Long noncoding RNA Lnc-DIF inhibits bone formation by sequestering miR-489-3p

Summary Osteoporosis has become a high incident bone disease along with the aging of human population. Long noncoding RNAs (LncRNAs) play an important role in osteoporosis incidence. In this study, we screened out an LncRNA negatively correlated with osteoblast differentiation, which was therefore named Lnc-DIF (differentiation inhibiting factor). Functional analysis proved that Lnc-DIF inhibited bone formation. A special structure containing multiple 53 nucleotide repeats was found in the trailing end of Lnc-DIF. Our study suggested that this repeat sequence could sequester multiple miR-489-3p and inhibit bone formation through miR-489-3p/SMAD2 axis. Moreover, siRNA of Lnc-DIF would rescue bone formation in both aging and ovariectomized osteoporosis mice. This study revealed a kind of LncRNA that could function as a sponge and regulate multiple miRNAs. RNA therapy techniques that target these LncRNAs could manipulate its downstream miRNA-target pathway with significantly higher efficiency and specificity. This provided potential therapeutic insight for RNA-based therapy for osteoporosis.


INTRODUCTION
Osteoporosis has become an emerging threat to human health. Its incidence is elevating along with the aging of the population. Osteoporosis can be caused by several factors, including heredity, aging, and postmenopausal hormone disorder (Clarke and Khosla, 2010). Although impacted by multiple factors, the formation and development of osteoporosis is largely attributed to reduced bone formation that resulted from decreased osteoblast differentiation (Marie, 1999). The differentiation of osteoblast is a multistep process controlled by a variety of genetic and epigenetic signaling pathways. Long noncoding RNAs (LncRNAs) have been proved as an essential regulator to osteoblast differentiation among all epigenetic pathways (Hassan et al., 2015;Peng et al., 2018aPeng et al., , 2018b, and it glittered as a highlight in recent bone research. LncRNA is a kind of transcript composed of more than 200 nucleotides and usually with no coding potential. LncRNAs have been reported to regulate multiple physiological and pathological processes including development, metabolism, cancer, and musculoskeletal system function (Batista and Chang, 2013;Geisler and Coller, 2013, Meng et al., 2015. Several mechanisms for LncRNA regulation have been characterized, including histone modification (Zhao et al., 2008), transcription factor regulation (Hung et al., 2011), alternative splicing (Tripathi et al., 2010), and competing endogenous RNA (ceRNA) of miRNAs (Salmena et al., 2011;Thomson and Dinger, 2016). By sponging miRNAs, the LncRNAs protect corresponding mRNA from being silenced. Researchers have identified several LncRNAs that regulated osteoblast differentiation by competing miRNA, including ODSM, H19, KCNQ1OT1, XIST, MALAT1, PGC1b-OT1, etc. (Feng et al., 2020;Li et al., 2020;Wang et al., 2018Wang et al., , 2019Wang et al., , 2020Wu et al., 2018;Wu et al., 2019;Yi et al., 2019;Yuan et al., 2019). These findings provided essential theoretical basis for the significance of ceRNAs in manipulating bone formation. However, most competing LncRNAs could only sequester multiple different sorts of microRNAs at random binding sites, which resulted in low efficiency and low specificity. Studies about osteoblastic LncRNA that could interact with one target microRNA with multiple binding sites were relatively limited.
In this study, we discovered an LncRNA that inhibited osteoblast differentiation and bone formation, which was therefore named Lnc-DIF (differentiation inhibiting factor). A special region of thirteen repeats with 53nt length in the trailing end part of Lnc-DIF sequence was illustrated. Lnc-DIF could efficiently sequester In addition, the correlation analysis showed that there was a negative relationship between AK138929 expression and osteogenic marker genes Col Ia1 (collagen type I), Alp (alkaline phosphatase), Ocn (osteocalcin), and Runx2 (runt-related transcription factor 2) in the femur tissue of different ages of C57BL/6 mice ( Figures 1E-1H). These data suggested that AK138929 might function as an osteogenic differentiation inhibiting factor, thus we named it as Lnc-DIF (differentiation inhibiting factor).

