The coordination of cell cycle progression and osteoclast differentiation by RANKL signaling was studied. Experiments with mouse genetic models revealed that RANKL promoted cell cycle withdrawal of osteoclast precursors dependent on the cyclin kinase inhibitor p27-KIP1, but that both p27-KIP1 and p21-CIP1 were required for osteoclast differentiation. These cyclin inhibitors may directly regulate osteoclast differentiation in addition to regulating cell cycle withdrawal.
Introduction: RANKL stimulates mononuclear precursor cells of the myeloid lineage to differentiate into multinuclear osteoclasts, thus providing a system to study the fundamental problem of coordination of cell cycle progression with cell differentiation.
Materials and Methods: Mice that lack expression of functional cyclin inhibitors p27KIP1and p21CIP1 were used to study cell cycle progression and differentiation of osteoclast precursors in vitro and in vivo.
Results and Conclusions: Experiments with cells derived from p27KIP1- and p21CIP1-deficient mice indicated that p27KIP1 function alone was necessary for RANKL-mediated cell cycle withdrawal by osteoclast precursors, but osteoclasts from mice with single mutations in either of these two genes differentiated normally. In contrast, p21/p27 double knockout mice developed osteopetrosis, with fewer osteoclasts that exhibited lower TRACP activity and abnormal cell morphology present in long bone. Moreover, isolated osteoclast progenitors from p21/p27 double knockout mice were defective in RANKL-mediated differentiation in vitro, expressing low levels of osteoclast-specific genes like TRACP and cathepsin K. Taken together, these data suggest p27KIP1 and p21CIP1 play roles in osteoclast differentiation in response to RANKL signaling distinct from their roles in promoting cell cycle withdrawal.
BONE DEVELOPMENT AND maintenance in vertebrate animals are carried out by the coordinated action of two cells types, osteoblasts and osteoclasts.(1) Osteoblasts are a mesenchymal cell type that form bone, whereas mononuclear precursors of myeloid origin fuse to form multinuclear osteoclasts capable of bone resorption.(1) In response to developmental or homeostatic signals, osteoblasts produce two factors, macrophage colony-stimulating factor1 (CSF-1) and RANKL, which in combination, are necessary and sufficient for osteoclast differentiation.(2–6) RANKL binds to its receptor RANK, a member of the TNF-receptor super family, to effect osteoclast differentiation.(7,8) Soluble RANKL, in the presence of CSF-1, induces the in vitro formation of multinuclear osteoclasts from macrophage/monocyte precursor cells. CSF-1 is essential for the proliferation of the precursor cells and for the survival of differentiated osteoclasts.(9,10) Mice with targeted deletions in RANKL or RANK exhibit osteopetrosis caused by the lack of differentiated osteoclasts.(3,4,8)
Previous studies have indicated that between days 15 and 17 of embryonic metatarsal development in mice, osteoclast progenitors shift from a population of actively proliferating cells to that of committed, postmitotic mononuclear precursor cells. These precursors start to fuse on day 18 to become multinuclear osteoclasts.(11,12) Osteoclast progenitors grown in co-culture with osteoblasts actively proliferate for the first 4 days in co-culture. However, 80% of these precursors withdraw from cell cycle by day 5 in co-culture.(9) Thus, osteoclast progenitors seem to be postmitotic at the time of fusion and terminal differentiation. However, the mechanism by which osteoclast precursors exit the cell cycle remains unknown.
