PTH regulates osteogenesis and suppresses adipogenesis through Zfp467 in a feed-forward, PTH1R-cyclic AMP-dependent manner

Conditional deletion of the PTH1R in mesenchymal progenitors reduces osteoblast differentiation, enhances marrow adipogenesis, and increases zinc finger protein 467 (Zfp467) expression. In contrast, genetic loss of Zfp467 increased Pth1r expression and shifts mesenchymal progenitor cell fate toward osteogenesis and higher bone mass. PTH1R and ZFP467 could constitute a feedback loop that facilitates PTH-induced osteogenesis and that conditional deletion of Zfp467 in osteogenic precursors would lead to high bone mass in mice. Prrx1Cre; Zfp467fl/fl but not AdipoqCre; Zfp467fl/fl mice exhibit high bone mass and greater osteogenic differentiation similar to the Zfp467-/- mice. qPCR results revealed that PTH suppressed Zfp467 expression primarily via the cyclic AMP/PKA pathway. Not surprisingly, PKA activation inhibited the expression of Zfp467 and gene silencing of Pth1r caused an increase in Zfp467 mRNA transcription. Dual fluorescence reporter assays and confocal immunofluorescence demonstrated that genetic deletion of Zfp467 resulted in higher nuclear translocation of NFκB1 that binds to the P2 promoter of the Pth1r and increased its transcription. As expected, Zfp467-/- cells had enhanced production of cyclic AMP and increased glycolysis in response to exogenous PTH. Additionally, the osteogenic response to PTH was also enhanced in Zfp467-/- COBs, and the pro-osteogenic effect of Zfp467 deletion was blocked by gene silencing of Pth1r or a PKA inhibitor. In conclusion, our findings suggest that loss or PTH1R-mediated repression of Zfp467 results in a pathway that increases Pth1r transcription via NFκB1 and thus cellular responsiveness to PTH/PTHrP, ultimately leading to enhanced bone formation.


9
almost evenly distributed in the cytoplasm and nucleus of +/+ cells but underwent partial 176 translocation to the nucleus in Zfp467 -/-cells (Fig. 4B). Zfp467 -/-cells also showed much higher 177 nuclear protein level of both p50 and GATA1 (Fig. 4C, D). 178 179 p50-Relb heterodimer may drive greater Pth1r transcription in Zfp467-/-cells 180 In order to confirm whether p50 or GATA1 could activate the specific P1 or P2-2 Pth1r promoter, 181 we co-transfected MC3T3-E1 cells with P1 and P2-2 dual-luciferase reporter and GATA1, p50 182 overexpression plasmids. Only the p50 overexpression group was able to significantly activate the 183 P2-2 promoter (Fig. 5A). ChIP results showed that the DNA was properly sheared and IP was 184 successfully conducted (Fig. 5B). ChIP-qPCR results showed that the first two parts of P2 were 185 properly enriched in our IP product (>0.5%) (Fig.5-figure supplement A), and the first part of P2 186 was approximately 20 fold more highly enriched in our p50 IP product than IgG (Fig. 5C, Fig.5-187 figure supplement B); this indicated that p50 binds to the P2 promoter, especially at the first 200 188 bp site. Subsequently, we treated COBs and BMSCs with p50 siRNA and found that p50 knock-189 down could significantly inhibit the expression of Pth1r in both +/+ and -/- COBs and BMSCs. 190 Importantly, p50 knock-down in -/-cells reverts the levels of Pth1r to the levels seen in +/+ cells 191 ( Fig. 5D-F). 192 In order to determine whether p50 could bind to the Pth1r P2-2 promoter directly, we performed 193 DNA pulldown assay using biotin-labeled Pth1r P2 promoter as a probe. As shown in Fig. 6A, the 194 biotin-Pth1rP2 group showed a specific band in both MC3T3-E1 nuclear extracts and purified p50 195 protein, suggesting a direct physical interaction between p50 and Pth1r P2 promoter. However, we 196 noticed that p50 does not have a transcriptional activation domain, so p50 must heterodimerize 197 10 with other transcription factors in order to increase gene transcription. Using String database and 198 checking published studies, we found 8 candidate that might heterodimerize with p50 to regulate 199 gene transcription: NFYC, NPAS1, Rel, AKAP8, RelA, RelB, ANKRD42 and HDAC1. Using 200 siRNA, we knocked down all these potential p50 partners (Fig. 6B), but the upregulated Pth1r 201 induced by p50 overexpression could only be dampened by Npas1 and Relb siRNA (Fig. 6C). 202 Further co-immunoprecipitation results confirmed that p50 could heterodimerize with RelB only 203 (Fig. 6D), which suggested p50-Relb heterodimer may drive greater Pth1r transcription in 204 Zfp467-/-cells. 205 206

