Application of microRNA in Human Osteoporosis and Fragility Fracture: A Systemic Review of Literatures

MicroRNAs (miRNAs) could serve as ideal entry points to the deregulated pathways in osteoporosis due to their relatively simple upstream and downstream relationships with other molecules in the signaling cascades. Our study aimed to give a comprehensive review of the already identified miRNAs in osteoporosis from human blood samples and provide useful information for their clinical application. A systematic literature search for relevant studies was conducted in the Pubmed database from inception to December 2020. We set two essential inclusion criteria: human blood sampling and design of controlled studies. We sorted the results of analysis on human blood samples according to the study settings and compiled the most promising miRNAs with analyzed diagnostic values. Furthermore, in vitro and in vivo evidence for the mechanisms of the identified miRNAs was also illustrated. Based on both diagnostic value and evidence of mechanism from in vitro and in vivo experiments, miR-23b-3p, miR-140-3p, miR-300, miR-155-5p, miR-208a-3p, and miR-637 were preferred candidates in diagnostic panels and as therapeutic agents. Further studies are needed to build sound foundations for the clinical usage of miRNAs in osteoporosis.


Introduction
Osteoporosis affects more than ten million Americans over age 50, leading to an estimated two million incident fractures and a total cost of $19 billion per year in the U.S. [1,2]. Often manifested as a silent disease, osteoporosis remains underdiagnosed and thus undertreated [1]. Currently, dual-energy X-ray absorptiometry (DXA) is still the most widely used technique for bone mineral density (BMD) assessment. However, the measurement by DXA could not provide enough clues to the underlying mechanisms resulting in osteoporosis. MiRNAs, in contrast, are widely studied in recent years and its measurement may provide vital information of the epigenetic environment [3,4].
MiRNAs are small non-coding endogenous RNA molecules (around 19-25 nucleotides in length), which regulate post-transcriptional gene expression [5]. The biogenesis of miRNAs is regulated at multiple levels, from initial miRNA transcription and processing by RNase Drosha in the nucleus, further processing by RNase Dicer and modification in the cytoplasm, loading onto the RNA-induced silencing complex (RISC) as a functional unit, to finally RNA decay [6]. After incorporation into the RISC, miRNAs usually act as a guide that base-pair with the 3 untranslated region (3 UTR) of their target mRNAs, and then the double-stranded miRNA-mRNA complex induces the subsequent translational repression and mRNA degradation, thereby silencing the target mRNAs [7]. The alteration of miRNA expression could therefore passively suggest the mechanism of a disease, and it is also possible to control the disease actively by producing a change to the regulation of miRNA.
It is challenging to measure the expression level of intracellular miRNAs. Instead, extracellular miRNAs are easily detectable, and the measurement is meaningful since they may serve as mediators for intercellular communication. Studies indicate that extracellular miRNAs come from three main routes [8,9]. First, they could be transported by extracellular vesicles, either by exocytosis or direct budding of the plasma membrane. Second, they could bind to specific proteins such as lipoprotein and ribonucleoproteins, then secreted in the form of protein-miRNA complexes. In addition, they may also come from damaged or dead cells.
In a previous study, the majority (>80%) of the studied miRNAs is found in various tissues, but there are still some miRNAs found to be tissue-specific [8]. In the domain of osteoporosis, we hope to identify the miRNAs involved in the pathways of bone homeostasis such as transforming growth factor-beta (TGF-β)/bone morphogenic protein (BMP) signaling for osteoblast differentiation and the osteoprotegerin (OPG)/receptor activator of nuclear factor Kappa-B ligand (RANKL)/RANK pathway for osteoclast differentiation [10,11]. In addition to identification, miRNAs could serve as ideal entry points to explore the full picture of deregulated gene expression in osteoporosis and further highlight the major dysfunctional pathways according to the etiology.
As more and more researchers are devoted to studying miRNAs in osteoporosis, evidence from in vitro, in vivo experiments and clinical trials is getting abundant. Reviews on this topic published in recent years started to focus on specific issues rather than giving a general description on the function of miRNAs. For, instance, a meta-analysis published in 2019 examined the miRNAs as potential biomarkers for postmenopausal osteoporotic patients, and a recent review particularly described the strength and weakness in practical usage of miRNAs [3,12]. For a step toward clinical applications, we focused on the studies that included human blood samples and demonstrated the research findings step by step. Demonstrating the results from the evidence of direct increased or decreased expression of the miRNAs in human blood samples, the in vitro evidence for the possible mechanisms of the above miRNAs, to the evidence that confirms the pathways in animal models, our study aims to target the miRNAs with top priorities for further clinical studies or usage.

