Circulating miRNAs as Diagnostic and Prognostic Biomarkers in Common Solid Tumors: Focus on Lung, Breast, Prostate Cancers, and Osteosarcoma

An early cancer diagnosis is essential to treat and manage patients, but it is difficult to achieve this goal due to the still too low specificity and sensitivity of classical methods (imaging, actual biomarkers), together with the high invasiveness of tissue biopsies. The discovery of novel, reliable, and easily collectable cancer markers is a topic of interest, with human biofluids, especially blood, as important sources of minimal invasive biomarkers such as circulating microRNAs (miRNAs), the most promising. MiRNAs are small non-coding RNAs and known epigenetic modulators of gene expression, with specific roles in cancer development/progression, which are next to be implemented in the clinical routine as biomarkers for early diagnosis and the efficient monitoring of tumor progression and treatment response. Unfortunately, several issues regarding their validation process are still to be resolved. In this review, updated findings specifically focused on the clinical relevance of circulating miRNAs as prognostic and diagnostic biomarkers for the most prevalent cancer types (breast, lung, and prostate cancers in adults, and osteosarcoma in children) are described. In addition, deep analysis of pre-analytical, analytical, and post-analytical issues still affecting the circulation of miRNAs’ validation process and routine implementation is included.


Introduction
Cancer represents a significant global cause of death, so early diagnosis with a constant checkup of disease progression and patient response to chemotherapy and surgery are essential to efficiently manage subjects with cancer. Up to now, tissue biopsies have been considered one of the most valuable methods for cancer diagnosis, but they are an invasive procedure. To overcome these limits, great efforts have been made in the past few years to find novel, minimally invasive, fast, reproducible, and low-cost techniques for cancer management, such as circulating biomarkers in human biofluids (i.e., blood and derived plasma or serum). Currently, circulating microRNAs (miRNAs) are promising markers for cancer diagnosis, prognosis, for monitoring the treatment response, and as powerful tools for personalized approaches.

MiRNA Biogenesis and Biological Functions
MiRNAs, a category of non-coding small RNA (~22 nucleotides) [1,2], act as modulators of gene expression in human cells and tissue, and their potential as biomarkers for human pathologies has been highlighted by their detection in 12 types of human fluid [3][4][5]. In addition, miRNA expression has been characterized in 61 types of tissue [6]. MiRNA genes can be sited within both protein-coding or non-coding gene sequences in all human chromosomes, and their biogenesis includes multiple stages [7][8][9][10][11]. First, in the nucleus, miRNAs are transcribed and processed into a hairpin precursor (pre-miRNA), by the DROSHA-DGCR8 complex, which is translocated into the cytoplasm and cleaved by DICER obtaining, a double-stranded miRNA. Both strands are, or only one strand is, selected for incorporation into the RNA-induced silencing complex (RISC) to act as mature miRNAs and, thus, to bind the "seed" region of the target mRNA. MiRNAs are involved in a wide range of regulatory functions in both biological and pathological processes: cell/tissue development, differentiation, metabolism, homeostasis, proliferation and apoptosis, immune responses, and cancer development, progression, and response to therapy [12]. MiRNA biogenesis and functions can be affected by: (i) epigenetic modifications, (ii) alteration in protein activity involved in this process [13][14][15][16], (iii) single nucleotide polymorphisms (SNPs) in miRNA genes [17,18], (iv) hereditary mutations in the miRNA seed region [19][20][21]; (v) somatic mutation in miRNA genes [22,23], and (vi) somatic copy-number variation of miRNA genes [24].

Circulating miRNAs as Biomarkers
In cancer, a biomarker can give diagnostic, prognostic, or predictive information about the considered pathological condition, and it must be highly specific, sensitive, and easily measurable. In addition, a biomarker must be validated by clinical studies and should be useful for patient management [27]. Circulating miRNAs show both strengths and weaknesses as biomarkers in each stage of their validation process (preanalytical, analytical, and postanalytical). Considering the preanalytical phase, miRNAs are detectable in human biofluids, collected with a minimally invasive procedure, and their transport into extracellular vesicles or binding to proteins makes them stable in several conditions (extreme pH values, repetitive freeze and thaw, up to 24 h at room temperature, or for decades at <−70 • C) [4,5,[28][29][30][31][32][33]. On the other hand, this stage of validation can be affected by several biases concerning patients' daily habits (smoking, physical activity, kidney pathologies, diet, and circadian rhythm) [34][35][36][37][38], sample collection (type of tubes and anticoagulant additive as well as the phlebotomy technique), and handling (sample centrifugation force and time) [39][40][41][42]. For instance, our newest results demonstrated that plasma from ethylendiaminotetraacetate (EDTA) salts-anticoagulated tubes with a gel separator resulted in higher detectability and stability of a wide panel of miRNA than standard EDTA plasma and platelet-poor EDTA plasma [43]. In the analytical phase, reliable methods for miRNA measurement are available, with real-time quantitative polymerase chain reaction (RT-qPCR) considered as the gold-standard technique, but significant interplatforms (PCR-based, microarray, and next-generation sequencing (NGS)) differences have been highlighted [44,45]. In this phase, circulating miRNA validation can be affected by non-circulating miRNA contamination (skin and blood cells, and activated platelets) and by hemolysis [28,33,39,42,[46][47][48][49]. Finally, issues regarding the postanalytical phase mainly concern the absence of a standardized data-normalization process (single or multiple exogenous or endogenous genes, global mean of all evaluated miRNAs) [32,50,51]. Therefore, the use of circulating miRNAs as biomarkers in patient management requires standardized protocols and procedures for each phase of their validation [28,42,50].

