*Article* **Identification of miR-29c-3p as a Robust Normalizer for Urine microRNA Studies in Bladder Cancer**

## **Julia Oto 1,**†**, Emma Plana 1,2,**†**, Álvaro Fernández-Pardo 1, Fernando Cana 1, Manuel Martínez-Sarmiento 3, César D. Vera-Donoso 3, Francisco España <sup>1</sup> and Pilar Medina 1,\***


Received: 17 September 2020; Accepted: 19 October 2020; Published: 22 October 2020

**Abstract:** Bladder cancer (BC) is among the most frequent malignancies worldwide, being the most expensive cancer to treat and monitor and the most lethal urological cancer. Urine microRNAs (miRNAs) have been proposed as novel non-invasive biomarkers to early diagnose and monitor BC patients in order to avoid the performance of current aggressive diagnostic techniques. However, huge discrepancies arise among studies mainly due to the lack of standardization in the normalization, a crucial step in all miRNA studies. Our aim was to identify the best miRNA normalizer for miRNA studies in urine of BC patients. We evaluated the performance of 110 candidate miRNAs in urine of 35 BC patients and 15 healthy controls by Real Time quantitative Polymerase Chain Reaction (RT-qPCR) followed by a stability analysis with RefFinder. In this screening stage, miR-29c-3p arose as the most stably expressed miRNA in BC and controls, with a good expression level. Stability of miR-29c-3p expression was validated in an independent cohort of 153 BC patients and 57 controls. Finally, we evaluated the robustness of miR-29c-3p as normalizer in the expression study of miR-200c-3p, a potential diagnostic marker for BC. We propose miR-29c-3p as a normalizer for miRNA studies in BC urine. This is the first study that characterizes a reliable normalizer that may allow the comparison of future urine miRNA studies as non-invasive biomarkers for BC diagnosis and monitoring.

**Keywords:** bladder cancer; urine; miRNA; reference miRNA; normalizer; RT-qPCR

#### **1. Introduction**

Bladder cancer (BC) is among the most frequent malignancies worldwide. Indeed, BC accounts for 3% of all malignant tumors in adults and is the most lethal urological tumor type. According to the GLOBOCAN 2018 study, BC accounts for more than 540,000 new cases worldwide and almost 200,000 deaths annually [1,2]. Moreover, BC largely increases health expenses since it is the most expensive cancer to treat and the one that accounts for the highest monitoring costs. In fact, the cost of the muscle-invasive subtype reaches \$150,000 per capita [3,4] and, by the end of the decade, BC is expected to account for >3% of all cancer-related medical expenses [3].

Bladder ultrasound, computerized tomography (CT) scan, cystoscopy or cytology are presently the gold standard techniques for BC diagnosis. However, cystoscopy is a highly invasive procedure that causes a high discomfort in patients and urine cytology, while being non-invasive, is unable to detect low-grade bladder tumors. Furthermore, none of both achieve a high sensitivity and

specificity [5]. These last two techniques are routinely used as follow-up methods. As a consequence, novel non-invasive biomarkers are being explored to early and accurately diagnose and monitor BC patients in order to avoid the performance of this aggressive technique while reaching or even exceeding the sensitivity and specificity of cystoscopy [6]. Urine represents the most accessible source of biological markers for the analysis of BC and other urological tumors since it is rapid and easily obtained by the patient, avoids patient discomfort and potential complications from an invasive procedure, is available in copious amounts and re-sampling is easily achievable [7]. Thus, urine represents the ideal sample to develop novel tools to diagnose and monitor the prognosis of BC patients.

microRNAs (miRNAs) are small non-coding RNAs that regulate protein expression. They have been found to play a role in the different processes of tumor development [8,9] and have been proposed as regulatory molecules and biomarkers in virtually all cancer types [10–12]. miRNAs are known to be released from tissues into biological fluids such as urine and have been proposed as biomarkers for many disorders. Indeed, several profiles of miRNAs have been proposed as diagnostic and prognostic tools for BC [13–24]. However, strikingly only a small number of miRNAs are shared by these studies, what is probably due to the lack of standardization in the protocols used among laboratories. Particularly results are often not reproducible among publications because of different criteria used for selecting patients and controls, the use of different procedures for sample processing, RNA isolation and miRNA quantification but, mainly, it may be caused by different normalization strategies used, what certainly represents a crucial step in any miRNA study. Several RNA species have been proposed as normalizers for miRNAs studies: small nuclear RNAs (snRNAs), nucleolar RNAs (snoRNAs), ribosomal RNAs (rRNAs), miRNAs or even exogenous synthetic RNAs. However, some of these proposed normalizers differ from miRNAs in length (150 nt for snRNAs and 60–200 nt for snoRNAs, compared to 20–24 nt for miRNAs) and also in structure [25]. This disparity in lengths and structure can affect the isolation efficiency, reverse transcription and amplification. Consequently, the use of stably expressed miRNAs for the normalization of miRNAs studies is the best option rather than the use of other small RNA species [26]. Likewise, the addition of exogenous synthetic RNAs is intended to track the isolation and reverse transcription efficiency, not to normalize for amplification. To date, no consensus exists on internal reference miRNAs for BC studies performed in urine samples by Real Time quantitative Polymerase Chain Reaction (RT-qPCR)

In the present study, we aimed for the first time to ascertain the best miRNA normalizer for miRNA studies in BC performed in urine samples. The discovery of a good and reproducible internal miRNA normalizer will eliminate the current inconsistency among studies and will finally allow the comparison of results obtained in urine of BC patients. This is an inexorable requirement in order to apply this technique to clinical practice.

