Circulating microRNA as potential diagnostic and prognostic biomarkers of well-differentiated thyroid cancer: A review article

The most frequently referenced microRNA in the available literature is miR-222 and its mature isomer miR-222-3p. A total of 10 independent research papers cited this microRNA as being differentially expressed in PTC. Eight studies demonstrated an increase of miR-222 expression by 1.61–3.42 fold in PTC when compared to healthy controls, with a diagnostic area under the curve (AUC) of 0.852–0.882 as well as a sensitivity and specificity of 90.9–94.3% and 70.0–70.5% respectively [27, 28, 29, 30]. The AUC represents a comparable value of diagnostic accuracy as it describes the sensitivity of a test over all possible specificity values and vice versa [31].

One study was unable to detect miR-222 at all in their PTC serum samples using microarray analysis [32] whilst another study found the raise in miR-222 could not achieve significance [33]. The findings are more divisive when comparing PTC to benign goitre samples (BG), 5 studies observed raised miR-222 & miR-222-3p levels in PTC whilst another 4 studies found no significant changes between PTC and BG [27, 28, 34, 35]. In this comparison the diagnostic utility was less reliable with an AUC range of 0.587–0.906, sensitivity and specificity ranged from 47.7%–81.1% and 84.2%–92.5% respectively. All studies which measured miR-222 & miR-222-3p levels post thyroidectomy found them to be significantly decreased ( 0.05).

Many prognostic correlations were identified, as miR-222 was found to be expressed 13x higher in PTC patients with BRAF gene mutation ( p = 0.012) [27]. MiR-222 was also associated with lymph metastases ( 0.05), TNM stages III & IV ( 0.05), multifocal lesions ( p = 0.002), tumour size 2 cm ( 0.05) and bilateral lesions ( p = 0.005). MiR-222 was associated with extrathyroidal extension ( p = 0.026) in a 2018 publication, however another study found no such correlation [28, 33].

Only one paper studied miR-222-3p in medullary thyroid cancer (MTC), and found it to be expressed more in MTC versus healthy controls but also in BG. An AUC of 0.858 was calculated for differentiating MTC from BG suggesting miR-222 may provide significant diagnostic utility for that condition [35].

The use of miR-222-3p as a diagnostic marker for various cancers has been supported by evidence published by Wang et al.’s review in 2021. Associations were found between upregulated miR-222-3p and cancers of the bladder, breast, GI tract, endometrium, liver, skin, ovaries, bones and thyroid [36].

2.2 miRNA-146

In the miR-146 family, 7 studies measured levels of miR-146b and three other studies looked at miR-146a. Together, this family of microRNAs is the second most cited to be dysregulated in thyroid cancer. MiR-146b was shown to be expressed higher in PTC vs BG by four studies with a diagnostic AUC of between 0.653–0.9, sensitivity and specificity of 77.14%–79.5% and 52.6%–78.95% respectively [27, 28, 30, 33]. Only one study found a relative significant decrease of 2.4x when comparing PTC to BG ( 0.001) [37]. Three studies found no significant difference in the miR-146b levels in PTC vs healthy controls, whilst in 2018 Zhang et al. in fact found a higher level in PTC ( 0.001). All studies measuring post-operative miR-146b levels in PTC patients found them to have significantly decreased as much as 5.1x ( p = 0.032) [27, 29]. Serum miR-146a was studied in four papers and in three of those, found to be decreased in both PTC and BG groups when compared with healthy controls.

As a diagnostic marker, miR-146b was found to be consistent with traditional thyroglobulin assay results in 85% of participants in one study. Strong prognostic correlation was found with BRAF mutation signified by a 35x higher expression level ( p = 0.0025) [27]. Levels were higher in patients with lymph metastases ( 0.05), bilateral lesions ( p = 0.001), multifocal lesions ( p = 0.005) as well as tumours $>$ 2 cm in size ( p = 0.007). Levels were also higher with extrathyroidal extension but did not achieve statistical significance ( p = 0.103).

