MicroRNA-27a is essential for bone remodeling by modulating p62-mediated osteoclast signaling

The ability to simultaneously modulate a set of genes for lineage-specific development has made microRNA an ideal master regulator for organogenesis. However, most microRNA deletions do not exhibit obvious phenotypic defects possibly due to functional redundancy. MicroRNAs are known to regulate skeletal lineages as the loss of their maturation enzyme Dicer impairs bone remodeling processes. Therefore, it is important to identify specific microRNA essential for bone homeostasis. We report the loss of miR-27a causing severe osteoporosis in mice. MiR-27a affects osteoclast-mediated bone resorption but not osteoblast-mediated bone formation during skeletal remodeling. Gene profiling and bioinformatics further identify the specific targets of miR-27a in osteoclast cells. MiR-27a exerts its effects on osteoclast differentiation through modulation of Squstm1/p62 whose mutations have been linked to Paget’s disease of bone. Our findings reveal a new miR-27a-p62 axis necessary and sufficient to mediate osteoclast differentiation and highlight a therapeutic implication for osteoporosis.

and miR-24-2. Aberrant regulation of miR-23a and miR-27a has been associated with osteoporotic patients and increased bone fracture risks 15,16,17 . The effect of miR-23a and miR-27a on the differentiation of osteoblast or osteoclast cells has been shown by in vitro overexpression studies 16,18,19 . Interference of miR-23a or miR-27a by the use of inhibitor/sponge has implied their role in osteoblast and osteoclast cells 18,20 . However, due to cross-reactivity of the RNA inhibitor among family members that share common seed motif 21,22 , the inhibitor assay may not truthfully reflect the function of the target miRNA. Therefore, the genetic loss-of-function study remains the most rigorous method for determining the endogenous function of miR-23a and miR-27a as well as testing the removal of miR-23a~27a sufficient to cause bone loss.
We have performed mouse genetic analyses to definitively assess the requirement of miR-23a and miR-27a for skeletogenesis and homeostasis. Surprisingly, the skeletal phenotypes developed in newly established loss-of-function mouse models reveal findings different from previous reports based on the gain-of-function analyses. Severe loss of bone mass developed in mice with the deletion of miR-23a~27a or miR-27a.
MiR-23a~27a is dispensable for osteoblast-mediated bone formation. However, compelling evidence support that miR-27a is essential for osteoclastogenesis and osteoclast-mediated bone resorption during bone remodeling. Gene expression profiling and bioinformatics analyses further identify osteoclast-specific targets of miR-27a. We demonstrated that miR-27a exerts its effects on osteoclast differentiation through modulation of a new and essential target Squstm1/p62. MiR-27a is necessary and sufficient to mediate osteoclast differentiation, and as a biomarker and therapeutic target for osteoporosis.

The loss of miR23a~27a in mice causes low bone mass phenotypes
To determine the requirement of miR-23a~27a for skeletal development and maintenance, we created a new mouse model with the deletion of miR-27a and miR-23a.
The CRISPR/Cas9 gene edition method was used to establish the mouse strain carrying the ∆miR-23a~27a allele (Fig. 1A). PCR analysis of the miR-23a cluster demonstrated the sgRNA-mediated deletion and the reduction of 500 bp in the wild-type to 303 bp in the mutants (Fig. 1B). Sequencing analysis also confirmed the expected genome editing (data not shown). Next, we tested if the deletions affect the expression of other microRNA molecules, generated from the same RNA precursor, within the same cluster.
Mice heterozygous for ∆miR-23a~27a are viable and fertile. Intercross between the heterozygotes successfully obtained the homozygous mutants without any noticeable skeletal deformity, suggesting that miR-23a~27a are not required for the developmental processes (Fig. S1). Next, we examined if their deletions affect the homeostatic maintenance of the bone in adults. At 3 months, von Kossa staining and threedimensional (3D) micro-computed tomography (µCT) analyses of the ∆miR-23a~27a femurs revealed significant loss of the trabecular bone volume in both sexes (Fig. 1D-G; BV/TV, n=3, mean + SD; student t-test). Much more severe osteoporotic defects were detected in the 7-month-old mutant females of ∆miR-23a~27a ( Fig. S2; BV/TV, n=3, mean + SD; student t-test). However, cortical bone thickness does not seem to be affected by the deletion of miR-23a~27a (Fig. S3). Next, we examined if similar bone loss phenotypes can be detected in the vertebrae where age-related changes in the trabecular architecture are minimal. Therefore, we examined the vertebrae of ∆miR-23a~27a and identified drastic reductions in vertebral bone mass associated with the mutations (Fig. 1H). These data demonstrated that miR-23a~27a is required for homeostatic maintenance of the bone.