#### **2. Experimental Section**

#### *2.1. Study Subjects*

BC can be subdivided in two types, superficial or non-muscle-invasive bladder cancer (NMIBC) (70% of total), which comprehends Ta and T1 lesions, and muscle-invasive bladder cancer (MIBC) (30% of total), which comprehends T2, T3 and T4 lesions. Additionally, grading indicates the degree of cellular differentiation, being G1 well differentiated and less likely to spread, G2 moderately differentiated and G3 poorly differentiated and more likely to spread. This can be better envisaged in Figure 1, where the study workflow is also described.

**Figure 1.** Graphical description of BC subtypes and study workflow.

For the initial screening stage, 35 BC patients (10 TaG1, 8 TaG3, 5 T1G3 and 12 T2G3) were recruited at La Fe University and Polytechnic Hospital (Valencia, Spain). Fifteen healthy volunteers (control group) with similar age and sex who underwent an ultrasound scan to rule out the presence of urological malignancies or other alterations were also recruited. For the validation stage, an independent cohort of 153 BC patients (33 TaG1, 13 TaG3, 29 T1G3, 54 TaG2, 9 T1G2, 4 T2G2, 10 T2G3 and 1 T3G3) and 57 healthy controls were additionally recruited.

Pre-operative clinical staging was performed through physical examination, urine cytology and CT scans of the chest, abdomen and pelvis. The tumor histological classification was done according to both the 1973 and 2004/2016 WHO classifications. Demographic and clinical data were collected.

The exclusion criteria were lack of informed consent, absence of histological confirmation and presence of other malignancies.

Informed consent was obtained from all participants according to protocols approved by the ethics review board at La Fe University and Polytechnic Hospital (2014/0314 and 2017/0474). The study was performed according to the declaration of Helsinki, as amended in Edinburgh in 2000.

#### *2.2. Urine Collection*

A first morning urine sample of 25 to 50 mL was collected in sterile containers from all participants and kept at 4 ◦C until processing. Urine was centrifuged at 805× *g* for 5 min at 4 ◦C to remove cellular debris and supernatant was aliquoted and frozen at −80 ◦C until analyzed. The concentration of creatinine in urine was measured by clinical laboratory standardized methods.

#### *2.3. RNA Isolation and cDNA Synthesis from Urine*

Total urine RNA (including miRNAs) was isolated from 600 μL of urine using the miRNeasy Mini Kit (Qiagen, Hilden, Germany) following manufacturer's instructions with several modifications optimized by our group [27]. Briefly, 200 μL of cell-free urine were transferred to a tube with 1 mL

Qiazol (Qiagen) and 1 μL carrier (1 μg/μL yeast RNA, Invitrogen, ThermoFisher Scientific, Waltham, MA, USA). This step was done in three independent tubes for each sample. In one of the three tubes 1 μL spike-in mix (UniSp2/4/5, Exiqon, Vedbaek, Denmark) was also added and each tube was gently mixed. After a 5 min incubation at room temperature, 200 μL chloroform were added to each tube, and centrifuged at 12,000× *g*, 15 min at 4 ◦C to allow phase separation. Ethanol in a proportion of 1.5:1 (volume:volume) was added to the liquid phase. The three tubes containing the urine sample from the same individual were pooled in one single column in order to increase the final RNA yield. Then, 4 cleaning steps with the buffers supplied in the kit were performed. Total RNA was finally eluted in 50 μL of DNase/RNase-free sterile distilled water.

The concentration and purity of the RNA was assessed by spectrophotometric quantification with the NanoDrop ND-1000 (Thermo Fisher Scientific). RNA was stored at −80 ◦C until used.

In the initial screening stage where predesigned panels were used, cDNA was obtained from 5 μL of urine RNA with the miRCURY LNA RT Kit (Qiagen) according to the supplied protocol in a final reaction volume of 25 μL. Due to the addition of a RNA carrier during the isolation, the final RNA yield includes the RNA isolated from urine plus the carrier RNA. Therefore, urine RNA retrotranscription was based on volume (μL) rather than RNA quantity (ng), according to the suppliers' recommendations. In the validation stage, where the expression level of selected miRNAs was conducted, cDNA was obtained from 2 μL of urine RNA using the same technology (final reaction volume 10 μL). In all cases, the reaction mix containing RNA, enzyme, buffer, RNAse-free water and UniSp6 RNA spike-in template, was incubated 60 min at 42 ◦C followed by 5 min at 95 ◦C for reverse transcriptase inactivation. Reactions were carried out in a thermocycler TC-412 (Techne, Minneapolis, MN, USA).

#### *2.4. miRNAs Quantification*

In the screening stage, 35 urine cDNA samples from BC patients and 15 from healthy controls were analyzed. In them, a total of 179 miRNAs were quantified using the commercially predesigned Serum/plasma miRNA PCR Panel V5 (Qiagen). This panel contains 179 miRNAs commonly found in human plasma and serum according to the manufacturer's in-house analyses of miRNA expression in blood, serum and plasma samples, as well as on the limited number of peer-reviewed published papers available. The list of all the quantified miRNAs is detailed in Table S1. miRCURY LNA SYBR Green Master Mix (Qiagen) was used as a fluorophore, according to manufacturer´s indications. Briefly, cDNA (dilution 1/40), water and PCR master Mix (which includes SYBR Green) were added to a 384-well PCR plate supplied that includes the LNATM primer sets in a final reaction volume of 10 μL. Furthermore, each panel included the following internal controls: 5 synthetic RNAs of the RNA Spike-in kit aimed to monitor the RNA isolation and cDNA synthesis, and an inter-plate calibrator in triplicate and a negative control to evaluate qPCR performance. qPCR reactions were performed as follows: a polymerase activation/denaturation cycle of 2 min at 95 ◦C followed by 55 cycles of 10 s at 95 ◦C and 1 min at 56 ◦C with a ramp-rate of 2.2 ◦C/s. All RT-qPCR reactions were conducted in a LightCycler 480 II (Roche, Mannheim, Germany). In the validation stage, selected miRNAs were quantified using specific LNA PCR primer sets (Qiagen) in a total of 153 urine samples from BC patients and 57 healthy controls.