Searching Strategy
We systematically searched the Pubmed database to screen relevant articles from inception to December 2020, without restriction to language. The following key terms with Boolean operators were adopted to search articles: ("micro RNA" OR miRNA OR miR) AND (osteoblastogenesis OR "osteoblast differentiation" OR osteoclastogenesis OR "osteoclast differentiation" OR osteoporosis OR osteoporotic) AND (serum OR blood sample OR circulating) AND (patient OR participant). Titles and abstracts of all the identified articles in the database were screened for potential studies. Next, full texts of the potential studies were further examined. The screening procedure was performed independently by two reviewers (Y.Z.W. and C.H.C.). Any discrepancy was solved by discussion until the two reviewers reached a consensus.

Inclusion and Exclusion Criteria
Studies were included if the following criteria were met: (1) clinical trials that examined miRNA in the field of osteoporosis; (2) a study population with at least a group of osteoporotic patients and a group of controls, without restriction to gender; (3) diagnosis of osteoporosis confirmed by DXA or clinical low-impact fracture (fragility fracture); (4) collected samples with at least human blood samples. The exclusion criteria were as follows: (1) studies that included other fields such as osteonecrosis; (2) a control group with other known bone diseases such as osteoarthritis.

Study Selection
The flow diagram of the selecting process is shown in Figure 1. Initially, 130 relevant articles were identified from the Pubmed database. Titles and abstracts of all the articles were screened, and 64 of them were considered as eligible. We further reviewed the full texts of the remaining studies, of which 51 studies met our inclusion criteria and were included in our study.

Study Characteristics
The main characteristics of the included studies are shown in Table 1. A least 3834 patients were included in this study (four studies did not provide sample size), and 89% (2475/2784) of the patients were female according to the available data. Mean age of the participants ranged from 34.0 to 85.8, which varied widely due to the study setting. Extracted outcomes included the analysis on human blood samples, as well as the results of in vitro and in vivo experiments.

Serum miRNAs as Clinical Potential Biomarkers for Human Osteoporosis
To be a useful clinical examination, the following three elements are usually required: availability (i.e., convenience), appropriateness (i.e., fulfilling indications), and diagnosability. For availability, we set human blood sampling as an essential inclusion criterion for its convenience in clinical sampling compared to bone biopsy. For appropriateness, we arranged the studied miRNAs according to their study settings. For diagnosability, diagnostic values of specific analyzed miRNAs were presented, and we recommended usage of multiple miRNAs in combination for a more comprehensive clinical judgement.
In our included studies, up to 851 miRNAs were detected in a single study for screening. Most of the studies made the detection of potential miRNAs in one round, while five studies carried out a discovery analysis with a small sample size using microarray or quantitative real time polymerase chain reaction (qPCR) to recognize the most deregulated miRNAs at first, followed by a validation analysis that tested the potential miRNAs with a larger sample size for confirmation [14][15][16]21,29]. Some studies further assessed the diagnostic values of selected miRNAs, and a receiver operating characteristic (ROC) curve analysis was often performed.
All included studies had a group of osteoporotic patients and a control group, but several distinct additional controlled variables were adopted by individual study, including menopause, fracture, and advanced age. The settings were designed by the researchers to identify the target miRNAs precisely under specific conditions, but we should still interpret the results with caution for at least the following two reasons. One is that circulating miRNAs in the serum are the combination of the physiological and pathological actions from all kinds of organs and tissues in the body. The other one is that the presentation may vary with different stages of the pathological actions, such as a compensatory effect in an unbalance homeostasis (not the primary cause). Trying not to over-interpret the results of a single miRNA, we recommended the usage of a combination of identified miRNAs for diagnosis, which was often referred to as miRNA signature, and tissue-specific miRNAs were preferred.
We used Venn diagrams to demonstrate the relationships of the main independent variables (menopause, fracture, aging) and dependent variable (osteoporosis) in Figure 2. The border between two areas is representative of the difference between them. The ultimate goal is to figure out the: (1) outer border of osteoporosis: distinguish osteoporotic patients from normal people who have a similar risk factor (2) inner demarcation within osteoporosis: recognize the etiologies among osteoporotic patients Figure 2. Schematic diagram of the de-regulated miRNAs by clinical settings. Take the red curve, for instance; it demarcates the difference in miRNA expression between postmenopausal and premenopausal groups in osteoporotic women; that is, the red curve represents the changes by estrogen deficiency leading to osteoporosis.