Osteoblast-mediated bone formation is not affected by the loss of miR-23a~27a
Proper maintenance of the skeleton requires balanced bone formation and resorption during bone remodeling. The bone loss phenotypes caused by the deletion of miR-23a~27a are likely to be associated with an imbalanced bone formation and resorption mediated by osteoblasts and osteoclasts, respectively. Therefore, we examined if the bone formation and osteoblast activities are affected by the loss of miR-23a~27a.
New bone formation was analyzed by double labeling with alizarin red and calcein at 3 months. Quantitative analyses did not reveal a significant difference in bone formation rate per unit of bone surface (BFR/BS) caused by the mutation (Fig. S4A). In addition, the numbers of osteoblast cells positive for type 1 collagen (Col1) and Osteopontin (OPN) lining the trabecular bone surface remain comparable between the wild-type and homozygous littermates (Fig. S4B), indicating that osteoblast-mediated bone formation is not affected by the miR-23a~27a deletion. The results suggested that miR-27a and miR-23a are not required for osteoblastogenesis and osteoblasts-mediated bone formation.

MiR-23a~27a regulates osteoclast differentiation
To determine if the loss of miR-27a affects bone resorption, we first examined osteoclast number by tartrate-resistant acid phosphatase (TRAP) staining. An increase of TRAP+ osteoclasts was detected in the 3-month-old ∆miR-23a~27a males and females (Fig. 2). When the number of TARP+ osteoclast cells in the total bone area (N.Oc/T.Ar), the ratio of TRAP+ bone surface (Oc.S/BS), and the number of TRAP+ osteoclast cells lining the bone surface (N.Oc/BS) were measured, we found that these parameters associated with bone resorption are significantly elevated in the mutants ( Fig. 2; n=5, mean + SD; student t-test). In addition, there is a ~3-fold increase in the number of Cathepsin K-expressing osteoclast cells lining the trabecular bone surface ( Fig. 2; n=5, mean + SD; student t-test). These results support that the loss of miR-23a~27a stimulates osteoclastogenesis, leading to an elevation of bone resorption.
Next, to determine the role of miR-23a~27a in osteoclastogenesis, we analyzed cell populations associated with the differentiation of the osteoclast cells. During hematopoiesis, a common myeloid progenitor gives rise to monocytes that are precursors of several cell types, including dendritic cells, macrophages, and osteoclasts 23 .
Osteoclast precursors are known to derive from a monocyte population positive for CD11b and negative for Gr-1 24 . Therefore, we examined CD11b+ and Gr-1-monocytes to see if the miR-23a~27a deletion affects the osteoclast precursor population. FACS analysis revealed that the CD11b+ and Gr-1-population was not affected in the ∆miR-23a~27a bone marrow of both male and female mice (Fig. 3A, B). The CD11b+/CD11c+ dendritic cell population, derived from the monocytes, was also unaffected by the miR-23a~27a deletion (Fig. 3A, B). Although the precursors are not affected, miR-23a~27a may play a role in osteoclast differentiation.
To test osteoclast differentiation affected by the loss of miR-23a~27a, an ex vivo analysis was performed with cultures of cells seeding at two different densities. Cells isolated from the bone marrow were cultured in the presence of M-CSF to obtain bone marrow-derived macrophages (BMMs), followed by differentiation into osteoclasts with the treatment of RANKL. TRAP staining was then used to assess the extent of osteoclast differentiation. The number of TRAP+ cells was significantly increased by the loss of miR-23a~27a ( Fig. 3C; n=5, mean + SD; student t-test).