Zfp467 -/-cells have increased PTH signaling and higher extracellular acidification rates 207 (ECAR) 208
As the PKA pathway is considered to be one of the major downstream pathways of the PTH 209 signaling network, it would be important to know whether -/-cells have higher PKA activation 210 levels due to the higher expression of PTH1R. PTH increased cAMP within 10 minutes ,measured 211 by ELISA in +/+ COBs, and -/-cells had a higher levels of cAMP expression than +/+ cells 212 (p=0.0007 for vehicle, p<0.0001 for 50nM, p=0.0008 for 100nM). In addition, -/-BMSCs also had 213 significantly higher levels of intracellular cAMP after 10-60 min exposure of PTH (p<0.0001 for 214 10 min-100nM, p=0.0341 for 30 min-100nM, p<0.0001 for 60 min-100nM) (Fig. 7A, B). 215 Importantly, -/-COBs and BMSCs showed a greater magnitude of increase of cAMP after PTH 216 treatment (p=0.0053 for interaction in COBs, p=0.0010 for interaction in BMSCs), which resulted 217 from higher PTH1R in -/-cells. Additionally, as CREB was one of the major downstream targets 218 of PKA, we found higher phosphate ratio of CREB in -/-COBs and BMSCs whereas the total 219 11 protein level of CREB showed no difference (Fig. 7C, D). 220

221
In a previous study, PTH was shown to enhance aerobic glycolysis, which is a major source of 222 ATP for osteoblast differentiation(16). We measured the oxygen consumption and extracellular 223 acidification in both COBs and BMSCs pre-osteogenic differentiation and 3 days after osteogenic 224 differentiation. Cellular respiration measurements showed that  BMSCs had significantly increased ECAR, but no difference was found for pre-differentiated 226 COBs in respect to either ECAR or oxygen consumption rate (OCR) (Fig. 7E, F). However, both -227 /-COBs and BMSCs cells had significantly higher ECAR levels after 3 days of osteogenic 228 differentiation, although no difference was found regarding OCR between genotypes (Fig. 7G, H). 229 These data suggest that -/-COBs or BMSCs may have higher glycolysis as a result of higher 230 PTH1R expression. 231 232 Zfp467-/-cells showed increased sensitivity to PTH and enhanced pro-osteogenic as well as 233

