Krox20/EGR2 deficiency accelerates cell growth and differentiation in the monocytic lineage and decreases bone mass

The family of early growth response (EGR) genes consists of 4 members, EGR1/Krox24, EGR2/Krox20, EGR3, and EGR4.¹  Their expression is rapidly induced by serum and growth factors, such as nerve growth factor, epidermal growth factor, and platelet-derived growth factor.^2-4  The 4 EGR proteins are transcription factors that share a highly conserved DNA-binding domain composed of 3 zinc fingers, which recognize G:C-rich DNA motifs^1,5-7  in the promoters of target genes such as Hox-1.4,⁸  Hoxa-2 and Hoxb-2,⁹  thymidine kinase,^8,10  and synapsin I and II.^11,12  The transactivation activity of EGR factors is modulated by co-activators such as host cell factor C1¹³  and corepressors such as NAB1,¹⁴  NAB2,¹⁵  and Ddx20.¹⁶ 

Early studies highlighted the role of EGR genes in cell proliferation.⁴  In cancer cells of various origins, Krox20 plays a role downstream of the PTEN tumor suppressor, and its deletion is associated with cancer cell proliferation.^17,18 Krox20 knockout mice have severe abnormalities in hindbrain development^19-21  and impaired peripheral nerve myelination²²  attributable to the role of Krox20 in regulating Schwann cells growth and differentiation.^22-24  Krox20 has also been implicated in fate determination in the myeloid lineage. Specifically, it inhibits neutrophil- and activates macrophage-specific genes such as c-fms, encoding the macrophage colony-stimulating factor (M-CSF) receptor.^25,26  Krox20 interacts with PU.1, a master transcription factor critical in macrophage differentiation, to enhance transcription of c-fms.²⁶  Krox24 stimulates monocytic differentiation in myeloid cell lines and murine marrow progenitors at the expense of granulopoiesis.^27-30  Krox20 may also favor monocytic fate determination at the expense of granulopoiesis by repressing transcription of Gfi-1, a negative regulator of macrophage genes required for neutrophil cell differentiation.²⁵  Krox24 and Krox20 have redundant roles in monocytic maturation.²⁵ 

In vertebrates, bone mass and shape are determined by continuous remodeling executed by the concerted action of osteoclasts, the bone resorbing cells, and osteoblasts, the bone forming cells. Therefore, both decreased bone formation and increased bone resorption may result in bone loss. Little is known about the role of EGRs in bone. Krox24-deficient mice have low bone mass, attributable to increased expression of M-CSF in stromal cells and increased bone turnover.³¹  In addition, retinoic acid and prostaglandin E2, which stimulate osteoblastic cell proliferation and differentiation, increase the expression of Krox24.^32,33  Krox20 is largely expressed in long bone at the cartilage-trabecular bone transition line, and its absence results in failure of embryonic trabecular bone formation, which has been attributed to a differentiation block in the chondroosteoblast lineage.³⁴  Recently, we reported that Krox20 expression was up-regulated in vitro during late stages of osteoblast differentiation and was repressed by glucocorticoids.³⁵  In vivo investigation of the role of Krox20 in postnatal bone metabolism is difficult because Krox20-null mice die within days after birth, presumably due to neuronal defects.²¹  In this study, we employed the viable mice haploinsufficient for Krox20 to assess its role in postnatal bone metabolism. We demonstrate that these mice have a low bone mass (LBM) phenotype attributable to increased bone resorption and that Krox20 attenuates osteoclast growth and differentiation in a cell autonomous manner.

All reagents were purchased from Sigma-Aldrich unless otherwise specified. α-minimal essential medium and Dulbecco modified Eagle medium were purchased from Gibco, fetal bovine serum (FBS) from Hyclone, and culture plates from Corning. In splenocyte cultures, we used supernatant from cardiomyogenic 14-12 cells, containing 1.3 μg/mL M-CSF.³⁶ 

All experiments were approved by the Institutional Animal Care and Use Committee of the University of Southern California. Krox20^+/− female mice were compared with their littermate female controls. The mice were either on a pure C57BL/6 background (The Jackson Laboratory) or on a C57BL/6 × 129 background. Genotyping was performed as described previously²¹  using primers listed in Table 1. To assess bone formation, mice were injected intraperitoneally with calcein 4 and 1 day before sacrifice.³⁷ 

