A miRNA-Based Prognostic Model to Trace Thyroid Cancer Recurrence

Simple Summary Some thyroid tumors elected for surveillance remain indolent, while others progress. The mechanism responsible for this difference is poorly understood, making it challenging to devise patient surveillance plans. Early prediction is important for tailoring treatment and follow-up in high-risk patients. The aim of our study was to identify predictive markers for progression. We leveraged a highly sensitive test that accurately predicts which thyroid nodules are more likely to develop lymph node metastasis, thereby improving care and outcomes for cancer patients. Abstract Papillary thyroid carcinomas (PTCs) account for most endocrine tumors; however, screening and diagnosing the recurrence of PTC remains a clinical challenge. Using microRNA sequencing (miR-seq) to explore miRNA expression profiles in PTC tissues and adjacent normal tissues, we aimed to determine which miRNAs may be associated with PTC recurrence and metastasis. Public databases such as TCGA and GEO were utilized for data sourcing and external validation, respectively, and miR-seq results were validated using quantitative real-time PCR (qRT-PCR). We found miR-145 to be significantly downregulated in tumor tissues and blood. Deregulation was significantly related to clinicopathological features of PTC patients including tumor size, lymph node metastasis, TNM stage, and recurrence. In silico data analysis showed that miR-145 can negatively regulate multiple genes in the TC signaling pathway and was associated with cell apoptosis, proliferation, stem cell differentiation, angiogenesis, and metastasis. Taken together, the current study suggests that miR-145 may be a biomarker for PTC recurrence. Further mechanistic studies are required to uncover its cellular roles in this regard.


