combinations (Figure 6C) and transfected these reporters into p53_WT mBMSCs. This assay revealed that the triple mutation (QC1, 2, 3) had the highest relative luciferase activity (Figure 6D). We then co-transfected the vectors expressing miR34s together with the reporter of 3’ UTR of Runx2 or the reporter of 3’ UTR containing the triple mutation into p53_KO mBMSCs. The expression of miR34 only decreased the relative luciferase activity of the wild type 3’ UTR reporter but not of the reporter of the 3’ UTR with the triple mutation (Figure 6E), demonstrating that miR34s suppress the 3’ UTR of Runx2 through these three target sites.

Figure 6. miR-34s suppresses the translation of Runx2.

A, Schematics of the reporter of 3’ UTR of mouse Runx2, pSiCheck2-mRunx2-3UTR. The ratio of Renilla luciferase activity versus Fire Fly luciferase activity was used as the unit of relative luciferase activity. B, Relative luciferase activity of pSiCheck2-mRunx2-3UTR in p53_KO mBMSCs and p53_WT mBMSCs. C, Schematics of mutating the 3’ UTR of mouse Runx2 using Quikchange (QC). Asterisks indicate the mutated sites. The numbers after “QC” indicate the position of putative miR34 targeting sites. D, Luciferase assay of the 3’ UTR of mouse Runx2 and various mutated version in p53_WT mBMSCs. E, Luciferase assay of the 3’ UTR (left panel) and 3’ UTR_QC1, 2, 3 (right panel) of mouse Runx2 without or without co-transfection of vector expressing miR34a, miR34b, or miR34c. Error bars are ± SEM, n=6. For all panels, **, p<0.01; *, p<0.05; n.s., not significant.

p53 loss in osteosarcoma cells is linked to enhanced osteogenic differentiation

To explore whether p53 has a role in osteogenic differentiation of osteosarcoma cells, we first examined the protein levels of p53 and Runx2 in four different human osteosarcoma cell lines, SAOS2, U2OS, TE85-MNNG, and Hu09-M112 (Figure 7A). SAOS2 and Hu09-M112 cells are p53 null. U2OS cells express low levels of wild type p53 while TE85-MNNG cells carry high basal levels of mutant p53 [43] (Figure 7A). We observed high levels of Runx2 in SAOS2 and Hu09-M112 cells, medium levels in TE85-MNNG cells, and low levels in U2OS cells, suggesting a reverse correlation of the protein levels of p53 and Runx2. We then performed in vitro osteogenic differentiation using these osteosarcoma cells and found that both SAOS2 and Hu09-M112 readily differentiated while TE85-MNNG and U2OS cells did not (Figure 7B). Therefore, p53 loss in osteosarcoma cells is associated with enhanced osteogenic differentiation.

Figure 7. The roles of p53 in mBMSCs are relevant to human osteosarcomas.

A, Western blot analyses showing the protein levels of Runx2, p53, and beta-actin in four different human osteosarcoma cell lines. Adriamycin (0.5 µM) treatment (8 hours) showing the statuses of p53 in cells: for wild type p53, Adriamycin treatment increases p53 levels (U2OS); for mutant p53, Adriamycin treatment does not alter the high basal levels of p53 (TE85_MNNG). SAOS2 and Hu09-M112 cells have been shown to be p53 null. B, Alizarin red staining of osteogenic differentiation of human osteosarcoma cells at different time points as indicated. C, Heatmap showing the expression of transcripts, measured by RNA-seq, directly or indirectly regulated by p53 in mBMSCs. D, Gene set enrichment analysis (GSEA) testing the enrichment of the gene set down-regulated in p53_WT mBMSCs (p53_WT down) and the gene set up-regulated in p53_WT mBMSCs (p53_WT up) in genes associated with human osteosarcoma versus mesenchymal stem cells (GSE42352). E. A model of the roles of p53 in the regulation of the osteogenic differentiation of mBMSCs. Bold, high levels; dash lines, low activity or levels.