General Biomarkers for Distinguishing Osteoporotic Patients from Non-Osteoporotic Controls
The following two study settings were categorized in this group. Without additional controlled variables, the identified miRNAs in this group could hardly provide clues for specific etiology.

•
Comparison between osteoporotic patients and controls [16,20,22  Schematic diagram of the de-regulated miRNAs by clinical settings. Take the red curve, for instance; it demarcates the difference in miRNA expression between postmenopausal and premenopausal groups in osteoporotic women; that is, the red curve represents the changes by estrogen deficiency leading to osteoporosis.

General Biomarkers for Distinguishing Osteoporotic Patients from Non-Osteoporotic Controls
The following two study settings were categorized in this group. Without additional controlled variables, the identified miRNAs in this group could hardly provide clues for specific etiology.
It is noteworthy that conflicting results of miR-21 regulation were found by different studies, and it is the same case in the clinical setting comparing postmenopausal osteoporotic patients to postmenopausal non-osteoporotic controls. Further studies are needed to investigate its regulation.

miRNAs That Are Potentially Associated with Estrogen
Estrogen is a well-known important regulator in female osteoporosis, and the interactions between miRNAs and estrogen or its receptor have been studied extensively. For instance, miR-18a, miR-22, and miR-206 are found to target estrogen receptor (ER) α, and 17β-estradiol also regulates the expression of various miRNAs by several ER-mediated signaling pathways [64].

•
Comparison between postmenopausal osteoporotic patients and controls [25,28,37,39,46,60] • Comparison between postmenopausal osteoporotic patients and premenopausal controls [21,48] There were six significantly up-regulated and five significantly down-regulated miR-NAs in osteoporotic patients compared to the control group identified and listed in Table 3. To confirm the relationship between the identified miRNAs and bone metabolism further, miR-28, miR-373, and miR-101 were also found to be down-regulated in the bone tissue by qPCR analysis, and the expression of miR-133a in the serum was found to be negatively correlated with lumbar BMD [25,37].

Studies That Aim to Attenuate the Influence of Estrogen Itself
The design of this setting contains at least two purposes. One is to offset the regulation of miRNAs influenced by menopause; the other one is to identify the most common pathogenesis in the postmenopausal women.

•
Comparison between postmenopausal osteoporotic patients and postmenopausal non-osteoporotic controls [13,15,17,33,42,43,58] There were 12 miRNAs found to be significantly up-regulated and three miRNAs significantly down-regulated in osteoporotic patients compared to the control group. Tang et al. 2019 conducted an enzyme-linked immunosorbent assay (ELISA) to detect serum levels of Sfrp1 and TNF-α, and a similar trend with miR-144 was observed [43]. In Suarjana et al. 2019, analysis revealed that miR-21 was positively correlated with serum RANKL level and RANKL/OPG ratio, and it also negatively correlated with TGF-β1, OPG, and BMD in the postmenopausal osteoporotic group. They further carried out a linear regression analysis, and the following relationship was documented [42]. BMD = 1.373 − 0.085 × Ln. miR-21 − 0.176 × Log 10 RANKL (1)

Studies That Aim to Attenuate the Osteogenic Effect After a Recent Fracture
The design of this setting has two basic considerations. One is to cancel out the regulation of miRNAs by a common osteogenic effect after a recent fracture; the other one is to distinguish the reactions to fractures in osteoporotic patients from the controls.