Femora (1 per mouse) and third lumbar vertebrae were examined as reported previously³⁷  using Scanco microcomputed tomography μCT40 (Scanco Medical AG). Briefly, scans were performed at a 20-μm resolution in all 3 spatial dimensions. The mineralized tissues were differentially segmented by a global thresholding procedure.³⁸  Trabecular parameters were measured in the secondary spongiosa of the distal femoral metaphysis and in the entire trabecular bone compartment of the third lumbar vertebrae body. Cortical parameters were determined in the mid-diaphyseal region.

Dynamic histomorphometric measurements and osteoclast counting were performed as described previously using Media Cybernetic ImagePro Plus Version 6.0 software.³⁷  The terminology and units used for these measurements were according to the convention of standardized nomenclature.³⁹ 

Both femora from 4 wild-type (WT) mice were removed, fixed in 4% paraformaldehyde, and decalcified in 10% EDTA (ethylenediaminetetraacetic acid) before paraffin embedding. Five-micrometer sections separated by 40 μm were incubated with Krox20 antibody (Covance) and labeled with Avidin-Biotin. This was followed by tartrate-resistant acid phosphatase (TRAP) staining and couterstaining with methyl green (Vector Inc). Sections were analyzed using a Zeiss Axioskop (Zeiss), with a 40× objective and a SPOT digital camera (Diagnostic Instruments). Metaphyseal TRAP-labeled cells located at least 200 μm from the growth plate were enumerated as either positive or negative for Krox20.

Blood was drawn from 6-hour fasting animals at approximately 2 PM. Levels of serum C-terminal telopeptides (CTX) were determined by enzyme-linked immunosorbent assay using RatLaps (Immunodiagnostic Systems) according to the manufacturer's instructions.

Splenocytes were extracted from 1-day-old pups, and preosteoclasts were obtained after 3 days of culture in “standard” medium (α-MEM, 10% FBS, and 1% penicillin/streptomycin) supplemented with 100 ng/mL M-CSF. For differentiation assay, adherent preosteoclasts were plated in 96-well plates with standard medium supplemented with 20 ng/mL M-CSF and 100 ng/mL receptor activator for nuclear factor κ B ligand (RANKL). On day 7, multinucleated osteoclasts were stained using a TRAP kit, and TRAP-positive surface was measured using ImageJ Version 1.41 software (National Institutes of Health). Alternatively, splenocytes were extracted, plated in 96-well plates and treated from the first day with 20 ng/mL M-CSF and 100 ng/mL RANKL for 8 days before TRAP staining. For proliferation assay, preosteoclasts were plated with standard medium supplemented with 100 ng/mL M-CSF, and cell number was assessed using the thiazolyl blue tetrazolium bromide (MTT) assay. Calvarial osteoblasts cultures were prepared from 1-day-old pups as described previously.³⁵  Cells were cultured on 12-well plates for reverse-transcription quantitative polymerase chain reaction (RT-qPCR; see next paragraph) and mineralization assays. For the latter, osteogenic medium containing ascorbic acid (50 μg/mL) and β-glycerophosphate (10mM) was initiated at confluence and alizarin red staining was performed at day 20. Mesenchymal stem cells were cultured and analyzed as described previously.³⁷ 

RNA from cells was extracted and processed as described previously.³⁷  mRNA levels were corrected for ribosomal protein L10A (rpL10A) mRNA. Primers used for PCR are listed in Table 1. For Western blot analyses, protein extraction, electrophoresis, and transfer were performed essentially as previously described,³⁵  and the Pierce Fast Western Blot kit (Thermo Scientifics) was used for antibody binding and chemiluminescence. Antibodies were purchased from Cell Signaling (anti–phospho extracellular-signal-regulated kinase [ERK]1/2 Thr202/Tyr204, anti-Akt, and anti–phospho Akt Ser473), Promega (anti-ERK1/2), and Sigma-Aldrich (anti-cFms). The mouse monoclonal anti-Tubulin antibody, developed by Dr Charles Walsh, was obtained from the Developmental Studies Hybridoma Bank under the auspices of the National Institute of Child Health and Development and The University of Iowa, Department of Biological Sciences, Iowa City, IA.

