Dicer ablation in osteoblasts by Runx2 driven cre-loxP recombination affects bone integrity, but not glucocorticoid-induced suppression of bone formation

Glucocorticoid-induced osteoporosis (GIO) is one of the major side effects of long-term glucocorticoid (GC) therapy mediated mainly via the suppression of bone formation and osteoblast differentiation independently of GC receptor (GR) dimerization. Since microRNAs play a critical role in osteoblast differentiation processes, we investigated the role of Dicer dependent microRNAs in the GC-induced suppression of osteoblast differentiation. MicroRNA sequencing of dexamethasone-treated wild-type and GR dimer-deficient mesenchymal stromal cells revealed GC-controlled miRNA expression in a GR dimer-dependent and GR dimer-independent manner. To determine the functional relevance of mature miRNAs in GC-induced osteoblast suppression, mice with an osteoblast-specific deletion of Dicer (DicerRunx2Cre) were exposed to glucocorticoids. In vitro generated Dicer-deficient osteoblasts were treated with dexamethasone and analyzed for proliferation, differentiation and mineralization capacity. In vivo, abrogation of Dicer-dependent miRNA biogenesis in osteoblasts led to growth retardation and impaired bone formation. However, subjecting these mice to GIO showed that bone formation was similar reduced in DicerRunx2Cre mice and littermate control mice upon GC treatment. In line, differentiation of Dicer deficient osteoblasts was suppressed to the same extent as wild type cells by GC treatment. Therefore, Dicer-dependent small RNA biogenesis in osteoblasts plays only a minor role in the pathogenesis of GC-induced inhibition of bone formation.

monomer binding sites and GR dimer binding sites in the genome during GC exposure 11 . Strikingly, for inducing osteoporosis GR monomer function is sufficient 6 . Upon GC exposure, mice with a GR dimerization defective receptor exhibit inhibition of bone formation and impairment of osteoblast differentiation to a similar extent as wild type mice. In these GR dim knock-in mice, a point mutation in the GR abrogates the dimerization interface in the DNA binding domain of the GR resulting in strongly reduced dimeric binding of the GR, and subsequent reduced GR dimer dependent transcription 11,12 . However, transcriptional regulation by the GR monomer is preserved in these mice 11,12 . As GIO was retained in GR dim mice 6 , transcriptomes executed by GR monomers are decisive to decrease bone mass during GIO.
MicroRNAs are small non-coding, single-stranded RNA molecules that induce post-transcriptional gene silencing through base pairing with their target mRNAs. They are approximately ~22 nucleotides in length and they are processed from double-stranded hairpin precursors by two RNase III proteins, Drosha and Dicer (reviewed in ref. 13). Dicer is responsible for the final step in miRNA processing and germline deletion of Dicer1 is embryonically lethal in mice underscoring the essential role of Dicer in physiological processes including embryonic development 14 . Emerging evidences indicate that numerous miRNAs are also involved in the regulation of bone homeostasis as inactivation of Dicer leads to bone defects and impaired mineralization during embryonic stage 15 . In line, several miRNAs have been identified with enhancing or inhibitory functions in osteoblast differentiation [16][17][18][19] . Recent studies also demonstrated a regulation of miRNAs in various GC-mediated cellular processes such as cell proliferation, cell differentiation and apoptosis [20][21][22][23] . Furthermore, a few miRNAs have been shown to be involved in GC-induced suppression of osteogenesis 24-26 . Thus, miRNA regulation by GCs might be also decisive in glucocorticoid-induced inhibition of osteoblast differentiation and bone formation. Therefore, the aim of this study was to assess the role of miRNA biogenesis in the pathogenesis of GIO, in particular focusing on the suppression of osteoblast function and the consequent inhibition of bone formation.