Aim
Based on the potentialities of miRNAs as biomarkers, enormous efforts have been made in studying and in defining the existing relationships between miRNA-altered expression and human pathological conditions, particularly in oncology, as demonstrated by the huge number of articles published in the last few years on this topic. Due to this growing interest, the aim of this review was to comprehensively collect available data (published until the end of 2018) about current and potential next uses of circulating miRNAs as biomarkers in the diagnosis and prognosis of a subset of solid tumors. Specifically, we focused our attention on high-incidence tumors and on those tumors in which miRNA-based diagnostics has gained the most promising results. In particular: lung cancer, due to high mortality in both genders and its increasing incidence in women; prostate and breast cancers, due to their high incidence in males and females, respectively; and osteosarcoma, since it represents one of the most common solid tumors in children. Information was obtained from each paper claiming diagnostic, prognostic, and/or predictive potential of circulating miRNAs, also including any available information about pre-analytical precautions adopted, quantification platforms used, and normalization strategy applied. Where available, we also collected information about the sensitivity and specificity of specific circulating miRNA signatures to evaluate their clinical potential as biomarkers to assess the presence of a disease and, at the same time, the absence of the same pathology in healthy individuals. In addition, the plot of specificity vs. sensitivity, namely, the receiver-operating-characteristic (ROC) curve, and the respective area under the curve (AUC), which represents the degree of separability and is an effective accuracy measure, were considered [52]. An AUC of 1 refers to perfect separability, whereas a value of 0.5 shows no capacity to discriminate the datasets. Articles with methodological deficiencies were excluded from the review.
Besides the described distinction between free (mainly protein-associated) and exosome-/ microvescicle-/LDL-associated circulating miRNAs and their distinct functions, from biomarker-like (free circulating miRNAs) vs. hormone-like (encapsulated circulating miRNAs) [53], based on the aim of this review, only free-circulating miRNAs were considered.

MiRNAs as Tumor Biomarkers
Aberrant expression of circulating miRNAs as diagnostic and prognostic biomarkers in several types of cancer is a topic of interest. Altered levels of circulating miRNAs have been associated with tumor growth, progression, metastasis, and drug resistance, suggesting their potential use as a tool to optimize patients' therapies. As a general rule, oncogene miRNAs are upregulated in tumor samples, and consequently in patients' blood, whereas tumor-suppressor miRNAs are downregulated.

Circulating miRNAs and Lung Cancer (LC)
Worldwide, lung cancer (LC) is a leading cause of cancer mortality in both males and females. LC can be ascribed at two main types: small-cell lung cancer (SCLC) accounting for 15% of cases, and 85% are represented by non-small-cell lung cancer (NSCLC). NSCLC is characterized by two predominant subtypes: adenocarcinoma (ADC) and squamous cell carcinoma (SCC). To date, methods for early LC diagnosis include chest X-rays, low density computed tomography (LDCT), and other imaging techniques. In addition, several potential circulating biomarkers have been identified: the carcinoembryonic antigen (CEA), cytokeratin-19 fragment (CYFRA21-1), neuron-specific enolase (NSE), cancer-associated antigens CA125 and CA19-9, and chromogranin A, together with molecular signatures as mutations in KRAS and TP53 genes. However, these methods have shown limited specificity, sensitivity, and reproducibility for LC detection.

Circulating miRNAs as Biomarkers for LC Diagnosis and Prognosis
One of the first papers aimed at discovering novel circulating miRNAs as biomarkers for LC, starting from a screening of 880 mature miRNAs, identified miR-1254 and miR-574-5p as potential serum biomarkers for early-stage NSCLC due to their upregulation in NSCLC compared to healthy subjects. Their combination gave an AUC of 0.75 with 73% sensibility and 71% specificity for early-stage NSCLC [54]. MiR-21, previously reported as an oncogene in several human cancers, was identified by microarray analysis out of 427 miRNAs as the most upregulated miRNA in serum from NSCLC patients compared to healthy volunteers. Its levels are associated with worse prognosis and adverse clinicopathological features (grade and metastasis) [55]. MiR-21 was found in NSCLC tumor samples, other than in association with the tissue upregulation of miR-141 and miR-200c [56], and the downregulation of miR-486 in both tissue and plasma [57]. MiR-21 tissue expression is correlated with tumor size, while its circulating levels are correlated with stage and lymph-node metastasis. ADC is associated with higher miR-21 and lower miR-200c tissue expression than SCC. Moreover, tissue expression levels of miR-21, miR-141, and miR-200c as well as serum miR-21 levels, are inversely correlated with overall survival (OS) [56]. In addition, plasma miR-21 increase and miR-486 decrease gave an AUC of 0.740 and 0.857, respectively, in LC that increased to 0.901 if the miR-21-to-miR-486 ratio was considered [57]. MiR-21 upregulation in NSCLC samples is also correlated with a poor response to platinum-based chemotherapy after tumor resection [58].