Other studies showed that decreased levels of miR-146a are linked with systemic lupus erythematosus, rheumatoid arthritis, Sjögren’s syndrome whilst elevated levels in the brain are linked with Alzheimer’s and age-related macular degeneration [38].

2.3 miRNA-221

There are ten publications examining the utility of miR-221 as a diagnostic/prognostic blood biomarker for thyroid cancer, and five papers report either undetectable or insignificant changes in expression. Another four papers however, found significant raises of miR-221 in PTC vs healthy controls. Differentiation between PTC and BG is difficult, with only one study citing raised miR-221 levels with a diagnostic AUC of 0.65 with sensitivities and specificities of 77.14% and 50% respectively [32]. MiR-221 was found to be 6.5x higher in patients with BRAF mutation ( p = 0.005), TNM stage $>$ III ( p = 0.004) and lymph metastases ( p = 0.0363). No correlations were found with tumour size or extrathyroidal extension [27, 30, 33, 39].

MiR-221 has also been examined in the context of other disease and some associations have been made with glioblastoma, bladder cancer, prostate cancer, pancreatic cancer, colorectal cancer and gastric cancer [40].

2.4 Let-7 (family)

Collectively, the various let-7 microRNA act broadly as onco-suppressors by suppressing cellular proliferation, inducing chemosensitivity as well as modulating cell-signalling pathways [41]. Blazeau et al. published a review which collated evidence on the Let-7 family of microRNA which have subsequently been associated with over 20 different types of cancers [42].

The most frequently studied member of the let-7 microRNA group is Let-7e followed by Let-7i and Let-7b. In total there were four studies examining Let-7e of which two found a significantly increased level in PTC compared to BG, whilst the remaining studies could not detect sufficient levels for analysis using [29, 32, 35, 39]. The only prognostic correlation found was a higher Let-7e level in patients with multifocal lesions ( 0.001) [39].

Let-7i was found by Li et al. to be increased in PTC by 5.72x ( 0.016) and 5.10x ( 0.019) respectively, when compared to BG and healthy controls respectively [44]. Interestingly, Zhang et al. found stable expression levels of Let-7i across all samples and subsequently used it in normalizing expression data [45].

Let-7b was also raised in PTC and BG by 3.9x ( 0.001) and 2.7x ( p = 0.01) when compared to healthy controls [37].

2.5 miRNA-21

Dysregulated miR-21 has been implicated in over 29 different neoplastic and benign conditions ranging from myocardial infarctions to breast cancer [46]. More recently, studies have begun to examine the correlation with thyroid cancer.

MiRNA-21 levels were significantly raised (15.21x, 0.001) in PTC compared to healthy controls, a result which was independently replicated by only three research teams [28, 47, 48]. AUC for detecting PTC from BG was 0.95 with sensitivity and specificity of 88.57% and 89.47% [28]. However, four other studies in 2016, 2015 & 2013 could not find any significant differences in miR-21 levels between PTC and BG groups. In fact, one study found higher miR-21 levels only in BG vs PTC & HC ( p = 0.002) [34]. Samsonov et al. proposed some diagnostic utility for discerning FTC from BG, as significantly higher levels were observed in the FTC cohort [47].

MiRNA-21 was associated with poorer prognosis due to links with lymph metastases ( 0.03), bilateral ( p = 0.015) and multifocal lesions ( p = 0.005), advanced TNM stage ( p = 0.008) as well as extrathyroidal extension ( p = 0.021) [28, 37].

2.6 miRNA-451

Results from two studies found a raised miR-451 level in PTC compared to healthy controls (5.08x, 0.05), only one of which found a significant difference between PTC and BG (3.69x, 0.05)[44, 45]. The calculated AUC was 0.857 with sensitivity and specificity of 88.9% and 66.7% respectively. A post-operative miR-451 level decrease in PTC cases ( p = 0.004) was also found in that study. Using microarray detection, Yu et al. found no significant difference in miR-451 level in PTC or benign goitre. Another study found a significant decrease in PTC vs BG as well as in lymph metastases ( p = 0.017) [49]. A result which was not replicated in 2020 as Jiang et al. found no significant difference with lymph metastases ( p = 0.5) [39].