ll
Rescue effect of Lnc-DIF siRNA on bone formation was measured using 24-month-old mice (aging osteoporosis model). Lnc-DIF siRNA in vivo transfection significantly decreased Lnc-DIF expression level in BMSCs ( Figure S2D, p < 0.05). The transfection also upregulated bone formation markers Alp and Ocn by 42.5% (p < 0.01) and 657.6% (p < 0.001), respectively, compared with negative control ( Figure S5B). MAR and BFR/BS of si-Lnc-DIF transfection in femoral bone trabecula was increased by 49.9 and 56.7%, respectively (Figures 2F, S5C, and S5D, p < 0.001). MicroCT analysis also proved si-Lnc-DIF recovered BMD and BV/TV in aging osteoporosis mice (Figures 2H and 2J, p < 0.05). These results provided evidences that Lnc-DIF played a role in inhibiting osteoblast differentiation and bone formation in vitro and in vivo.
Lnc-DIF was a potential competing endogenous RNA of miR-489-3p We further explored the mechanism on how Lnc-DIF regulated osteoblast differentiation. Previous studies have revealed that LncRNAs may act as sponges to bind miRNAs and affect their function (Salmena et al., 2011). Intracellularly, Lnc-DIF was distributed in cytoplasm, as detected by fluorescence in situ hybridization (FISH) ( Figures 3A and S6), filling the requirement for ceRNA. The analysis of Lnc-DIF sequence revealed 13 sequence repeats in the trailing end part of Lnc-DIF. This region started at the 604th nucleotide of Lnc-DIF sequence and lasted until its end. Repeat sequence contained 53 nucleotides with the sequence ''CCTGTTCTTGTTAATACTGTATACTACGCATAGATGTTATATGCAGATGTTAT,'' and slight differences were observed in different repeats. Eleven out of the thirteen repeat sequences were predicted to bind with miR-489-3p by RNA22 V2 (sequences see Table 1) (Miranda et al., 2006). The 604 nucleotides that did not contain repeat sequence were designated as ''head'' region, whereas the repeated sequence as ''tail'' region ( Figure 3B).
To determine which domain of Lnc-DIF was responsible for its effect on osteoblast differentiation and miR-489-3p, expression plasmids containing Lnc-DIF head region (head) and Lnc-DIF tail region (tail) were constructed and transfected into MC3T3-E1 cells, with Lnc-DIF overexpression plasmid and empty pCDNA3.1 (+) plasmid used as positive and negative control, respectively. Transfection of Lnc-DIF head region, which had no miR-489-3p binding site, had no effect on miR-489-3p level. In contrast, Lnc-DIF tail region, which contained repeated binding sites of miR-489-3p, significantly decreased miR-489-3p ( Figure 3E, p < 0.001). Luciferase reporter assay and RNA pull-down were performed to evaluate the direct binding between Lnc-DIF and miR-489-3p. For luciferase reporter assay, reporter plasmids containing either a WT or a mutant miR-489-3p binding repeat sequence of Lnc-DIF were constructed, respectively. Reporter plasmids were transfected into MC3T3-E1 cells together with agomiR-489-3p or agomiR-NC. AgomiR-489-3p alone significantly reduced luciferase activity of the WT Lnc-DIF luciferase reporter plasmid ( Figure 3F, p < 0.01).

OPEN ACCESS
Then, luciferase reporter plasmids containing Lnc-DIF full length (Luc-FL), Lnc-DIF head region (Luc-head), and Lnc-DIF tail region (Luc-tail) were constructed and transfected into MC3T3-E1 cells, respectively, with empty pMIR-Report Luciferase plasmid (Luc-vec) used as control. AgomiR-489-3p or antagomiR-489-3p was also applied to these transfected cells with its corresponding control. Results showed that agomiR-489-3p significantly decreased luciferase activity of cells transfected with Luc-FL or Luc-tail. In the cells that were transfected with Luc-head, which had no miR-489-3p binding site theoretically, the luciferase activity was not affected ( Figure 3G, p < 0.001). Conversely, antagomiR-489-3p significantly increased luciferase activity of cells transfected with Luc-FL and Luc-tail. No differences were observed for cells transfected with Luc-head ( Figure 3H, p < 0.001). As for RNA pull-down, interactions of miR-489-3p with both sense and antisense Lnc-DIF were detected, with empty biotin beads (beads) and total RNA (input) as control. Result revealed that Lnc-DIF sense sequence sequestered most of miR-489-3p, compared with other groups ( Figure 3I, p < 0.001).
Regulatory effects of Lnc-DIF to other miRNAs were also investigated. Osteogenic miRNAs miR-20a-5p and miR-210-3p were both predicted to bind with the repeat sequence of Lnc-DIF. Results showed that si-Lnc-DIF had no effect on the expression levels of miR-20a-5p and miR-210-3p. Lnc-DIF overexpression plasmid, as well as plasmid containing Lnc-DIF tail region only showed very slight inhibition to miR-20a-5p and miR-210-3p levels ( Figure S7). All these results suggested that Lnc-DIF might act as a ceRNA of miR-489-3p; it binded with miR-489-3p through its tail region and downregulates intracellular level of miR-489-3p.