Circulating miRNAs as Markers for Metastatic LC, Tumor Recurrence, and Response to Adjuvant Therapies
Some papers specifically focused their attention on the identification and validation of circulating miRNAs as biomarkers for metastatic NSCLC. miR-222, miR-183, and miR-126 were significantly downregulated in serum samples of Stage IV NSCLC patients compared to healthy controls, but only miR-183 and miR-126 maintained this trend when compared to Stage I/II patients, thus indicating that these two miRNAs could be useful as serum biomarkers for more advanced NSCLC [129]. Conversely, plasma miR-422a upregulation discriminated LC patients with lymphatic metastasis from non-metastatic patients (AUC = 0.79) [130].
Half of the NSCLC patients that undergo complete tumor resection may suffer local or distant recurrence. Consequently, to reduce this risk, patients are treated with adjuvant therapies. The whole miRNome of NSCLC patients monitored after tumor resection could be useful to reveal metastasis occurrence [131]. Early-stage ADC patients with tumor recurrence, metastasis, and not submitted to adjuvant therapy showed increased serum levels of miR-142-3p and miR-29b compared to the non-recurrence group that is associated with reduced OS [132]. Of these, only miR-142-3p, alone or in combination with the tumor stage, remained an accurate serum biomarker for recurrence risk in patients with poor outcome that received adjuvant therapy (AUC of 0.64 and 0.78, respectively). In addition, miR-486, miR-221, miR-30d, and miR-1 altogether are correlated with shorter DFS in NSCLC patients after tumor resection [133].
Platinum-based chemotherapy combined with radiotherapy represents the first-line treatment for locally advanced NSCLC. Circulating miRNAs have been investigated over their predictive potential of response to chemotherapy. High levels of miR-125b have been found in the serum of NSCLC patients resistant to cisplatin-based chemotherapy compared to nonresistant, and its levels have been associated with cancer stage and differentiation status [134]. Plasma miR-150 and miR-29a levels are inversely correlated with administered radiotherapy dose in Stage IIIA NSCLC patients [135]. Dynamic changes of plasma miR-125b and miR-19b expression levels, and miR-125b/miR-19b ratio are valuable markers for discriminating NSCLC patients responsive to chemotherapy [136]. Another study suggested that elevated plasma miR-613, miR-495-3p, miR-302e, and miR-98-5p levels might mark sensitivity to radiotherapy in NSCLC [137]. Serum let-7 levels change in response to radiotherapy in NSCLC patients, and up-or downregulation of this miRNA is correlated with cancer-cell proliferation rate during therapy [138]. Eleven serum miRNAs, miR-205-5p, miR-200b-3p, miR-155-5p, miR-145-5p, miR-134miR-126-3p, miR-125b-5p, miR-92a-3p, miR-34a-5p, miR-22-3p, and miR-10b-5p, combined with clinical factors, have been used to create a predictive dose response score (DRS) for NSCLC patients treated with radiotherapy [139]. Low DRS and high-dose radiation therapy are associated with longer OS, and reduced risk of distant metastasis and tumor progression, compared to patients treated with a standard dose of radiation.

LC miRNome
A valuable retrospective study investigated the potential of serum miRNome as a prognostic and diagnostic tool for LC. The authors compared the serum miRNA profile of LC patients to those obtained immediately or many years before diagnosis. The serum miRNA profile of samples taken after LC diagnosis was more comparable to that from samples collected closest prior to diagnosis than that collected years before. Consequently, the serum miRNA profile seemed to change with the development of the tumor. Moreover, the serum miRNA signature of healthy controls was more similar to that of patient samples collected many years before diagnosis (four deregulated miRNAs) compared to that closer to or after diagnosis (19 and 29 deregulated miRNAs, respectively) [142]. Through NGS, 32 known miRNAs and 7 novel miRNAs were identified as significantly altered in WB from LC patients compared to healthy subjects [143]. Using a similar approach, a 34-serum miRNA signature discriminated both asymptomatic early-stage and advanced-stage NSCLC patients from healthy smokers (AUC: 0.89 for Stage I NSCLC, and 0.88 for Stage II-IV NSCLC) [144]. A 24-plasma miRNA panel able to discriminate early-stage NSCLC patients from healthy controls was identified and validated (AUC, sensibility, sensitivity: 0.92, 83%, 80%). The diagnostic accuracy of this panel increased when considered in combination with age, sex, and smoking habits (0.94, 83%, 84%) [145].
Five, 10, and 50 blood-based miRNA sets discriminated with elevated accuracy, specificity, and sensitivity NCSLC patients from healthy controls (95% accuracy and AUC of 0.98 for five markers; 95% accuracy and AUC of 1.00 for 10 markers; 98% accuracy and AUC of 0.99 for 50 markers) [146]. In addition, the whole miRNome of NSCLC patients monitored after tumor resection could be useful to reveal metastasis occurrence [131].
For LC, one of the most studied circulating miRNAs as diagnostic and prognostic biomarker is miR-21. Interestingly, all studies that evaluated the circulating levels of this miRNA showed increased levels of miR-21 in LC patients vs. the control group, with AUC from 0.59 to 0.87, despite the different employed preanalytical protocols (matrix: plasma vs. serum; centrifugation speed and length) and analytical methods (RT-qPCR, NGS, microarray followed by RT-qPCR).
All information about the circulating miRNAs associated with LC are summarized in Table 1.