Zhang et al. similarly found no significant difference in the microRNA level in MTC compared with benign goitre [45].

A review published in 2019 found that dysregulated miR-451 may be associated with hepatocellular carcinoma, GI-tract cancers, urinary tract cancers, ovarian and cervical cancers as well as lung and bone cancers [50].

2.7 miRNA-151

Raises in the serum levels of miR-151 in patients with PTC were found in two studies compared to healthy controls, however a difference in expression between PTC and BG was only found by one research team (2.17x, 0.001) [34, 43]. For PTC vs BG an AUC of 0.78 was calculated along with specificity and sensitivity of 59.4% and 89.5% respectively. Two other studies found no significant difference between miR-151 levels in PTC vs BG with one study unable to obtain sufficient signal for analysis. Both Yu et al. and Yoruker et al. found prognostic association of miR-151 with lymph metastases ( p = 0.012) and tumour size $>$ 2 cm ( 0.001).

Raised levels of miR-151-3p have been previously found to be linked with colon adenocarcinoma by one study in 2020 [51].

2.8 miRNA-190 and miRNA-95

Levels of these microRNAs have been detected by 3 independent research groups. Cantara and Pilli found elevated miRNA-190 in PTC vs BG ( 0.001) and decreased miRNA-95 for the same group ( 0.001) [52]. They found that for diagnosis of PTC combining the two markers into a panel yields an AUC of 0.99, with sensitivity and specificity 94.9% and 98.7% respectively. Two other teams however could not detect sufficient levels for statistical analysis, by both microarray and PCR methods [33, 28].

Dysregulated levels of miR-190-5p have also been linked with bladder cancer, pancreatic cancer, meningioma, gastric cancer, breast cancer, prostate cancer and others [53]; whilst miR-95-3p has been loosely linked with hepatocellular carcinoma and colorectal cancer [54].

2.9 Other miRNAs

A large selection of other microRNAs has been documented by various individual research teams, these have not yet been thoroughly re-examined and have not had results reproduced. MiR-31 was identified by Samsonov et al. to be elevated in PTC vs BG, Yoruker et al. found no significance in the change in expression they observed, and Lee JC et al. had a low detection signal unsuitable for statistical analysis [27, 34, 47]. Cantara and Pilli identified in their study decreased levels of miR-29b and miR-579 in PTC vs BG ( 0.001) [52].

Yu et al. examined plasma expression levels of miR-9-3p, miR-25-3p, miR-124-3p, miR-196b-5p and miR-4701. Raised levels of both miR-9-3p and miR-124-3p were seen in PTC vs BG ( 0.05), with AUC, sensitivity and specificities of 0.823, 80%, 73.7% and 0.831, 88%, 76% respectively [32]. Raises in miR-9-3p and miR-124-3p were also higher in patients younger than 45 years of age, and for miR-124-3p they were associated with tumour sizes larger than 2 cm. MiR-196b-5p was decreased in PTC vs BG, an AUC of 0.781 as well as sensitivity and specificity of 74% and 66% were calculated. For miR-25-3p no significant change was found in PTC vs BG, however Li et al. indeed found a 2.4-fold increase. Yu et al. found no change in MiR-4701 expression between all cohorts including PTC, BG and HC.

Graham et al. examined levels of miR-150-5p, miR-342-3p, miR-191-5p and miR-93-5p but found no significance in any changes found comparing PTC to BG ( $p>$ 0.05) [37].

Dai et al. found significant raises ( 0.05) in PTC vs BG in all plasma microRNAs that it studied which included miR-376a-3p, miR-485-3p, miR-4306 and miR-4433a-5p [15]. The only correlation found was for miR-485-3p which was noted to be associated with tumour size >1cm, advanced TNM stage, extrathyroidal extension, lymph metastases and BRAF mutation.

Zhang et al. found no significant changes in expression in PTC vs BG for miR-17-5p, miR-132-3p and miR-183-3p. Notably, a significantly higher expression of serum miR-17-5p in MTC vs BG ( 0.01) was observed by the researchers representing an AUC of 0.84.