Introduction
Among cancers, thyroid carcinoma (TC) is the most prevalent endocrine tumor with the largest increasing annual incidence rate in the United States (U.S.). TC is projected to surpass colorectal carcinoma as the fourth most common cancer overall by 2030 [1]. Owing to its low mortality rate and prolonged follow-up management, TC imposes a high healthcare burden and financial cost on the expanded survivor pool [2][3][4]. Papillary thyroid carcinomas (PTC) are the most common histological subtype, comprising 85% of all differentiated thyroid cancers. Although PTCs are the least aggressive thyroid cancers and have an 80-95% five-year overall survival rate, these tumors can metastasize into lymph nodes and various organs, leading to poorer prognosis. Up to 30% of patients experience recurrence after initial treatment, leading to an over 60% increase in mortality rate [5][6][7][8][9]; thus, a key issue in managing patients with PTC is minimizing the morbidity and mortality associated with disease recurrence.
Clinicopathological risk stratification by the American Thyroid Association (ATA) has been used to classify well-differentiated thyroid cancer (DTC) cohorts to guide the management of patients [10,11]. However, it is a one-time assessment at the time of initial diagnosis and relies on genomic mutation and the final pathological results of tissue biopsies. Like other cancers, repeated tissue biopsy is not feasible following operative excision. The current follow-up modalities, including basal and recombinant human thyrotropin (rhTSH)-stimulated serum thyroglobulin measurements, as well as neck ultrasonography, offer only a snapshot of the disease at a single point in time, have limited sensitivity and specificity, and are poor predictors of progression [12]. Radioiodine whole-body scintigraphy (WBS), which takes advantage of the high avidity of radioiodine in the functioning thyroid tissues, has been used for the detection of DTC; however, radioiodine uptake is not specific for thyroid tissue, with the potential of false-positive results [13]. Additionally, postablative non-stimulated Tg have been proposed as a valuable prognostic marker during follow-up [14], with postoperative decline showed good prognostic efficacy for a tumor-free status [15]. Nevertheless, undetectable serum thyroglobulin levels are found in more than 40% of DTC patients with residual or structural recurrence [16][17][18]. Thyroglobulin has been reported as a predictor for treatment efficacy during ablative radioiodine treatment; however, some patients still exhibit elevated levels even after receiving adjuvant radioactive iodine therapy. Moreover, thyroglobulin autoantibodies, found in~25% of patients and known to interfere with the measurement of thyroglobulin, render tests unreliable [19,20]. Conversely, unnecessary treatment by radioactive iodine or high doses of thyroid hormone replacement can elicit secondary malignancies [21][22][23][24][25][26][27][28][29], adverse cardiovascular or skeletal outcomes [30][31][32], and mortality [31]. Risk stratification to decide which patients should be treated with adjuvant radioiodine therapy or extensive postoperative surveillance following surgery remains highly challenging, as it is difficult to accurately predict which tumors will remain indolent. Hence, there is an urgent need to establish a well-validated ancillary molecular predictive marker that accurately identifies patients at risk of progression. In the absence of such an assay, prognostication will continue to be a challenging practice that puts clinicians at risk of overtreating low-risk patients or undertreating high-risk patients.
Despite decades of research, a simple and universal test to identify cancer recurrence at its earliest stages remains elusive. The accurate and timely prognostication of cancer is a prerequisite for efficient postoperative surveillance and is critical for improving disease-free and overall survival [33]. The ideal test would eliminate the need for biopsies and other invasive and risky procedures to expose cancers in the most inaccessible parts of the body. Recent advances have suggested that liquid biopsies have the potential to detect most cancers at early stages when treatment is more likely to be effective [34]. Liquid biopsies have transformed clinical practice by detecting minimal residual disease after surgery and complementing imaging to monitor progression and treatment response [35,36]. The procedure is non-invasive, easily repeatable, detects tumor evolution and heterogeneity, and can predict clinical outcomes [35].
Epigenetics is inextricably linked to the malignant phenotype [37]. Several molecular studies have investigated the roles of various non-coding RNAs along with genetic and epigenetic modifications in cancer recurrence and progression [37]. Cancer reprogramming confers drastic changes in chromatin structure and transcriptomic profiles, which instruct both cell fate and state transition. Cancer-related epigenomic abnormalities arise in advance of morphological alterations [35]; therefore, TC cells with invasive potential exhibit modifications in their general gene expression pattern before morphological alterations of the lesion are even visible. Testing molecular biomarkers before changes are clinically or radiologically evident would facilitate the early detection of minimal disease or recurrence.
Small non-coding microRNAs (miRNAs) are essential for post-transcriptional regulation of gene expression and are attracting increasing attention because of their association with tumor progression. Enrichment and depletion of miRNAs lead to deregulated co-expression of miRNA target genes and disrupted cellular biological processes in cancer [38,39]. They can be secreted into circulation and exist in remarkably stable forms under extreme conditions such as RNase exposure, multiple freeze-thaw cycles, and extreme pH [34,40]. Like intercellular miRNAs, circulating miRNAs participate in the regulation of numerous biological processes and are expressed aberrantly under pathological status. Deregulation of circulating miRNAs is associated with the initiation and progression of cancer; therefore, high-efficacy, low-cost detection of miRNAs may aid in non-invasive diagnostics and prognostics.
The aim of this paper is to create a miRNA-based model to track long-term prognosis and recurrence of PTC. We used microRNA sequencing analysis in TCGA and GEO datasets to explore miRNA expression profiles of PTC patients. The results were validated using quantitative real-time PCR (qRT-PCR). We found miR-145 to be significantly downregulated in tumor tissues and blood. Deregulation was significantly related to clinicopathological features of PTC patients including tumor size, lymph node metastasis, TNM stage, and recurrence. Both in vivo and in vitro studies were reviewed to determine the putative role of miR-145 in tumor recurrence, and our results suggest that miR-145 may be a biomarker for PTC recurrence. Transcriptomic sequence data from 495 thyroid cancer patients were retrieved from the Genomic Data Commons (GDC) data portal for The Cancer Genome Atlas thyroid cancer dataset (TCGA-THCA) (https://www.cancer.gov/about-nci/organization/ccg/research/ structural-genomics/tcga, accessed on 1 November 2020). A total of 1035 microRNAs from miRNA-seq and 20,531 lncRNAs and mRNAs from RNA-seq were included. Clinical and pathological data, molecular landscape, and survival information were obtained from cBio-Portal for Cancer Genomics (https://www.cbioportal.org, accessed on 10 February 2020) and FireBrowse (http://www.firebrowse.org/, accessed on 10 February 2020). Outcomes of interest included disease recurrence and/or progression. Patients with incomplete recurrence data or unmatched microRNA and RNA sequencing data were excluded, leaving 467 matched cases for analysis. TCGA-THCA cohorts were classified according to 2015 ATA risk stratification for structural disease recurrence into low, intermediate, or high-risk groups. Demographic, oncologic, and clinicopathologic data were compared between recurrent and non-recurrent groups. The gene read counts were first filtered for low abundance and low variance transcripts. Mapped reads were summarized into a gene level count using the median estimates. EdgeR was applied to identify the differentially expressed microRNAs (DEmiRs) and RNA genes (DERs) in recurrent compared with non-recurrent groups. Raw data were normalized using log2-counts per million. The significance threshold was set at a false discovery rate (FDR, the q-value) < 0.05 and a |log2 fold change (FC)| > 1.0. Box plots for global transcriptomic signature and volcano plots for genes were generated. Spearman's correlation analysis was performed to examine co-expression between significant DEGs and DEmiRs, and a correlation matrix was generated.

Functional Enrichment Analysis
The Database for Annotation, Visualization, and Integrated Discovery (DAVID) online tool v6.8 (https://david.ncifcrf.gov/, accessed on 1 July 2021) and DIANA-mirPath v3.0 (https://dianalab.e-ce.uth.gr/html/mirpathv3/index.php?r=mirpath, accessed on 1 July 2021) were utilized for Gene Ontology (GO) analysis to annotate genes and to define the functions of DEG into three domains: cellular component, biological process, and molecular function. Subsequently, KEGG enrichment analysis was conducted to discover the potential signaling pathways in which DEGs were involved. The cut-off value of significant GO terms and KEGG pathway was p < 0.05.