•
Comparison between osteoporotic patients with fracture and non-osteoporotic controls with fracture [14,30,38,54] There were 14 miRNAs found to be significantly up-regulated and one significantly down-regulated in the osteoporotic groups compared to control groups, and they were listed in Table 5. Seeliger et al. 2014 and Wang et al. 2018 further tested the identified miRNAs in bone tissues, and the results showed that an up-regulation of miR-21, miR-23a, miR-24, miR-24-3p, miR-27a-3p, miR-100, miR-122a, and miR-125b was consistently observed in both blood samples and bone tissues [14,30].

miRNAs That Potentially Associate with Osteoporosis in Elderly People
Aging is one of the important risk factors resulting in osteoporosis. From the cellular aspects, cell senescence leads to telomere shortening, a change in gene expression, epigenetic regulators, and protein processing [66]. Many miRNAs have been identified in other age-related diseases. MiR-146, miR-155, miR-21, and miR-126, for example, are found to be helpful for differentiating patients with cognitive impairment from the age-matched controls [67].
This setting aimed to offset the ubiquitous influence by advanced age; meanwhile, the difference in the miRNA expressions may represent the common deregulated pathways in elderly people.
MiR-96 and miR-107 were up-regulated in osteoporotic patients compared with healthy controls, and the serum level of miR-96 was also significantly higher in the elderly group compared to the young group [23,26].

miRNAs with High Diagnostic Value for Osteoporosis
ROC analysis was performed in nine studies for assessing the diagnostic value of the identified miRNAs with the greatest potential as osteoporotic biomarkers. The result is shown in Table 6. Besides using a single miRNA as a biomarker for osteoporosis, the results of Mandourah et al. 2018 suggested that miR-122-5p and miR-4516 be used together to increase the diagnostic value [27]. In Kocijan et al. 2016, the results of a multivariate model revealed that the highest predictive power was reached by using a combination of miR-155-5p, miR-181c-5p, miR-203a, miR-330-3p, and miR-942-5p (AUC: 0.97) [19].
In Shuai et al. 2020, two indices were developed to distinguish osteoporosis patients from healthy controls, shown as follows [56]:

Mechanisms of the Identified miRNAs
To study the mechanism of the potential miRNAs identified by blood sample analysis, the role of the miRNAs was explored step by step.
In an in vitro experiment, several types of cells were commonly used. Mesenchymal stem cells (MSCs) such as bone marrow mesenchymal stem cells (BMSCs) or adiposederived mesenchymal stem cells (ADSCs) were cultured for studying the differentiation of osteoblast, and peripheral blood mesenchymal cells (PBMCs) for osteoclast differentiation. The first step was to determine whether either the osteoblastic or osteoclastic pathway was the selected miRNA involved. This was usually established by transfecting the cells with miRNA mimics or inhibitors to alter expression of the miRNA, followed by detecting the activities of the cells related to bone metabolism (ex. alizarin red stain and alkaline phosphatase (ALP) staining) and the expression of specific transcriptional regulators (dualluciferase reporter gene assay) or proteins (western blot analysis). Next, the target of the selected miRNA was usually predicted by tools such as Targetscan, miRanda or miRWalk software. The final step was to verify the hypothesized pathway. It was usually performed by transfecting the cells with miRNA mimics or inhibitors, and co-transfection with siRNA for the target gene, followed by detecting the expression of involved molecules and making a deduction from the results.
In an in vivo experiment, animals were usually injected with miRNA mimics or inhibitors, and tools such as micro-CT and many bone parameters were used for assessing the difference in bone architecture in the experiment group compared to the control group.