Lnc-DIF inhibited osteoblast differentiation and bone formation through sequestering miR-489-3p
Although miR-489 has been widely reported as an inhibitor for multiple cancers (Lin et al., 2017;Yuan et al., 2017), its function on osteoblast differentiation was still unclear. Palmieri et al. reported that miR-489 expression was increased in osteoblast-like cell line MG63 co-cultured by PerioGlas (a material that could enhance osteogenic differentiation) (Palmieri et al., 2008), which implied that miR-489 may have positive regulatory function for osteoblast differentiation. In this study, we transfected MC3T3-E1 cells with ago-miR-489-3p to increase miR-489-3p expression level. Upon transfection, osteoblast differentiation markers  The role of miR-489-3p on bone formation was also investigated by intra-femoral injection. Mice treated by agomiR-489-3p presented enhanced mineral apposition rate and bone formation rate as compared with normal control ( Figures 4E and S9A). Bone formation markers Alp and Ocn expression were also enhanced ( Figure 4G). Consistently, the mice treated by antagomiR-489-3p showed decreased mineral apposition rate, bone formation rate, and bone formation marker expression ( Figures 4F, 4H, and S9B).

Target gene
Subsequently, to further verify the necessity of the miR-489-3p binding site for the function of Lnc-DIF, expression plasmids with Lnc-DIF head and tail region were transfected to MC3T3-E1 cells, and their effects on osteoblast differentiation were investigated. Col Ia1 and Runx2 mRNA expressions in Lnc-DIF tail-region-transfected cells were decreased by 33.1% (p < 0.001) and 52.0% (p < 0.001), respectively, compared with cells transfected by Lnc-DIF head region ( Figures 5B and 5C). ALP activities and mineralized nodules in Lnc-DIF tail-region-transfected cells were also significantly diminished (Figures 5A and S10). The results indicated that Lnc-DIF inhibited osteoblast differentiation by binding with miR-489-3p via its tail region.
Functions of Lnc-DIF head and tail region were further tested in vivo. RT-PCR result revealed that after Lnc-DIF tail region injection into the bone marrow of mice femur, miR-489-3p was decreased by 57.6% (Figure S11A, p < 0.001), and bone formation markers Alp and Ocn in BMSCs were downregulated by 41.8% (p < 0.001) and 85.8% (p < 0.001), respectively, compared with Lnc-DIF head region ( Figure S11B and S11C). The MAR and BFR/BS of femoral bone trabecula in Lnc-DIF tail-region-transfected mice were decreased by 56.9% and 54.3%, respectively, compared with Lnc-DIF head region ( Figures 5D, 5E, and S12, p < 0.01, p < 0.001). The results indicated that Lnc-DIF tail region inhibited bone formation through binding with miR-489-3p.
Regulatory effects of Lnc-DIF to SMAD2 were also investigated by transfecting Lnc-DIF overexpression plasmid and si-Lnc-DIF into miR-489-3p knockout MC3T3-E1 cells (miR-489-KO), along with CRISPR-Cas9 control cells. In control cells, transfection of Lnc-DIF overexpression plasmid caused a 54.6% increment of Smad2 mRNA expression (p < 0.001) as well as protein level and phosphorylated SMAD2 level. This result demonstrated that Lnc-DIF overexpression enhanced SMAD2 expression and activity. However, in miR-489-KO MC3T3-E1 cells, SMAD2 and phosphorylated SMAD2 levels were not dramatically changed by Lnc-DIF ( Figure 7B). On the contrary, transfection of si-Lnc-DIF resulted in decrease of Smad2 mRNA expression by 44.1% (p < 0.001) as well as the protein, suggesting again the positive regulatory effect of Lnc-DIF on SMAD2. However, silencing Lnc-DIF did not influence SMAD2 expression and activity in miR-489-KO cells ( Figure 7C). All these results suggested that Lnc-DIF upregulated SMAD2, and the positive regulation effect was dependent on miR-489-3p.
Subsequently, we transfected expression plasmids with Lnc-DIF head and tail region to MC3T3-E1 cells. Expression level of SMAD2 and phosphorylated SMAD2 levels in Lnc-DIF tail-region-transfected cells were both increased compared with cells transfected by Lnc-DIF head region, and the in vivo result also supported our conclusion ( Figure S19).