Predictive Performance of DEG and DEmiR
Association analysis was performed for the 8 miRNAs and 16 genes using Mann-Whitney U test in the TCGA cohort. Receiver operating characteristics (ROC) curve was employed to estimate their prognostic performance with various clinical, genomic mutation, and pathological data. The optimum threshold level was selected using the Youden index. Area under the curve (AUC), test accuracy, sensitivity, specificity, and diagnostic odds ratio measures were calculated. DEmiRs/DEGs with AUC > 0.75 and p < 0.05 were selected for further downstream analysis. Univariate and multivariate Cox proportionate hazards regression analyses were employed to predict disease recurrence. Hazard ratios and 95% confidence intervals were reported. p-values were corrected for multiple-hypothesis testing using the Benjamini-Hochberg correction, with a significance threshold of FDR < 0.05. Suez Canal University School of Medicine, Egypt (#4344). The study followed "Declaration of Helsinki" guidelines.

Human Specimens
A total of 484 thyroid specimens (242 cancer tissues and their paired non-cancer adjacent tissues) were analyzed. These included 64 paired fresh frozen samples from the Louisiana Cancer Research Center (LCRC), New Orleans, Louisiana, United States, and 178 paired formalin-fixed paraffin-embedded (FFPE) archival samples from three Pathology Labs of (i) El-Bayan, Portsaid City; (ii) Mansoura University, Mansoura City; and (iii) Suez Canal University, Ismailia City, Egypt. In addition, 111 blood samples (64 thyroid cancer, 36 benign non-toxic multinodular disease, and 11 control subjects) were recruited from Tulane Medical Center, New Orleans, LA, United States. Tissue samples were obtained during surgical resection, and fasting blood samples were collected before thyroidectomy. Deidentified samples were processed and linked to their demographic, pathological, and clinical information.
The current study investigated patients diagnosed with well-differentiated thyroid cancer (papillary TC or follicular TC) according to the International Classification of Oncological Diseases, 4th edition, who underwent thyroidectomy and/or lobectomy and did not receive any treatment before operative resection. There was no restriction for age, sex, race, ethnicity, national origin, language, religion, pregnancy, socioeconomic status, or disability. The exclusion criteria included (a) poorly differentiated thyroid carcinoma, anaplastic (undifferentiated) carcinoma, medullary thyroid cancer, Hürthle cell thyroid carcinoma, thyroid lymphoma, thyroid cancer arising from a thyroglossal duct cyst, thyroid cancer in malignant struma ovarii, or secondary carcinomas; (b) patients with a history of primary cancer in other organs; (c) patients with incomplete follow-up or missing clinical data; (d) cohorts with unmatched paired tissue samples; or (e) no available archived paraffin blocks and/or blocks with insufficient tissue to perform immunohistochemical studies.

Clinical Assessment and Outcomes
Baseline demographic and clinicopathological characteristics were obtained from electronic medical records. Demographic data included age at diagnosis, sex, race, ethnicity, and year of diagnosis. For risk assessment, body mass index (BMI), smoking and alcohol intake, history of radiation exposure, occupational hazards, and family history of cancer or previous thyroid diseases were evaluated. Pathological characteristics included tumor side and laterality, histopathological subtype (PTC or FTC) and variant, TNM (tumor, node, and metastasis) stage, and extrathyroidal extension (ETE). Sonographic features and genomic mutation reports were collected if available. Patients with concomitant Hashimoto's thyroiditis were identified. Laboratory results for differential white blood cell count, hepatitis C virus, thyroglobulin, and autoantibodies were collected. The type of surgical resection (unilateral, subtotal, or total thyroidectomy), extent of neck dissection (central, unilateral, or bilateral block neck dissection), thyroxine intake, and use of radioactive iodine or other adjuvant treatment modalities (e.g., external beam radiotherapy) were reported. Additional lines of treatment received during the postoperative surveillance period were obtained.
Post-thyroidectomy, patients were monitored every six months. They underwent clinical examination, imaging studies, and serum thyroglobulin level estimation. Relapse, recurrence, progression, and death were reported at follow-up ( Figure 1). Patients fulfilled the 2015 ATA intermediate-risk criteria: exhibiting aggressive histology (tall cell, solid, hobnail), microscopic extrathyroidal extension (peri-thyroid fat, strap muscles), vascular invasion, or >5 involved lymph nodes (0.2-3 cm). Our primary outcomes were the presence of persistent or recurrent disease. Persistent disease was identified as residual tumor in the area of the primary tumor or regional nodes. Clinical recurrence was defined as the reappearance of malignancy in the thyroid bed (local recurrence), lymph node (regional recurrence), or distant metastasis (distant recurrence) following initial surgery. No clinical Cancers 2022, 14, 4128 6 of 30 evidence of disease (NCED) was defined as the absence of disease at the end of the followup period, based on physical examination and imaging studies. Progression-free survival (PFS) was defined as the time from initial diagnosis to the progression of tumor (distant metastasis, local recurrence, and regional or distant recurrence) or death by any cause. Disease-free survival measured the period from initial surgery to the time of recurrence (local, regional, or distant) or death by any cause, while overall survival referred to the time from initial surgery to death by any cause, regardless of the presence or absence of evidence of recurrence/progression. Outcomes were assessed as binary and time-to-event variables. hobnail), microscopic extrathyroidal extension (peri-thyroid fat, strap muscles), vascular invasion, or >5 involved lymph nodes (0.2-3 cm). Our primary outcomes were the presence of persistent or recurrent disease. Persistent disease was identified as residual tumor in the area of the primary tumor or regional nodes. Clinical recurrence was defined as the reappearance of malignancy in the thyroid bed (local recurrence), lymph node (regional recurrence), or distant metastasis (distant recurrence) following initial surgery. No clinical evidence of disease (NCED) was defined as the absence of disease at the end of the followup period, based on physical examination and imaging studies. Progression-free survival (PFS) was defined as the time from initial diagnosis to the progression of tumor (distant metastasis, local recurrence, and regional or distant recurrence) or death by any cause. Disease-free survival measured the period from initial surgery to the time of recurrence (local, regional, or distant) or death by any cause, while overall survival referred to the time from initial surgery to death by any cause, regardless of the presence or absence of evidence of recurrence/progression. Outcomes were assessed as binary and time-to-event variables.

