The Roles of Long Non-Protein-Coding RNAs in Osteo-Adipogenic Lineage Commitment

Osteoblasts and adipocytes share a common mesenchymal progenitor in the bone marrow. This implies that a reciprocal relationship exists between osteogenic and adipogenic differentiation. Further, cells of osteoblast lineage transdifferentiate into adipocytes under some circumstances. Dysregulation of osteo-adipogenic fate-determination leads to bone diseases such as osteoporosis, accompanied by an increase in bone marrow adipose tissue. Thus, the fine-tuning of osteo-adipogenesis is necessary for bone homeostasis. Osteo-adipogenic progression is governed by a complex crosstalk of extrinsic signals, transcription factors, and epigenetic factors. Long non-protein-coding RNAs (lncRNAs) act in part as epigenetic regulators in a broad range of biological activities, such as chromatin organization, transcriptional regulation, post-translational modifications, and histone modification. In this review, we highlight the roles of epigenetic regulators, particularly lncRNAs, in the osteo-adipogenic lineage commitment of bone marrow mesenchymal stem cells and the adipogenic transdifferentiation of osteoblasts.


Introduction
Bone marrow mesenchymal stem cells (BMSCs) differentiate into multiple different cell types such as osteoblasts, adipocytes, and chondrocytes. BMSCs are also known to play an essential role in maintaining the hematopoietic stem cell niche by providing modulatory signals [1,2]. The bone marrow cavity is filled with active hematopoietic cells (red marrow) in newborns, and these cells are replaced with BMSC-derived adipocytes with age, resulting in the formation of marrow adipose tissue (MAT) [3]. MAT also accumulates in reaction to chemotherapy, glucocorticoid therapy, and radiotherapy [4]. Increased MAT is associated with bone pathologies including decreased cortical bone, low bone mineral density, decreased bone volume, and decreased bone formation rate [5]. Conversely, lipodystrophic mice that carry a hypomorphic allele of PPARγ, a master regulator of adipogenesis [6], show increased bone mass, with a reduction in bone marrow cavity [7]. The current understanding of the mechanism underlying MAT accumulation is that it occurs due to the increased expression of PPARγ which causes the commitment of BMSCs to the adipogenic lineage [8,9]. Thus, an inverse correlation exists between osteogenesis and adipogenesis, and the fine-tuning of the osteo-adipogenic fate-choice is crucial for bone and bone marrow health. Recent discoveries of non-protein-coding RNAs (ncRNAs) have opened an exciting new field of the epigenetic mechanisms involved in fate-choice decisions and differentiation processes in many different cell types [10,11]. For many years, ncRNAs had been considered as accessory components to aid protein activities. Non-coding XIST RNA was first identified to be directly involved in the inactivation of the X chromosome for dosage compensation [12]. Thereafter, multiple approaches, especially computational biology and RNomics, have led to the identification of numerous ncRNAs, such as microRNAs (miRNAs), lncRNAs, small nucleolar RNA, enhancer RNAs, RNAs [10,11]. In this review, we highlight recent studies describing the epigenetic processes by which lncRNAs regulate osteogenic versus adipogenic differentiation in the bone marrow.

