Dicer-dependent pathways regulate chondrocyte proliferation and differentiation

Dicer Deficiency in Chondrocytes Causes Defects in Skeletal Development.

To investigate the physiological role of Dicer-dependent small RNAs in skeletal development, we deleted the Dicer gene in cartilage by crossing mice containing a floxed Dicer allele (Dicer^fl/fl) (9) with transgenic mice expressing Cre recombinase under the control of a Col2a1 promoter (Col2-Cre). This Cre allele exhibited efficient Cre-recombinase activity in chondrocytes (10). Col2-Cre:Dicer^fl/fl mice showed a significant growth defect and mostly die by the time of weaning. Skeletal preparation demonstrated relatively proportional reduction in skeletal size (Fig. 1A). However, the growth defect of the skull and maxilla in Col2-Cre:Dicer^fl/fl mice caused relative overgrowth of the mandible and lower incisors (Fig. 1B). This is likely due to growth defects in the nasal cartilage, because we observed histological changes similar to those of limb growth plates (data not shown). Histological examination of postnatal 12-day-old (P12) mice shows a delay in formation of the secondary ossification center and flattening of the tibial epiphysis of Col2-Cre:Dicer^fl/fl mice (Fig. 1C). The chondrocyte density in Col2-Cre:Dicer^fl/fl growth plates was reduced because of the reduction in the number of columnar proliferating chondrocytes (1,656 ± 49/mm³ vs. control 2,266 ± 108/mm³; P < 0.05 by ANOVA, n = 3 per each group) (Fig. 1D). To assess the efficiency of Dicer removal, we quantified the mRNA level of Dicer mRNA by quantitative RT-PCR (qRT-PCR) using RNA isolated form microdissected hindlimb cartilage of 3-day-old mice. The Dicer mRNA level of Col2-Cre:Dicer^fl/fl cartilage was reduced by 87% of that of control cartilage (Fig. 1E). As expected, the expression levels of ubiquitous miRNAs, let-7a, and miR-30c were significantly reduced in Col2-Cre:Dicer^fl/fl mice (Fig. 1F).

Analysis of fetal and neonatal Col2-Cre:Dicer^fl/fl bones revealed a reduction in bone width and an expansion of the hypertrophic region of the growth plate (Fig. 2 A and B). The reduction in bone width was already evident at embryonic day (E)14.5, and the expansion of the hypertrophic region was present after the initiation of bone marrow formation at E15.5 (data not shown). At E18.5, the hypertrophic region, indicated by Col10a1 expression, was expanded, whereas the Mmp13 expression domain of the cartilage was not expanded in Col2-Cre:Dicer^fl/fl mice (Fig. 2 B and C). Because Mmp13 is expressed only by terminally differentiated hypertrophic chondrocytes in the growth plate, these observations suggest that the expansion of the hypertrophic region was caused by an acceleration of hypertrophic differentiation of proliferating chondrocytes rather than a reduction in cartilage resorption by bone cells, which would cause an increase in terminally differentiated hypertrophic chondrocytes. Expansion of the hypertrophic region can also occur as a consequence of stimulation of chondrocyte differentiation at earlier steps (11, 12). To test whether Dicer deficiency directly stimulated hypertrophic differentiation of proliferating chondrocytes, we deleted Dicer only in late proliferating chondrocytes using Osx-Cre transgenic mice (13). Osx-Cre mice show Cre recombination activity in late-proliferating chondrocytes of the growth plate and in osteoblasts (Fig. 2D). Osx-Cre:Dicer^fl/fl mice showed an expansion of hypertrophic region without affecting the size of the periarticular region (Fig. 2E). We also observed increased Col10a1 expression in primary rib chondrocytes upon Dicer deletion in vitro [supporting information (SI) Fig. 6]. These observations strongly suggest that the loss of Dicer in proliferating chondrocytes directly stimulates their hypertrophic differentiation, leading to the expansion of the hypertrophic region during fetal and neonatal stages.