Histopathological Assessment
FFPE blocks of tissue specimens were reviewed by a pathologist to distinguish between cancer and non-cancer regions and assess the percentage of tumor cells and normal cells. The following microscopic parameters were assessed via hematoxylin and eosin (H & E) staining: histopathological variant, TNM tumor stage according to the American Joint Committee on Cancer (AJCC, 8th edition), focality, ETE, lymphatic invasions, extranodal extension, perineural invasion, and safety margin. FFPE sections were cut of 4-micron in thickness in separate Eppendorf tubes (cancer and non-cancer) for molecular analysis. Fresh frozen tissues were cryosectioned and stained with the Histogene™ laser capture microdissection (LCM) frozen section staining kit and micro-dissected using the Arcturus LCM instrument. Only tissues containing <20% necrosis were subjected to nucleic acid extraction. Following xylene deparaffinization, total RNA (including small RNAs) was isolated from 178 paired cancer and non-cancer FFPE tissues using Qiagen miRNeasy FFPE Isolation kit (catalog number 73504, Qiagen, Germantown, MD, USA). Total RNA including miRNA was isolated from the 64 paired frozen samples using Arcturus ® PicoPure ® Frozen RNA Isolation Kit (catalog number KIT0204, ThermoFisher, Waltham, MA, USA), according to the manufacturer's protocol. Circulatory miRNA was isolated from 183 archived peripheral plasma samples using total exosome RNA and protein isolation kit (catalog number 4478545, Thermo Fisher Scientific, Waltham, MA, USA) and Qiagen miRNeasy Serum/Plasma Kit (catalog no. 217184, Qiagen). The total RNA quality and quantity were assessed via absorbance spectrophotometry on a Nanodrop-1000 spectrophotometer at the wavelengthdependent extinction coefficient of 33 (Thermo Scientific, Wilmington, DE, USA) and Qubit™ fluorometer (Invitrogen, Waltham, MA, USA).
RNA (10 ng) was converted to complementary DNA (cDNA) using TaqMan MiRNA Reverse Transcription (RT) kit (P/N 4366596, Applied Biosystems, Foster City, CA, USA); Thermo Fisher, CA, USA), and 5× specific stem-loop primers or endogenous control primers (RNU6B, catalog number 001039) for normalization were used separately. Reverse transcription (RT) was carried out in a T-Professional Basic, Biometra PCR system (Biometra, Goettingen, Germany) at the following amplification conditions: 16 • C for 30 min, 42 • C for 30 min, and 85 • C for 5 min, and then held at 4 • C. For quality control assessment, miR-23a, miR-30c, miR-103, miR-191, and miR-451 were tested for hemolysis, and the RNA spike-ins UniSp101, UniSp100, and UniSp6 were assayed to screen the inhibition of enzymatic reactions.
Real-time quantitative polymerase chain reaction (PCR) was followed using Universal Master Mix (catalog number 4440042) and TaqMan assay for miR-145-5p (Assay ID: 002278, Applied Biosystems, Thermo Fisher Scientific Inc.) in StepOne™ Real-Time PCR System (Applied Biosystems) and QuantStudio 6 Flex (Applied, Foster City, CA, USA). Reactions were run in triplicate under the following conditions: 95 • C for 10 min, followed by 45 cycles of 92 • C for 15 s and 60 • C for 1 min. Appropriate negative controls were included in each run (no template and no enzyme samples), and duplicate PCR runs were performed. The Minimum Information for Publication of Quantitative Real-Time PCR Experiments (MIQE) guidelines were followed during the experiments. Fold changes were estimated via the Livak method based on the quantification (threshold) cycle (Cq or CT) value: relative gene expression = 2 −∆∆Cq . Standard deviation >2.0 was set as an outlier.

Systematic Review
A systemic review on miR-145 in thyroid cancer was performed on 26 June 2022 using PubMed, Embase, Google Scholar, and genomic public repositories following PRISMA guidelines. The keywords "(miR-145 OR microRNA-145 OR MIR145) AND (thyroid)" were used in the search process. All human, in vitro, and in vivo studies were included, without time or language restriction. Manual searches of relevant bibliographic articles were performed. Data abstraction was performed in Microsoft Excel (Muhib MH and ET) and stratified by the functional role of a biomarker.