Results
MicroRNA sequencing of GC treated mesenchymal stromal cells revealed GR dimerization and GR monomer dependent miRNAs. We have previously shown that GCs mediate the suppression of bone formation in GIO mainly via the GR monomer in osteoblasts 6 . In order to determine miRNAs expression dependent on GR monomer function during osteoblast differentiation, we first aimed to identify identical expressed miRNAs in GC treated wildtype and GR dimer deficient osteoblasts by miRNA sequencing. We therefore cultured mesenchymal stromal cells derived from wild type and GR dimer deficient GR dim mice under osteogenic conditions for 6 h with or without Dexamethasone, constructed a miRNA library and sequenced it using the Illumina technology. Venn analyses revealed sixteen differentially expressed miRNAs in wildtype primary mesenchymal cells that were treated with dexamethasone ( Fig. 1A), of which eleven were up regulated (let-7 family, miR-125b, miR-146a, miR-148a + b, miR-152, miR-423) (Table S1) and five were down regulated (miR-1724a, miR-23a+ b, miR-24-1,-2, miR-29a) (Table S1). In GR dim derived primary osteoblasts we found 34 miRNAs differentially expressed upon Dex-treatment, of which 24 were up regulated and 10 down regulated (Fig. 1A + Table S1). In the total pool of differentially expressed miRNAs in Dex-treated wild type and GR dim mesenchymal cells we found eight identical miRNAs that were up regulated (let-7a, let-7c, let-7f, let-7g, let-7i, miR-148a, miR-148b and miR-152) indicating GR monomer dependent GC regulation of miRNA expression (Table S1).
Thus, we observed miRNA regulation by GCs in mesenchymal stromal cells in a GR dimer and GR dimer independent manner. Ablation of dicer in osteoblasts causes growth retardation, low bone density and impairment of bone formation during postnatal development. To clarify to which extent miRNA regulation by GCs affects inhibition of bone formation we generated osteoblast-specific Dicer knockout mice. Therefore Dicer floxed mice were crossed with transgenic mice expressing Cre recombinase under the regulation of the Runx2 promoter. As confirmed by genomic PCR the Dicer encoding gene is almost completely ablated (Fig. S1A). We further confirmed abrogated miRNA biogenesis in Dicer Runx2Cre mice by real-time PCR analysis of selected bone-related miRNAs in femurs of E15.5 embryos. MiRNA let-7a, miR-27a, miR-101b as well as miR-143 were all down regulated ( Fig. 2A-D). Thus, miRNA expression is severely attenuated in Dicer Runx2Cre mice.
Conditional deletion of Dicer alleles in pre-osteoblasts and osteoblasts results in significant growth retardation starting three weeks postnatally indicated by significant reduced torsal length as well as reduced body weight in Dicer Runx2Cre mice compared to littermate control mice ( Fig. S1B and S1C). Skeletal radiography also revealed shorter femur length in Dicer Runx2Cre mice starting 3 weeks after birth. (Figure 2E,F). Furthermore trichrome staining for the distal part of femur depicted closure of the growth plate in Dicer Runx2Cre mice at the age of 10 weeks (Fig. 2G).
Further dynamic histomorphometry analysis following dual calcein labeling in femurs of Dicer flox and Dicer Runx2Cre mice revealed a significantly reduced bone formation rate in Dicer Runx2Cre mice compared to littermate control mice (Fig. 2L,M). In line, numbers of osteoblasts and osteoblast surface were significantly lower in Dicer Runx2Cre mice (Fig. 2N,O). Numbers of osteocytes was slightly but not significantly changed in the femora of Dicer Runx2Cre mice (Fig. S1D).
Osteoclast parameters and osteoblast-osteoclast crosstalk were not influenced by Dicer ablation in Runx2-expressing osteoblasts ( Fig. S1E-S1G). To explore whether Dicer ablation in osteoblast influences osteoclasts, we cultured primary osteoblasts derived from calvaria of Dicer flox mice that were crossed to mice expressing ubiquitously cre-ERT2 fusion (Dicer gtRosa26CreERT2 mice), allowing an inducible deletion of Dicer upon tamoxifen treatment.

Ablation of Dicer in osteoblasts does not affect dexamethasone inhibited proliferation and differentiation in vitro.
Next, we investigated whether the absence of Dicer dependent mature miRNAs affects osteoblast proliferation and differentiation under GC treatment in vitro. Proliferation of preconfluent preosteoblasts was similar in both cultures of Dicer flox and Dicer gtRosa26CreERT2 cells after 6 days (Fig. 3A). Treatment with dexamethasone strongly prevented expansion of wild type and Dicer gtRosa26CreERT2 cells (Fig. 3A). In contrast to proliferation of preconfluent osteoblasts, osteoblast differentiation reflected by alkaline phosphatase staining at day 10 of osteogenic induction and mineralization reflected by Alizarin Red staining at day 20 was decreased in cells derived from Dicer gtRosa26ERT2 mice (Fig. 3B, 3C). ALP activity and matrix mineralization were inhibited by Dex treatment in both, control and Dicer ablated cells (Fig. 3B,C). In line, the mRNA expression of the osteoblastic markers Col1a1 and Bglap (Fig. 3D,E) tended to be reduced upon Dex treatment in Dicer deficient and wild type cells.