Circulating miRNAs and Breast Cancer (BC)
Breast cancer (BC) is the most prevalent tumor in women, and mammography is considered the gold-standard method for its diagnosis; however, mammography presents some limitations, including the use of ionizing radiation and the possibility of false-positive results. Clinical decisions about BC treatment depend on its molecular features, such as the estrogen receptor (ER), the progesterone receptor (PR), and the human epidermal growth factor receptor 2 (HER2). To date, the most used marker in clinical routine for detecting BC is CA 15-3, but it presents low sensitivity. Other serum markers, such as CEA and TPS, are even less sensitive than CA 15-3. Based on these facts, novel biomarkers for BC diagnosis and prognosis are desirable.
Previous evidence about deregulated miRNAs found in other cancer types, in BC tumor tissue, and BC-derived cell lines has given the basis for novel investigations in the field of BC diagnosis and prognosis. One of the first reports validated the upregulation of miR-195 and let-7a as discriminating BC cases from healthy controls. However, only miR-195 showed the same altered regulation in BC tissue compared to the non-cancerous surroundings. Moreover, blood miR-195 is correlated with tumor stage and lymph-node metastasis. Higher circulating levels of miR-10b and miR-21 have been observed in ERcompared with ER + BC patients [154]. BC tissue and serum are characterized by the upregulation of miR-155, miR-106a, and miR-21, and the downregulation of miR-335, miR-199a, and miR-126. Their relative expression, but not miR-106a, is strictly associated with tumor grade and ER/PR expression [155]. MiR-155 is highly accurate in discriminating BC patients from healthy controls, with AUC, sensitivity, and specificity of 0.80, 65%, and 81% [156].
The BC luminal-A subtype is ER + and/or PR + , HER2 − and Ki67 < 14%. These patients are characterized by better prognosis but are less responsive to chemotherapy than other BC subtypes. Elevated serum levels of miR-205 and miR-19a in luminal-A BC patients treated with NCT (epirubicin and paclitaxel) are correlated with chemotherapy-resistance onset, and the combination of these two miRNAs can effectively predict chemosensitivity and discriminate resistant patients from the sensitive group [193]. A 5-serum miRNA panel (miR-328-3p + miR-199a-3p + miR-195-5p + miR-25-3p + let-7i-5p) diagnosed luminal-A BC with an AUC of 0.94 [190]. Plasma miR-155, miR-21, miR-10b (upregulated compared to healthy controls) and let-7a (downregulated) turned out to be potential biomarkers for luminal-A BC monitoring. In fact, after surgery, chemotherapy, and radiotherapy, their levels normalized [194].

BC miRNome
The BC blood miRNome analysis can be a future powerful approach for early identification of BC. A prospective study provided complete serum miRNA screening from women that later developed BC or remained cancerfree, and identified a panel of 21 differently expressed miRNAs [219]. Another study identified a 41-tissue miRNA signature able to predict BC with an accuracy of 63%-83%. Of these, 20 miRNAs were detected and validated in serum, and turned out to predict BC occurrence within 18 months from blood collection with an accuracy of 53% [220].
In the case of BC, one of the most studied circulating miRNAs as diagnostic and prognostic biomarker is miR-21. All BC studies that evaluated the circulating levels of miR-21 showed increased levels of this miRNA in BC patients vs. control groups, with an AUC from 0.69 to 0.93, despite the different employed preanalytical protocols (matrix: plasma vs. serum; centrifugation speed and length) and analytical methods (RT-qPCR, microarray followed by RT-qPCR).
All information related to the above-described circulating miRNAs is reported in Table 2.