Functional Role of miR-145
Cancer Hallmarks Analytic tools (https://chat.lionproject.net) were used to define miR-145-related hallmarks based on online publications. Gene targets of miR-145-5p were identified using miRTargetLink 2.0 (https://ccb-compute.cs.uni-saarland.de/mirtargetlink2). Starbase software (http://starbase.sysu.edu.cn/mirLncRNA.php) was used to identify competing endogenous RNA regulatory network sponging miR-145. Diana miRPath v3.0 (https://dianalab.e-ce.uth.gr/html/mirpathv3/) and ingenuity pathway analysis software (https://reports.ingenuity.com/) were used to investigate enriched KEGG pathways and Gene Ontology terms. Harmonizome database (https://maayanlab.cloud/Harmonizome/) was used to identify CHEA transcription factor binding sites and targets of the promoter of . The estimated power is 90% for a minimum of 228 TC patients to predict disease progression, medium effect size = 0.5, and alpha error probability = 0.05, using G*Power version 3.1.9.2 (http://www.gpower.hhu.de/) (accessed on 10 February 2022). For continuous variables, means and standard deviations or medians with interquartile ranges were calculated. For categorical variables, absolute numbers with percentages were recorded. Categorical variables were compared using chi-square (χ 2 ) or Fisher's exact tests where appropriate, while the Student's t-test, Wilcoxon matched-pairs signed rank test, or Mann-Whitney U test were used to compare continuous variables. Changes in miRNA expression were calculated via the LIVAC method based on the Cq value; 2 −∆∆Cq formula. Wilcoxon matched-pairs signed-rank test was used for comparison between cases and controls. Correlations between blood and tissue results were compared using Spearman's correlation test. Diagnostic statistics such as specificity, sensitivity, positive and negative predictive values, ROC analysis with AUC estimation, DeLong test, and F-test were performed using pROC R package to determine the performance of the miRNA in blood and tissues. Multivariate regression analysis was used to build a predictive model for recurrence/progression in TC patients at the time of surgery using selected clinical, pathological, and molecular data. Biological variability by age, gender, and race was adjusted. The prognostic risk scores for each patient were calculated based on a linear combination of the miRNA expression level and demographics weighted by the regression coefficient derived from the regression. Prediction accuracy was assessed using Harrell's concordance index (C-index), Brier score (BS), and Hosmer-Lemeshow test using DescTools R package and Bootstrap 1000. Discrimination and calibration were evaluated. Decision curve analyses (DCAs) were performed to test the clinical utility of the Cox nomogram model. Principal component analysis and hierarchical clustering were employed for risk stratification. Disease-free and overall survival analysis was performed, and Kaplan-Meier curves were plotted. Log rank (Mantel-Cox) test was used. The Survminer package (https://cran.r-project.org/web/packages/survminer/index.html, accessed on 17 April 2022) for R was applied to determine the optimal cut-off value to divide patients into two groups, high or low gene expression, based on receiver operating characteristic. A two-tailed p-value of < 0.05 was considered significant.  Table 1. Patients in the recurrent group were more likely to be older (p = 0.040) and white (p = 0.018). A higher prevalence of recurrence was found in patients with tumor stage T3/T4 (p = 0.009), distant metastasis (p < 0.001), and TERT mutation (p = 0.011). The median overall survival was 31.0 months (IQR = 17.4-51.9), with longer follow-up of the recurrent group (p = 0.006).

Functional Enrichment Analysis
As depicted in Figure 2D, the top biological processes included (1) Figure 2E) revealed the involvement of the extracellular matrix receptor interaction, TGF-beta signaling pathway, and adherens junction, and miR-145-5p was enriched in almost all significant pathways.

Association of DEG and DEmiR with Clinicopathological Characteristics and Survival Analysis
Marker selection for the 8 miRNAs and 16 genes was performed using four statistical tests: Mann-Whitney U test for association analysis, area under the curve (AUC) estimation for receiver operating characteristic analysis to assess prognostic performance, and univariate and multivariate Cox regression analyses to identify the role of each deregulated gene on recurrence. Association analysis with recurrence was significant in 18 markers. ROC analysis showed significance (AUC > 0.75) in 2 miRNAs and 13 genes. The highest AUC was for miR-145-5p, with an AUC of 0.78 (95%CI = 0.71-0.85; 64% sensitivity and 80% specificity), and the best discriminative genes for recurrence were FZD9 (AUC = 0.87, 95%CI = 0.71-0.99), AGTR1 (AUC = 0.85, 95%CI = 0.74-0.95), and NPR3 (AUC = 0.82, 95%CI = 0.71-0.93) ( Figure 3A). Univariate regression analyses showed significant results in seven markers ( Figure 3B). After adjustment for age, sex, race, and BRAF mutation, multivariate regression analysis revealed an association between miR-145-5p and disease-free survival ( Figure 3C). Cancers 2022, 14, x FOR PEER REVIEW 13 of 31