Indian hedgehog (Ihh) regulates chondrocyte differentiation and proliferation through PTHrP-dependent and -independent pathways (12, 14). The expression of Ihh and the transcriptional target of Ihh signaling, Patched (Ptch1), was preserved in Col2-Cre:Dicer^fl/fl mice, suggesting that the acceleration of hypertrophic differentiation was caused by defects either downstream or independent of Ihh signaling. Another critical signaling system downstream of Ihh that negatively regulates hypertrophic differentiation is the PTHrP signaling pathway. Because the basal expression level of PTHrP in cartilage was too low to reliably detect its possible down-regulation in Col2-Cre:Dicer^fl/fl growth plates, we took advantage of transgenic mice expressing a constitutively active PTHrP receptor (Col2-caPPR) to test whether possible impairment of PTHrP signaling was involved in the accelerated hypertrophic differentiation in Col2-Cre:Dicer^fl/fl mice. Col2-caPPR transgenic mice were able to successfully rescue growth plate abnormalities caused by loss or impairment of PTHrP signaling in vivo (11, 15, 16). The expansion of the hypertrophic region in Col2-Cre:Dicer^fl/fl mice was not reversed in compound mutant mice, Col2-caPPR:Col2-Cre:Dicer^fl/fl mice, suggesting that the acceleration of hypertrophic differentiation in Col2-Cre:Dicer^fl/fl chondrocytes was caused by a defect either independent or downstream of PTHrP receptor signaling (SI Fig. 7A). Because Dicer deficiency would affect multiple pathways, it is also possible that Dicer deficiency indeed affected PTHrP signaling, but because defects in other pathway played a dominant role, caPPR overexpression could not rescue the phenotype.

Dicer Is Required for Maintenance of Proliferating Chondrocytes in the Growth Plate.

The skeletal growth defect and reduced width of the growth plate suggested a decrease in number of chondrocytes in Col2-Cre:Dicer^fl/fl mice. Indeed, the reduction in proliferating chondrocytes in Col2-Cre:Dicer^fl/fl mice was particularly well demonstrated in the growth plate between the basisphenoidal and basioccipital bones in the skull base (Fig. 3). This bidirectional growth plate had two Col10a1 domains flanking the region containing proliferating chondrocytes. The basisphenoidal–basioccipital growth plate of Col2-Cre:Dicer^fl/fl mice was thinner in width, and the domain of proliferating chondrocytes progressively became smaller (Fig. 3A). The diminished number of proliferating chondrocytes at P3.5 in Col2-Cre:Dicer^fl/fl mice was shown by the fused Col10a1 domains and the reduction in the BrdU-positive domain (Fig. 3). Overexpression of caPPR failed to restore the mass of proliferating chondrocytes, again supporting that defective PTHrP receptor signaling is not the major cause for the phenotype (SI Fig. 7B). Loss of Dicer has been reported to increase cell death in multiple types of cells (9, 17–21). We did not find an overt increase in cell death assessed by the TUNEL assay (SI Fig. 8C), and therefore it was unlikely that cell death contributed to the decrease in proliferating chondrocytes. In addition, we found a dramatic decrease in chondrocyte proliferation in Col2-Cre:Dicer^fl/fl mice (Fig. 4). These findings suggest that loss of Dicer results in a reduction in the number of proliferating chondrocytes through two distinct mechanisms: acceleration of differentiation into postmitotic hypertrophic chondrocytes and reduction in chondrocyte proliferation.

miRNA and Gene Expression Profiling Suggests Limited Direct Regulation of mRNA Expression by Chondrocytic miRNAs.