In general, progenitor cells cease proliferating and enter into a state of quiescence to achieve the specialized characteristics of a terminally differentiated cell. Proliferation in mammalian cells is controlled primarily by events that lead to the activation and inactivation of cyclin-dependent kinases (CDKs).(13,14) The checkpoint for entry into the S-phase of the cell cycle is controlled by signals that control the activity of specific CDK complexes. Thus, the activities of cyclinD/cdk4, cyclinD/cdk6, cyclinE/cdk2, and cyclinA/cdk2 are rate limiting for entry into S-phase by mammalian cells.(13,14) The INK (p15INK4a, p16INK4b, p18INK4c and p19INK4d) and the Cip/Kip (p21CIP1, p27KIP1 and p57KIP2) families of CDK inhibitors (CDKIs) regulate the activities of the CDKs and the cyclin-CDK complexes during the G1-S transition.(13)
The CDKIs are known to play key roles in coordinating cell cycle withdrawal with differentiation in many different tissue and cell systems.(15) In addition to controlling cell proliferation, mounting evidence suggests that CDKIs may play a more direct role in cell differentiation.(15) In particular, genetic studies in mice indicate that p21CIP1, p57KIP2, and p27KIP1 alone or in combination play a direct role in differentiation.(15) For example, in mice deficient for both p21CIP1 and p57KIP2, terminal differentiation of skeletal muscle and lung epithelial cells are affected.(16) Similarly, it seems that p27KIP1 and p21CIP1 play distinct and overlapping roles during oligodendrocyte differentiation.(17) Okahashi et al.(18) showed that p21CIP1 and p27KIP1 are transiently induced in primary osteoclast progenitor cells on RANKL and TNFα treatment and that osteoclast formation in these cultures could be blocked by the addition of a combination of antisense oligonucleotides. These workers suggested that the CDKIs played a role other than cell cycle inhibition during osteoclast differentiation, because after RANKL stimulation, they observed stimulation of cell growth, and not cell cycle arrest, in osteoclast precursor cells.(18)
In this study, we have examined the role of CDKIs in the terminal differentiation of osteoclasts both in vitro and in vivo using mouse genetic models that lack either p27KIP1 or p21CIP1 genes singly or both genes in combination. Our results indicate that mice lacking both p27KIP1 and p21CIP1 exhibited osteopetrosis and had fewer osteoclasts with lower TRACP activity and with abnormal cell morphology. RANKL-dependent in vitro differentiation of osteoclast-like cells from p27−/−p21−/− double knockout mice was impaired, and the expression of the osteoclast differentiation markers TRACP and cathepsin K was affected. These data suggest that p21CIP1 and p27KIP1 may play a role in RANKL-mediated osteoclast terminal differentiation in addition to their role in cell cycle regulation.
MATERIALS AND METHODS
p27KIP1−/− (C57BL/6J) mice were a gift from Dr Andrew Koff.(19) p21CIP1−/− (C57Bl/6J) mice were a generous gift from Dr Ming You (Ohio State University).(20) PCR-based genotyping of both mutant and wildtype alleles was performed as described.(19,20) To generate mice for experiments reported here, mice heterozygous for both p21 and p27 knockout alleles were mated to generate littermates with p27WTp21WT (wildtype), p27KOp21WT (p27KO), p27WTp21KO (p27KO), and p27KOp21KO (DKO) genotypes. Animal studies were conducted under a protocol approved by the Institutional Animal Use and Care Committee.
In vitro osteoclast cultures
Bone marrow cells from murine femurs were cultured for 3 days in DMEM medium with 10% heat-inactivated FBS containing 50 ng/ml recombinant colony-stimulating factor (CSF)-1 (proliferation medium) on nontissue culture petri dishes (Lab Tek/Nunc 4021). CSF-1 was produced by Chiron Corp. and was a gift from Dr David Hume (Queensland University). At this point, >95% of these cells express MAC-1 and c-fms, as determined by flow cytometry, and thus represent a highly enriched population of bone marrow-derived macrophages. After 3 days of culture, the cells were switched to 25 ng/ml of CSF-1 and 50 ng/ml of recombinant soluble RANKL (differentiation medium). Soluble RANKL was prepared as previously described.(21) Differentiation medium was replaced after every 2 days of culture.