Expression Pattern of miR-145 in Tumor Tissues of Cancer Patients
Downregulated miR-145-5p was validated in other public datasets for thyroid cancer compared with benign (GSE116196) and adjacent non-cancerous tissues (GSE64912, GSE83520, GSE97070), as well as in aggressive versus non-aggressive tumors (GSE48953) ( Figure 4A). In a pan-cancer perspective, 145 datasets were found with available miR-145-5p expression data and 24 with an absolute log fold change below 1.0 were excluded. Fifteen studies were from TCGA, and miR-145 was under-expressed in all of them. The other 106 GEO experiments showed downregulation in 88.6% of the datasets, ranging from −11.21-in lung cancer to −1.03-fold change in colorectal cancer ( Figure 4B). Other datasets showed lower expression patterns in aggressive compared with non-aggressive samples ( Figure 4C).

Expression Pattern of miR-145 in Tumor Tissues of Cancer Patients
Downregulated miR-145-5p was validated in other public datasets for thyroid cancer compared with benign (GSE116196) and adjacent non-cancerous tissues (GSE64912, GSE83520, GSE97070), as well as in aggressive versus non-aggressive tumors (GSE48953) ( Figure 4A). In a pan-cancer perspective, 145 datasets were found with available miR-145-5p expression data and 24 with an absolute log fold change below 1.0 were excluded. Fifteen studies were from TCGA, and miR-145 was under-expressed in all of them. The other 106 GEO experiments showed downregulation in 88.6% of the datasets, ranging from −11.21in lung cancer to −1.03-fold change in colorectal cancer ( Figure 4B). Other datasets showed lower expression patterns in aggressive compared with non-aggressive samples ( Figure 4C). Cancers 2022, 14, x FOR PEER REVIEW 14 of 31

Expression Pattern of miR-145 in Tissues and Blood
As depicted in Figure 5, analyzing thyroid cancer tissue specimens showed downregulation of miR-145-5p in cancer tissues compared with their corresponding paired non-cancer tissues, with median and interquartile values of −1.

Association with Clinicopathological Characteristics
Baseline characteristics of the study population are summarized in Table 3. Downregulation of miR-145 was significantly associated with recurrence. Expression levels were lower in tissues of the recurrent cohort (median: −5.    Furthermore, tissue miR-145 was associated with clinical stage (p = 0.001), larger tumor size (p = 0.006), and lymph node metastasis (p = 0.008). Circulatory miR-145 was associated with Hashimoto disease (p = 0.030) and TERT mutation (p = 0.004). We did not find an association with other clinical or pathological features.

Predictive Accuracy of miR-145
Receiver operator characteristic (ROC) curve analysis demonstrated a good discrimination ability of miR-145 expression level to differentiate between cancer and non-cancer tissues (AUC = 0.734 at cut-off value of −2.37, specificity% = 73.7%, p < 0.001) and blood (AUC = 0.718 at cut-off value of −2.06, sensitivity = 87.5%, specificity = 73%, p = 0.047) ( Figure 7A). Log fold changes in miR-145 in tissues were split into high and low expression groups based on the optimum cutoff value (−2.37) of ROC analysis.

Systemic Review on the Functional Role of miR-145 in Thyroid Cancer
Systematic search; duplicate removal; and screening of titles, abstracts, and full text yielded eight eligible articles [41,[43][44][45][46][47][48][49]. The expression of miR-145 was significantly reduced in thyroid cancer compared with normal cells [41]. In contrast, its overexpression in thyroid cancer cell lines resulted in inhibited cell differentiation, proliferation, migration, and invasion. It suppressed VEGF secretion and E-cadherin expression and inhibited the PI3K/Akt pathway by targeting AKT3 [41]. In vivo, miR-145 overexpression decreased tumor growth and metastasis in a xenograft mouse model [41]. Similarly, miR-145 was  Figure 7D).