anti-adipogenic effects 234
To determine the sensitivity of Zfp467-/-cells to PTH, we treated COBs with osteogenic 235 differentiation media and PTH for 7 or 14 days simultaneously. A dose response to PTH led to a 236 significant increase in ALP staining in -/-COBs, and in +/+ cells. Furthermore, 100nM PTH in -/-237 cells produced remarkably higher positive-stained cells than vehicle group (p=0.0001) while there 238 was no statistical significance seen among +/+ groups (p=0.6536) (Fig. 8A, B). Importantly, -/-239 COBs showed a more significant response to PTH regarding ALP staining (p=0.0201) (Fig. 8B). 240 Alizarin red staining showed a parallel trend as ALP staining; an increase in PTH dose resulted in 12 an increase in mineralization for -/-COBs only (Fig. 8A). 242 In osteogenic media, calvarial progenitors can differentiate into adipocytes. We noted significantly 243 less lipid droplets in -/-adipocytes cultured in osteogenic media compared to untreated +/+ 244 samples (Fig. 8C). Additionally, a reduction in adipogenesis was observed with PTH exposure at 245 50nM and 100nM in both +/+ and -/-cells, and -/-cells showed much less lipid droplet formation 246 compared to +/+ group (Fig. 8C, D). However, the magnitude of decrease after PTH treatment is 247 almost identical in +/+ and -/-cells (Fig. 8D). These results were then confirmed by qRT-PCR 248 after 7 days osteogenic differentiation which indicated higher expression of osteogenic 249 differentiation related genes such as Alp, Sp7, Rankl and Igf-1 in -/-compared to +/+ COBs with 250 PTH exposure (Fig. 8E, F). It is noteworthy that there is a statistically significant interaction 251 between genotype and PTH treatment, suggesting an increased sensitivity to PTH in -/-cells. To determine whether the increase in osteogenic differentiation seen in Zfp467-/-cells is due to 256 higher PTH1r levels, we knocked down Pth1r by siRNA and found that Pth1r knock down led to 257 decreased osteogenic differentiation in both +/+ and -/-COBs. Although -/-COBs still show 258 slightly higher Alp staining (Fig. 9A), Pth1r knock down in -/-cells dampens the increase in Alp 259 and Sp7 gene expression during osteogenic differentiation compared to +/+ cells, indicating that 260 this increase is associated with the up-regulation of Pth1r seen in -/-cells (Fig. 9B). 261 Furthermore, a PKA inhibitor could also significantly decrease osteogenic differentiation in COBs 262 (Fig. 9C). The staining results showed ALP activity was much higher in -/-cells in the absence of 13 the PKA inhibitor, but no difference was found with the PKA inhibitor between genotypes (Fig.  264   9C). In addition, we observed that with the PKA inhibitor, upregulated osteogenic genes in -/-265 cells including Alp and Sp7 could be totally reversed (Fig. 9C). Similarly, treating COBs with the 266 PKA inhibitor during PTH treatment simultaneously for 7 days led to a suppression of osteogenic 267 differentiation in both +/+ and -/-cells ( Fig. 9-figure supplement). qPCR results showed that 268 PKA inhibitor could totally reverse the upregulation of Alp, Sp7 and Rankl in PTH treated -/-cells 269 ( Fig. 9-figure supplement A, B). We then next asked what were the downstream targets of Zfp467. In an unbiased analysis, we used 273 RNAseq in pre-and post-differentiated calvarial osteoblasts from +/+ and Zfp 467 -/-cells to 274 assess potential regulatory pathways and differentially expressed genes ( Fig. 10A-D). The PI 3-K 275 and MAPK signaling pathways were differentially up regulated in Zfp467-/-cells whether pre-or 276 post-differentiated when compared to WT (Fig. 10D). There were several highly expressed genes 277 in the -/-cells related to osteogenesis, including Wdfy1, Sox10, and Ngfr ( Fig. 10B) These results 278 were confirmed