Bone marrow-derived macrophages were pulse-labeled for 30 minutes with 10 μM 5-bromodeoxyuridine (BrdU; BD Biosciences) along with 0.4 μM 5-fluoro-2′deoxyuridine to block endogenous thymidine synthesis after RANKL treatment. BrdU-labeled cells were detected using αBrdU antibody (BD Biosciences) and fluorescein (FITC)-conjugated secondary antibody (Jackson Immuno Research Laboratories). Cells were counterstained with 0.04 μg/ml of propidium iodide (PI) and analyzed by fluorescent microscopy and digital photography. For each experiment, 4000 cells were counted, and the ratio of BrdU+ to total PI+ cells was determined. Each experiment was repeated three times.
Immunoblotting and kinase assays
Western blotting of whole cell extracts was performed with polyclonal antibodies against p27(C-19), p21 (M-19), p18 (M-20), p19 (M-167), CDK2 (M2), and ERK1 (all from Santa Cruz Biotechnology). The intensity of the individual signals was quantified using the Lumi-Imager (Roche). For the signaling experiments, 10 μM SB203580 or 40 μM PD98059 (both from Calbiochem) or 10 μM LY294002 (AG Scientific) were added to the cells together with RANKL for indicated time-points. For all Western blots, the results of three independent experiments were analyzed and quantified. To calculate fold-induction, the ratio of the signal from p27 or p21 to the ERK kinase signal at a specific time-point was compared with the same ratio at time zero (before addition of RANKL). CDK2 kinase assays were performed as previously described.(21) The kinase reactions were run on 12% SDS-PAGE, blotted onto nitrocellulose membrane, and quantified with a Molecular Dynamics Phosphoimager.
TRACP staining of differentiated osteoclasts and calcium phosphate resorption assays
Osteoclast-like cells were washed twice with PBS, fixed in 3.7% paraformaldehyde (Polysciences), assayed for TRACP activity using Leukocyte Acid Phosphatase kit (Sigma), and counterstained with 10 μM bis-benzamide (Sigma). The number of multinuclear osteoclasts was measured using fluorescent microscopy and digital photography. An equal number of bone marrow-derived macrophages for each genotype tested was plated in differentiation medium onto BD Biocoat Osteologic 16-well multitest slides (BD Biosciences) according to manufacturer's protocol. The area, number, and perimeter of the pits were measured using Bioquant Nova software (R&M Biometrics).
cDNA was prepared from 1 μg of total RNA using 24U of Reverse Transcriptase (Roche Applied Science) according to manufacturer's protocol. For real-time PCR, the following primers and Taqman probes were used: TRACP forward: 5′ GATCTCCAAGCGCTGGAACTT3′; TRACP reverse: 5′CAGTTATGTTTGTACGTGGAATTTTGA3′; TRACP probe: 6Fam-CCCAGCCCTTACTACCGTTTGCGC-Tamara; cathepsin K forward: 5′ACCCAGTGGGAGCTATGGAA3′; cathepsin K reverse: 5′TCCCAAATTAAACGCCGAGA3′; cathepsin K probe: 6Fam-CATCCACCTTGCTGTTATACTGCTTCTGGTGA-Tamara. GAPDH primers and probes were purchased from Applied Biosystems. Reactions were performed using Taqman Universal PCR mix (Applied Biosystems) according to manufacturer's directions. Thermocycling was performed using the Cepheid Smart Cycler System. Fold induction was calculated as the difference in cycle threshold (CT) values (ΔCT) between control and RANKL-treated samples raised to the power of 2 (2ΔCT). These were further normalized against ΔCT values for GAPDH or 18S RNA.
Radiography and histomorphometric analysis
The mouse skeletons were fixed in 3.7% paraformaldehyde at 4°C for 48 h and transferred to 70% ethanol. Radiography was performed using a Faxitron X-ray machine (model 43855A; Hewlett Packard) at 35 kvp for 2 minutes. Glycol methacrylate infiltration and embedding of femurs were performed using the JB-4 embedding kit (Polysciences). Histomorphometric analysis on TRACP-stained sections was performed using Bioquant Nova software (R&M Biometrics).