#### *2.5. Selection of Candidate miRNA Normalizers and Analysis of Their Stability*

To normalize the expression level of each miRNA, the best candidate with the highest stability and the lowest biological variance over the entire range of samples being investigated (BC and controls) was selected. To that aim, all miRNAs with a mean Ct < 35 in BC and controls were scrutinized. To select the best normalizer, the comprehensive tool RefFinder was employed which integrates the computational programs Genorm [28], BestKeeper [29], the comparative Delta Ct method [30] and NormFinder [31] (https://www.heartcure.com.au/for-researchers/) [32].

#### *2.6. Statistical Analysis*

Continuous variables were presented as median and interquartile range, and categorical variables as count and percentage. The analysis of variance (ANOVA) with the Tuckey Post Hoc test and unpaired t-test were used to identify significant differences in miRNA expression levels between BC patients and healthy controls. The statistical analysis was performed using the GraphPad Prism software v.8.0.1 (GraphPad software Inc., La Jolla, CA, USA). The Venn diagram was performed using the online tool available on http://bioinformatics.psb.ugent.be/webtools/Venn/. *p*-Values < 0.05 were considered statistically significant.

#### **3. Results**

#### *3.1. Clinical Characteristics of the Study Subjects*

A total of 188 BC patients were prospectively recruited together with 72 healthy volunteers (control group) with similar age and sex. The clinical characteristics of the study subjects are depicted in Table 1. The patients studied in the screening stage (*n* = 35) were: 10 patients (28.57%) with a TaG1 BC, 8 patients (22.86%) with TaG3, 5 patients (14.29%) with T1G3 and 12 patients (34.29%) with T2G3. The patients studied in the validation stage (*n* = 153) were: 33 patients (21.57%) with TaG1, 13 patients (8.5%) with TaG3, 29 patients (18.95%) with T1G3 and 15 patients (9.80%) with T2G2/T2G3/T3G3. Additionally, in order to validate the proposed normalizer in the whole spectrum of BC patients, 2 more groups of patients were included in the validation stage: TaG2 (54 patients, 35.29%) and T1G2 (9 patients, 5.88%).


**Table 1.** Clinical characteristics of the BC patients and healthy controls studied.

Continuous variables are presented as median and interquartile range and categorical variables are presented as count and percentage.

#### *3.2. Quality Internal Control with Synthetic Spike-in RNAs*

To ensure that miRNA quantification was not influenced by technical and interpersonal variability, synthetic non-human spike-in RNAs are frequently used. We assessed the RNA isolation step by adding the synthetic spike-in 2 and spike-in 4 RNAs during all RNA isolations, and the retrotranscription efficiency by adding the spike-in 6 RNA in all retrotranscription reactions. No differences were observed in any spike-in studied among the study groups (Figure 2), thus indicating a proper performance of isolation and retrotranscription steps. To evaluate the qPCR performance of all reactions, the inter-plate calibrator spike-in 3 RNA in triplicate and a negative control were included in each panel. No differences were observed in any comparison made.

**Figure 2.** Differences in expression levels of three synthetic spike-in RNAs among BC patients with different stages and grades and healthy controls. Spike-in 2 and spike-in 4 monitor the RNA isolation step, spike-in 6 monitors the retrotranscription efficiency and spike-in 3 functions as inter-plate calibrator. Expression levels are represented as Ct values and error bars represent standard deviation.

#### *3.3. Selection of Candidate miRNA Normalizers and Analysis of Their Stability*

Of the 179 miRNAs quantified in each sample of the screening stage, we obtained high quality signals in 110 miRNAs (mean of the Ct < 35) both in BC patients and controls, thus they were included in the analysis with RefFinder. This tool comprehends the computational algorithms Genorm, BestKeeper, Delta Ct and NormFinder. Figure 3 shows the best 10 reference miRNAs selected by each algorithm. The stability analysis conducted with Genorm revealed that the greatest stability was reached by the combination of let-7e-5p and let-7a-5p (Figure 3a). BestKeeper revealed that the most stable miRNA was miR-2110 (Figure 3b). The Delta Ct method and NormFinder agreed with the most stable miRNA being miR-29c-3p (Figure 3c,d). Finally, the recommended comprehensive ranking that integrates all the previous analyses, rendered miR-29c-3p as the most stable miRNA (Figure 3e), being aligned with the results of the Delta Ct method and NormFinder. Next, we represented in a Venn diagram the overlap among the best 10 reference miRNAs selected by each algorithm (Figure 4). Three miRNAs were shared by the Delta Ct method, Genorm and NormFinder: miR-29c-3p, miR-26a-5p and miR-361-5p. Likewise, 7 miRNAs were shared by Delta Ct Method and NormFinder. In contrast, none of the 10 most stable miRNAs rendered by BestKeeper were selected by any other algorithm, thus none of them were included in the comprehensive ranking provided by RefFinder and consequently discarded as potential normalizers in our study.