by qRT-PCR (Fig. 10E). When Zfp 467 was over-expressed in MC3T3-E1 cells 279 those three genes were significantly suppressed relative to Gfp overexpression. In this paper we characterize the role of Zfp467 in the molecular, cellular and biochemical 283 responses to PTH in osteoblast progenitors. Based on our data, we propose a novel pathway 284 by which PTH can enhance its own anabolic actions in osteoblast progenitors by repressing 285 Zfp467, a negative regulator of PTH1R expression. Zfp467 was originally isolated from 286 mouse hematogenic endothelial LO cells as an OSM-inducible mRNA, which encodes protein 287 ZFP467(17), also called EZI. Zinc finger motifs are involved in protein-protein interactions 288 and protein-DNA binding (10,18,19). Based on the zinc finger domain, zinc finger proteins are 289 classified into C2H2, C3H and C4(20). ZFP467 belongs to C2H2 zinc finger protein family 290 whose zinc finger domain consists of two cysteine and two histidine residues. ZFP467 was 291 initially found to cooperate with STAT3 and augment STAT3 activity by enhancing its nuclear 292 translocation in a kidney cell line(17). ZFP467 is expressed ubiquitously and may be an 293 important mediator of bone marrow cell differentiation into the adipogenic or osteogenic 294 lineages as well as functioning in other tissues in distinct ways (12,13,21). 295 In our previous work we found that deletion of pth1r in mesenchymal cells resulted in low 296 bone mass, high marrow adiposity, and upregulation of Zfp467(8). In contrast, the absence of 297 Zfp467 resulted in a significant increase in trabecular bone volume fraction, accompanied by 298 a marked reduction in peripheral and marrow adipose tissue and improved glucose 299 tolerance(14). Most importantly for the current study, whole bone marrow gene expression by 300 qRT-PCR revealed a ~40% increase in Pth1r expression and greater protein levels in the 301 Zfp467 -/-mice compared to controls(14). Therefore, we hypothesized that PTH1R and 302 ZFP467 could be involved in a feedback loop whereby the suppression of Zfp467 mediated by 303 PTH leads to an increase of PTH1R, and therefore an enhanced response to PTH treatment. 304 However, due to the global nature of the Zfp 467 deletion, we could not exclude a cell non-305 autonomous effect. In the present study therefore we tested that hypothesis both in vivo and 306 in vitro, and sought to determine the cellular and biochemical mechanisms involved in this 307 15 novel regulatory pathway. 308 First, we were able to show that conditional deletion of Zfp 467 by PrrxCre led to an increase 309 in osteogenesis and bone formation, a finding that mirrors our results with the global deletion 310 of Zfp 467. On the other hand, using an AdipoCre driven system, we found no effect of 311 conditional deletion in adipocytes on body composition, fat mass, bone mass or total weight. 312 Taken together these data would suggest that Zfp467 is an early mesenchymal transcriptional 313 factor that regulates lineage allocation in early osteoblast but not adipocyte progenitors. Second, in 314 this report, we found constitutive up-regulation of Zfp467 when we genetically knocked down 315 Pth1r in calvarial osteoblasts and bone marrow stromal cells. In addition, much like the study of 316 Quach et al. (12), we noted that acute PTH treatment could significantly suppress gene expression 317 of Zfp467 in both COBs and BMSCs, and the suppression could be partly rescued by both a PKA 318 pathway inhibitor and a PKC inhibitor (12). Similarly, Forskolin, a PKA pathway activator could 319 also inhibit the expression of Zfp467, with a more sustained effect. These data indicated that PTH 320 might suppress Zfp467 expression via activation of the PTH1R through predominantly PKA 321