Next we cultured femora derived from E15.5 embryos under osteogenic conditions with or without Dexamethasone for 3 days. Gene expression analysis of osteoblastic markers revealed a significant downregulation of Col1a1 and Bglap2 in embryonic bones of Dicer Runx2Cre mice compared to bones of Dicer flox mice (Fig. 4A,B). Dexamethasone treatment reduced mRNA expression of Col1a1 in femora of Dicer flox and Dicer Runx2Cre mice (Fig. 4A). Bglap expression was also suppressed by Dexamethasone in femora of Dicer flox mice, but only by trend in Dexamethasone-treated bones of Dicer Runx2Cre mice (Fig. 4B). Taken together, the absence of Dicer-dependent miRNA biogenesis does not overall affect the inhibitory effects of pharmacological doses of Dex on osteoblast differentiation markers in primary osteoblasts and organ culture.

Glucocorticoids inhibit bone formation independent of dicer ablation in osteoblast.
To define the role of Dicer dependent miRNAs in GC-induced suppression of bone formation in vivo, we treated Dicer Runx2Cre mice with prednisolone for 14 days. Bone formation rate was similarly repressed in both Dicer flox and Dicer Runx2Cre mice upon prednisolone exposure (Fig. 4C,D). In line the mineral apposition rate was reduced in GC treated Dicer Runx2Cre mice and in littermate control animals (Fig. 4F). In addition, prednisolone treatment reduced the mineralized surface in Dicer flox and Dicer Runx2Cre mice (Fig. 4E), indicating reduced numbers of osteoblasts on the femoral bone surface upon GC exposure.
Prednisolone-induced reduction of bone formation was accompanied by reduced numbers of osteoblasts, osteoblast surfaces and osteocytes in Dicer flox mice as dynamic bone histomorphometry of femoral bone revealed ( Fig. 4G-I). We also observed a non-significant trend of reduction of these parameters in Dicer Runx2Cre mice ( Fig. 4G-I) upon prednisolone treatment. Osteoclast parameters were similar in all four groups, independently of GC treatment, indicating that dicer ablation in osteoblasts may not change osteoblast-osteoclast cross-talk (Fig. 4J, K). In summary, glucocorticoids inhibit bone formation and mineralization independent of dicer ablation in osteoblasts. Thus, our in vitro and in vivo data support the findings that Dicer dependent miRNA biogenesis only plays a minor role in the pathogenesis of GC-induced osteoporosis.

Discussion
Pharmacological application of GCs in the treatment of chronic inflammation causes severe complications on the skeletal system leading to glucocorticoid-induced osteoporosis mainly due to GC-induced suppression of osteoblast function 27 . MicroRNAs play a critical role in osteoblast differentiation processes. Even though there is evidence that miRNAs regulated by GCs influence osteoblast function [28][29][30] , the role of Dicer dependent microRNA biogenesis in GC-induced suppression of osteoblast differentiation has not been investigated in vivo yet. Here we show by in vivo analysis of pre-osteoblast specific deletion of Dicer that Dicer dependent miRNA biogenesis in early pre-osteoblasts is indispensable for normal bone growth, osteoblast differentiation and bone formation. We further revealed that Dicer ablation in pre-osteoblasts has only a minor impact on GC-induced suppression of bone formation.  The significant role of Dicer generated miRNAs in osteogenesis was also previously shown in different mouse models where Dicer was bone-specific deleted 17,18,31 . Whereas deletion of Dicer in osteoclasts 31 decreased bone resorption and bone formation, deletion in the osteoblast lineage affected bone formation at different levels. Conditional deletion of the Dicer enzyme in osteoprogenitors by Col1a1Cre mice induced embryonic lethality and impaired osteoblast formation 17 . Similarly, Dicer deletion in more committed osteoprogenitor cells by OCN-Cre also prevented osteoblast differentiation and mineralization at early age. This phenotype was reversed at 2 to 4 month of age to a higher bone mass 17 . Dicer OsxCre mice exhibited a mortality of 30% at 8 weeks and surprisingly, trabecular bone mass was not altered at 6 weeks but cortical thickness and texture was affected 18 . X-Ray radiography and micro-CT analysis of Dicer flox mice crossed to Runx2Cre mice used in this study here (Dicer Runx2Cre mice) revealed a significant retardation of skeletal growth and a severe reduction in bone mass. We did not observe lethality in Dicer Runx2Cre mice until week 32. Since we detected a strong decreased expression level of several microRNAs and no impaired survival the Dicer Runx2Cre mouse is an excellent model to study the effects of impaired microRNA generation starting from early-differentiated osteoblasts throughout the osteoblastic lineage, Specifically, deletion of Dicer in Runx2 expressing cells does not exert any effects on osteoclastogenesis as our co-cultures with osteoblasts and osteoclast precursors revealed.