Most p53-dependent genes in mBMSCs are indirectly regulated by p53

To gain a more global view of the roles of p53 in mBMSCs, we compared the transcripts that have at least two-fold expression change between p53_KO mBMSCs and p53_WT mBMSCs (Figure 7C). We identified 507 up-regulated transcripts (p53_WT up) and 538 down-regulated transcripts (p53_WT down) in p53_WT mBMSCs (Supplementary Table 2). KEGG pathway analysis showed that the p53 signaling pathway is enriched in the p53_WT up-regulated transcripts while several pathways, such as biosynthesis of unsaturated fatty acids (p=1.2e–04), extracellular matrix (EMC)-receptor interaction pathway (p=1.7e–04), glycine, serine and threonine metabolism pathway (p=2.4e–03), cell cycle (p=5.0e–03) and aminoacyl-tRNA biosynthesis (p=8.1e–03), are enriched in the p53_WT down-regulated genes (Figure S7D). To test whether p53 directly regulates the expression of these genes, we also combined ChIP-seq analysis in p53_WT mBMSCs and assigned p53 peaks to the 507 up-regulated and 538 down-regulated transcripts. To our surprise, only 52 up-regulated genes and 6 down-regulated transcripts have associated p53 peaks, suggesting that the majority of p53-dependent transcripts in mBMSCs are indirectly regulated by p53 (Figure S7E).

The lineage-controlling role of p53 in BMSCs is relevant to osteosarcomagenesis

To test the relevance of up-regulated genes and down-regulated genes to human osteosarcomas, we performed gene set enrichment analysis (GSEA) using the 507 up-regulated genes (p53_WT up) and 538 down-regulated genes (p53_WT down) as two separate sets and tested their association with human osteosarcoma tumor tissues and human BMSCs (hBMSCs). We found that the genes in the p53_WT up list are significantly associated with normal hBMSCs (p=0.031, 2000 permutations) while those in the p53_WT down list are associated with human osteosarcomas (p=0.036, 2000 permutations) (Figure 7D). To further test this association, we performed the Fisher’s exact test as a complementary approach (Supplementary Table 3). The significant association to osteosarcoma of the p53_WT down-list was corroborated by this analysis (p-value = 7.961e–06, odds ratio 2.02). The Fisher’s exact test of the association between the genes in the p53_WT up list and the genes up-regulated in hBMSCs also found that these two sets of genes are tightly linked (p-value = 1.494e–05, odds ratio 1.89). Of note, gene expression microarray cannot precisely determine p53 status (deletion or mutation) in these human osteosarcomas. But given the high frequency of p53 inactivation (deletion or mutation) in human osteosarcomas, it is possible that the association observed in GESA is partially attributed to p53. In summary, genes up-regulated in p53_KO (down in p53_WT) mBMSCs are increased in human osteosarcomas, suggesting that p53 controls the expression of a set of genes critical for osteosarcomagenesis.

Discussion

The inhibitory role of p53 in the osteogenic differentiation of mBMSCs

Our results propose a model of regulation of osteogenesis by p53 in mBMSCs (Figure 7E), in which p53 activates miR34s that in turn suppress the levels of Runx2 protein in p53_WT mBMSCs. p53 loss led to decreased miR34s, higher levels of Runx2, and faster osteogenesis of p53_KO mBMSCs. This model provides new molecular and cellular insights into the in vivo observation that the loss of p53 increases bone density [17]. These findings are outwardly contradictory to the general concept that loss of p53 inhibits differentiation of stem cells, such as ES cells and neural stem cells (NSCs). However, the underlying molecular mechanisms are similar, that is, p53 directly or indirectly represses transcription factors that play important roles in these stem cells. In ES cells, p53 represses the transcription of ES cell master regulators, the loss of which causes the differentiation of ES cells [23, 44, 45]. In the context of NSCs, p53 suppresses the expression of c-Myc, an important anti-differentiation factor in NSCs [46]. Runx2, however, is a pro-osteogenic transcription factor in BMSCs. In mBMSCs, p53 appears to act as one of the gate-keepers for osteogenesis by repressing Runx2 through miR34s. Therefore, the repression of master transcription factors critical for stem cell function may be one of the general mechanisms used by p53 to control the differentiation or cell fate specification of stem cells. The ultimate biological outcome of p53-mediated repression, however, depends on the specific roles of these master transcription factors in a given stem cell type. In differentiated cells, p53 is a cell fate regulator that directs cells to undergo cell cycle arrest or apoptosis depending on the types and strength of stresses [37]. Our results show that p53 plays a much broader role in regulating cell fate; i.e., lineage choice.