Rescue effect of Lnc-DIF siRNA on bone formation in OVX mice
In view of the results presented earlier, we moved forward to investigate the rescue effect of Lnc-DIF siRNA on osteoporosis mice. The OVX mice were separately treated with si-Lnc-DIF or si-NC carried by an osteoblast-targeting delivery system, which was a targeting system involving dioleoyl trimethylammonium propane (DOTAP)-based cationic liposomes attached to six repetitive sequences of aspartate, serine, and serine [(AspSerSer)(6)] for delivering siRNAs specifically to bone-formation surfaces . The animals received three consecutive injections via tail vein every week during the OVX process ( Figure 8A). We found that the expression of Lnc-DIF in mice BMSCs was enhanced by OVX surgery but was strongly downregulated upon siRNA treatment ( Figure 8B). In the contary, miR-489-3p expression was iScience Article downregulated in BMSCs of OVX mice and recovered by si-Lnc-DIF injection ( Figure 8C). In addition, the MAR of femoral cortical bone was decreased by 48.3% (p < 0.001) after the ovariectomy, whereas it was recovered by 31.0% (p < 0.01) after the exposing to si-Lnc-DIF; BFR/BS showed a similar tendency as MAR ( Figures 8D, 8F, and S20). Similarly, microCT results showed that si-Lnc-DIF revealed the deteriorated changes of trabecular bone mass and trabecular microarchitecture caused by OVX ( Figure 8E). Further, bone parameter analysis showed that bone mineral density (BMD), bone volume to tissue volume (BV/TV), and trabecular number (Tb.N) were significantly decreased in OVX mice compared with sham, whereas this decrease was revealed by si-Lnc-DIF. Conversely, the trabecular separation (Tb.Sp) was higher in OVX groups than sham group and lower after treating with si-Lnc-DIF ( Figures 8G, 8H, and S20). These data suggested that si-Lnc-DIF could rescue the minus consequence of menopause, therefore promoting mice bone formation.