For the real-time PCR assays, an ANOVA was performed using a model that assigns a unique mean CT value for each gene, genotype, and time-point tested from individual experiments, and whether the difference in the fold changes between samples was significant using a 95% CI for each fold estimate was tested. For all other experiments, the significance of the differences observed between wildtype and mutant samples was tested by Student's t-test, using Statview for Windows software, version 5.0.1 (SAS Institute). In all experiments, differences of p < 0.05 were deemed as significant.
RANKL signaling stimulates cell cycle withdrawal and expression of cyclin-dependent kinase inhibitors p27KIP1 and p21CIP1 in bone marrow-derived macrophages
RANKL is the key cytokine required for osteoclast differentiation both in vivo and in vitro.(2,4,6) However, whether RANKL affects cell cycle progression before osteoclast differentiation has not been extensively studied. To determine whether RANKL coordinates cell cycle withdrawal with osteoclast differentiation, we pulse labeled bone marrow-derived macrophages from 15- to 20-day-old wildtype mice with 5-BrdU at 0, 24, and 48 h in proliferation (CSF-1 only) or in differentiation medium (CSF-1 + RANKL; Fig. 1). A total of 4000 cells was counted at each time-point, and each experiment was repeated three times. Micrographs representative of the data collected are presented in Fig. 1A, and the cumulative results are presented graphically in Fig. 1B. The percentage of cells in the S-phase of the cell cycle as measured in this assay decreased from 40.9 ± 1.3% (n = 3) before adding RANKL (time-point 0 h) to 6.3 ± 2% (n = 3) by 48 h of adding RANKL (Fig. 1B, hatched bars). For cells maintained in proliferation medium, there was no significant differences in the percentage of cells in S-phase after 48 h (Fig. 1B, white bars). Identical results were obtained when flow cytometry was used to analyze cell cycle parameters after 24 h of RANKL treatment, with a decrease of 40-10% S-phase observed (data not shown).
The INK and Cip/Kip families of cyclin-dependent kinase inhibitors (CDKIs) are known to regulate cell cycle withdrawal based on their intracellular concentrations(13); therefore, expression of the CDKIs in bone marrow-derived macrophages after RANKL treatment was analyzed by Western blotting (Fig. 2). Three independent experiments showed that p27KIP1 (p27) levels are elevated by ∼4- to 5-fold 12 h after adding RANKL and by 7- to 8-fold 24 h after adding RANKL (a representative experiment is presented in Fig. 2A and quantification of the Western blot in Fig. 2B). In the same set of experiments, p21CIP1 (p21) levels were also elevated in the osteoclast progenitors by ∼2- to 3-fold 24 h after adding RANKL (Figs. 2A and 2B). The expression levels of p18ink4c or p19ink4d remained unchanged after treatment with RANKL in these experiments (Figs. 2A and 2B).
p27 and p21 block entry of cells into S-phase by binding to and inactivating the Cdk2-cyclin E complex, thus preventing phosphorylation of the retinoblastoma protein.(13) To investigate whether RANKL stimulation of p27 and p21 levels coincided with decreased Cdk2 activity, the activity of Cdk2 complexes were measured in immune kinase assays using Histone H1 as substrate (Fig. 2C). In two independent experiments, Cdk2 kinase activity was diminished by 5- to 7-fold after a 24-h RANKL treatment, despite the presence of similar amounts of Cdk2 protein in all of the immune complexes (Fig. 2C).