We used Sigma Mission shRNA against Krox20 (shKrox20-1, Clone 567s1c1; shKrox20-2, Clone 788s1c1) or nonspecific (NS) shRNA as control (SHC002). Lentiviral particles were generated by transfecting HEK293T cells with the plasmids carrying the shRNA sequence.⁴⁰  For silencing, transduction of M-CSF–treated splenocytes was performed at approximately 30% confluence. Medium was changed after 16 hours, and cells were allowed to recover for 48 hours before further analysis.

For cell-cycle analysis, preosteoclasts were harvested by trypsinization 16 hours after medium change and analyzed with Fluorescence-activated cell sorting (FACS) as previously described.⁴¹  To determine the monocytic cells distribution in bone marrow, 1 femur per animal was cut at the mid-diaphysis and the epiphysis removed. Then, bone marrow cells were flushed out from the distal cut end and red blood cells were lysed using a 10μM Tris-HCl buffer containing 0.74% NH₄Cl. Then, an aliquot of 1 × 10⁶ cells was stained with allophycocyanin-conjugated anti–mouse CD115 (eBiosciences). Stained cells were analyzed and sorted on a FACSCalibur machine (Becton Dickinson). For analysis of apoptosis, trypsinized cells were stained with fluorescein isothiocyanate (FITC)–conjugated annexin V and propidium iodide according to the manufacturer's instructions (Clontech). Cells positive for annexin V and/or propidium iodide were recorded by flow cytometry.

Values are expressed as mean ± SEM. Values beyond the mean ± 1.5 × SD range were excluded from the analysis. Difference between groups was defined as P < .05 using Student t test and 1-way analysis of variance when 2 and 3 groups were compared, respectively.

Although a role for Krox20 in embryonic endochondral bone formation was suggested 2 decades ago,³⁴  its potential postnatal role in the skeleton has not been determined. Here, we investigated the postnatal skeletal role of Krox20 using heterozygous mice, which display no overt phenotype and whose total body weight is similar to that of sex/age-matched WT controls. Microcomputed tomographic analysis of 4- and 8-week-old Krox20-haploinsufficient female mice revealed a LBM phenotype compared with littermate controls (Figure 1 and Table 2). The trabecular bone volume density measured in the distal femur of Krox20^+/− mice was 32% and 42% lower than the WT controls at 4 and 8 weeks of age, respectively (Table 2 and Figure 1A). No difference was observed in the femoral length, mid-diaphyseal diameter, and cortical thickness (Table 3). The trabecular bone volume density was also lower in the body of the third lumbar vertebra in 4-week-old mice (Table 2), suggesting a global positive role for Krox20 in the postnatal regulation of trabecular bone mass.

In pursuit of the cellular mechanism underlying the LBM phenotype of Krox20^+/− mice, we measured indicators of bone resorption and bone formation in vivo. First, we performed TRAP staining of histologic sections obtained from distal femoral metaphyses of 8-week-old mice. As shown in Figure 2A-B, Krox20^+/− mice had a significant 32% higher osteoclast number compared with WT controls, suggesting that the Krox20^+/− LBM phenotype is attributable to increased osteoclast activity. Further implicating bone resorption in the LBM phenotype, Krox20^+/− mice exhibited a significant 44% increase in serum CTXs, a biochemical marker of bone resorption (Figure 2C). Because osteoclasts derive from monocytes,⁴²  we used FACS to compare CD115+ monocytes in bone marrow of WT versus Krox20^+/− mice. However, the monocyte count in the mutant mice was normal (Figure 2D), suggesting that Krox20 haploinsufficiency does not exert its effect during early monocyte differentiation but rather later during the development of the osteoclast phenotype.

Because previous work with mouse embryos and cell culture models suggested a possible role for Krox20 in cells of the osteoblast lineage,^34,35,43  we also tested whether the Krox20^+/− LBM phenotype was associated with reduced bone formation. We used vital double calcein labeling to assess bone matrix apposition in vivo in the distal femur of 8-week-old mice, but, as shown in Figure 2E-G, the Krox20^+/− mice had normal mineralizing perimeter (a surrogate for osteoblast number), normal mineral apposition rate (a surrogate for osteoblast activity), and therefore normal bone formation rate. Thus, the LBM phenotype observed in Krox20-heterozygous mice is attributable to increased bone resorption rather than decreased bone formation.