We found that p53 inhibits osteogenic differentiation using primary BMSCs in vitro, and that in an in vivo differentiation assay, p53_KO BMSCs demonstrated accelerated bone formation (as demonstrated by the amount of bone at early time points) and maturation (marked by increased adiposity) of ectopic ossicles compared with p53_WT BMSCs (Figure 2). In the study using mature osteoblasts from adult mice [17], p53 inhibits osteogenesis by repressing Osterix, but not Runx2. In our study, p53 indirectly represses Runx2 through miR34s. These differences are very likely caused by the different cell types used in these studies and therefore highlight the need for using well-characterized cell types (stage of maturation), particularly in studying differentiation regulation by p53. Indeed, we have previously shown that p53 regulates different sets of genes in different cell types, for example, mES cells and MEFs [25]. Future studies should address how the epigenetic landscape within a given cell type shapes the cellular outcome of p53 activation.

The indirect repression of Runx2 by the p53/miR34 axis in mBMSCs

It is not fully understood how p53 represses gene expression. Compared to p53-mediated activation, p53-mediated repression has multiple proposed mechanisms. In general, these mechanisms fall into two categories: direct repression and indirect repression. During direct repression, p53 can recruit co-repressors, such as Sin3a and LSD1, or interfere with distal enhancers to down-regulate its target genes [23, 47, 48], while in indirect repression, p53 regulates an intermediate factor, such as a transcription factor or miRs, which in turns repress p53 target genes [45, 49]. Our data showed that the repression of Runx2 by p53 in the osteogenic lineage control of mBMSCs is indirect through miR34s. A similar mechanism involving miR34s was also observed during the retinoic acid-induced differentiation of human ES cells, in which p53 induced miR34s to repress Oct4, Sox2, Lin28a, and Klf4 [45]. Although the regulation of miR34s by p53 is conserved across different cell types, the differential expression of master regulators in different cell types could change the outcome of the p53/miR34 axis.

Potential connection of p53-regulated osteogenesis and osteosarcoma suppression

Inactivation of p53 in uncommitted bone marrow stromal cells preferentially increased the incidence of osteosarcomas over that of chondrosarcomas and liposarcomas [4], coinciding with our observation that p53 loss mainly influences the osteogenic differentiation of mBMSCs. It is worth noting that Runx2, known as a master regulator for osteogenesis, is also a context-dependent oncogene in osteosarcoma [50]. These data suggest that osteogenic regulation and osteosarcomagenesis are two closely related processes. The p53/miR34/Runx2 axis may be one of the major pathways governing these two processes. Our GSEA and Fisher’s exact test showed that genes up-regulated in p53_KO mBMSCs are also increased in human osteosarcoma, suggesting that p53-regulated osteogenesis and osteosarcoma suppression are connected. Our result, however, does not show that these hundreds of p53-dependent genes in mBMSCs have direct roles in osteosarcomagenesis. To test this, future studies need to investigate the function of these genes, individually or as sub-groups, in osteosarcoma biology. Since most of p53-dependent genes are indirectly regulated by p53 (Figure S7E), it will be interesting to know whether Runx2 and/or other transcriptional factors mediate the regulation of some of these genes. Our characterized primary p53_WT mBMSCs and p53_KO mBMSCs will be invaluable for achieving these goals.