Circulating miRNAs and Prostate Cancer (PC)
One of the most prevalent malignancies in males is prostate cancer (PC). Currently, the combination of prostate-specific antigen (PSA) quantification and digital rectal examination are considered the gold standard for the early detection of PC. Unfortunately, the diagnostic ability of PSA is limited by low specificity that frequently gives false-positive results in patients with benign prostatic hyperplasia (BPH). Based on this consideration, novel biomarkers with higher specificity and sensibility are needed.
Evidence suggests that single circulating miRNAs may have poorer diagnostic power than PSA in PC; hence, combined panels or specific ratios have been studied. The serum miR-519c-5p + miR-345 + miR-19a/b miRNA signature effectively discriminated high-and low-risk PC (AUC = 0.94) [227], while the combination of WB miR-155 + miR-145 + miR-141 + let-7a gave an AUC of 0.78. Compared to BPH, PC patients displayed an upregulation of oncomirs miR-155, miR-145, and miR-141, and a downregulation of tumor-suppressor let-7a; their tissue deregulation is associated with increased risk of PC development [228]. The combination of plasma let-7c, miR-375, miR-141, miR-30c (downregulated in PC), and PSA better discriminated PC from BPH and healthy controls (AUC: 0.78 and 0.88, respectively) than PSA alone [229]. Plasma miR-375 and miR-21, which are upregulated in PC compared to BPH, when combined with PSA have higher diagnostic potential (AUC, sensitivity, specificity: 0.88, 88%, 75%) than those given by each single miRNA (0.76, 75%, 75% for miR-375; 0.80, 88%, 75% for miR-21). Of these, miR-21 is correlated with tumor stage and metastasis [230]. Interestingly, in these two studies, the levels of miR-375 were found both upregulated and downregulated in PC vs. BPH patients despite their evaluation being performed in the same matrix (plasma) and with similar analytical and postanalytical protocols.
The same plasma miR-21 upregulation has been associated with miR-106 and miR-20a upregulation, and miR-223 downregulation. MiR-106a/miR-130b and miR-106a/miR-223 ratios, alone or in combination, effectively discriminate between lPC and BPH patients (AUC: of 0.81, 0.77, and 0.84, respectively) with specificity higher than that of PSA [231]. A subsequent study identified another plasma miRNA panel for PC diagnosis, composed of miR-4289, miR-326, miR-152-3p, and miR-98-5p, all upregulated in PC patients compared to healthy controls. ROC analysis revealed that each miRNA of this panel has diagnostic potential, but accuracy is highly increased for the whole panel (AUC = 0.82-0.95) [232]. However, according to the Cancer Genome Atlas (TCGA) dataset, miR-152-3p expression in PC tissue may be downregulated compared to adjacent normal tissue. Serum miR-326 upregulation, in accordance with its tissue expression, marks the risk for tumor recurrence, together with miR-185-5p and miR-221-3p, after radical prostatectomy [233].
Specific circulating miRNAs associated with other cancer types or deregulated in PC tissue have been tested in PC patients in order to identify potential diagnostic and prognostic circulating biomarkers. MiR-139-5p was upregulated in WB from PC patients compared to BPH, accurately discriminated PC patients from BPH and healthy subjects (AUC: 0.94 and 0.92, respectively), and is correlated with advanced and more aggressive tumor stages [234]. Serum miR-410-5p levels are higher in PC patients compared to healthy controls or non-PC patients, and are correlated with high-risk PC development. ROC analysis revealed that elevated miR-410-5p accurately discriminated PC from non-PC/healthy subjects (AUC = 0.81), and high-risk from low-risk subjects (AUC = 0.71) [235]. WB miR-18a, a previously identified oncomir, discriminated PC patients from both BPH and healthy groups (AUC = 0.81), and its high circulating levels are correlated with more aggressive tumor behavior and advanced clinical stage [236]. Another set of serum miRNAs (let-7c, let-7e, let-7i, miR-26a-5p, miR-26b-5p, miR-25-3p, miR-18b-5p), found altered in PC tissue, discriminate PC patients from BPH. Of these, miR-25-3p, miR-18b-5p, and their combination gave the highest sensitivities and specificities in predicting PC (AUC: 0.79, 0.87, and 0.92, respectively) [237]. Based on the TCGA database, circulating miR-200b and miR-200c have been validated as potential diagnostic markers for PC (AUC: 0.57 and 0.62, respectively). Compared to healthy controls, in plasma from PC patients, miR-200b was downregulated and is correlated with bone metastasis, whereas miR-200c was upregulated and is associated with more aggressive behavior [238]. Serum upregulation of miR-375, miR-141-3p, miR-106b, miR-34a-5p, and miR-21 also discriminated between PC patients and healthy subjects. Among these, miR-375 and the combination miR-375 + miR-141-3p + miR-21 have been identified as the most accurate diagnostic tools for PC (AUC, sensitivity, specificity: 0.91, 100%, 75% for miR-375 alone, and 0.86, 93%, 63% for the combination) [239].
A combination of interacting plasma miRNAs (miR-17/miR-192) and of three other independent miRNAs (miR-181a, miR-150a, and miR-22) discriminated between aggressive and nonaggressive PC. Interestingly, low miR-192 and high miR-17 circulating levels are the most likely to present aggressive PC due to a synergic effect of the inhibition of tumor-suppressing target genes (by miR-17) and induction of oncogenes (by miR-192) [240]. Later, it was highlighted that both up-and downregulation of circulating miR-17 are correlated with more aggressive PC, thus underlining the importance of this miRNA in PC behavior [241].
The standard chemotherapy for CRPC is based on docetaxel; serum miR-21 is elevated in docetaxel-resistant CRPC patients [249]. A 6-serum/plasma miRNA panel (miR-301b, miR-222, miR-200c, miR-200b, miR-146a, miR-20a) is correlated with the response to docetaxel in CRPC. Particularly, increased serum levels of miR-200 family members before docetaxel treatment or lower/unchanged levels of miR-17 family members after docetaxel treatment are associated with poor response to chemotherapy and reduced OS [250]. After the validation stage, high plasma levels of miR-429, miR-375, miR-200a, miR-200b, miR-200c, and miR-132 before chemotherapy have been correlated with shorter OS without any association with the response [251].
In the past few years, active surveillance for an early detection of PC patients has increased, but conventional markers do not effectively differentiate PC stages. A 3-serum miRNA panel (miR-375, miR-223, miR-24,) discriminates indolent from aggressive PC (AUC: 0.69 increased to 0.70 when combined with PSA) [252].
All information about circulating miRNAs in PC is reported in Table 3.

Circulating miRNAs and Osteosarcoma
The most common human primary malignant bone tumor in young adults and children is osteosarcoma [253].