DISCUSSION
Osteoporosis is a high incidence bone disease in aging population. The main symptoms of osteoporosis include reduced bone mass, deterioration of bone microstructure, and decreased bone strength, which may increase the risk of fractures (Clarke and Khosla, 2010). Osteoporosis has become an emerging threat to human health in recent years. The current therapeutic strategies for osteoporosis are still limited. Therefore, developing a therapeutic target for osteoporosis would have great significance for prevention and treatment of osteoporosis.
An essential mechanism is that LncRNAs can function as sponges for miRNAs and further inhibit miRNA function (Salmena et al., 2011;Thomson and Dinger, 2016). For example, Wang et al. recently showed that LncRNA ODSM binds with miR-139-3p and promoted osteoblast function and bone formation through miR-139-3p/ELK1 axis . Feng et al. evaluated the LncRNA XIST's function in periodontal ligament stem cells and showed that XIST sponged miR-214-3p to promote osteogenic differentiation (Feng et al., 2020). Zhang et al. reported LncRNA NEAT1/miR-29b-3p/BMP1 Axis promoted osteogenic differentiation of human bone-marrow-derived mesenchymal stem cells . However, functional LncRNAs as ceRNA might have low efficient. In this study, we have identified long noncoding RNA Lnc-DIF that functioned as an inhibitor of osteoblast differentiation and bone formation. Lnc-DIF contained a special structure of repeat sequences in its trailing end, and this repeat sequence can sequester multiple miR-489-3p and efficiently inhibit the positive effect of miR-489-3p/SMAD2 axin on osteoblast differentiation and bone formation. The study provided an important hint that some LncRNAs may contain endogenous repeats and potentially sequester miRNAs. Manipulating the expression levels of these LncRNAs might be a potential therapy for aging and postmenopausal osteoporosis.
On the basis of our previous study, we have further screened LncRNA AK138929 negatively in association with osteogenic differentiation and therefore named it as Lnc-DIF (differentiation inhibiting factor). The functions of Lnc-DIF were further investigated. Experiments in vitro confirmed that Lnc-DIF inhibited osteoblast differentiation (Figure 2), same as the function of Lnc-DIF in vivo. So far, only a few studies reported the function of LncRNA in bone formation. In our previous studies, siRNA of AK016739 and AK045490 were injected subcutaneously over the calvarial surface, and the calvarial bone formation was analyzed Yin et al., 2019). However, aging and postmenopausal osteoporosis mostly occur in weight-bearing bones, especially femur and tibia. In Yuan's study, siRNA was administrated into the bone marrow of mice femur to investigate the function of PGC1b-OT1 on femoral bone formation (Yuan et al., 2019). In this study, Lnc-DIF overexpression plasmid or Lnc-DIF siRNA was injected into LncRNAs could bind with miRNAs and further sequester them. This kind of LncRNAs was defined as ceRNAs (Salmena et al., 2011;Thomson and Dinger, 2016). So far, most LncRNAs that function as ceRNA only have one binding site for one sort of miRNA, which means one LncRNA can only sequester one miRNA. This resulted in low efficiency and low specificity. In our study, we found 13 sequence repeats with slight differences in the trailing end part of Lnc-DIF, 11 of 13 repeats were predicted to bind with miR-489-3p, which suggested that Lnc-DIF may act as an efficient miRNA sponge, and one single molecular of Lnc-DIF would potentially sequester 11 miR-489-3p. The plasmids containing Lnc-DIF repeat sequence were constructed, and the repeat sequence was proved to bind with miR-489-3p and would efficiently decrease miR-489-3p level, which was similar to the function of Lnc-DIF full length (Figure 3). Regulatory effects of Lnc-DIF to other miRNAs were also investigated. miR-20a-5p and miR-210-3p were both predicted binding with Lnc-DIF repeat sequence and promoting osteogenic differentiation. Result showed both Lnc-DIF and its repeat sequence do not affect miR-20a-5p and miR-210-3p levels. The result proved the specificity of Lnc-DIF repeat sequence binding and regulating miR-489-3p.
The function of miR-489 had been widely reported as an inhibitor for cancer (Lin et al., 2017;Yuan et al., 2017), whereas only limited researchers reported its osteogenic functions (Palmieri et al., 2008). In this study, we have firstly proved that miR-489-3p enhanced osteoblast differentiation (Figure 4). With the absence of miR-489-3p, neither Lnc-DIF overexpression nor Lnc-DIF knockdown could affect osteoblast differentiation ( Figure 5). These results confirmed that the function of Lnc-DIF as an osteoblast differentiation inhibitor relied on its sequestering of miR-489-3p. We also firstly proved SMAD2 as the target gene of miR-489-3p. SMAD2 inhibited osteoblast differentiation (Matsumoto et al., 2012), and Lnc-DIF enhanced SMAD2 level through miR-489-3p (Figure 7). These results revealed the mechanism of Lnc-DIF that function as a ceRNA that sequestered miR-489-3p and impeded its inhibiting effect to SMAD2 and therefore inhibited osteoblast differentiation and bone formation.
RNA-based therapy has become an emerging tendency in exploring more methods to rescue bone metabolic diseases. The development of RNA delivery system or bone-specific aptamer would transport RNA sequences to its target region with high efficiency Liu et al., 2015;Wang et al., 2013;Zhang et al., 2012). The technique had made the siRNA of Lnc-DIF as a potential therapy of osteoporosis. We investigated the rescue effect of si-Lnc-DIF to postmenopausal osteoporosis by intravenously injecting si-Lnc-DIF together with an osteoblast-targeting delivery system. The delivery system was developed previously and have been used to deliver nucleic acid specially to osteoblast with low side effects and toxicity Zhang et al., 2012). The results showed that si-Lnc-DIF significantly enhanced bone formation and trabecular microarchitecture in osteoporosis mice (Figure 8), which further proved Lnc-DIF as a potential therapeutic target of osteoporosis.
As an important regulatory factor to bone formation, LncRNAs has received increasing attention of related researchers. However, due to its complicated mechanisms and low homology between different species, LncRNAs were seldom considered as a direct therapy for osteoporosis. In this study, a special repeat sequence was found in Lnc-DIF. This repeat sequence was proved to bind with miR-489-3p and inhibit osteoblast differentiation and bone formation. This implied us that there are more LncRNAs that exist as endogenous sponges for osteogenic miRNAs. Manipulating the expression levels of these sponges would have an amplified effect for regulating their target miRNA levels and would efficiently regulate bone formation. A representative example was our study, in which we utilized the osteoblast-targeting delivery system  to transport si-Lnc-DIF and treated ovariectomized osteoporosis mice. Moreover, one more miR-489-3p sponge has been detected in human genome, and its function has already been determined in vitro. This discovery provide an ideality to RNA-based therapy and might be developed as potential therapeutic strategy of osteoporosis in our further studies.
Taken together, this study has identified Lnc-DIF as an inhibitor for osteoblast differentiation and bone formation. Lnc-DIF was also identified as a ceRNA that sequestered multiple miR-489-3p by its repeat sequence and inhibited osteoblast differentiation and bone formation through manipulating miR-489-3p/SMAD2 axis. This study has revealed a special LncRNA function as high efficient miRNA sponge, which provided more ideas and potential therapeutic strategies for osteoporosis. iScience Article