MiR-27a is an essential regulator for osteoclast-mediated skeletal remodeling
Using the miRPath Reverse-Search module, we searched and ranked miRNAs whose targets are accumulated in osteoclast differentiation-related genes based on the enrichment of the targets in the Kyoto Encyclopedia of Genes ad Genomics (KEGG: mmu04380) 25,26 . Among them, miR-27a was predicted as the top 3 candidate to regulate osteoclast differentiation (Fig. S5, p < 2.0 x 10 -71 ). Its sister gene miR-27b contains the same seed sequences also ranked in the top 3. However, miR-23a had a lower estimated rank suggesting that miR-27a alone may be sufficient to exert osteoclast regulation (Fig. S5, p < 4.9 x 10 -29 ). To test this hypothesis, we created another mouse strain with the deletion of only miR-27a using the CRISPR/Cas9 genome editing (Fig. 4A). PCR analysis revealed the sgRNA-mediated deletion causes the reduction of 500 bp in the wild-type to 474 bp in the mutants (Fig. 4B). Sequencing of the PCR products confirmed the genomic deletion (data not shown). Next, semi-quantitative RT-PCR analysis indicated that only miR-27a are disrupted in the ∆miR-27a mutants, suggesting the deletions does not affect the expression of other microRNA generated from the same RNA precursor (Fig. 4C). These results demonstrated our success in establishing mouse models deficient for miR-27a.
Mice heterozygous and homozygous for ∆miR-27a are viable and fertile similar to the miR-23a~27a deletion. As anticipated, there was no noticeable skeletal deformity associated with the loss of miR-27a suggesting its dispensable role in the developmental processes. However, von Kossa staining and 3D micro-computed tomography (µCT) analyses of the 3-month-old male and female femurs of ∆miR-27a revealed significant loss of the trabecular bone volume (  Fig. S8; n=5, mean + SD; student t-test), supporting that the loss of miR-27a stimulates osteoclastogenesis, leading to elevated bone resorption. While the CD11b+/Gr-1-and CD11b+/CD11c+ osteoclast precursor populations were not affected by the miR-27a deletion (Fig. S9A, B), osteoclast differentiation was significantly increased by the loss of miR-27a ( Fig. 5B; n=5, mean + SD; student t-test). Furthermore, the loss of a single miR-27a recapitulates the osteoporotic phenotypes caused by the double deletion of miR-23a and miR-27a, suggesting that miR-27a is responsible for skeletal maintenance through the modulation of bone remodeling processes. The results suggested that miR-27a functions as a negative regulator in osteoclast differentiation. To test this possibility, we overexpressed miR-27a in cells undergoing osteoclast differentiation. High levels of miR-27a significantly reduce the number of differentiated osteoclast cells (Fig. 5C). Our findings demonstrated that miR-27a is necessary and sufficient to modulate osteoclast differentiation. Osteoclastogenesis mediated by miR-27a is essential for bone remodeling and homeostasis.