miRNAs Involved in Osteoblastogenesis
The relationship between the miRNAs and the involved reactions is illustrated in Figure 3 [68] and Table 7. We demonstrated the miRNAs from the involvement in important signaling pathways to transcriptional regulators, and the promotive or inhibitive effects on osteoblastogenesis were described after every single miRNA.
Int. J. Mol. Sci. 2021, 22, 5232 18 of 30 cells with miRNA mimics or inhibitors to alter expression of the miRNA, followed by detecting the activities of the cells related to bone metabolism (ex. alizarin red stain and alkaline phosphatase (ALP) staining) and the expression of specific transcriptional regulators (dual-luciferase reporter gene assay) or proteins (western blot analysis). Next, the target of the selected miRNA was usually predicted by tools such as Targetscan, miRanda or miRWalk software. The final step was to verify the hypothesized pathway. It was usually performed by transfecting the cells with miRNA mimics or inhibitors, and co-transfection with siRNA for the target gene, followed by detecting the expression of involved molecules and making a deduction from the results.
In an in vivo experiment, animals were usually injected with miRNA mimics or inhibitors, and tools such as micro-CT and many bone parameters were used for assessing the difference in bone architecture in the experiment group compared to the control group.

miRNAs Involved in Osteoblastogenesis
The relationship between the miRNAs and the involved reactions is illustrated in Figure 3 [68] and Table 7. We demonstrated the miRNAs from the involvement in important signaling pathways to transcriptional regulators, and the promotive or inhibitive effects on osteoblastogenesis were described after every single miRNA.   Wnt proteins comprise a family of secreted glycoproteins that play a central role in osteoblast differentiation. When Wnt binds to its principal receptor (frizzled protein) and co-receptor, multiple intracellular signaling cascades are activated, including the canonical β-catenin-dependent pathway and noncanonical β-catenin-independent pathway. In the canonical β-catenin-dependent pathway, β-catenin translocates into the nucleus and interacts with several transcriptional factors, thereby stimulating gene expression. Besides, Wnt signaling is regulated by various antagonists, such as Dickkopf 1 (Dkk1), Dkk2, and Sclerostin (Sost) proteins; secreted frizzled-related protein 1 (Sfrp1) inhibits the formation of Wnt-frizzled complexes by directly binding to the Wnt ligand [73].
In our included studies, there were eight miRNAs found to be involved in the regulation of Wnt signaling pathways.

TGF-β Pathway
TGF-β is secreted and stored in the extracellular matrix. Activated TGF-β binds to the tetrameric receptor complex, which is composed of a TGF-β type I receptor (TβRI or ALK5) and type II receptor (TβRII). Downstream signaling included canonical (Smad-dependent) and non-canonical (non-Smad-dependent) pathways. In the canonical pathway, R-Smad (Smad2 or 3) form complexes with Smad4 and regulate gene expression. Smad7, as a regulatory molecule, competes against Smad2 or 3 for binding to Smad4 [10,76].
In our included studies, there was one miRNA found to be involved in the regulation of the TGF-β signaling pathway.

•
MiR-300 targets Smad3, inhibiting osteogenic differentiation. Moreover, micro-CT for evaluation in a rat model showed that miR-300 injections led to lower bone parameters (BMD and BV/TV) compared to sham and negative control groups [52].

BMP Pathway
BMP is vital for embryonic skeletal development and bone homeostasis after birth. When BMP binds to its ligands, type II receptors will form a complex with type I receptors, leading to transphosphorylation of the type I receptors and signal transduction. Most BMPs activate the canonical BMP pathway (Smad-dependent), and BMP-2, 4, 5, 6, 7, and 9 are found to induce osteoblast differentiation and bone formation actively [10,76].
In our included studies, there was one miRNA involved in the regulation of the BMP pathway.

•
MiR-410 targets BMP-2, inhibiting osteoblast differentiation. Besides, the up-regulation of miR-410 was found in both postmenopausal osteoporotic patients compared to healthy controls and in the OVX mice group compared to the sham group [46].

Common Transcriptional Pathway Runt-Related Transcription Factor 2 (RUNX2)
RUNX2 is an essential transcriptional factor in the osteoblast differentiation. The effect of RUNX2 depends on its interaction with other DNA sequences or proteins that may bind to the various domains of RUNX2 [68]. Studies revealed that Runx2 could up-regulate the expression of genes encoding bone matrix proteins, such as Col1a1, Spp1, Ibsp, bone gamma-carboxyglutamate protein (Bglap), and Fn1 [77].
In our included studies, four miRNAs were found to target RUNX2.