In 2020, Jiang et al. studied changes in levels of miR-30a-3p, miR-204-5p and miR-138-5p in patients with metastatic and non-metastatic PTC, they found that only miR-204-5p was correlated with lymph metastases ( p = 0.015). Ye et al. compared miR-423-5p in metastatic and non-metastatic PTC finding it raised when compared to healthy controls ( 0.01). Other microRNAs found increased in PTC vs BG were miR-140-3p (23.8-fold increase, 0.05) [43], miR-155 ( p = 0.009), miR-22 ( 0.001) [33]

Further, two other microRNAs were found to have no significant differences in PTC vs BG expression which included miR-181b [27, 47], miR-100 [45].

Finally, for PTC recurrence miR-203 was uniquely found raised vs non-recurrent PTC (AUC 0.755, 0.01) and it was also associated with higher TNM stage ( p = 0.014), lower survival rate as well as perineural invasion [55].

3. Concluding remarks

Biomarker research has been gaining popularity over the last decade and our resultant knowledge has continued to expand along with it. Although tissue biomarker research has become more widely studied, “liquid biopsy” serum/plasma biomarkers are still in their infancy for many diseases. Such is the case with microRNAs in thyroid cancer, the bulk of research focusing on PTC but at present little or no research has been done into the other thyroid cancers.

Although the body of evidence is still small, some microRNAs seem to show promise for their diagnostic and prognostic utility. Whilst the currently used prognostic biomarkers are limited as calcitonin is in MTC, microRNA appear to be linked to various different prognostic factors. Associations with tumour size, tumour invasiveness, lymph node involvement and metastases, further research could examine an assorted panel of microRNA which may provide the clinician with important insights into these risks. It may therefore be beneficial to understand if any of these prognostic microRNA are present in borderline cases when it comes to decide whether or not to proceed with surgery. MiR-222 has been reproducibly shown to be positively correlated with PTC diagnosis as well as many associated prognostic factors and as a potential diagnostic test for MTC. To a lesser extent, in the case of PTC evidence seems to show some diagnostic and prognostic utility for miR-146b, Let-7 family, miR-221, and miR-21. Comparing sensitivities and specificities for current biomarkers such as thyroglobulin shows that generally individual microRNA have higher sensitivities but lower specificities. Thyroglobulin had a sensitivity of around 57% and a sensitivity of 98% whilst 5 of the most commonly cited microRNA (ie. miR-222, miR-146, miR-221, miR-21, miR-451) have sensitivities ranging from 77-94% and specificities ranging from 50-89%. Since microRNA are involved in cell level metabolic processes present in many different malignancies this is manifested as low test specificities in multiple studies. Further research is however required to reproduce many of the individual finding claims and to identify panels of interest. In particular miR-21 was suggested for further study in FTC diagnosis by Samsonov et al [47]. Other potential miRNAs of interest are miR-190, miR-95 which together could represent a strong diagnostic tool for PTC with an AUC of 99% as found by Cantara et al. [52]. Further evaluation of the proposed diagnostic power of miR-451 and miR-17-5p for MTC is needed.

Major hurdles remain in assuring accurate, sensitive, reproducible and unbiased data as some of the existing study results appear to be contradicting. Possible reasons for this are differences in the used lab protocols and PCR kits, especially as the examined publications span from 2013 to 2021. The development of more consistent microarray testing technology would eliminate selection bias when studying microRNAs. There is also a need to establish a widely used and comprehensive protocol for sample collection and processing with emphasis on data normalization for PCR to further reduce process variables. Enrolment of sufficient study participants to achieve good statistical power is a challenge for the less common cancers like FTC and MTC. Further research is needed to better understand the roles of different microRNAs and their changes in expression in human physiology and pathophysiology related to common non-oncologic conditions as this could account for some of the discrepant findings between studies.

Progress is being made to better understand the mechanics of microRNA expression, with further research they hold potential as powerful tools for easier, cheaper, and quicker diagnosis as well as improved mortality and morbidity.