Circulating miRNAs as Biomarkers for Osteosarcoma Diagnosis and Prognosis
Several studies, aimed at evaluating novel circulating miRNAs as potential biomarkers for the diagnosis, prognosis, and treatment of osteosarcoma patients, started from the validation of miRNAs that were previously demonstrated to play a crucial role in the cell differentiation, proliferation, and tumorigenesis of various cancer types. MiR-21 is one of the most studied miRNAs due to its association with several types of cancer. MiR-21 is upregulated in serum samples from osteosarcoma patients compared to their healthy counterparts, and it is also correlated with advanced Enneking stages, reduced tumor response to chemotherapy, and short OS [254]. This finding is in accordance with increased plasma levels of miR-21 in another osteosarcoma cohort. The same study also identified miR-199a-3p and miR-143 as downregulated in osteosarcoma subjects compared to healthy subjects, and ROC analysis revealed that the combination of these three plasma miRNA signatures successfully discriminated between osteosarcoma and healthy subjects with AUC of 0.95, 91% sensitivity, and 94%, specificity [255]. Significantly higher serum levels of miR-199a are also found in preoperative osteosarcoma patients compared to healthy subjects, which returned to normal values after tumor resection. MiR-199a shows high diagnostic accuracy for osteosarcoma (AUC: 0.86, 88% sensitivity, 77% specificity) [256]. A case-control study identified that miR-34b plasma levels in osteosarcoma patients was significantly decreased in comparison to healthy subjects, as well as in metastatic vs. non-metastatic patients; therefore, its expression is associated with increased osteosarcoma risk [257]. Serum miR-9 was also identified as a potential diagnostic and prognostic marker, since it is upregulated in osteosarcoma compared to healthy subjects and is correlated with aggressiveness features, advanced tumor stage, size, and metastasis presence [258]. Members of the miR-29 family are aberrantly expressed in several types of human cancer. MiR-29a, miR-29b, and miR-29c are upregulated in osteosarcoma patients' sera compared to normal control levels, and circulating miR-29a and miR-29b expression is associated with tumor grade, recurrence, metastasis presence, and OS, thus indicating their possible use as biomarkers for human OS prognosis, development, and progression [259]. Two other miRNAs are correlated with clinicopathological features and unfavorable prognosis: miR-196a and miR-196b, which belong to the human miR-196 cluster. The serum levels of these miRNAs are higher in osteosarcoma patients' sera compared to healthy subjects, and are associated with tumor grade, metastasis development, and recurrence, reduced OS, and DFS. Furthermore, their coexpression is a potentially independent biomarker for patient survival and prognosis [260]. ROC curve analysis identified plasma miR-148a as a circulating biomarker upregulated in osteosarcoma patients compared to healthy subjects (AUC, sensitivity, specificity: 0.78, 70%, 83%), and is correlated with reduced OS and DFS, tumor size, and metastasis [261]. MiR-195, another potential diagnostic and prognostic serum biomarker for OS, with 88% sensitivity and 83% specificity, is associated with an advanced clinical stage, distant metastasis formation, and reduced OS and DSF [262]. Using a different approach, a panel of four upregulated miRNAs (miR-374a-5p, miR-320a, miR-199a-3p, miR-195-5p), out of 739 screened, was identified in the plasma samples of osteosarcoma patients (AUC, sensitivity, specificity: 0.96, 91%, 94% in distinguishing osteosarcoma patients from healthy controls). MiR-199a-3p and miR-195-5p levels are also associated with metastasis and decrease after tumor removal, whereas miR-320a and miR-199a-3p are related with osteoblastic subtypes [263]. Results about miR-199a-3p and miR-195-5p are in contrast with two previous studies that observed their downregulation in osteosarcoma patients' blood samples [255,262]. This discordance can be due to the different analytical (RT-qPCR alone vs. microarray + RT-qPCR) and postanalytical (normalization: cel-miR-39 or snRNA U6) approaches used for miRNA evaluation and data elaboration. The profile of 752 plasma miRNAs in osteosarcoma patients revealed the downregulation of miR-205-5p and the upregulation of miR-574-3p, miR-335-5p, and miR-214, compared to healthy subjects, with good discrimination potential (AUC: 0.70, 0.88, 0.78, and 0.80). Furthermore, plasma miR-214 levels are associated with the presence of metastasis [264]. MiR-27a aberrant expression is associated with several types of cancer. Osteosarcoma patients' serum samples show increased expression levels of this miRNA that is also correlated with aggressive clinicopathological features: advanced clinical stage, distant metastasis, and poor response to chemotherapy. Moreover, serum miR-27a levels can distinguish osteosarcoma patients from control subjects with an AUC of 0.87, 70% sensitivity, and 98% specificity [265]. Wang and colleagues identified miR-191 as able to distinguish between healthy and osteosarcoma subjects with an AUC of 0.86, 74% sensitivity, and 100% specificity. MiR-191 is significantly overexpressed in osteosarcoma patients' serum compared to healthy subjects, and it is associated with negative clinicopathological features and prognosis in osteosarcoma patients: advanced stage, large tumor size, metastasis presence, as well as reduced OS and DFS [266].
Several miRNAs aberrantly expressed in other cancer types have been evaluated as potential markers in osteosarcoma diagnosis and for identifying individuals with negative prognosis. MiR-152, for instance, gives AUC of 0.96, 93% sensitivity, and 97% specificity. It is downregulated in serum samples from osteosarcoma patients compared to healthy subjects, and is correlated with distant metastasis and shorter survival [267]. MiR-221 plays a crucial role in osteosarcoma occurrence and progression, and it is significantly upregulated in osteosarcoma patients' serum compared to healthy subjects. High miR-221 serum levels are correlated with stage, metastasis, and shorter survival. Moreover, it reveals osteosarcoma presence with elevated AUC, sensitivity, and specificity (0.84, 66%, and 100%) [268]. Circulating miR-451a, miR-425-5p, miR-139-5p, miR-106a-5p, miR-25-3p, miR-20a-5p, and miR-16-5p are all downregulated in OS patient samples (AUC: 0.80, 0.78, 0.71, 0.73, 0.80, 0.85, and 0.77). Unfortunately, these data are less robust than those of previous studies due to missing information about their correlation with clinicopathological features [269]. The potential oncogenic role of miR-421, miR-17, miR-542-3p, miR-300, miR-222, and the tumor-suppression role of miR-223, miR-326, miR-95-3p, miR-491-5p, miR-375, miR-101, and miR-let-7a, whose circulating expression levels are associated with several types of cancer, have been studied in osteosarcoma patients' samples. Plasma miR-421, upregulated in osteosarcoma, is associated with poor OS [270], similarly to serum miR-17, which is also associated with worse OS prognosis [271]. Plasma miR-542-3p displays high discriminatory capacity between osteosarcoma patients and healthy subjects (AUC, sensitivity, specificity: 0.84, 78%, 94%), and survival analysis also revealed strong correlation between miR-542-3p levels and DFS and tumor stage [272]. Oncogenic behavior was also revealed by circulating miR-300, whose serum levels are increased in osteosarcoma patients; it effectively identifies osteosarcoma (AUC, sensitivity, specificity: 0.89, 84%, 89%; positive-negative predictability: 90% and 85%). Furthermore, it is associated with advanced clinical stage and metastasis, and independently predicts poor prognosis (reduced OS and DFS) [273]. Another unfavorable independent predictor marker for osteosarcoma OS and DFS has been identified in serum miR-222. It is strongly upregulated in osteosarcoma patients compared to healthy subjects and is associates with metastasis, tumor size, and clinical stage (AUC, sensitivity, specificity: 0.81, 67%, 84) [274]. The downregulation of serum miR-223 in osteosarcoma patients compared to healthy subjects is inversely correlated with clinical stage, metastasis development, and directly with OS (AUC, sensitivity, specificity: 0.92, 86%, 97%) [275]. Serum miR-326 downregulation in osteosarcoma patients compared to healthy subjects (AUC = 0.90, 84% specificity, 95% sensitivity) is associated with negative clinicopathological features (distant metastasis and advanced clinical stage) and poor prognosis [276]. Analogously, serum miR-497 downregulation in osteosarcoma (AUC = 0.85) is associated with reduced OS, metastasis presence, advanced tumor clinical stage, and poor response to chemotherapy [277]. Circulating miR-95-3p can also discriminate between osteosarcoma patients and healthy subjects with an AUC of 0.86. In addition, low serum levels of miR-95-3p independently predict OS and are associated with osteosarcoma development and progression [278]. Serum miR-491-5p upregulation discriminates osteosarcoma and healthy subjects with high accuracy (AUC, sensitivity, specificity: 0.83, 72%, 86%). Moreover, serum miR-491-5p levels are strongly associated with clinical stage and distant metastasis. Survival analysis reveals that reduced serum miR-491-5p expression is significantly associated with negative prognosis and independently predicts osteosarcoma [279]. Convincing evidence reveals that circulating miR-101 may be a moderately accurate biomarker for osteosarcoma diagnosis and prognosis. It is dramatically downregulated in serum samples from osteosarcoma patients and is powerfully associated with negative clinicopathological features (advanced tumor stage and metastasis) and worse prognosis (shorter OS and DFS). After tumor removal and therapeutic treatment, patients without metastasis showed higher miR-101 values, thus indicating that this circulating miRNA can be useful for monitoring patients' chemotherapy response. The diagnostic potential of serum miR-101 for osteosarcoma is underlined by ROC analysis (AUC = 0.85, 76% sensitivity, and 83% specificity), while survival analysis revealed that miR-101 may be an independent and negative prognostic factor [280]. MiR-let-7a, quantified in total blood of osteosarcoma patients, is considered a potential diagnostic (AUC = 0.90) and negative prognostic biomarker for osteosarcoma [281]. In a small Mexican cohort, microarray analysis of 648 serum miRNAs revealed that only miR-642a-5p and miR-215-5p are upregulated in osteosarcoma, with AUC of 0.84 and 0.87 [282].