**Figure 3.** Selection of candidate miRNA normalizers and analysis of their stability conducted with the comprehensive tool RefFinder. Each graph represents the best 10 reference miRNAs selected by each algorithm: (**a**) Genorm, (**b**) BestKeeper, (**c**) Delta Ct method, (**d**) NormFinder and (**e**) Comprehensive ranking. The lower the stability value, the higher the stability of each miRNA.

**Figure 4.** Venn diagram presenting the overlap among the best 10 reference miRNAs selected by each algorithm. Delta Ct method, NormFinder, Genorm and BestKeeper.

#### *3.4. Di*ff*erences in Expression Levels of Candidate miRNA Normalizers between BC Patients and Controls*

A crucial characteristic of a good normalizer in miRNAs studies is the stable expression among the samples analyzed. Thus, we compared the mean Ct values of the best 10 reference miRNAs selected by the comprehensive ranking of RefFinder between BC patients and controls. No significant differences were observed in the expression of miR-29c-3p, let-7e-5p, miR-26a-5p, let-7a-5p and let-7c-5p (unpaired t-test, *p* > 0.05). In contrast, we found significant differences in the expression levels of let-7b-5p (*p* = 0.029), miR-181-5p (*p* = 0.026), miR-361-5p (*p* = 0.025), miR-107 (*p* = 0.012) and miR-151-5p (*p* = 0.015) (Figure 5). As expected, those miRNAs with the highest stability value had similar expression levels between BC and controls.

**Figure 5.** Differences in expression levels of the candidate miRNA normalizers selected by the comprehensive ranking of RefFinder between BC patients and controls. Expression levels are represented as Ct values and error bars represent standard deviation. Unpaired T-test: \* *p* < 0.05.

#### *3.5. E*ff*ect of Di*ff*erent Normalization Strategies on the Relative Quantification of a miRNA Closely Related with BC*

miR-200c-3p has been previously proposed as urinary diagnostic biomarker for BC [24,33]. Thus, we evaluated the performance of the 10 most stable miRNAs, selected by the comprehensive ranking of RefFinder, as normalizers for miR-200c-3p quantification in the screening cohort. As seen in Figure 6, significant differences among the BC groups studied and controls were observed when any of the normalizers were used, although the results obtained were not always comparable in magnitude and trend.

**Figure 6.** Relative expression of miR-200c-3p normalized by each candidate miRNA selected by the comprehensive ranking of RefFinder. Normalization was performed by the 2−ΔΔCt method. Error bars represent the standard error of the mean. ANOVA with the Tuckey Post Hoc test: \* *p* < 0.05; \*\* *p* < 0.01; \*\*\* *p* < 0.001.

In consideration of the aforementioned results, we selected miR-29c-3p as the best normalizer for the following reasons: (1) It was proposed as the most stable miRNA in the comprehensive analysis of RefFinder, and in two of the main algorithms independently tested (Figure 2); (2) Its expression level was optimal for a proper quantification (mean Ct = 28.4) (data not shown); and (3) Significant differences in the expression of miR-200c-3p, a miRNA related to BC, were observed when it was used as normalizer (Figure 6).

#### *3.6. Validation of the Performance of miR-29c-3p as Normalizer in an Independent Cohort of BC Patients and Controls*

We verified the stability of miR-29c-3p in an independent cohort of BC patients and healthy controls. As in the screening cohort, no significant differences were observed in the expression of miR-29c-3p between BC patients (mean Ct = 27.65) and healthy controls (mean Ct = 28.13) (*p* = 0.37). Next we validated the robustness and efficacy of miR-29c-3p as endogenous control for BC in the validation cohort using the 2−ΔΔCt method by analyzing the expression level of the BC-related miR-200c-3p in every group of BC patients and controls. As occurred in the samples studied in the screening stage, we observed significant differences in the expression of miR-200c-3p among the different clinical groups studied in the validation cohort (*p* < 0.001), with an increase of miR-200c-3p with the increase in BC stage (Figure 7). As occurred in the screening cohort, the biggest differences were found in the T1G3 BC group. As the patients with mildest stage of BC seem to have a lower expression level of miR-200c-3p, we grouped all the patients with the Ta stage and repeated the analysis. Figure S1a confirms that the expression of miR-200c-3p significantly increases with the severity of NMIBC, being T1G3 the BC type with the highest expression. Next, we grouped all NMIBC patients (TaG1+TaG2+TaG3+T1G2+T1G3) and compared the expression of miR-200c-3p to that of to MIBC patients (T2G2+T2G3+T3G3) and healthy controls. While NMIBC patients still showed the highest expression level of miR-200c-3p, no significant differences were observed when compared to MIBC or healthy controls (Figure S1b), probably because the highest difference occurs within the NMIBC group.

**Figure 7.** Relative expression of miR-200c-3p normalized by miR-29c-3p in the validation cohort. Normalization was performed by the 2−ΔΔCt method. Error bars represent the standard error of the mean. ANOVA with the Tuckey Post Hoc test: \*\* *p* < 0.01; \*\*\* *p* < 0.001.

#### **4. Discussion**

New non-invasive markers are presently being under study to circumvent several drawbacks in BC diagnosis, monitoring and prognosis. Urine miRNAs are non-invasive promising biomarkers that have been previously proposed for BC diagnosis [13–17,19–21,23,24]. However, huge discrepancies arise among miRNA studies, to a great extent due to nonexistence of standardized procedures. Although the populations studied, sample processing, and RNA isolation and miRNA quantification methods are partly responsible for these inconsistencies; the normalization strategy used may represent the main hurdle. In fact, to minimize the effect occasioned by methodology-related factors on miRNA expression levels, an accurate data analysis ought to be performed using appropriate normalizers for external and internal variation correction [26]. These normalizers should be chosen from a selection of candidates that are expected to be stably expressed over the entire range of samples being investigated, since miRNAs can be affected by the condition under study. As an alternative, the mean expression value of all commonly expressed microRNAs in a given sample has been proposed for normalization [34]. Although this strategy presents good and robust results it implies that a large number of miRNAs have to be always profiled, which may not be possible or cost-effective in all studies [26,35]. Thus, as a general guideline suggested by the manufacturer, the use of the global mean for normalization is limited to the use of PCR panels that contain a larger number of microRNA assays, and it cannot be used for studies analyzing less than 20–50 different miRNAs.