Limitations of the study
Our study discovered a long noncoding RNA that inhibited bone formation via miR-489-3P/SMAD2; it provided potential therapeutic insight for RNA-based therapy for osteoporosis. However, Lnc-DIF is a micederived LncRNA inhibiting bone formation, which means neither Lnc-DIF sequence nor its siRNA would be used as a therapeutic RNA for human osteoporosis. However, this study has provided us the idea that screening LncRNAs with special repeat sequences in human genome would be a potential way for osteoporosis diagnosis and treatment.

STAR+METHODS
Detailed methods are provided in the online version of this paper and include the following:

Lead contact
Further information and requests for resources should be directed to and will be fulfilled by the lead contact, Chong Yin (yinchong42@nsmc.edu.cn).

Materials availability
All reagents used in this study will be made available on request to the lead contact.

Data and code availability
Microarray data is deposited in a publicly accessible data base and the accession code for this data is E-MTAB-11426 (https://www.ebi.ac.uk/arrayexpress/experiments/E-MTAB-11426) for LncRNA microarray data, and mRNA microarray data is E-MTAB-11425 (https://www.ebi.ac.uk/arrayexpress/experiments/ E-MTAB-11425). Data reported in this paper will be shared by the lead contact upon request. This paper does not report original code. Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request . Cell cultures were maintained at a humidified, 37 C, 5% CO 2 incubator (Thermo Fisher Scientific, Waltham, MA). For osteogenic differentiation treatment, MC3T3-E1 cells at density of 100% were induced by osteogenic medium with a-MEM, 10% FBS, 1% b-glycerophosphate (Sigma, G9422), 1% ascorbic acid (Sigma, A7631) and 1% L-glutamine. The cell cultures were maintained at 37 C with 5% CO 2 , and medium was replaced every 2 days.

EXPERIMENTAL MODEL AND SUBJECT DETAILS
Aging and ovariectomized (OVX) mice was adopted to construct the osteoporosis model. All mice were purchased from the Laboratory Animal Center of the Fourth Military Medical University (Xi'an, China). For aging mice model, 6-month-old male C57BL/6 mice were maintained under standard animal housing conditions (12-h light, 12-h dark cycles and free access to food and water). Mice which were kept until 18 month old were selected as aging group. Mice were euthanized and femurs were collected and processed for bone marrow mesenchymal stem cells isolation.
For OVX mouse model, 2-month-old female C57BL/6 mice were maintained under standard animal housing conditions. The mice were ovariectomized or sham-operated at 3 months of age. Mice were euthanized 38 days after surgery (4 months of age) and femurs were collected. Euthanasia was performed by CO 2 . Mice were euthanized and femurs were collected and processed for bone marrow mesenchymal stem cells isolation. For Lnc-DIF plasmid in vivo transfection, female C57BL/6 mice (4 month old) were randomly divided into four groups (vector, Lnc-DIF, head, tail). For each group, mice were injected into medullary cavity of femur with plasmids formulated with Entranster TM In Vivo Transfection Reagent (Engreen Biosystem Co. Ltd., 18668-11-2, Beijing, China) at the dosage of 40 mL according to the manufacturer's instructions. All mice received the same standard diet during the experimental period. 3 mice from each groups were euthanized 3 days after treatment and femurs from mice were processed for BMSC isolation and RT-PCR. All other Mice were euthanized 15 days after treatment and femurs were processed for histomorphometric analyses .
For Lnc-DIF siRNA in vivo transfection, male aging C57BL/6 mice (24 month old) were randomly divided into two groups (si-NC, si-Lnc-DIF). In the si-Lnc-DIF group, mice were injected into medullary cavity of ll Ltd., 18668-11-2, Beijing, China) at the dosage of 40 mL according to the manufacturer's instructions. In the si-NC group, mice were injected with negative control siRNA mixed with the same technique. All mice received the same standard diet during the experimental period. 3 mice from each groups were euthanized 3 days after treatment and femurs from mice euthanized 15 days after OVX were processed for BMSC isolation and RT-PCR. All other Mice were euthanized 12 days after treatment and femurs were processed for histomorphometric analyses and MicroCT.
For miR-489-3p agomir and antagomir in vivo transfection, 4 month old female C57BL/6 mice were injected into medullary cavity of femur by the same method as Lnc-DIF injection. 3 mice from each groups were euthanized 3 days after treatment and femurs from mice were processed for BMSC isolation and RT-PCR or SMAD2 luciferase reporter assay. Other 3 Mice were euthanized 15 days after treatment and femurs were processed for histomorphometric analyses.