Circulating miRNAs as Markers for Metastatic Osteosarcoma, Tumor Recurrence, and Response to Adjuvant Therapies
The expression pattern of miR-21 in osteosarcoma patients has been deeply investigated before and after chemotherapy in order to explore its correlation with chemosensitivity [283]. Consistent with other studies [254,255], miR-21 is remarkably upregulated in serum from osteosarcoma patients in comparison with healthy subjects and, after tumor removal, its serum level before and after chemotherapy is significantly lower in nonresponders than in responders. Moreover, PDCD4, a miR-21 target, is upregulated in serum from patients after chemotherapy [283]. The combined downregulation of miR-133b and miR-206 is associated with aggressive tumor progression, metastasis formation and recurrence, and impaired response to chemotherapy [284]. Low serum miR-375 expression level in osteosarcoma patients is associated with advanced clinical stage, important tumor size, metastasis formation, and poor response to chemotherapy. It discriminates osteosarcoma patients from control subjects (AUC, sensitivity, specificity: 0.89, 82%, 75%), and poor from good tumor drug response (AUC, sensitivity, specificity: 0.83, 84%, 84%) as well as a negative prognostic index [285].

Osteosarcoma miRNome
Fujiwara et al. screened the serum miRNome in osteosarcoma patients in comparison to nonosteosarcoma patients with other benign tumors and healthy volunteers. The serum concentration of miR-25-3p in osteosarcoma patients is higher than in non-osteosarcoma and healthy groups, while serum concentrations of miR-17-5p was increased only in relation to healthy volunteers [286], which was consistent with a previous observation [271]. ROC analysis revealed that both serum miR-25-3p and miR-17-5p levels strongly discriminated osteosarcoma patients from healthy individuals (AUC, sensitivity, specificity: 0.87, 71, 92% and 0.72, 64%, 85%, respectively). Elevated miR-25-3p levels predict negative prognosis and OS, and are correlated with metastasis at diagnosis [286].