pathways. 322
To test the validity of our hypothesis, we investigated how deletion of Zfp467 affected the 323 expression of the PTH1R. We first examined which transcript of Pth1r was upregulated and 324 further confirmed that via dual-fluorescence reporter assay that both P1 and P2 promoters of Pth1r 325 were activated in Zfp467 null cells; however P2 was more activated than P1. Using three different 326 transcription factor prediction databases including PROMO, JASPAR and Animal TFDB, we 327 found several candidate transcription factors that might be involved in the regulation of Pth1r in 328 Zfp467 -/-cells via activation of the P2 promoter. After overexpressing or knocking down each 329 16 candidate transcription factor, we found that only p50 could up-regulate Pth1r via activation of the 330 P2-2 Pth1r promoter. Further confocal immuno-florescence and nuclear protein detection 331 indicated that the nuclear translocation of p50 was much higher in Zfp467-/-cells. Moreover, 332 ChIP-qPCR results showed that p50 could bind to the P2 promoter of the Pth1r. Taken together, 333 these data suggested that the deletion of Zfp467 resulted in higher nuclear translocation of p50 334 which bound to the P2 promoter of Pth1r and promoted its transcription. 335 p50 is one of the DNA binding subunits of the NF-kappa-B (NFκB) protein complex. NFκB is a 336 transcriptional regulator that is activated by various stimuli including cytokines, bacteria and 337 oxidation. The NFκB pathway is involved in several biological processes including inflammation, 338 bone resorption, aging and cancer (22). Activated NFκB translocates into the nucleus and 339 stimulates the expression of an array of genes. However, we found no previous studies that p50 340 may regulate the gene expression of Pth1r or other osteogenic related genes. Nevertheless, it is 341 clear that p50 is involved in osteoclastogenesis and bone resorption, hence it plays a significant 342 role in bone remodeling. It is also conceivable that p50 could associate with histone deacetylase-1 343 (HDAC1) or be regulated by a PKA catalytic subunit which is also downstream of PTH signaling 344 (23). Moreover, it is likely that other proteins like RelB that might bind to p50 and enhance its 345 effect on the transcriptional regulation of Pth1r play a critical role in PTH mediated osteogenesis. 346 Osteoblasts require substrates for energy utilization during collagen synthesis and mineralization. 347 Previous reports by Lee et al. and Guntur et al. demonstrated that glycolysis is a major source of 348 ATP for differentiating osteoblasts(9,16). Consistently, we found that Zfp467 -/-cells showed 349 higher cAMP levels in response to PTH as well as higher glycolytic activity (i.e. ECAR). COBs 350 from Zfp 467-/-mice also showed greater differentiation in osteogenic media but less 351 17 differentiation into adipocytes compared to controls. Higher rates of osteogenesis, enhanced Rankl 352 and Igf1 expression, increased glycolysis and decreased adipogenesis are likely related to greater 353 activation of the PTH1R, possibly due to its higher endogenous expression level. 354 Since it was reported that the downstream effects of PTH could be partially blocked by PKA 355 inhibitors (24), we used Pth1r siRNA and PKA inhibitors to block the effect of PTH1R. We 356 confirmed that PTH1R and its downstream pathways were involved in the deletion of Zfp467  Despite uncovering a novel regulatory circuit through p50, we recognize there are some 371 limitations to our study. First, it should be noted that we also found that Zfp467-/-cells have 372 higher nuclear translocation of GATA1 and overexpression of GATA1 could promote Pth1r 373 18 expression. However overexpression of GATA1 failed to activate the P2-2 promoter of Pth1r. 374 Hence, we believe that GATA1 might be another of the regulating transcription factors between 375 ZFP467 and Pth1r, but it likely regulates Pth1r expression via binding sites other than those 376 present on P2-2. 377 Second, we have not determined how Zfp467 binds with p50-RelB heterodimer, and we have not 378 identified the exact genomic sequence required for p50-RelB heterodimer binding at the P2 379 promoter binding site. Further studies will be needed to address this important mechanistic 380 question. Hence the precise mechanism of action of Zfp467, and the downstream consequences of 381 p50-RelB heterodimer binding on the presumed switch between adipocytes and osteoblasts is still 382 not fully defined. 383 Last, to our knowledge there must be a break on the activation of a feed forward system, but we 384 have to acknowledge that we have not delineated whereby the fast forward system interacts with 385 β-Arrestin or any other turns off signaling. 386 In summary, taken together, we demonstrated the importance of the zinc finger protein ZFP467 for 387 lineage allocation in vitro and in vivo, as well as responsiveness to PTH in osteoblast progenitors. 388 Our data also support a novel feed-forward regulatory loop whereby suppression of Zfp467 389 mediated by PTH and its downstream PKA pathway leads to an increase of PTH1R via p50, and 390 subsequent enhanced responsiveness to PTH treatment. These findings have significant 391 implications for our understanding of the anabolic effects of PTH on bone. Zfp467 fl/fl on the C57BL/6 J background was generated by Cyagen ( Fig. 1 supplement 1). Exons 407 2~4 were selected as conditional knockout region (Transcript: Zfp467-001 408 ENSMUST00000114561). The targeting vector, homology arms and cKO region were generated 409 by PCR using Bacterial Artificial Chromosome (BAC) clone RP24-144J8 and RP23-24K23 from 410 the C57BL/6J library as template. To generate mice lacking Zfp467 in limb mesenchymal stem 411 cells, Prrx1Cre; Zfp467 fl/fl were generated by crossing Prrx1Cre transgenic mice to Zfp467 fl/fl mice 412 (Prx-Cre Zfp467 fl/fl ) and · · · · · · · · · · · · · · · · · · · · · · · · · fl/fl mice were used as controls. To generate mice lacking Zfp467 in