Epigenetics and ncRNAs
Lineage commitment of stem cells proceeds via multi-step, hierarchical processes under the control of transcription factors. With the onset of differentiation, stem cell plasticity becomes restricted due to acquisition of epigenetic modifications, such as DNA methylation and histone modifications, which impose cellular identity [13]. DNA methylation is a reversible biochemical process involving the addition of a methyl (-CH3) group to the C-5 position of cytosine in CpG dinucleotides [14]. This process is catalyzed by a family of DNA methyltransferases, including the maintenance methyltransferase DNMT1 and de novo methyltransferases DNMT3A and DNMT3B [14]. Histone modifications are also reversible processes where the N-terminal tails of histone proteins are subjected to post-translational modifications, such as acetylation, methylation, phosphorylation, and ubiquitylation [15]. These modifications are catalyzed by specific enzymes such as histone acetyltransferases, histone deacetylases (HDACs), histone methyltransferases, histone demethylases, and so on [15]. DNA methylation and histone modifications influence one another to form and stabilize higher-order chromatin structure [16].
Recent studies have uncovered the important role of ncRNA transcripts in epigenetic gene regulation. ncRNAs are primarily classified into two categories, short ncRNAs (<200 nucleotides) and long ncRNAs (lncRNAs, >200 nucleotides) [10,11]. Short ncRNAs include miRNAs that are approximately 22 nucleotides in length and usually bind to 3′ untranslated regions of target mRNAs to repress transcription and translation (see [17][18][19][20] for review). The targets of miRNAs are not only protein-coding RNAs but also lncRNAs [21]. DNA methylation regulates the expression of miRNAs, such as miR-10a and miR-200, in human HCT116 cells [22]. In the opposite direction, the miR-290 family of miRNAs controls de novo DNA methylation in mouse embryonic stem cells [23]. LncRNAs regulate gene expression either in a cis-or trans-acting manner through multiple mechanisms ( Figure 1): (1) serve as reservoirs of miRNAs [24,25], (2) act as miRNA sponges (competitive inhibitors of miRNAs) [26], (3) bind to complementary DNA sequences and form RNA-DNA triplex structures that can block the transcriptional process [27], (4) bind to pre-mRNA or mRNA and modulate posttranscriptional events such as splicing, degradation, and translation [28], and (5) recruit chromatin remodeling complexes to specific DNA loci [26,29,30]. (See other excellent reviews for more details, [31][32][33]) Thus, ncRNAs, DNA methylation, and histone modifications are mutually required to regulate their modifiers and chromatin signature, leading to a cascade of complex gene regulation events.
In contrast to the above-mentioned lncRNAs, lncRNA HOXA-AS3 has been identified as a promoter of adipogenesis [92]. The number of HOXA-AS3 transcripts is found to increase gradually during adipogenesis of human MSCs. It interacts with EZH2 to recruit it to the RUNX2 promoter and causes subsequent suppression of RUNX2 and osteogenesis [92]. EZH2 catalyzes the methylation of H3K27 to form a repressive chromatin structure [81], and to repress WNT genes to facilitate adipogenesis [79]. HOXA-AS3 may act on this process.
Thus, complex crosstalk among signaling molecules, ncRNAs, DNA methylases, and histone modifiers is crucial for achieving a balance between osteoblast and adipocyte lineage commitment.

Epigenetics in Osteo-Adipogenic Transdifferentiation
It was believed that differentiation was a unidirectional process and a terminal event to yield a stable cellular identity. However, as demonstrated by induced pluripotent stem cells, terminally differentiated cells can be experimentally reprogramed to function as pluripotent cells [102]. This conversion of cell fate is also observed in vivo during the healing process of tissues, such as intestine, liver, and so on [103]. Bone injury in zebrafish induces mature osteoblasts to dedifferentiate into osteoprogenitor cells that become a source of new reparative bone [104]. Mammalian mature osteoblasts appear not to behave as undifferentiated cells during bone healing processes [105]. However, rat calvaria-derived osteoblasts show a marked plasticity of phenotypes even at the maturation stage [106], suggesting that osteoblasts have the potential to switch their cell fates to other cell lineages by inductive extracellular cues. Indeed, we and other groups have shown fate shifts of osteoblast-lineage cells into adipocytes under certain conditions [107][108][109][110][111]. For instance, a subset of relatively mature rat osteoblasts expressing PPARγ become adipocytes when cultured with the synthetic PPARγ agonist rosiglitazone [108]. Loss of osteogenic WNT/β-catenin signaling in early osteoblast-lineage cells in vivo increases the number of adipocytes [109]. Recently, a subpopulation of CXC chemokine ligand 12-abundant reticular (CAR) cells that expresses mature osteoblast markers was identified [112]. CAR cells are mesenchymal progenitor cells having osteo-adipogenic potential [113]. Thus, a subpopulation of osteoblasts may transdifferentiate to the adipogenic lineage. These findings raise a question about how cell-intrinsic epigenetic programs participate in osteo-adipogenic transdifferentiation.
Epigenetic modifications undoubtedly serve as important regulators of cell plasticity and transdifferentiation potential [114]. Among multiple histone modifications, bivalent modifications of histone H3, through both active and repressive marking by H3K4me3 and H3K27me3, respectively, are thought to play a vital role in keeping genes poised for rapid activation in response to appropriate signals [115][116][117][118]. The presence of these modifications in a subset of genes is thought to correlate with cell plasticity [115][116][117][118]. Recent genome-wide analyses of histone modifications in mouse BMSCs reveal that the most highly expressed genes at stages of osteogenic commitment allow bivalent modifications before expressing them [119]. H3K27me3 removal activates specific genes that bestow osteogenic identities upon cells and reduce their plasticity [119]. Consistent with this and as mentioned above, histone demethylases of H3K27me3 such as KDM4B/5B/6A/6B play a pivotal role in osteogenic commitment [76,77,120,121]. Interestingly, the genes downregulated during osteogenesis rarely gain H3K27me3 [119], suggesting the absence of active mechanisms for repression during osteogenic differentiation. In contrast, EZH2 mediated H3K27me3 is vital for adipogenic lineage specification [120]. Genome-wide comparison of chromatin profiles during the osteo-adipogenic differentiation of BMSCs suggests a default preference for osteogenic differentiation [111]. Gene activation-associated histone modifications in BMSCs are also found in osteogenic differentiated cells, but not in adipogenic differentiated cells, indicating that osteoblasts encode innate plasticity through the epigenome [111]. Thus, the chromatin state of osteoblasts may allow them to switch their cell fate relatively easily.
Further studies are warranted to understand the mechanisms by which epigenetic modifiers, including ncRNAs, regulate the plasticity of osteoblasts. There is an increasing body of evidence suggesting that histone variants and nuclear architectures are involved in cell plasticity [114,122]. The epigenetic program may be even more complicated than has been discovered so far for determining osteo-adipogenic commitment.