Conclusions
In tumor biology, miRNAs can act as oncogenes or tumor suppressors and their altered tissue expression, as well as their altered circulating levels, has been correlated with the occurrence of several cancers. Therefore, miRNAs are considered novel potential molecular diagnostic and/or prognostic non-invasive tools. However, besides the (sometimes confliting) evidences discussed above, this correlation has been reported for multiple but not for all tumors. Indeed, the level of diaseaseand tumor-associated circulating miRNAs is affected by several factors including, but not limited to, age, gender, ethnicity, diseases and comorbidities, drugs, smoking, diet, physical activity, and lifestyle [36,289,290]. For instance, miR-122 is elevated in 70% of hepatocellular carcinoma (HCC) cases, but it is also increased in case of hepatitis B infection or liver injury [291] and it fluctuates in response to therapy in chronic hepatitis C [292]. Similarly, platelets significantly contribute to the circulating miRNA profile; therefore, any intervention therapy that targets platelets, including common antiplatelet therapies for cardiovascular disorders, can alter the circulating miRNA profile [47]. On the same basis, given the biological variability (as both intra-and inter-individual variability) of circulating miRNA and the plethora of variables affecting their levels (including the stage of the treatment, commonly used drugs as NSAIDs corticosteiroids), the current evidence suggests that miRNAs may not yet be usable as universal markers for response to chemotherapy [293]. Another main issue is the lack of their cancer type-specificity, i.e., the same miRNA is altered in multiple cancers. Circulating miR-21 is altered in several cancers (e.g., head and neck, breast, colorectal, hepatocellular, pancreatic, other than in those mentioned in this review) however, miR-21 levels are also associated with inflammatory and wound-healing responses. miR-92 levels are altered in both ovarian and colorectal cancers and it also marks the surgical response [87]. Furthermore, both miR-21, miR-92a behave differently in different cancers. Only, the detailed definition of their biological role in the neoplastic transformation process will lead to the identification of a putative marker, selected throughout a rigorous literature search taking into account as many aspects as possible, as a definitive biomarker [293]. These concerns must not discourage the path towards the search for tumor-associated circulating miRNAs. Indeed, as reported in the conclusive parts of each section of this review, promising markers are available at least for breast, lung, prostate cancers, and osteosarcoma. Their priceless value resiedes in their association with fine molecular mechanisms taking place in a group of cells or in a tissue assayable without the need of any invasive procedure and this is of particular interest for those tissues in which biopsy can be difficult (e.g., bone [294], pituitary gland [295]).
On the other hand, there are fields in which research investigating miRNA-based circulating biomarkers is strongly needed. Pituitary adenomas, for instance, are among the most frequent intracranial tumors with an incidence rate of 10-15%. Although mostly begnign, they manifest through several morbidities mainly associated with the compression produced on the adjacent structures and/or hormone disturbances. Although the expression profile and role of tissue miRNAs are quite well investigated, information about the diagnostic potential of pituitary adenoma-associated circulating miRNAs are still lacking. Indeed, if hormone monitoring represents the best way to follow tumor growth and function and, therefore, the role of diagnostic miRNAs could be less important, in the case of non-functional pituitary adenomas a blood-based miRNA biomarker could help the diagnosis and patient follow-up after surgery. Potential diagnostic markers such as miR-26a, miR-126, miR-300, miR-329, miR-381, miR-665 in GH-secerning adenomas, miR-26a in corticotropic adenomas, miR-432 and miR-410 in prolactinomas, miR-106b in non-functioning adenomas and metastatic pituitary carcinomas, are currently the most promising [296].
Another important point not addressed by this review that deserves attention is the role of extracellular vesicle (EV)-associated miRNAs (exosomes and microvesicle-associated miRNAs). Rather than biomarkers, miRNAs contained into EVs act as hormone-like mediators since they target more or less specifically cells/tissues different from those of origin, thanks to the interaction between surface proteins expressed on the EV and the target cell. Once fused with (or endocytosed by) the target cell, the EV releases its content into the cell. Consequently, the EV-associated miRNA species act in this cell by modulating the gene expression and, hence, the cell function. By contrast with free extracellular miRNAs which, generally, mirror the cell content since being passively released by the cell (e.g., membrane leaking, autophagy, cell death), EV-associated miRNAs are actively and specifically loaded into an EV in order to be addressed to a target cell and, hence, their intra-vesicle concentration does not necessarily mirror the intracellular concentration [297]. The pathogenic role of EVs is, currently, a hot topic in biomedical sciences. Indeed, as carriers of information (e.g., miRNAs), EVs are fundamental modifiers of the evolution of a pathological condition as demonstrated in different fields such as heart disease [298], osteoporosis [294] and sarcopenia [299], diabetes [300], and cancer [297]. The role of EVs in cancer have been established since the demonstration of their implication in virtually every aspect of tumorigenesis. EVs can, indeed, mediate the "transfer" of a phenotype and, for instance, vesicles generated by highly aggressive cancer cells can stimulate growth, survival, and migration ability to less malignant cancer cells. This phenomenon is particularly relevant in primary tumors that often consist of heterogeneous populations of cells with different levels of aggressiveness. EVs are also key determinants of the tumor microenvironment and of the promotion of invasion in a two-way fashion that sees tumor cells and surrounding normal cells exchanging information that finally drives the process. Similarly, the recipient tissue of a metastasis is "prepared" by the primary tumor and circulating cancer cells also throughout EVs [297]. In this process, EV-associated miRNAs play a pivotal role. Yu and colleagues have recently demonstrated that HCC cells under hypoxic conditions (i.e., those found into a tumor mass) showed an increased exosomal production and these exosomes, in turn, enhanced proliferation, migration, and invasiveness and induced the epithelial-to-mesenchymal transition in HCC cells under normoxic conditions. EV-derived miR-1273f, an activator of Wnt/β-catenin signaling pathway, was identified as the main mediator of this intercellular information transfer [301]. Exosome-associated miRNAs also drive the complex modulatory effects exerted by mesenchymal stem cells (MSCs) on tumor growth and progression, as recently investigated by Che and coworkers. They demonstrated that MSC-derived exosomal miR-143 suppressed proliferation, migration, invasion, and tumor growth and induced apoptosis in prostate cancer cells by targeting trefoil factor 3 (TFF3) [302].
In conclusion, considering all the evidence reported in this review, altered circulating miRNA levels may be correlated with cancer development and progression (tumor stage and metastasis presence), thus underlining their potential as diagnostic and prognostic biomarkers. MiRNAs may also be useful tools for monitoring patients during chemotherapy treatment as well as after surgery, and can be considered as prognostic markers for patient survival. In some cases, circulating miRNA panels, miRNA ratios, or a combination of circulating miRNAs levels with currently used cancer biomarkers strongly increase diagnostic and prognostic performance compared with single miRNAs or standard markers alone. Obviously, the presence of conflicting evidences about the regulation of specific circulating miRNAs in the studied type of cancer, together with the bias affecting the miRNA validation process, underline the necessity to standardize all preanalytical, analytical, and postanalytical protocols in order to obtain reliable biomarkers for cancer diagnosis and as potential tools for personalized treatments.