Isolation of bone marrow mesenchymal stem cells (BMSCs)
After sacrifice, mice femoral bones were immediately harvested and attached soft tissues were carefully removed. Bone marrow was washed and collected by flushing several times with phosphate buffered saline (PBS) using a 25G syringe needle. The collected PBS with bone marrow were centrifuged (1200 g, 8 min) and mechanically dissociated by culture medium (a-MEM, Gibco supplemented with 10% fetal bovine serum, 1% L-glutamine, 1% penicillin and streptomycin) using a 29G syringe needle. Then, the suspension was cultured in a 60 mm plate for 3 hours (37 C, 5% CO 2 ) followed by carefully wash with culture medium. Cells were cultured for another 36 hours with culture medium changed every 12 hours. The cells were transferred into a new plate as the 1st-passage cells. Third passage cells were used for characterization and experiments.

mRNA, LncRNA microarray
To investigate the mRNAs and LncRNAs changed with MACF1 (a cytoskeletal protein positively regulate osteoblast differentiation and bone formation via multiple osteogenic transcription factors), RNA from MACF1 knockdown MC3T3-E1 preosteoblasts and negative control cells were used for mRNA and LncRNA microarray. Microarray was performed at RiboBio Co. Ltd. . The fold change of each deferentially expressed mRNA and lncRNA was obtained by log 2 (normalized intensity of treat/normalized intensity of control). Quantile normalization method was used and average of repeated data from the same sample was taken. The p-values were calculated with ANOVA Method.

Screening of osteogenic LncRNAs
The LncRNA-array results may contain hundreds or thousands of lncRNAs, and these lncRNAs were screened by certain standards. LncRNAs that longer than 3500 nt might be hard to insert into vectors and transfect into cells. While, LncRNAs that shorter than 800 nt might be hard to design siRNAs, which made it unable to determine its function. Thus, we only select LncRNAs with length between 800-3500nt. Then, to select lncRNAs that highly correlated with osteogenic differentiation, only LncRNAs that were significantly changed in MACF1-knockdown MC3T3-E1 cell were chosen. RT-PCR was established in normal MC3T3-E1 cells first, and the chosen LncRNAs with a CT value higher than 42 were weeded out. Remaining LncRNAs were selected for co-expression network analysis.

Co-expression network analysis
Co-expression network analysis was adopted to screen the LncRNAs correlated with osteogenic mRNAs . Overexpressed LncRNAs in the microarray data were briefly screened by over-expression fold change and LncRNA length. Then the co-expression metrices were created by computing Pearson's correlation coefficient (PCC) between each screened LncRNA and mRNAs related to osteogenic pathways (Wnt signaling pathway, bmp signaling pathway, TGF-beta signaling pathway, HIF-1 signaling pathway) . The average correlation values of each LncRNA were calculated and LncRNAs with most significant correlation value was selected (Shannon et al., 2003).

Real time PCR
RT-PCR was used to assess expression levels of selected LncRNAs and osteogenic genes. Total RNA was extracted from mouse tissues or cultural cells using Trizol reagent.