Concluding Remarks and Outlook
Many epigenetic factors, including ncRNAs, play an important role in the fate choice of BMSCs and enable crosstalk among multiple signaling pathways. Genome-wide studies show aging-associated depletion of DNA methylation in humans and mice [123][124][125][126]. Since DNMT1 haploinsufficiency promotes age-related health problems, including decreased bone mineral density [127], the genome-wide age-related decrease in DNA methylation may be attributed to a reduction of DNMT1 activity [128]. Similarly, aging-associated alterations are observed in the context of chromatin organization [129]. The histone deacetylase SIRT1 is known to function as a longevity factor, and its deficiency leads to premature aging, in parallel with global enrichment of acetylation at lysine 16 of histone H4 [130]. SIRT1 osteoblast-specific knockout mice show a significant decrease in trabecular bone volume and trabecular thickness [131]. Downregulation of HDAC1 and HDAC2 is also observed during senescence of human MSCs, followed by downregulation of EZH2 and upregulation of KDM6B [132]. These age-related alterations of histone-modifying enzymes contribute to perturbation in heterochromatin structure with aging [128]. Changes in the expression of ncRNAs are also believed to be associated with age-associated diseases and degenerative diseases [133,134]. Cell identity is mediated by epigenetics, so epigenetic aberrations may affect the plasticity of both stem and differentiated cells [135]. This idea is supported by recent evidence that biological senescence and tissue injury promotes epigenetic cellular reprogramming in vivo [136]. Therefore, age-related increases in PPARγ and/or adipogenic miRNA expression may affect not only the fate-choice decisions of BMSCs, but also the plasticity of the osteoblast population [8,137]. Whether osteoblasts convert their fates into adipocytes in vivo still remains controversial. Lineage tracing studies demonstrate the potential of cells in the mesenchymal lineage to transdifferentiate, for instance, hypertrophic chondrocytes transdifferentiate into osteoblasts during the endochondral bone formation and fracture-healing processes [138,139]. Forced expression of transcription factors and ncRNAs bestows the ability upon cells to directly convert from one cell type to another in vivo [140,141]. Given these observations, epigenetic alterations with age may induce transdifferentiation of osteoblasts into adipocytes, which may explain age-associated MAT accumulation. We believe that functional studies of ncRNAs and their interacting partners during osteo-adipogenesis may open new prospects for the prevention of MAT accumulation and bone loss in the future.

Conflicts of Interest:
The authors declare no conflict of interest.