Systemic Review on the Functional Role of miR-145 in Thyroid Cancer
Systematic search; duplicate removal; and screening of titles, abstracts, and full text yielded eight eligible articles [41,[43][44][45][46][47][48][49]. The expression of miR-145 was significantly reduced in thyroid cancer compared with normal cells [41]. In contrast, its overexpression in thyroid cancer cell lines resulted in inhibited cell differentiation, proliferation, migration, and invasion. It suppressed VEGF secretion and E-cadherin expression and inhibited the PI3K/Akt pathway by targeting AKT3 [41]. In vivo, miR-145 overexpression decreased tumor growth and metastasis in a xenograft mouse model [41]. Similarly, miR-145 was sponged by the oncogenic lncRNA n384546, which is highly expressed in PTC tissues and cell lines, leading to cancer progression and metastasis. Anti-miR-145 could partially reverse the effects of n384546 knockdown in vitro by regulating AKT3 expression [47]. In another study, miR-145 inhibited the migration and invasion of papillary thyroid carcinoma cells through nuclear factor-κB (NF-κB) pathway regulation [49]. In human tissues, miR-145 was decreased in cancer compared with normal adjacent tissues. Overexpression of miR-145 led to downregulation of its gene target dual-specificity phosphatase 6 (DUSP6) at the protein level and inhibited cell proliferation in TPC1 cells [43]. Furthermore, miR-145 functions as a tumor suppressor in PTC by inhibiting RAB5C [44]. RAB5C is a member of the RAS oncogene family that regulates cell invasion and motility. Zhang et al. found that miR-145 was downregulated in PTCs and negatively correlated with PTC progression and metastasis. MiR-145 inhibited PTC migration and proliferation and promoted apoptosis by directly suppressing RAB5C, suggesting that MiR-145 and RAB5C could serve as therapeutic targets against aggressive PTC cases [44].
Non-coding RNAs (ncRNAs) include long non-coding RNAs (lncRNAs), miRNAs, and circular RNAs (circRNAs). Intracellular circRNAs suppress miRNA by sponging and binding to the miRNA response element (MRE). Overexpressed circNUP214 has been negatively correlated with downregulated miR-145 levels in PTC compared with paratumor tissues [45]. In vitro studies showed that circNUP214 induced tumor cell migration and invasion through miR-145 sponging, releasing the inhibitory effect on zinc finger E-box binding homeobox (ZEB2) [45]. Moreover, miR-145 has been reported to target the lncRNA taurine upregulated 1 (TUG1), a tumor oncogene associated with various human cancers. The deregulated ceRNA network of the TUG1/miR-145/zinc finger E-box binding homeobox 1 (ZEB1) signaling pathway increased cell proliferation, enhanced migration capacity, and promoted EMT formation in three thyroid cancer cell lines (the ATC cell lines SW1736 and KAT18 and the FTC cell line FTC133) [46]. In contrast, metastasis and proliferation were suppressed in cells transfected with miR-145 mimics. Overexpression of miR-145 may reverse the EMT phenotype of thyroid cancer cells; relative levels of miR-145 in thyroid cancer tissues were significantly lower than in the corresponding normal tissues [46]. Moreover, Wang et al., testing genetic variants flanking the miR-145 gene, found significant association between rs4705342*T single nucleotide polymorphism and a higher risk of papillary thyroid carcinoma [48].
In thyroid cancer, research has indicated that miR-145 is a downregulated tumor suppressor miRNA. Functional and preclinical studies identified AKT3, ZEB1, ZEB2, DUSP6, and RAB5C as direct targets of miR-145 [41,[43][44][45][46]. Akt is a crucial protein of the PI3K/Akt signal transduction cascade that plays a central role in cell proliferation, apoptosis, and motility in thyroid cancer [66]. ZEB1 functions to regulate important transcriptional networks necessary for cell differentiation, maintenance, and function [67]. It can modulate the PD-1/PD-L1 checkpoint [68,69], promote cell proliferation and migration, and inhibit The transcription factors FLI1, GATA1, PPARG, RAD21, RCOR3, SIN3B, SPI1, SREBF2, STAT3, STAT5A, and TP63 were found to target the promoter of the MIR145 gene. We did not find any of these markers to be deregulated in the thyroid cancer TCGA dataset. Using ingenuity pathway analysis, gene ontology analysis of the MIR145 gene showed subcellular localization in extracellular vesicular exosomes and RNA-induced silencing complex in cytoplasm. Some biological processes included actin cytoskeleton organization, cell differentiation, negative regulation of angiogenesis, negative regulation of cell migration, negative regulation of inflammatory response, positive regulation of canonical Wnt receptor signaling pathway, positive regulation of cardiac vascular smooth muscle cell differentiation, positive regulation of fibroblast migration, positive regulation of macrophage activation, positive regulation of macrophage differentiation, regulation of collagen biosyn-thetic process, regulation of ERK1 and ERK2 cascade, transforming growth factor beta receptor signaling pathway, and vascular smooth muscle cell differentiation.