The RANKL-RANK signaling cascade activates several signaling pathways including the MAPK and phosphatidylinositol-3 (PI 3) kinase pathways. Inhibitor-based studies have shown that the p38 MAPK pathway is involved in the differentiation of osteoclast-like cells in vitro(22) and in regulating osteoclast-specific gene expression.(23) To determine the downstream RANKL pathway(s) that might regulate CDKI activity, we treated bone marrow-derived macrophages either with RANKL alone or with RANKL in the presence of pharmacologic inhibitors of specific signaling pathways (Figs. 2D and 2E). The p38 MAPK inhibitor SB203580 inhibited the increase of p27 protein levels normally induced by RANKL at both 12 and 24 h by ∼3-fold (compare lanes 2 and 4 and lanes 3 and 5, respectively, in Fig. 2D; quantified in Fig. 2E). Inhibitors for p42/p44 MAPK (PD98059) or PI-3 kinase (LY294002) pathways had little effect on RANKL stimulation of p27 levels (Fig. 2D, lanes 6-7 and 8-9, respectively). The levels of p21 were not changed by any of the drugs used in this experiment (Figs. 2D and 2E).
Cells lacking p27KIP1 are defective in withdrawal from the cell cycle in response to RANKL
To more directly test the role of p27 and p21 in RANKL-mediated cell cycle withdrawal, bone marrow-derived progenitor cells prepared from p27−/−, p21−/−, or p21/p27 double knockout (DKO) mice were studied (Figs. 1C and 1D). In these experiments, RANKL treatment resulted in cell cycle arrest in 95% of both wildtype and p21−/− cells after 48 h of treatment (Fig. 1C), while ∼38% of both wildtype and p21−/− cells remained in the cell cycle after 48 h of treatment with CSF-1 alone (Fig. 1D).
In contrast, 33 ± 5% of cells from p27−/− mice remained in S-phase after 48 h of RANKL treatment. Similarly, after 48 h of RANKL treatment, 41 ± 5% of DKO cells remained in the cell cycle (Fig. 1C). These numbers were not significantly different than those observed after 48 h of treatment with CSF-1 alone (41 ± 6% for p27−/− and 49 ± 7% for DKO cells; Fig. 1D). After 72 h of RANKL treatment, both p27−/− (42 ± 3%) and DKO (43 ± 3%) cells remained in the cell cycle, again very similar to cells treated with CSF-1 alone. Thus, p27−/− and DKO cells don't efficiently exit the cell cycle even after 72 h of RANKL treatment.
After 96-120 h of treatment, cells of all genotypes had withdrawn from the cell cycle, regardless of whether treatment was with differentiation or proliferation medium (Figs. 1C and 1D). Proliferating cells of all genotypes withdrew from the cell cycle after six to eight cell cycles, although the culture medium was renewed every other day, and this cell cycle exit was RANKL independent. This withdrawal from the cell cycle may reflect a response to the culture conditions or to intrinsic properties of the monocyte/macrophage lineage as has been suggested for other cell types that behave in similar fashion.(24)
Osteopetrosis in long bones of p21/p27 DKO mice
To determine whether the defects on cell cycle withdrawal observed in vitro correlated with in vivo osteoclast function, we examined both single CDKI knockout and p21/p27 DKO mice. The phenotypes of p21−/− and p27−/− single mutant mice have previously been described.(19,20) The DKO mice were born according to expected Mendelian ratio and were healthy and viable. The DKO mice had a phenotype similar to that reported for the p27 single knockout mice; specifically, the DKO mice were larger than littermates and there was organomegaly, but no pathological abnormalities could be detected in organs examined, and only males were fertile.(19)
The long bones of the single knockouts and the DKO mice were examined by radiography (Fig. 3A). Mice that were 3-4 days old were chosen for this initial analysis because the growth of the long bone is very rapid at this age, and osteoclasts are abundant, so that any defects in osteoclast differentiation and function are more likely to be observed at this stage of development. The long bones p21−/− and p27−/− single mutant mice were not significantly different than wildtype, shown by the modeling of bone marrow cavities evident in the femurs of these mice (Fig. 3A). However, 13/20 DKO null pups examined (65%) had obvious sclerotic lesions in the femur mid diaphysis, a phenotype characteristic of severe osteopetrosis (Fig. 3A, arrowhead).