Osterix
Osterix, also known as Sp7, is a transcriptional factor specific for osteoblast. It plays an important role in regulating the gene expression during differentiation of the pre-osteoblast into mature osteoblast. On the other hand, it was regulated by the BMP2 signaling pathway and insulin-like-growth-factor (IGF) pathway. The BMP2/Smad pathway activates RUNX2, which then activates the expression of osterix, whereas the IGF pathway activates osterix in a RUNX2-independent manner [78].
In our included study, there were three miRNAs found to target osterix.
• MiR-27a-3p, miR-96, and miR-637 target osterix, inhibiting osteogenic differentiation [26,59]. The experiment of a mice model showed that repetitively agomiR-96-injected young mice had significantly decreased BMD compared with vehicle-treated mice, and aged mice treated with antagomir-96 had higher bone strength compared to controls [26]. In addition, the signal transducer and activator of transcription 3 (Stat3) was found to be a pseudo-target of miR-637 by biological experiments [70].

Other Reactions in Osteogenic Differentiation
Histone Deacetylase (HDAC) HDACs are the component of transcriptional co-repressors complexes that regulate gene expression. HDAC4 is regulated by PTH, and the HDAC4 inhibitor may facilitate osteoblast differentiation [79,80].

Phosphatase and Tensin Homolog (PTEN)/Phosphoinositide 3-Kinases (PI3K)/AKT Signaling Pathway
AKT is inhibited in osteoblastic cells due to the abundance of PTEN within it. Although the PTEN/PI3K/AKT pathway is not a dominant signaling pathway in osteoblastic differentiation, inhibition of PTEN activity leads to increased AKT activation and subsequent cell proliferation [81].

•
MiR-19b targets PTEN, promoting osteogenic differentiation. Moreover, bone parameters including BMD, bone volume, and trabecular number were significantly higher in ovariectomized mice injected with agomiR-19b than those of the negative control group [57].

ATF3
Activating transcription factor 3 (ATF3) is a transcriptional factor belonging to the ATF/cAMP response element-binding protein (CREB) family. ATF3 expression could be up-regulated by TNF-α through the JNK signaling pathway, and overexpression of ATF3 would in turn inhibit osteoblast differentiation [82].

Glutaminase (GLS)
Glutamine could serve as a material in protein synthesis and also an energy source. Our included study showed that the uptake of L-glutamine increased with time during osteogenic differentiation induction at 7 and 14 days, suggesting that glutamine plays an important role in the differentiation [41].

•
MiR-200a-3p targets glutaminase, inhibiting osteogenic differentiation [41] Activin A Receptor Type I (ACVR1) A study found that ACVR1-null mice had decreased expression levels of Wnt inhibitors Sost and Dkk1; canonical Wnt signaling was then increased and facilitated osteogenic differentiation. A hypothesis of the BMP7-ACVR1-SOST/DKK1 axis in osteoblasts was then proposed [83]. However, our included study showed opposite results. ACVR1 was found to correlate positively with BMP2, and downregulation of ACVR1 therefore led to suppression of osteoblastogenesis [71].
Proprotein Convertase Subtilisin/Kexin Type 5 (PCSK5) PCSK belongs to a family of subtilisin-like serine proteinases and activates various precursor proteins and peptides. PCSK5-knockout mice have VACTERL (vertebral, anal, cardiac, trachea-esophageal, radius or renal, limb) syndrome-like malformations, and a recent study indicates that PCSK5 is expressed in mice osteoblast with osteopontin (OPN) as one of its substrates [84,85].

miRNAs Involved in Osteoclastogenesis
The relationship between the miRNAs and the involved reactions is illustrated in Figure 4 [11] and Table 8. To assess the activity of osteoclast, tartrate-resistant acid phosphatase (TRAP) staining is often used in the studies.