Although Dicer has been shown to process RNA species besides premiRNAs in mammalian cells (22), it is believed that the primary role of Dicer in mice is to generate mature miRNAs. Therefore, we hypothesized that the reduction in levels of chondrocytic miRNAs caused the physiologic changes in Dicer-deficient chondrocytes. To understand the roles of miRNAs in chondrocytes, first we performed miRNA profiling using a bead-based array (23). miRNA profiles from different chondrocyte sources showed modestly different patterns, whereas the miRNA profile of calvarial cells showed greater dissimilarity from those of chondrocytes (SI Fig. 9 and SI Table 1). The majority of miRNAs abundantly expressed in hindlimb chondrocytes and calvarial osteoblasts were evolutionary conserved and relatively ubiquitous ones such as members of the let-7 family, miR-16, and miR-26 (Fig. 5A). The “cartilage-specific” miRNAs, miR-140 and miR-140*, are expressed at high levels, a result consistent with previous reports (24, 25) (Fig. 5A). We also found that several miRNAs were preferentially expressed in chondrocytes compared with calvarial cells (Fig. 5B and SI Fig. 9B). miRNA profiling comparing control and Col2-Cre:Dicer^fl/fl mice showed that most miRNAs were reduced by ≈60–70% in neonatal Col2-Cre:Dicer^fl/fl hindlimb growth plates (Fig. 5C and SI Table 2). We noticed that some miRNAs, such as miR-133 and miR-1, showed no reduction. This is probably due to contaminating muscle tissues into microdissected cartilage specimens, because these miRNAs are expressed at very high levels in muscle (26, 27). The observation that certain amounts of miRNAs were still detectable in the Col2-Cre:Dicer^fl/fl sample was unexpected. Considering the relatively proportional reduction in the majority of miRNAs and the efficient elimination of Dicer mRNA, it is unlikely that contamination of noncartilaginous tissues or inefficient Dicer deletion is the reason for the remaining miRNAs. Because mature miRNAs appear quite stable (28), these miRNAs in mutant growth plates are probably a carryover of preexisting miRNAs.

Next, we performed microarray analysis to identify genes whose expression was altered in Dicer-deficient chondrocytes. RNA isolated from whole growth plates of P3.5 control and Col2-Cre:Dicer^fl/fl hindlimbs was subjected to microarray analysis. Of the 45,036 probes, 2,658 showed >1.5-fold expression changes (SI Table 3). To test whether the reduction of miRNAs directly influenced RNA levels of miRNA-target genes in Dicer-deficient chondrocytes, we examined expression levels of predicted miRNA-target genes that had binding sites for five or more different seed sequences of the 30 most-abundant chondrocytic miRNAs using the microarray data. These miRNAs had 17 different seed sequences (seven-nucleotide sequences in positions two to eight of mature miRNAs) that were used to predict target genes by the computer program, TargetScanS (29). A total 4,310 genes were predicted as potential target genes of these miRNAs abundantly expressed in chondrocytes (SI Table 4). Two hundred and fourteen genes (5.0%) had binding sites for five or more different seed sequences. Among these genes, analysis of 71 genes whose expression data were available showed that most were either not expressed or had unchanged expression (SI Fig. 10A). Only Hmga2 was found up-regulated in the Col2-Cre:Dicer^fl/fl growth plate, a result subsequently confirmed by quantitative RT-PCR and in situ hybridization (SI Fig. 10B). This finding is consistent with the reports that Hmga2 transcripts were destabilized by miRNAs, let-7 and miR-98 (30–32). We also examined the 3′-UTR sequences of the 1,437 genes up-regulated in the Col2-Cre:Dicer^fl/fl growth plate for potential binding sites of chondrocytic miRNAs. The frequencies of sequences complementary to the 17 miRNA seed sequences in 3′ UTRs were calculated by using the MotifADE (33). We did not find significant enrichment in miRNA-binding sites in genes up-regulated in Col2-Cre:Dicer^fl/fl chondrocytes (data not shown). This finding may suggest that Dicer deficiency in chondrocytes has limited effects in miRNA-target gene expression at the RNA level.

Mice.

Col2-Cre transgenic mice (10), floxed Dicer mice (9), Osx-Cre transgenic mice (13), and Col2-caPPR transgenic mice (15) were described. Genotyping of Cre transgenic mice was performed by PCR using primers detecting the Cre sequence (11). The floxed and wild-type Dicer alleles were detected by using primers, P1: 5′-AGTGTAGCCTTAGCCATTTGC-3′ and P2: 5′-CTGGTGGCTTGAGGACAAGAC-3′. These primers amplify the region spanning the downstream loxP sequence. Littermates were used as control. The Col2-caPPR allele was genotyped by PCR by using primers specific to the transgene, P3: 5′-TAGTTGGCCCAGGTCCTGT-3′ and P4: 5′-TAACCATGTTCATGCCTTCTTC-3′.