Different RNA species have been proposed as normalizers (snRNAs, snoRNAs, rRNAs, miRNAs or exogenous synthetic RNAs); however, substantial differences in length and structure to that of miRNAs generate a high variability in the results [26]. Regarding the use of exogenous synthetic RNAs as normalizers, these are meant to track isolation and reverse transcription efficiency in order to eliminate deviations in the experimental process and make the results more reliable. However, it is important to remark that their use would never correct for deviations in sampling, in the quality of samples or in the amplification process. In fact, age, sample collection, preparation or storing can modify miRNA expression levels, which may be caused by cell lysis or miRNA degradation [26].

Several molecules have been proposed as normalizers for miRNAs studies in different disorders, mainly in blood [36], plasma [37–39], cell cultures studies [25,40], tissue [41] and urine [35,42–44]. However, there is no consensus, even in the same sample type, regarding which is the best normalizer for miRNA studies. Indeed, no consensus exists on a robust normalizer for urine studies. A recent study proposed miRNA miR-193a and miR-448 as normalizers [35]. However, in this study, only urine from healthy donors was analyzed and these results could vary when urine from cancer patients is investigated. In prostate cancer, miR-191-5p showed the lowest degree of variation and the highest stability value [42]. In our study, miR-191-5p ranked 29 out of 110 according to the comprehensive ranking of RefFinder, thus it is not a reliable normalizer for BC studies in urine. miR-16 was identified as the most stable endogenous reference miRNA in chronic kidney disease, making it a suitable normalizer for urinary exosome-derived miRNA [43]. Conversely, in our data set, miR-16-5p ranked 106 out of 110 according to the comprehensive ranking of RefFinder, turning it an ineffectual normalizer for urine BC studies. Finally, U6 has been proposed as normalizer for miRNA studies in urinary sediment of IgA nephropathy [44]. Although U6 was widely used as normalizer in countless studies at the origins of miRNA investigation, it is a member of the larger small RNA species which have a different biogenesis pathway (originate from the nucleus), may not be secreted or protected in cell-free biofluids in the same way that microRNAs are and may also behave differently during RNA isolation. Thus, U6 is an unreliable normalizer for urine miRNA studies. Other studies have employed the combination of two stable miRNAs as normalizers however, this strategy presents several drawbacks: it may increase technical variability, it implies a higher economic cost since the number of miRNAs to be quantified by RT-qPCR increases and it is more time consuming. Altogether, the use of a combination of miRNAs as normalizers hampers the direct translation of miRNA studies to daily clinical practice with diagnostic/staging purposes.

In the present study, we set for the first time the aim to ascertain the best miRNA normalizer for miRNA studies in urine of BC patients in order to avoid future inconsistencies among studies. We evaluated the performance of 110 candidate miRNAs with the comprehensive tool RefFinder that integrates 4 programs (Genorm, BestKeeper, Delta Ct method and NormFinder) in 35 BC patients and 15 healthy controls. We selected miR-29c-3p as the best normalizer for miRNA studies in urine of BC. It was the most stable miRNA according to the comprehensive analysis of RefFinder among the 110 studied, and also according to the Delta Ct method and NormFinder. Moreover, it had a good

expression level in urine (mean Ct value in BC patients and healthy controls = 28.4) and no differences were observed between BC patients and controls, both in the screening and the validation cohorts.

Urine miR-200c-3p has been previously related to BC and it has been proposed as diagnostic and staging marker [24,33]. miR-200c-3p appears to control the epithelial-to-mesenchymal transition process through BMI-1 in BC cells, and it inhibits their proliferation by down-regulating E2F3 [45]. Thus, we selected this miRNA to test the robustness of miR-29c-3p as endogenous normalizer. We found significant differences in miR-200c-3p among the different clinical groups studied both in the screening and validation cohorts, with a trend in the increase of miR-200c-3p with the severity of NMIBC. The evaluation of additional stably expressed miRNAs proposed by RefFinder may have rendered other potential normalizers for urine miRNA studies in the context of BC, what represents a limitation of our study. Nonetheless, our results confirm previous findings and reinforce the use of miR-29c-3p as normalizer.

#### **5. Conclusions**

In summary, our study is the first report characterizing a reliable normalizer for the analysis of urine miRNAs in BC patients. miR-29c-3p, being one of the most stably expressed miRNAs in urine of BC patients and healthy individuals, arises as an optimal reference miRNA that may allow the comparison of future urine miRNA studies as non-invasive biomarkers for BC diagnosis and monitoring.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2227-9059/8/11/447/s1, Table S1: miRNAs and internal controls included in the Serum/plasma miRNA PCR Panel V5 (Qiagen); Figure S1. Relative expression of miR-200c-3p normalized by miR-29c-3p in the validation cohort. (a) Comparison of the mildest stage of NMIBC patients (TaG1+TaG2+TaG3) and the other clinical groups. (b) Comparison of NMIBC patients (TaG1+TaG2+TaG3+T1G2+T1G3), MIBC (T2G2+T2G3+T3G3) and healthy controls. Normalization was performed by the 2−ΔΔCt method. Error bars represent the standard error of the mean. ANOVA with the Tuckey Post Hoc test: \* *p* < 0.05; \*\* *p* < 0.01; \*\*\* *p* < 0.001.