We next investigated the role of Krox20 in osteoclasts and osteoblasts in vitro. M-CSF–treated splenocytes were used to study the effects of Krox20 on osteoclast growth and differentiation. The 2-fold decrease in Krox20 expression in the preosteoclast cultures derived from the heterozygous animals (Figure 3A) was associated with a significant 70% increase in proliferation rate, as determined by MTT assay (Figure 3B). We then tested whether the increased proliferation of the osteoclast precursors led to more differentiated osteoclasts. Splenocytes from WT and Krox20^+/− mice were plated and treated with M-CSF and RANKL to promote differentiation into preosteoclasts and then mature osteoclasts in the same culture. Consistent with the increased proliferation (Figure 3B), the Krox20^+/− splenocytes gave rise to significantly more osteoclasts than did their WT counterparts (Figure 3C). Next, we specifically tested the role of Krox20 in the growth and differentiation of monocytes that have already committed to the osteoclast phenotype. To this end, splenocytes were treated with M-CSF for 3 days, and equal numbers of the resulting preosteoclasts from Krox20^+/− and WT mice were replated and induced to further differentiate by the addition of RANKL. Following this protocol, we again observed significantly more TRAP⁺ osteoclasts in cultures derived from Krox20-haploinsufficient mice (Figure 3D-E). These results suggest that the in vivo Krox20^+/− LBM phenotype is attributable to increased growth and differentiation of the osteoclast precursors.

The role of Krox20 in the osteoblast lineage was investigated using osteoblast cultures derived from either newborn mouse calvariae or adult bone marrow. As in preosteoclasts, Krox20 expression was reduced, by as much as 3.5-fold, in osteoblast cultures derived from calvariae of Krox20^+/− compared with Krox20^+/+ newborn mice (Figure 3F). However, this decrease was not associated with compromised mineralization as assessed by alizarin red staining of day-20 calvarial osteoblast cultures (Figure 3G-H). In mesenchymal stem cell cultures derived from the bone marrow of Krox20 heterozygous mice, we again did not observe a decrease in the potential for bone-like tissue formation. In fact, the number of osteoblastic colony-forming units (CFU-Ob) was surprisingly increased in Krox20^+/− cultures (Figure 3I). Further cell proliferation in the osteoblast lineage, assessed by the average size of the osteoblastic colonies, was unaffected by the Krox20 genotype (Figure 3J). Thus, consistent with the in vivo data (Figure 2), the results of our ex vivo and culture assays indicate that the Krox20^+/− LBM phenotype is attributable to increased bone resorption and not to decreased bone formation.

To determine the mechanism by which Krox20 attenuates preosteoclasts proliferation, we initially profiled Krox20^+/+, Krox20^+/−, and Krox20^−/− cells for their distribution among the phases of the cell cycle. As shown in Figure 4A, reduction in the Krox20 gene dosage was associated with a progressive, up to 3-fold, increase in the percentage of cells in the S and G2/M phases, indicating a negative control of the preosteoclast cell cycle by Krox20. The role of Krox20 in preosteoclast survival was then assessed based on annexin V staining and propidium iodide retention. As shown in Figure 4B, reduction in the Krox20 gene dosage had a mild positive effect on preosteoclast survival, attributable to decreased apoptosis, measured by annexin V staining. Taken together, our results indicate that Krox20 attenuates preosteoclast proliferation primarily by restraining cell-cycle progression.

Although osteoclast growth and differentiation was increased in the relatively pure splenocyte cultures from Krox20-deficient mice (Figure 3B-D), these results do not necessarily indicate a cell autonomous role of Krox20 because the splenocytes were extracted from mice that were Krox20-compromised from conception and in many cell types, including, for example, spleen stromal cells and myeloid early precursors. To conclusively address the potential cell autonomous effect of Krox20 in osteoclasts, we first confirmed the expression of Krox20 in osteoclasts in vivo. Histologic sections of mature bone were subjected to TRAP staining combined with Krox20 immunohistochemistry as described in “Methods” (Figure 4C), and 263 TRAP⁺ cells were scored as either Krox20⁺ or Krox20⁻. We found that 94% of the osteoclasts (248 cells) expressed Krox20, indicating that Krox20 can have a cell autonomous role in osteoclasts in vivo.