Because Col2-Cre:Dicer^+/+, Col2-Cre:Dicer^fl/+, and Dicer^fl/fl mice were indistinguishable from wild-type mice in growth, growth plate morphology, and chondrocyte marker expression (data not shown), either Col2-Cre:Dicer^fl/+ or Dicer^fl/fl littermates were used as control in this study.

Skeletal Preparation, Histology, and Rosa-26R Cre Reporter Assay.

Alizarin red and alcian blue staining was performed by using a modified McLeod's method (50). Carcasses were fixed in 95% ethanol, stained with alcian blue and alizarin red, cleared in 1% KOH, and kept in 50% glycerol. For histological analysis, mice were dissected, fixed in 10% formalin, decalcified in 10% EDTA, paraffin-processed, cut, and subjected to hematoxylin/eosin staining, in situ hybridization, and BrdU staining. X-gal staining for the Rosa 26-R Cre reporter assay was performed as described (51).

In Situ Hybridization.

In situ hybridization was carried out as described (52). Probes for Col10a1, MMP13, Ihh, and Ptch1 mRNAs were described (11).

BrdU Labeling and Detection.

For BrdU labeling, 50 μg of BrdU per gram of body weight was given to mice i.p. 2 h before death. Tissues were fixed in 10% formalin solution, processed, and sectioned by using standard procedures. BrdU was detected by using the BrdU-staining kit (Zymed). The BrdU-labeling index was calculated as the ratio of BrdU-positive nuclei over total nuclei in each region of the growth plate.

miRNA Expression Profiling.

Total RNA was extracted from growth plates and calvariae of neonatal C57BL/6 mice. Limbs were freshly frozen in OCT media, cryosectioned with 60-μm thickness, and dissected under a dissecting microscope using 28-gauge needle tips to obtain the desired populations of chondrocytes. Chondrocytes from the periarticular, columnar, and hypertrophic regions were separately isolated from the distal femur by using TRIzol (Invitrogen). RNA was purified also from whole growth plates of the femur and tibia (hindlimb chondrocytes), the humerus (forelimb chondrocytes), and calvarial cells. miRNA profiling was performed by using the Luminex bead-based array (23).

Northern Blot Analysis.

Small RNA was isolated from whole growth plate samples of the femur and tibia of 3-day-old mice by using the mirVana miRNA isolation kit (Ambion). RNA was separated on a denaturing 15% polyacrylamide gel containing 8M urea and 1× TBE and electroblotted onto a nylon membrane. Synthesized antisense DNA oligo probes were labeled with digoxigenin by using the DIG oligonucleotide tailing kit, second generation (Roche Applied Science). Hybridization and washing were performed by using the ULTAHyb-Oligo Hybridization Buffer (Ambion) according to the manufacturer's protocol. Signal detection was performed by using the DIG Luminescent Detection Kit (Roche Applied Science).

Quantitative RT-PCR.

Total RNA was extracted from growth-plate cartilage of the distal femur and proximal tibia of P3.5-old mice by using RNeasy mini kit (Qiagen). cDNA synthesis was performed by using random hexamers with the Protoscript First Strand cDNA Synthesis Kit (New England Biolabs), and quantitative PCR was performed by using the DNA Engine Opticon 2 Continuous Fluorescence Detection System (Biolab) and the Sybr green mix (Applied Biosystems). Signals were normalized to β-actin. Each reaction was performed in quadruplicate and repeated on different sample sets to confirm the results. Primer sequences were: Dicer1-F, 5′-AATTGGCTTCCTCCTGGTTAT-3′ and Dicer1-R, GTCAGGTCCTCCTCCTCCTC-3′; β-actin-F, 5′-GCACTGTGTTGGCATAGAGG-3′ and β-actin-R, 5′-GTTCCGATGCCCTGAGGCTC-3′.

Microarray Analysis.

RNA was isolated from microdissected whole growth-plate cartilage of the distal femur of neonatal mice by using the RNeasy mini Kit (Qiagen). Up to 10 limbs were pooled for a group. Two samples per each group were separately prepared. The biotynylated cRNA was hybridized to the Affymetrix MOE430 v2 chip. Normalization and comparison between groups were performed by using the dChip (53). Genes considered to be absent in more than two sample sets based on the Affymetrix detection (present/absent) call were removed from the comparison analysis.