By analyzing regulation of miRNA expression in wild type cells and cells expressing a GR with impaired dimerization 11 , we found several miRNAs up-regulated including let-7, miR-146a, miR-148a, miR-148b and miR-152 by GC treatment, some of them exclusively in wildtype cells, and not in GR dim cells. Among the GC-induced down-regulated miRNAs was miR-29a. miR-29a was shown to target Dkk-1, an inhibitor of the osteogenic Wnt signaling pathway, thereby reducing Dkk1 levels and enhancing Wnt signaling 32,33 . Along this line, Wang and colleagues suggested that miRNA-29a signaling protects against glucocorticoid-induced reduction of osteoblast differentiation and mineralization 20,28 . For several up and down regulated miRNAs intact GR dimerization is essential. However, since we previously demonstrated that GR monomer is sufficient for impaired osteoblast differentiation and bone formation by GCs 6 , we reasoned that miRNAs regulated by the GR monomer would be more likely to be involved in the deleterious effects of GCs. We identified eight miRNAs fulfilling this criterion that were up regulated in wild type and GR dimer deficient cells. Thus, we show here for the first time GR monomer dependent GC regulation of miRNA expression. Among the GR monomer dependent miRNAs were miR-148a and let-7i, whose expression was reported to be involved in differentiation of mesenchymal stromal cells towards osteoblasts 34 . Furthermore, we identified miR-152. Miao et al. showed that increased levels of miR-152 inhibited Wnt signaling, one of the most important pathways for maintaining bone homeostasis, by up regulating the negative Wnt regulator SFRP4 35 .
Despite these different molecular mechanisms of the GR to regulate miRNAs annotated for osteoblast function, we suggest that this regulation -at least of Dicer-dependent miRNAs -has a minor relevance for the deleterious effects on bone formation of GCs in vivo.
Our most striking finding was that osteoblast differentiation ex vivo and bone formation in vivo was similarly repressed by GC exposure in Dicer flox and Dicer Runx2Cre mice. The seemingly discrepancy to the function of GC-regulated miRNAs influencing osteoblasts revealed by 22,24-26 could be explained that these studies focused on proliferation, which was impaired by anti-miRNA oligonucleotide against single miRNAs, but did not include differentiation 26 . Furthermore these results were mainly obtained by studying isolated mesenchymal stromal cells (derived from different tissues) or osteoblasts ex vivo.
However our study cannot exclude the functional involvement of miRNAs that mature independent of Dicer 36 in the suppression of bone formation in GIO. Future analysis of in vivo loss and gain of functions studies of individual miRNAs will clarify the degree of Dicer independent miRNA effects in GIO. Thus, we cannot completely rule out that miRNAs play a role in glucocorticoid-induced changes of bone microarchitecture.
In conclusion, our results demonstrate that particular miRNAs are regulated by GCs in osteoblasts in a GR dimer and monomer dependent manner. Importantly, the abrogation of Dicer dependent miRNA regulation in vivo using Runx2Cre mice crossed to Dicer flox mice reveal a model with viable mice with impaired bone growth and bone mass, underscoring the role of miRNAs for bone integrity. Our study further reveals that the regulation of Dicer generated miRNAs is dispensable during GC inhibition of bone formation. We conclude therefore that Dicer generated miRNAs are no suitable targets to prevent GC induced osteoporosis during long term GC therapy.

Materials and Methods
Declaration of approval for animal experiments. All experiments involving animals were approved by the Regierungspräsidium Tübingen, Germany.