**Author Contributions:** J.O. processed samples, performed the research, analyzed the data and wrote the manuscript. E.P. analyzed the data and wrote the manuscript. Á.F.-P. performed the research and critically revised the manuscript. F.C. processed the samples and prepared the databases. M.M.-S. recruited the patients, revised the clinical records and critically revised the manuscript. C.D.V.-D. recruited the patients, revised the clinical records and critically revised the manuscript. F.E. designed and supervised the study and critically revised the manuscript. P.M. designed and supervised the study, analyzed the data and wrote the manuscript. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by research grants from Instituto de Salud Carlos III (PI17/00495), FEDER una manera de hacer Europa, Generalitat Valenciana (ACIF/2017/138), Sociedad Española de Trombosis y Hemostasia and Sociedad Española de Angiología y Cirugía Vascular.

**Acknowledgments:** We would like to thank María José Solmoirago for her technical support.

**Conflicts of Interest:** The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

#### **References**


**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Review* **MicroRNAs Regulating Hippo-YAP Signaling in Liver Cancer**

**Na-Hyun Lee 1,2, So Jung Kim 1,2 and Jeongeun Hyun 1,2,3,\***


**Abstract:** Liver cancer is one of the most common cancers worldwide, and its prevalence and mortality rate are increasing due to the lack of biomarkers and effective treatments. The Hippo signaling pathway has long been known to control liver size, and genetic depletion of Hippo kinases leads to liver cancer in mice through activation of the downstream effectors yes-associated protein (YAP) and transcriptional coactivator with PDZ-binding motif (TAZ). Both YAP and TAZ not only reprogram tumor cells but also alter the tumor microenvironment to exert carcinogenic effects. Therefore, understanding the mechanisms of YAP/TAZ-mediated liver tumorigenesis will help overcome liver cancer. For decades, small noncoding RNAs, microRNAs (miRNAs), have been reported to play critical roles in the pathogenesis of many cancers, including liver cancer. However, the interactions between miRNAs and Hippo-YAP/TAZ signaling in the liver are still largely unknown. Here, we review miRNAs that influence the proliferation, migration and apoptosis of tumor cells by modulating Hippo-YAP/TAZ signaling during hepatic tumorigenesis. Previous findings suggest that these miRNAs are potential biomarkers and therapeutic targets for the diagnosis, prognosis, and treatment of liver cancer.

**Citation:** Lee, N.-H.; Kim, S.J.; Hyun, J. MicroRNAs Regulating Hippo-YAP Signaling in Liver Cancer. *Biomedicines* **2021**, *9*, 347. https:// doi.org/10.3390/biomedicines9040347

Academic Editor: Francesca Lovat

Received: 9 March 2021 Accepted: 25 March 2021 Published: 30 March 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

**Keywords:** microRNAs; hepatocellular carcinoma; Hippo kinase; Yes-associated protein; liver disease; diagnosis; prognosis; cancer therapy

#### **1. Introduction**

Liver cancer is one of the leading causes of cancer-related death worldwide, with a 5-year survival rate as low as 30–40% [1,2]. Hepatocellular carcinoma (HCC) is the most common type of primary liver cancer and occurs in patients with chronic liver disease [3]. Despite remarkable advances in research to overcome viral hepatitis, which is the biggest cause of HCC, the incidence and mortality of liver cancer are still increasing along with the increased prevalence of nonviral steatohepatitis. However, the complex pathophysiology of liver cancer has limited the development of effective diagnosis and therapeutic intervention, prompting a comprehensive understanding of liver carcinogenesis.

Many studies have revealed that several developmental pathways, such as Wnt/βcatenin, Hedgehog (Hh) and Hippo/Yes-associated protein (YAP) signaling, contribute to hepatic carcinogenesis [4]. For example, Hh signaling is inactive in the normal liver of adult mice and humans but reactivated in chronic liver diseases and liver cancers, promoting liver fibrosis and hyperplasia and constructing a tumor-favorable microenvironment [5]. Likewise, overexpression of YAP in mice massively increases the size of the liver due to hyperproliferation of liver cells and ultimately promotes hepatic tumorigenesis [6]. The hepatocyte-specific induction of YAP nuclear localization dedifferentiates hepatocytes into ductal-like progenitor cells that are highly proliferative [7]. Similar results have been found after genetic ablation of one of the key components of the Hippo signaling pathway, including Neurofibromin 2 (Nf2), Mammalian STE20-like 1/2 (Mst1/2), and Large Tumor

Suppressor 1/2 (Lats1/2), which regulate the activity of YAP and transcriptional coactivator with PDZ-binding motif (TAZ), another downstream effector of Hippo signaling, by phosphorylation and subsequent proteasomal degradation [8,9]. For example, Nf2 conditional null mice show hyperproliferation of hepatic progenitor cells, known as a ductular reaction, and develop both HCC and intrahepatic cholangiocarcinoma (ICC) [10,11]. Notably, codeletion of Yap suppresses liver overgrowth, progenitor expansion and tumor development in Nf2 knockout mice, demonstrating that Yap drives hepatic tumorigenesis caused by inactivated Hippo signaling [10]. Accordingly, it has been observed that the expression of NF2 negatively correlates with the expression of YAP in liver tissues of patients with HCC or ICC [12,13]. Except for NF2, for which missense mutations have been discovered in 1.9% and 5.3% of human HCC and ICC cases, respectively [12], somatic or germline mutations in neither MST1/2 nor LATS1/2 genes have been reported in common cancers [14]. This suggests that post-transcriptional regulation or epigenetic silencing, rather than DNA mutations, of the Hippo signaling pathway may play central roles in the aberrant inactivation of Hippo signaling and hyperactivation of YAP/TAZ transcriptional coactivators in liver cancers. Nevertheless, whether and how the activity of the Hippo and YAP/TAZ signaling pathways are regulated post-transcriptionally has not been well investigated.