Femurs from mice of all three mutant genotypes and wildtype controls were embedded in glycol methacrylate, and sections prepared were stained for TRACP activity, a histochemical marker for osteoclasts, and counterstained with hematoxylin. The severely affected DKO mice had more unresorbed trabecular bone than littermates with single mutations or wildtype (Fig. 3B, arrowheads). Overall, the osteoclasts in all the DKO mice exhibited considerably weaker staining for TRACP activity compared with wildtype or single mutants, independent of the osteopetrotic phenotype of the mice (Figs. 3B and 4, arrowheads). The osteoclasts in DKO mice contained only small amounts of TRACP activity compared with the intensely stained osteoclasts in control samples. In addition, the osteoclasts from the DKO mice exhibited abnormal cell morphology (Fig. 4, arrowheads). These osteoclasts were heavily vacuolated compared with wildtype osteoclasts, and most of these vacuoles were localized around the nucleus, a phenotype similar to that reported for TRACP knockout mice.(25)
Histomorphometric analysis of these sections was performed using the Bioquant Nova Software system. Three parameters studied in this analysis are presented: the percentage of unresorbed trabecular bone volume in the femur relative to the total bone volume (BV/TV; Fig. 5A, histogram 1), the percentage of osteoclast surface to total bone surface, a measure of osteoclast size and function (Oc.S/BS; Fig. 5A, histogram 2), and the number of osteoclasts per total bone surface (N.Oc/BS; Fig. 5A, histogram 3). The results obtained for wildtype and p21−/− mice were identical, and only the results for wildtype are presented. The results presented for DKO mice were divided into two groups: the severely affected mice where bone lesions were detected by radiography (n = 13) and the mildly affected mice not readily detected by radiography (n = 7).
When wildtype (n = 7) and p27−/− (n = 7) are compared, all three parameters were found to be similar. The BV/TV values were identical, whereas osteoclast number and osteoclast surface were slightly lower in p27−/−, but these differences were not statistically significant (Fig. 5A, first two bars in each histogram, respectively). The severely effected DKO group is significantly different for all three parameters: BV/TV is over 3-fold higher, whereas osteoclast size and numbers are reduced by ∼3-fold (Fig. 5A, third bar in each histogram). The less severely affected DKO mice have significant differences for the two osteoclast parameters measured, also with an average 3-fold reduction in these parameters compared with wildtype (Fig. 5A, fourth bar in each histogram). However, BV/TV is, on average, only 1.4-fold higher in this group of mice compared with wildtype controls, a change that was not statistically significant.
Because p27KIP1 plays a role in osteoblast differentiation,(26) we wanted to investigate whether the osteopetrotic phenotype observed in the 3- to 4-day-old DKO mice could be caused by an increase in osteoblast size or number (Fig. 5B). This analysis indicated that there was no statistically significant difference in Ob.S/BS or N.Ob/BS parameters between the wildtype and severely osteopetrotic DKO mice (Fig. 5B).
In 20-day-old DKO mice, although all three osteoclast histomorphometric parameters were still significantly lower than those in the wildtype mice, the osteopetrotic phenotype is less severe compared with the 3- to 4-day-old mice (Fig. 5C). For example, BV/TV is an average of 1.5-fold higher in the DKO mice at 20 days of age. In adult DKO mice at 60 days of age, no obvious difference in BV/TV between wildtype and DKO samples persists (Fig. 5C), indicating that osteopetrosis is resolved with increasing age. However, differences in both osteoclast size and number that are statistically significant were still observed.
Impaired terminal differentiation in p21/p27 DKO osteoclast-like cells in vitro
The in vivo results described above suggested that osteoclast terminal differentiation was impaired in p21/p27 DKO mice. To determine if the defects in osteoclast differentiation and function were cell autonomous, we studied differentiation of DKO cells in vitro. Cells were plated in differentiation medium on either gelatin-coated dishes to test for the formation of TRACP+ multinuclear osteoclasts (representative experiment presented in Fig. 6A), or on calcium phosphate-coated wells (see the Materials and Methods section) to test for the formation of resorption pits (Fig. 6B). For these experiments, bone marrow cells from four 14-day-old mice (two males and two females) of each genotype were pooled and treated under standard differentiation conditions without determining the bone phenotype of the mice.