Committee. 424
No statistical methods were used to predetermine sample size. Mice label and measurements were 425 performed by two independent researchers and selected at random (by cage) into following 426 experiments. Masking was used during data collection and data analysis. Animals with ulcerative 427 dermatitis or other diseases were excluded from the study. 428

Dual-energy X-ray Absorptiometry (DXA) 429
Whole body composition evaluation of the head was performed using the PIXImus densitometer 430 (GE-Lunar, Fairfield, CT, USA). The PIXImus was calibrated daily with a phantom provided by 431 the manufacturer. 432

Brüttisellen, 435
Switzerland) was used to assess the trabecular and cortical bone microarchitecture, volume and 436 mineral density in mouse femurs. Scans were acquired using a 10.5 μm 3 isotropic voxel size, 70 437 kVp peak x-ray tube intensity, 114 mA xray tube current, 250 ms integration time, and were 438 21 subjected to Gaussian filtration and segmentation. All scans were analyzed using manufacturer

Marrow adipose tissue quantification by osmium tetroxide staining and μCT 452
At the time of sacrifice, tibiae were isolated and placed into 10% neutral buffered formalin 453 overnight at 4°C. Soft tissue was carefully removed to ensure that the fibula remained intact and 454 the bones were washed under continuous cold PBS for one hour, then stored in PBS at 4°C. 455 Quantification and visualization of marrow adipose tissue was performed as described 456 previously(32). Briefly, bones were decalcified in 14% EDTA (pH 7.4) for 14 days, with EDTA 457 changes every 3-4 days. Bones were then washed for 10 minutes in PBS (3 times) and stained with 458 a 1:1 mixture of 2% aqueous osmium tetroxide (cat# 23310-10, Polysciences, Inc., Warrington, 459 PA, USA) and 5% potassium dichromate for 48 hours. Stained bones were then washed with PBS 460 22 (pH 7.4) for 5 hours (3 times), and subsequently scanned by μCT. Bone marrow adipose tissue 461 (BMAT) content was calculated by determining the whole volume of second osseous center of 462 tibiae. 463

Primary cells isolation 464
The generation of Zfp467-/-and wild-type mice was previously described. Calvaria osteoblasts 465 (COBs) were isolated from calvarias of 3-5-day old +/+ and -/-neonates as described in the Cytoplasm and nuclear protein were extracted using Nuclear and Cytoplasmic Extraction Reagent 489 Kit (Thermo Scientific, MA, US). Antibodies used for western blot and Chromatin 490 Immunoprecipitation (ChIP) or co-Immunoprecipitation (co-IP) were listed in Table 1  DNA fragments were used as templates for PCR amplification. ChIP-PCR primers sequences were 516 listed in Table 2. PCR products were used for nucleic acid electrophoresis to avoid unspecific 517 25 amplification. 518 DNA pulldown assay was performed according to previously reported protocol (36). 5' biotin-519 modified Pth1r P2-2 dsDNA probe was generated using PCR with oligonucleotide primers 520 modified at its 5' end by Sangon Biotech (Shanghai, China). Nuclear protein were extracted using 521 Nuclear and Cytoplasmic Extraction Reagent Kit (Thermo Scientific, MA, US). p50 purified 522 protein was purchased from Proteintech (Rosemont, US). Briefly, the Pth1r P2-2 or oligo probe 523 was incubated with streptavidin-coupled Dynabeads (Invitrogen, US) at room temperature for 1 524 hour to generate probe-bound Dynabeads, and then the probe-bound Dynabeads were incubated 525 with MC3T3-E1 nuclear extracts or p50 purified protein at 4 °C overnight. The protein bound to 526 the probe and beads were eluted and used for gel electrophoresis.  Carlsbad, CA) was used for normalization. 546