RANK
RANKL is a well-known important factor for osteoclast differentiation. It binds to RANK and in turn induces recruitment of the tumor-necrosis-factor-receptor-associated factor 6 (TRAF6) protein, which activates downstream signaling pathways. NF-κB signaling was then activated, which is essential for osteoclast differentiation [89]. Nevertheless, OPG secreted by osteoblasts and other cells could bind to RANKL to prevent osteoclast formation, and MafB is a regulatory molecule that inhibits the RANKL pathway [90,91].
In our included study, there were four miRNAs found to be involved in the RANKL signaling pathway.

Colony Stimulating Factor-1 Receptor (CSF1R)
CSF1R, a type III receptor tyrosine kinase, plays an important role in the differentiation of myeloid cells, cancer development, and progression of various diseases [92]. When the colony stimulating factor (CSF) or interleukin-34 (IL-34) binds to CSF1R, it induces PI3K/AKT signaling [92,93]. Downstream transcription factors are in turn activated, including NFATc1 and NFkB, which are essential to osteoclast differentiation and function.
In our included study, there were three miRNAs found to be involved in the PTEN/PI3K/ AKT signaling pathway.
Inhibitor of Nuclear Factor Kappa-κ Kinase Subunit β (IKKβ) Gene The inhibitor of nuclear factor kappa-κ kinase (IKK) complex is crucial for the activation of classical NF-κB signaling pathways, which is usually induced by TNF, IL-1, or RANKL. In the classical NF-κB signaling pathway, two main components of the IKK complex are involved, IKKβ and IKKγ, which in turn degrade IκBα [97,98].

miRNAs with Both Documented Diagnostic Values as Serum Biomarker for Osteoporosis and Identified Underlying Mechanisms
Based on the results of included studies, miRNAs with a verified diagnostic value as a serum biomarker and identified mechanisms were listed in Table 9 and regarded as preferred candidates for diagnostic panel or therapeutic agents. There are six miRNAs meeting the criteria without conflicting results: miR-23b-3p, miR-140-3p, miR-155-5p, miR-208a-3p, miR-300, and miR-637. It is reasonable to use a combination of these miRNAs in a panel to aid the diagnosis of osteoporosis. However, the deregulation of a single miRNA is insufficient to ascertain the etiology of osteoporosis unless the miRNA is confirmed tissue-specific and not involved in other types of bone activities.
As potential therapeutic agents, the six miRNAs are preferred owing to their confirmed mechanisms and also their more noticeable deregulated expression level compared with all the other studied miRNAs. The higher the level of deregulated expression of miRNA, the more feasible the detection for diagnosis and monitoring after treatment.
It is noteworthy that Feurer et al. 2019, in a study with 682 women included, yield interesting findings. In the study, a number of selected miRNAs are found to be associated with fragility fractures, BTMs, BMD, and microarchitecture by comparing postmenopausal with premenopausal women, but the effect is negated after an age adjustment [99]. It is not disappointing because we can draw two inferences from the findings. First, the expression levels of some miRNAs are confirmed to be significantly different between postmenopausal and premenopausal groups despite the inability to recognize the etiology (estrogen or age). Second, the difference negated by the age adjustment means that the deregulated pathways are age-dependent, or, they may be related to both estrogen and age. The findings of the study underline the importance of study design (i.e., study setting), as the effort our study puts forth in part 3. If we interpret the results of studies with caution, every single finding could contribute to the final successful clinical usage.

Conclusions
With accumulating evidence verifying the association between deregulated miRNA expression and osteoporosis, it is necessary to analyze the collected data for further useful application in clinical settings. Previous studies mentioned the difficulty in integrating the results in clinical trials due to a lack of adequate controls or varied study settings despite the fact that there was already substantial experience and knowledge learned from in vitro or in vivo experiments [3,4]. Considering both the study setting and diagnostic value, our study demonstrated the most promising miRNAs as biomarkers for osteoporosis and evidence of mechanisms. We recommended miR-23b-3p, miR-140-3p, miR-300, miR-155-5p, miR-208a-3p, and miR-637 as preferred miRNAs for candidates in diagnostic panels and as therapeutic agents. In terms of the small sample size in each clinical study and little overlapping of the identified miRNAs among different studies, further studies are needed to build sound foundations and consensus for the clinical application of miRNAs.