MicroRNAs (miRNAs) are a group of small noncoding RNAs that are approximately 21–25 nucleotides in length [15]. miRNAs regulate the expression of approximately 30% of protein-coding genes at the post-transcriptional level by binding directly to target messenger RNA (mRNA), which results in translational suppression or degradation of target mRNAs [16]. Many studies have demonstrated that miRNAs play important roles in diverse biological processes, including cell proliferation, differentiation, and death; thus, abnormal regulation of miRNAs can lead to various pathological conditions, including liver cancers [17]. miRNAs have become promising candidates for biomarkers and therapeutic targets in many cancers due to their regulatory functions and detectable properties in various biological fluids, including blood, saliva, and urine.

Recently, several miRNAs have been reported to be associated with Hippo-YAP/TAZ signaling in the control of liver cancer cell behaviors [18,19]. Here, we introduce recent findings of the interaction between miRNAs and Hippo-YAP/TAZ signaling components in the development and progression of liver cancers. We also discuss what remains to be addressed in future studies to improve our knowledge of the underlying mechanisms whereby dysregulation of miRNA and Hippo-YAP/TAZ signaling contributes to hepatic malignancies. Finally, we propose the potential of these miRNAs in clinical applications for the diagnosis, prognosis, and treatment of liver cancers.

#### **2. Hippo-YAP/TAZ Signaling Pathway**

Hippo signaling, which was first discovered in *Drosophila*, is evolutionarily conserved in vertebrates as a key regulator of organ growth [20,21]. In mammals, two serine/threonine kinases, the STE20-like protein kinase MST1/2 (Hpo in *Drosophila*) and the NDR family protein kinase LATS1/2 (Wts in *Drosophila*), consist of a core kinase cascade of the Hippo signaling pathway [8,22] (Figure 1). The MST1/2 kinases form a complex with Salvador 1 (SAV1, Sav in *Drosophila*) to phosphorylate and activate LATS1/2 kinases, which in turn phosphorylate and sequester the two major downstream effectors of the Hippo pathway, YAP and TAZ (Yki in *Drosophila*), by promoting the association of YAP/TAZ with 14-3-3 proteins in the cytoplasm where ubiquitin-mediated degradation of YAP and TAZ occurs [8,23–25]. MST1/2 also interact with Mps one binder kinase activator-like 1A/B (MOBKL1A/B, Mats in *Drosophila*), which enhances the activity of LATS1/2 kinases [26,27]. When Hippo signaling activity is suppressed, YAP and TAZ are dephosphorylated, translocate into the nucleus, and function as transcriptional coactivators via interactions with various transcription factors, such as p73, TEA domain family member (TEAD, Sd in *Drosophila*), SMAD, and Runt-related transcription factor (RUNX), to induce the expression of genes promoting cell proliferation and inhibiting apoptosis [25]. For example, *AREG* [28], *BIRC5* [29], *CCNE1* [30], *CTGF* [31,32], *CYR61* [30,32], and *GLI2* [33] have been identified as direct target genes of both YAP and TAZ. The upstream regulators of the Hippo kinase cascade include FAT1-4 (Fat in *Drosophila*), Merlin (encoded by the *NF2* gene; Mer in *Drosophila*), KIBRA (Kibra in *Drosophila*), RASSF, and Ajuba [25]. MST1/2 kinases are activated by FAT1-4 through the apical protein FRMD6/Willin (Ex in *Drosophila*) that forms a complex with two other apically localized proteins, Merlin and KIBRA [34,35]. In contrast, both RASSF and Ajuba inhibit the Hippo signaling pathway by competing with SAV1 for binding with MST1/2 [36]. In addition, YAP and TAZ sense extracellular mechanical stimuli, such as extracellular matrix (ECM) stiffness and cell geometry, and integrate and convert them into intracellular molecular signals, resulting in changes in cellular behaviors, including cell proliferation, migration, and transdifferentiation [37].

**Figure 1.** MicroRNAs that regulate the core components of Hippo signaling pathway. Several microRNAs are involved in kinase cascade (MST1/2 and LATS1/2) and downstream effectors (YAP/TAZ) of the Hippo signaling pathway in liver cancer as either oncogenes or tumor suppressors. Abbreviations: NF2, Neurofibromin 2; MST1/2, Mammalian STE20-like 1/2; SAV, Salvador; LATS1/2, Large Tumor Suppressor 1/2; MOB, Mps one binder kinase activator; YAP/TAZ, Yes-associated protein and transcriptional coactivator with PDZ-binding motif; TEAD, TEA domain family member; VGLL4, Vestigial like family member 4; AREG, amphiregulin; CTGF, connective tissue growth factor; CYR61, Cysteine-rich angiogenic inducer 61.