Osteogenic differentiation and related measurements 547
In vitro osteoblast differentiation and measurement were done according to previously published 548 protocols(14). Briefly, BMSCs and COBs were plated at a density of 1x10 6 /well in 6-well plates. 549 When cells reached around 80% confluency, osteogenesis induction began using osteogenic 550 induction media which consisted of complete αMEM (MEM, 5% fetal bovine serum, and 1% 551 penicillin/streptomycin), 50ug/mL ascorbic acid, and 8mM beta-glycerophosphate (both were 552 27 purchased from Sigma, St. Louis, MO)]. MEM and penicillin/streptomycin were purchased from 553 Life Technologies (Carlsbad, CA), FBS was purchased from VWR (Radmor, PA). Medium was 554 changed every other day until cells were ready to stain for alkaline phosphatase (ALP) and alizarin 555 red (ARS) to assess osteoblasts and mineralization, respectively, around d7 and d14 after 556 differentiation. Additionally, RNA was extracted on day 7 after differentiation for real time PCR 557 analysis. 558

Alkaline Phosphatase and Alizarin Red Staining 559
Alkaline phosphatase staining was performed using ALP kit obtained from Sigma (St. Louis, MO) 560 according to the manufacturer's instructions at d4 and d7 after osteogenic differentiation for 561 BMSCs and COBs, respectively. ARS (Sigma, St. Louis) staining was done using 1% solution at 562 pH 4.2 at d14 after osteogenic differentiation for both BMSCs and COBs. Briefly, after fixation, 563 cells were stained for 30 minutes with ARS solution at room temperature. Cells were then washed 564 a couple of times with water before they were visualized under the microscope (Leica DM IRB, 565 TV camera). Five randomly fields were chosen to capture images using the ZEISS Efficient 566 Navigation, blue edition camera (Bloomfield, CT) per treatment group for quantification. 567

Oil Red O (ORO) staining 568
Differentiated adipocytes were stained using ORO solution at d9 after adipogenic diff. Briefly, 569 cells were fixed with 10% neutral buffer formalin (Sigma, St. Louis). After that, they were washed 570 with 60% isopropanol (Sigma, St. Louis, MO) before stained with ORO solution (Sigma, St. 571 Louis, MO) for 15 minutes at room temperature. Cells were then washed a couple of times with 572 water before they were visualized and pictures were taking using the Zeiss microscope. For 573 quantification, lipid from adipocyte droplets was extracted using isopropanol and the absorbance 574 28 was measured at 490 nm using a plate reader (MRX Dynex Technologies, Chantilly, VA). 575

RNA-seq and Analysis 576
RNA-seq was performed on samples from COBs isolated from Zfp467+/+ and -/-mice, pre-or 577 after-4 days' osteogenic differentiation. RNA-seq analysis was performed by BGI Group (UK, 578 China). Samples were de-identified, and the analysis was blinded to assignment. We employed 579 DAVID 6.8 to perform the functional annotation (Cellular Component [CC]) and GSEA 580 enrichment plots for related pathways. Protein coding genes were assessed with p<0.05, fold 581 change > 2.0 or < -2.0. The enriched CC was evaluated using a false discovery rate (FDR, 582 Benjamini-Hochberg method) of 0.1 and DEG number. We have uploaded our original 583 sequencing data to Sequence Read Archive database (PRJNA877934, 584 http://www.ncbi.nlm.nih.gov/bioproject/877934). 585

Statistical Analysis 586
All data are expressed as the mean ± standard deviation (SD) unless otherwise noted. Results were 587 analyzed for statistically differences using Student's t-test between two groups or 2-way ANOVA 588 followed by Bonferroni's multiple comparison post hoc test among three or more groups where 589 appropriate. All statistics were performed with Prism GraphPad 7.0 statistical software (GraphPad 590 Software, Inc., La Jolla, CA). Values of p<0.05 were considered statistically different. 591 conditional knockout region. The targeting vector was generated by PCR using BAC clone RP24-728 144J8 and RP23-24K23 from the C57BL/6J library as template. 729 shown as mean ± SD by unpaired Student's t test, n=7-10 per group. 733 BMSCs. Data shown as mean ± SD by one-way ANOVA, n=3 independent experiments for each 738 group. (C) qPCR results· of Zfp467+/+ in COBs with 2 hours PKA or PKC inhibitor treatment 739 prior to 10min of 100 nM PTH exposure, PKA but not PKC inhibitor was able to rescue the 740 suppression of Zfp467 induced by PTH. Data shown as mean ± SD by one-way ANOVA, n=3 741