Mice and glucocorticoid-induced osteoporosis models. All mice were kept under standardized conditions with water and food ad libitum in specific pathogen-free animal facilities. Procedures for performing animal experiments were in accordance with the approved license from the Regierungspräsidium Tübingen, Germany. Dicer Runx2Cre and Dicer gtRosaCreERT2 mice were generated by intercrossing Dicer flox mice 37 with Runx2Cre transgenic mice 38 and gtRosaCreERT2 mice (Taconic artemis, Köln, Germany). Prednisolone was applied for 14 days by subcutaneous implantation of slow-release pellets, resulting in a dose of 12.5 mg/kg/d (15 mg; 60 day release; innovative research of America, Inc.) in 14-week-old male mice on a mixed background (C57BL/6, 129SV). GR dim mice 39 were backcrossed for at least 5 generations to FVB/N background. Histomorphometry. Static and dynamic histomorphometry was performed on undecalcified and decalcified femoral sections of mice receiving double calcein (Sigma-Aldrich, St. Louis, USA) i.p. injections, as described previously 40,41 using Osteometrics system (Osteometrics, Decatur, USA).

Micro-computed tomography (Micro CT).
Femora were analyzed using a SkyScan 1174 compact micro CT (Bruker, Billerica, USA) equipped with an X-ray tube working at 80 kV/100 μ A. Resolution was 6.2 μ m, rotation step was fixed at 0.40°, and a 0.5 mm aluminum filter was used. For reconstruction of femora, region of interest was defined 0.3 mm apart from the distal growth plate into the diaphysis spanning 1.8 mm. Trabecular bone volume/tissue volume (BV/TV), trabecular thickness (Tb.Th.), trabecular separation (Tb. Sp.) and trabecular number (Tb.N.) were determined according to guidelines by ASBMR Histomorphometry Nomenclature Committee 41,42 . X-ray radiograph was obtained by 70 μ m resolution (La Theta, Aloka, Tokyo, Japan).
Embryonic bones were isolated at day E15.5 from Dicer Runx2Cre mice and cultured for three days in osteogenic induction medium (as described for the calvaria derived osteoblasts) with or without 1 μ M Dexamethasone (Sigma-Aldrich, St. Louis, USA). miRNA sequencing. Total RNA was isolated (Trizol reagent, Invitrogen) from wild type and GR dim derived mesenchymal stromal cells that were treated for 6h with osteogenic induction medium with or without 10 −6 M Dexamethasone. Quality check and quantification was performed with Agilent Bioanalyzer 2100 and Agilent RNA 6000 nano kit (Agilent Technologies). 1 μ g of total RNA was used for miRNA library preparation using the Illumina small RNA v1.5 sample kit. Library preparation was performed as described 43 . Sequencing was conducted with the Illumina sequencing platform 44 . Sequencing was done using a GAIIx (Illumina) in single-read, 36 cycle mode without multiplexing of libraries (sequencing one library per lane). The reads were extracted in FastQ format using tool GA-Pipeline v1.4 (supported by Illumina). Sequencing resulted in 10-12 mio reads per sample.
Analysis of sequencing data. Individual sequence reads with base quality scores were produced by Illumina sequencing. The data were analyzed by the use of CLC-workbench (CLCbio, Arhus, Denmark). After eliminating reads with low quality and trimming the 3′ adaptor sequence, the remaining 18-to 33-nt reads were grouped into unique sequence clusters. To exclude potential sequencing errors, only sequences with at least 30 occurrences were included. Annotation of sequence clusters was performed, allowing 2 mismatches and up to 2 additional bases at each end, by using the Mus musculus reference from miRBase v16.0. Expression values were normalized to reads per million (RPM). Differentially-expressed miRNAs were identified using Kal's Z-test for pairwise comparison. Venn-analyses were performed to find commonly regulated miRNAs between the three (WT), respectively two (GR dim ), independent replications of experiments.
Coculture experiments. Primary osteoblast isolated from Dicer gtRosaCreERT2 mice were seeded as 8,000 cells per well in 96-well plates and treated with 1μ M 4-OHT for 3 days as described above. Bone marrow cells from wild-type mice were isolated and 200,000 cells per well were added on top of primary osteoblasts cultivated in alpha-MEM supplemented with 10 nM 1,25-dihydroxyvitamin D3 (Sigma-Aldrich, St. Louis, USA). Medium was changed every 3 days. After 9 days, cells were fixed and stained for TRAP activity.