To test whether the antiproliferative effect of Krox20 in preosteoclasts is truly cell autonomous, we examined the effects of Krox20 silencing in vitro. WT preosteoclasts were transduced with lentiviruses encoding a shRNA against Krox20. To rule out off-target effects, we used 2 different shRNA sequences, shKrox20-1 and shKrox20-2, which reduced Krox20 mRNA levels by 75% and 81%, respectively, compared with cells transduced with a lentivirus encoding an NS shRNA (Figure 4D). As shown in Figure 4E, Krox20 silencing by shKrox20-1 and shKrox20-2 resulted in 39% and 74% increase in preosteoclast proliferation, respectively, indicating that the antiproliferative effect of Krox20 does not require involvement of other cells. Consistent with the cell-cycle profiles of Krox20^+/+, Krox20^+/−, and Krox20^−/− preosteoclasts (Figure 4A), the accelerated proliferation of the shKrox20-transduced cells was associated with a 60% increase in the percentage of cells in the S and G2/M phases of the cell cycle (Figure 4E-F). However, Krox20 silencing did not affect cell survival (Figure 4G). Because the genetic approach suggested only a minor role for Krox20 in preosteoclast apoptosis (Figure 4B), this and the silencing approaches collectively support the conclusion that Krox20 inhibits cell proliferation primarily by attenuating preosteoclast cell-cycle progression in a cell autonomous manner.

Preosteoclast proliferation is tightly regulated by binding of M-CSF to its cFms receptor and the subsequent activation of the phosphatidylinositol 3-kinase/Akt and MEK/ERK pathways.⁴⁴  We tested the hypothesis that Krox20 restrains cFms signaling in preosteoclasts by measuring cFms expression as well as M-CSF–mediated activation of the phosphatidylinositol 3-kinase/Akt and the MEK/ERK pathways in shKrox20-transduced preosteoclasts. As shown in Figure 5A-B, Krox20 silencing increased cFms expression at both the mRNA and protein levels. Furthermore, analysis of Akt and ERK phosphorylation after M-CSF treatment demonstrated an augmented response in the Krox20-depleted cells (Figure 5C). These findings suggest that Krox20 attenuates the proliferation of preosteoclasts by down-regulating their cFms levels and thus blunting their response to M-CSF.

For the first time, this work establishes a postnatal role for Krox20 in bone metabolism. Krox20-haploinsufficiency in preosteoclasts results in accelerated proliferation and cell-cycle progression, likely due to increased cFms expression and signaling. This leads to increased number of mature osteoclasts both in vitro and in vivo. The resulting acceleration of bone resorption underlies the observed LBM phenotype.

By in large, macrophages and preosteoclasts are closely related as they both differentiate from monocytes after M-CSF stimulation. Cellular events triggered by RANKL can then induce M-CSF–stimulated monocytes to further differentiate into mature osteoclasts. Here, we found that the number of bone marrow monocytes was not affected by the Krox20 gene dosage (Figure 2D). Krox20 is also not critical for M-CSF–related differentiation of liver monocytes into macrophages,⁴⁵  likely because of the redundant role of Krox24 in monocyte maturation and macrophage differentiation.²⁵  In addition, our data suggest that Krox20 does not play a significant role in terminal osteoclast differentiation, as the overall increase in osteoclastogenesis in Krox20^+/− preosteoclasts cultures did not exceed the increase in preosteoclast proliferation (Figure 3B-E). Therefore, our findings imply that the LBM phenotype observed in Krox20-haploinsufficient mice results from enhanced growth and differentiation of osteoclast precursors.