The number of TRACP+, multinuclear osteoclasts in p21/p27 DKO cells was significantly lower than wildtype after 9 days in differentiation medium (Table 1). While total numbers of p21−/− and p27−/− multinuclear cells were also lower in this assay, the overall differences were not statistically significant (Table 1).
Table Table 1.. Quantification of Multinuclear Osteoclasts (No. of Cells/mm2)
When plated on calcium phosphate wells, DKO cells showed a significant decrease in both the total number and the relative size of resorptive pits formed compared with wildtype after 12 days in differentiation medium (Table 2). The majority of the pits formed by the DKO osteoclasts were within an area range of 20-500 square pixels, whereas those made by the wildtype osteoclasts were within an area range of 500-20,000 square pixels (Table 2). In this functional assay, both p21−/− and p27−/− cells showed a significant decrease in total number of pits, in particular, in the formation of larger pits (5000 to >20,000; Table 2), although the effect was not as severe as for DKO cells.
Table Table 2.. Quantification of Resorption Pits
The expression of TRACP and cathepsin K, two genes expressed during osteoclast differentiation,(27,28) was analyzed by real-time RT-PCR in wildtype, p21−/−, p27−/−, and DKO osteoclasts at 72 h after treatment with RANKL (Fig. 6C). This analysis showed that, in wildtype cells, the expression of TRACP and cathepsin K increased 12.9 ± 1.8- and 6.7 ± 0.9-fold, respectively, compared with the control that did not receive RANKL (Fig. 6C). In contrast, in DKO cells, expression of these genes was not significantly increased by RANKL treatment; TRACP expression was 1.1 ± 0.5-fold and cathepsin K expression was 1.5 ± 0.5-fold, respectively (Fig. 6C). In p27−/− single knockout cells, TRACP and cathepsin K expression levels were at an intermediate level: TRACP expression was decreased 3-fold, whereas cathepsin K was decreased 2-fold compared with wildtype. In p21−/− cells, there was no significant effect on expression of these genes. The expression of c-fms, the gene encoding the receptor for CSF-1, is reduced by ∼2-fold in cells of all four genotypes and serves as an internal control, indicating that global changes in gene expression are not observed in the DKO cells (Fig. 6C). These data show that two markers of osteoclast terminal differentiation, TRACP and cathepsin K, are not induced in DKO osteoclast-like cells.
Withdrawal from the cell cycle, initiation of differentiation-specific gene expression programs, and coordinated morphological changes are key steps in the terminal differentiation of cells.(15) A single agent, RANKL, induces terminal differentiation of osteoclasts, thus providing an attractive mammalian model to study the coordination of cell cycle withdrawal and terminal differentiation. Results reported here show that mice lacking the cell cycle regulators p21 and p27 developed osteopetrosis in long bones, especially evident during the rapid growth phase of bone that occurs during the first weeks after birth. DKO mice had fewer osteoclasts on bone that had abnormal morphology, but the number and size of osteoblasts was similar to control samples. In contrast, p27−/− or p21−/− single mutant mice had no detectable bone phenotype. These results indicate that the combination of both p21and p27 are required during the terminal differentiation of osteoclasts and suggest that their effects are cell autonomous to the osteoclast.