Discussion
The MIR145 gene is located at the 5p32 chromosomal region (5:149, 430, 646-149, 430, 733 forward strand). Its expression is regulated by p53 and other transcriptional factors [56]. The tumor-suppressive role of miR-145 has been suggested in diverse types of cancers [57][58][59], such as breast cancer [60], gastric cancer [61], bladder cancer [62], colorectal cancer [63,64], and ovarian cancer [65]. It has been sponged by multiple circular RNAs and lncRNAs, leading to deregulation of downstream gene targets (Figure 9). Previous studies have indicated that miR-145 is associated with cell apoptosis, proliferation, stem cell differentiation, angiogenesis, and metastasis [50][51][52]54,55]. Recent studies demonstrated that molecular profiling of miR-145 is emerging as a key noninvasive tool for monitoring cancer invasion and migration [79]. Interestingly, miR-145 has been detected in peripheral blood and other body fluids [80]. Expression levels of miR-145-5p have been correlated with clinical parameters in cancer patients, as shown in Table 4. In the present study, the expression of miR-145 was analyzed in 284 paired tissues and 111 blood samples. We found that miR-145-5p was downregulated in cancer tissues and blood compared with counterparts. Additionally, a lower miR-145 level was associated with advanced clinical stage, larger tumor size, and lymph node metastasis. Similarly, in the TCGA data, low tissue expression of miR-145-5p was associated with a higher tumor stage, distant metastasis, extrathyroidal extension, and BRAF V600E mutation.  In thyroid cancer, research has indicated that miR-145 is a downregulated tumor suppressor miRNA. Functional and preclinical studies identified AKT3, ZEB1, ZEB2, DUSP6, and RAB5C as direct targets of miR-145 [41,[43][44][45][46]. Akt is a crucial protein of the PI3K/Akt signal transduction cascade that plays a central role in cell proliferation, apoptosis, and motility in thyroid cancer [66]. ZEB1 functions to regulate important transcriptional networks necessary for cell differentiation, maintenance, and function [67]. It can modulate the PD-1/PD-L1 checkpoint [68,69], promote cell proliferation and migration, and inhibit apoptosis [70]. It also plays a crucial role in the maintenance of cancer stem cells [71]. ZEB2 functions as a DNA-binding transcriptional repressor that interacts with activated SMADs [72]. ZEB2 can trigger the induction of genes associated with epithelialmesenchymal transition [73], activate the Wnt/beta-catenin pathway [74], and regulate invasion and metastasis [72]. Endosomal RAB5C was reported to prevent VEGFR2 degradation during angiogenesis [75], regulate focal adhesion-mediated cell migration [76], and enhance resistance to ionizing radiation [77]. DUSP6, a member of the MAPK phosphatase family, is highly expressed in PTC and plays a pro-oncogenic role in thyroid tumorigenesis by regulating the ERK1/2 pathway, cell proliferation, and invasiveness [78]. Studies investi-gating the mechanism responsible for the aberrant expression of miR-145 in thyroid cancer demonstrated the role of three ncRNAs (TUG1, n384546, and circNUP214) in suppressing miR-145 expression [45][46][47]. Additionally, miR-145 was found to modulate NF-κB pathway regulation in PTC cells [49].
Recent studies demonstrated that molecular profiling of miR-145 is emerging as a key noninvasive tool for monitoring cancer invasion and migration [79]. Interestingly, miR-145 has been detected in peripheral blood and other body fluids [80]. Expression levels of miR-145-5p have been correlated with clinical parameters in cancer patients, as shown in Table 4. In the present study, the expression of miR-145 was analyzed in 284 paired tissues and 111 blood samples. We found that miR-145-5p was downregulated in cancer tissues and blood compared with counterparts. Additionally, a lower miR-145 level was associated with advanced clinical stage, larger tumor size, and lymph node metastasis. Similarly, in the TCGA data, low tissue expression of miR-145-5p was associated with a higher tumor stage, distant metastasis, extrathyroidal extension, and BRAF V600E mutation. Several chemical compounds used as medications have been shown to influence the expression pattern of miR-145-5p. Propofol modulates the proliferation, invasion, and migration of bladder cancer cells through the miR-145-5p/TOP2A axis [95]. In glioma, ropivacaine could promote cell apoptosis and inhibit tumor growth, proliferation, migration, and invasion through the regulation of the circ-SCAF11/miR-145-5p axis [96]. Apoptosisrelated miR-145-5p enhanced the effects of pheophorbide a-based photodynamic therapy (Pa-PDT) in oral cancer [97]. Bupivacaine via targeting circ_0000376 and miR-145-5p upregulation could cause cell apoptosis in gastric cancer cells [98]. Therefore, assessment of expression of this miRNA is a potential way for prediction of the response to these medications in different conditions.

Conclusions
This work is the first to indicate putative regulatory networks and mechanisms regulating its suppressive role in tumorigenesis and progression. Taken together, these findings suggest that miR-145 is a key regulator of thyroid cancer growth, mediates its effect through multiple gene targets and pathways, is secreted by the thyroid cancer cells, and may serve as an adjunct biomarker for TC diagnosis and prognosis. Genome-wide screening of miR-145 gene targets in in vitro and in vivo studies will be necessary for further exploration of the potential role of miR-145 in TC progression. Further large-scale, multicenter studies are needed to better elucidate the dynamic changes in miR-145 during surveillance follow-up; however, our comprehensive analysis provides evidence that miR-145 expression is downregulated in TC tissues and blood, and its expression pattern is significantly negatively associated with poor prognosis.

Institutional Review Board Statement:
The study was conducted according to the guidelines of the Declaration of Helsinki and approved by the Institutional Review Board of Tulane University, USA, (protocol code 2020-1636, 12 August 2020 and 2021-1214, 6 January 2022) and the Ethical Committee of Suez Canal University, Egypt (protocol code 4344, 11 March 2020). Patient data were obtained from the hospital medical records, which were anonymized and de-identified before analysis. Informed Consent Statement: (a) Informed consent was obtained from subjects involved in the study who were recruited from the tumor biobank (LCRC). (b) Patient consent of archived formalin-fixed paraffin-embedded samples was waived.
Data Availability Statement: Data are available from the corresponding author upon reasonable request and obtaining the approval of the "Office of Technology Transfer and Intellectual Property Development, Tulane University, USA". Sources of in silico data analysis sections are available in public repositories through the links provided in the manuscript.