Unlike the inhibitory effect of Krox20 on the preosteoclast cell cycle, its role in apoptosis is more complex. We observed a mild proapoptotic effect of Krox20 by comparing preosteoclasts from Krox20^+/+, Krox20^+/−, and Krox20^−/− mice. This effect is probably indirect because apoptosis was not affected by Krox20 silencing in isolated WT cells. Interestingly, unlike our preosteoclast cultures, Bradley et al⁴⁶  reported a prosurvival role for Krox20 in mature osteoclasts in vitro. Be that as it may, the increased proliferation of Krox20-compromised preosteoclasts in vitro (Figures 3B and 4E) and the increased osteoclast number and bone resorption in Krox20^+/− mice in vivo (Figure 2A-C) demonstrate that inhibition of cell-cycle progression (Figure 4A,F) is the overall predominant effect of Krox20 in the osteoclast lineage.

The present study demonstrates a major physiologic role for osteoclastic Krox20 in postnatal bone mass control. However, we cannot rule out a role for Krox20 in cells of the osteoblast lineage as well. In fact, we report here that Krox20 haploinsufficiency increases bone marrow–derived CFU-Ob (Figure 3I). This could reflect a bona fide effect of Krox20 on early osteoblast progenitors, similar to its effect on osteoclast progenitors. However, such a putative effect does not result in increased net bone formation in Krox20^+/− mice (Figure 2E-G). On the other hand, because preosteoblasts significantly contribute to the bone marrow M-CSF pool, the increased CFU-Ob may contribute to the elevated osteoclastogenesis observed in the Krox20-compromised animals. Indeed, although RT-qPCR analysis of Krox20^+/− newborn calvarial osteoblast cultures demonstrated normal M-CSF expression (0.82 ± 0.27 vs 1.00 ± 0.25, respectively; n = 5; P = .64), total M-CSF expression was elevated in RNA extracts from Krox20^+/− compared with WT femora (2.07 ± 0.20 vs 1.00 ± 0.18, respectively, n = 5, P < .05). Furthermore, any increase in osteoblast-derived M-CSF in Krox20^+/− bones is expected to be amplified due to the cell-autonomous increase of cFms signaling in the Krox20^+/− preosteoclasts.

In contrast to our findings in preosteoclasts, Krox20 operates in concert with PU.1 to enhance cFms expression in myeloid progenitors.^25,26  Conceivably, the Krox20/PU.1-binding element at the cFms gene switches its function from a transcriptional enhancer to a transcriptional repressor as monocytes differentiate to preosteoclasts. Such switch can occur as the cellular milieu changes, resulting in replacement of Krox20 co-activators with corepressors.^15,25 

In summary, we established Krox20 as a significant antiproliferative transcription factor in preosteoclasts. Decreased levels of Krox20 led to increased osteoclast number, increased bone resorption, and low bone mass. Thus, maintenance or stimulation of Krox20 expression in preosteoclasts may present a viable therapeutic strategy for high-turnover osteoporosis.

We thank Dr Steve Teitelbaum and Dr Jennifer Reeve (Washington University, St Louis, MO) for sharing their expertise and for the generous gifts of RANKL and the M-CSF–containing conditioned medium. We also thank Dr Debbie Johnson and Dr Sandra Johnson (University of Southern California) for sharing their expertise and for the generous gifts of antibodies.

This work was supported by grant RO1 AR047052 from NIH/NIAMS. Y.G. was partly supported by a Meyer Young Investigator Fellowship and S.K.B. by a postdoctoral Innovative Chapter Research Award, both from the Arthritis Foundation Southern California Chapter. B.F. is the holder of the J. Harold and Edna L. LaBriola Chair in Genetic Orthopaedic Research at the University of Southern California.

National Institutes of Health

Contribution: Y.G., N.L., and B.F. designed research; Y.G., S.K.B., N.L., Y.S., A.E.K.-G., J.C., A.D., M.B., and L.G. performed research; S.K.B. contributed new reagents/analytic tools; Y.G., S.K.B., N.L., J.E.T., and B.F. analyzed data; and Y.G. and B.F. wrote the paper.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Baruch Frenkel, DMD, PhD, Institute for Genetic Medicine, University of Southern California Keck School of Medicine, 2250 Alcazar St, IGM/CSC-240, Los Angeles, CA 90033; e-mail: frenkel@usc.edu; or Yankel Gabet, DMD, PhD, Institute for Genetic Medicine, University of Southern California Keck School of Medicine, 2250 Alcazar St, IGM/CSC-240, Los Angeles, CA 90033; e-mail: gabet@usc.edu.