A previous study used an antisense approach to implicate p21 and p27 in osteoclast-like cell differentiation in vitro.(18) Our results, obtained with cells isolated from knockout mice, confirm and significantly extend these observations. RANKL signaling resulted in cell cycle arrest preceding overt differentiation of bone marrow-derived precursors, and p27 action alone was sufficient for the RANKL-mediated arrest. In contrast, both p27 and p21 were required for RANKL-mediated differentiation of osteoclast-like cells in vitro. In particular, RANKL stimulation of TRACP and cathepsin K expression was abrogated in cells from DKO mice, indicating that the CDKI genes might play a role in the regulation of differentiation-specific gene expression during osteoclast differentiation. Cells that lacked only p21 or p27 singly did form significantly lower numbers of multinuclear osteoclasts, especially as measured in the resorption assays, and osteoclasts derived from p27KIP1−/− mice showed a slight 2- to 3-fold reduction in expression of TRACP and cathepsin K. However, in contrast to the cell cycle results, these effects were much less severe compared with the effects observed with DKO cells. The in vitro data support the hypothesis that the bone phenotype observed in vivo is the result of a cell autonomous effect of p21CIP1 and p27KIP1 on the osteoclast.
Osteopetrosis evident in long bones of DKO mice was most severe in pups in the first week of age, but the phenotype was not fully penetrant, because 65% of the DKO mice had a severe bone phenotype. In addition, the bone phenotype resolved with age. However, at all ages studied, fewer osteoclasts that stained weakly for TRACP activity were present in bone. Interestingly, the bone phenotype of p21/p27 DKO mice is similar to that reported for the cathepsin K and TRACP knockout models, both of which develop mild osteopetrosis and contain osteoclasts with abnormal morphology and function.(26,29–31) Taken together, the results indicate that p21 and p27 may be necessary for certain aspects of osteoclast differentiation, including expression of genes like TRACP and cathepsin K, but that CDKI action is not sufficient for terminal differentiation. The osteoclast phenotype in p21/p27 DKO mice may be compensated by other genes expressed in osteoclasts or by systemic factors that affect osteoclast function or by genes that similarly affect osteoblast function. Understanding why the p21/p27 DKO phenotype is not fully penetrant may provide important clues to deciphering communication between osteoclasts and osteoblasts during dynamic processes of bone modeling and remodeling.
What is the precise role of the CDKI genes in osteoclast terminal differentiation? Our results would argue that cell cycle exit is not the only activity of these genes necessary for osteoclast differentiation. Several genetic studies in mice provide precedence, showing that p21CIP1, p57KIP2, and p27KIP1, either alone or in combination, play direct roles in cell differentiation independent of their roles in cell cycle regulation.(15) There may be several mechanisms by which CDKIs affect cellular differentiation. In the case of the combination of p21CIP1 and p57KIP2, the effect on differentiation may be indirect, through modulation of the activity of the Rb tumor suppressor, which in turn is required for the function of basic helix-loop-helix (bHLH) transcription factors like myogenin.(16) In contrast, studies in Xenopus suggest a direct role for the p27 ortholog p27Xic−1 in both neuronal and muscle differentiation through modulation of bHLH transcription factor stability.(32,33) Furthermore, these studies suggest that different domains of p27Xic-1 are required for cell cycle withdrawal and for promoting bHLH factor stability.(32,33)
Similarly, during osteoclast differentiation, the bHLHzip protein microphthalmia-associated transcription factor (MITF) may provide a target whose activity could be modulated by CDKIs. MITF regulates target genes like TRACP and cathepsin K by binding to a 7-bp conserved sequence, TCANGTG, on the promoters of these genes.(27,28) MITF is also directly phosphorylated by p38 MAPK in bone marrow-derived cells, and phosphorylation increases MITF transactivation potential.(23) In addition, the phenotype of mice homozygous for some MITF mutant alleles, in particular the oak ridge allele, is very similar to the phenotype of p21/p27 DKO mice: severe osteopetrosis that resolves with age and reduced numbers of osteoclasts with abnormal cell morphology.(34) The hypothesis that p27KIP1 and p21CIP might regulate MITF activity during osteoclast differentiation provides a potential explanation for the action of these CDKIs that can be directly addressed by future studies.
We thank the staff at the Ohio State University Laboratory Animal Resources facilities for help with maintaining the mouse colonies. This work was supported by National Institutes of Health Grant AR-0447129 (MCO).