MiR-27a regulates osteoclast differentiation through the modulation of p62
To elucidate the mechanism underlying OC differentiation regulated by miR-27a, we first used a bioinformatics approach to identify its potential targets (Fig. 6A). The TarBase computationally predicted 2312 target genes for miR-27a 27 . Furthermore, there were 154 genes associated with osteoclast differentiation based on the Kyoto Encyclopedia of Genes ad Genomics (KEGG: mmu04380) 25,26 . The miRPath software further identified 26 targets overlapping with the osteoclast-related genes 28 . Next, we examined the transcript level of these 26 targets in wild-type and ∆miR-27a osteoclast cells. Quantitative RT-PCR analyses revealed that 5 of these targets, Snx10, Map2k7, Ctsk, Tgfbr1, and Sqstm1, are significantly up-regulated by the loss of miR-27a (Fig. 6B, p < 0.05, n = 3; two-sided student t-test). To test if these genes were the direct targets of miR-27a, we performed 3'UTR-reporter assays. The expression of miR-27a significantly downregulated the luciferase activity associated with the 3'UTR of Sqstm1, Tgfbr1, Snx10, Map2k7, but not Ctsk (Fig. 6C, *; p < 0.01, n=3, mean + SD; two-sided student ttest). As Cathepsin K is a protease expressed in the mature osteoclast cells, its alteration at the transcript level is likely ascribed to indirect effects of increased osteoclastogenesis in ∆miR-27a. The data indicated Sqstm1, Tgfbr1, Snx10, and Map2k7 as direct targets of miR-27a.
Sqstm1 also known as p62 whose gain of function mutations were linked to Paget's disease of bone with disruption of bone renewal cycle causing weakening and deformity 29 . The deletion of p62 in mice also impaired osteoclast differentiation 30 .
Therefore, we performed a functional study to test the importance of the miR-27a-p62 regulatory axis during osteoblastogenesis. Cells isolated from the bone marrow were induced for osteoclast differentiation and the number of osteoclast cells positive for TRAP staining was counted to determine the outcome of the differentiation. The enhanced osteoclast differentiation in the ∆miR-27a culture was significantly alleviated by the shRNA-mediated knockdown of p62 (Fig. 6D, p < 0.05, n=3, mean + SD; twosided student t-test). To assess osteoclast function, we first performed phalloidin staining to ensure actin ring formation was not affected by miR-27a deletion and p62 knockdown in OC cells (Fig. S10). Bone resorption pit assay then revealed enhanced bone resorption activity of ∆miR-27a osteoclasts that can be alleviated by p62 knockdown (Fig. 6E, p < 0.05, n=3, mean ± SD; two-sided student t-test). Next, qRT-PCR of osteoclast markers was performed to decipher the regulatory process of miR-27a-mediated osteoclast differentiation. The deletion of miR-27a elevated the expression of Rank but this elevation could be alleviated by p62 knockdown (Fig. 6F, p < 0.05, n=3, mean ± SD; two-sided student t-test), indicating the importance of the miR-27a-p62 axis for RANKL signaling. The expression of Dcstamp, essential for cell-cell fusion during osteoclastogenesis, was increased in ∆miR-27a osteoclasts (Fig. 6F, p < 0.05, n=3, mean ± SD; two-sided student t-test). However, the increased level of Dcstamp in ∆miR-27a osteoclasts could not be affected by p62 knockdown, implying that miR-27a mediated regulation of osteoclast fusion is independent of p62. Mature osteoclast markers, e.g.
Ctsk and Calcr, were significantly upregulated in ∆miR-27a osteoclasts but reduced by p62 knockdown (Fig. 6F, p < 0.05, n=3, mean±SD; two-sided student t-test). The results demonstrated that miR-27a-dependent osteoclast differentiation is mediated through the regulation of p62. The miR-27a-p62 regulatory axis plays an important role in osteoclastogenesis during bone remodeling.

Discussion
The dysregulation of miRNA has been implicated in osteoporosis in menopausal women. Among 851 miRNAs tested miR-27a is one of the most significant genes downregulated in postmenopausal osteoporosis patients 17 . However, it's not clear whether the alteration of miRNAs is the cause or consequence of the disease. Our genetic study presented here has demonstrated the loss of miR23a~27a or miR27a results in significant bone loss. The findings suggest a single miRNA deficiency can lead to severe osteoporotic defects, indicating an essential role of miR-27a in bone remodeling.
Because osteoporosis is caused by an imbalance of osteoblast-mediated bone formation and osteoclast-mediated bone resorption, the conventional knockout model is ideal to decipher the regulatory processes underlying miR-27a-dependent pathogenesis. Our data also suggest that miR-27a is dispensable for osteoblast differentiation and bone formation as its deletion does not affect the number of osteoblast cells and bone formation rates.
The results do not agree with the previous gain-of-function study indicating an inhibitory role of miR-27a in osteoblastogenesis 18 . Therefore, the association of osteoporosis with both upregulation and downregulation is possibly mediated through distinct mechanisms underlying the regulatory process of the miR-23a cluster.
This study provides compelling evidence to first demonstrate that miR-27a is essential for regulating the bone resorption process through modulation of osteoclast differentiation. The loss of miR-27a in mice leads to elevated numbers of osteoclast cells as well as increases in bone resorption activity. The inhibitory function of miR-27a on osteoclastogenesis is also in agreement with previous in vitro culture data showing its crucial downregulation among miRNAs associated with osteoclast differentiation 31 .
Although the number of osteoclast progenitors is comparable between the control and mutant, the deletion of miR-27a strongly accelerates the process of osteoclastogenesis.