#### **3. The Roles of the Hippo-Yap/Taz Signaling Pathway in Hepatic Tumorigenesis**

Hippo-YAP/TAZ signaling is well known to control organ size during development and to mediate the expansion of tissue-specific progenitor cells during tissue regeneration and normal cell proliferation [38]. Accumulated evidence shows aberrant expression of Hippo kinases, YAP/TAZ and their partners in many human cancers, including liver cancers [39–41]. In normal livers, Hippo kinases act as tumor suppressors by inhibiting hepatocyte proliferation and maintaining the differentiated state of hepatocytes [42]. In contrast, loss of Hippo kinase activities, as in mice with a genetic deletion of Nf2 [10,43], Mst1/2 [43,44], Lats1/2 [45,46], or Sav1 [47], causes hepatomegaly and liver cancers, including HCC, ICC, and/or the HCC/ICC mixed form. YAP, on the other hand, functions as an oncogene [48]. Overexpression of YAP phenocopies Hippo signaling deficiency in mice, as shown by liver overgrowth, which is mediated by an increase in hepatocyte proliferation coordinated with a decrease in hepatocyte death [49]. Moreover, removal of Yap in mice with Hippo signaling components knocked out prevented hepatomegaly and hepatic tumor development [10], indicating that YAP is required for hyperplastic cell proliferation and oncogenic transformation of liver cells. It has also been revealed that YAP induces the epithelial-to-mesenchymal transition (EMT), suppression of apoptosis, growth factor-independent proliferation, and anchorage-independent growth of cancer cells, which are attributes of cancer stem cells that are responsible for the major causes of cancer mortality, such as chemoresistance, metastasis, and recurrence [40].

Recent studies have demonstrated that activation of YAP and TAZ in tumor cells also fosters a tumor-favorable microenvironment by communicating with neighboring stromal cells [50,51]. For example, activated YAP/TAZ in hepatocyte-specific Mst1/2 knockout mice creates an inflammatory tumor microenvironment by increasing the production of inflammatory cytokines to suppress immune clearance of transformed hepatocytes by recruiting tumor-associated macrophages and to promote liver cancer development [38,41]. In addition, both tumor cells and cancer-associated fibroblasts increase the stiffness of extracellular matrices, which activates YAP/TAZ mechanosensors to mediate the metabolic crosstalk between tumor cells and cancer-associated fibroblasts, providing them with sufficient nutrients for tumor growth and maintenance of a cancer-prone microenvironment [52–54]. Hence, Hippo-YAP/TAZ signaling has been targeted for developing anticancer therapeutics [21].

Most Hippo-YAP/TAZ signaling genes are rarely mutated in liver cancers, which strongly suggests that molecular events, such as epigenetic or post-transcriptional regulation in response to mechanical stresses from the tumor microenvironment, other than DNA mutations, may cause dysregulation of the Hippo-YAP/TAZ signaling pathway in in liver cancers [19,40].

#### **4. MiRNAs Interacting with the Hippo-Yap/Taz Signaling Pathway in Liver Cancer**

Dysregulation of biogenesis and expression of miRNAs affect the incidence and progression of liver cancers [55]. miRNAs can function as either tumor suppressors or oncogenes depending on their target mRNAs and resultant phenotypic changes in the cells [56]. In this section, we discuss the interplay between miRNAs and the components of the Hippo-YAP/TAZ signaling pathway that play important roles in hepatic cancer biology (Figure 1).

#### *4.1. MiRNAs as Tumor Suppressors*

#### 4.1.1. MiRNAs Targeting YAP or TAZ

Bioinformatic algorithms performed by Liu et al. [57] identified that miR-375 binds directly to the 3 untranslated region of YAP mRNA. In PLC/PRF/5 and MHCC-97L HCC cell lines, ectopic expression of miR-375 downregulated the expression level of YAP protein and CTGF mRNA, which can be transcribed by activated YAP [57]. Recently, Dinh and Jewell et al. [58] found that miR-375 is the most downregulated miRNA in primary fibrolamellar carcinoma (pFLC), a rare liver cancer that primarily affects adolescents and young adults, compared with nonmalignant human livers. The loss of miR-375 was induced by the presence of the DNAJB1-PRKACA fusion gene, a hallmark of FLC, although it is yet unknown how DNAJB1-PRKACA inhibits miR-375 expression [58]. As in HCC cells, the overexpression of miR-375 in FLC cells inhibited YAP and CTGF, mitigating the proliferative and migratory ability of tumor cells [58]. Another miRNA, miR-186 [59], also decreases the expression of YAP and CTGF by directly disrupting YAP mRNA, inhibiting the proliferation, migration, and invasion of HepG2, Hep3B and SNU398 HCC cell lines [59]. CTGF plays a central role in tissue remodeling and liver fibrosis, which increases the risk for cancer development and progression [60]. Moreover, CTGF is associated with tumor progression by forcing crosstalk between cancer cells and hepatic stellate cells to form a tumor-favorable microenvironment [60]. Thus, the miRNAs discussed above might have tumor suppressive effects not only by killing cancer cells but also by targeting the tumor microenvironment (Table 1).

Both miR-506 and miR-132 also have binding sites in the 3 untranslated region of YAP mRNA [61,62]. It was shown that the expression of miR-506 is significantly reduced in human HCC tissues and inversely correlates with the expression of the YAP gene [61]. It was experimentally revealed that miR-506 inhibits the growth of HepG2 and H7402

HCC cell lines by downregulating YAP expression [61]. miR-132 promotes apoptosis and suppresses proliferation and invasion of Huh-7 and HepG2 HCC cell lines through direct inhibition of YAP [62]. When miR-132 was cotransfected with miR-520-3p, another tumor suppressive miRNA modulating GPC3, into the Huh-7 HCC cell line, the antiproliferative and proapoptotic functions of both miRNAs on the HCC cell line were shown to be enhanced by minimizing the level of YAP [63]. Similarly, Piao et al. [64] reported that miR-424-5p decreases YAP expression, attenuates proliferation and induces the apoptosis of ICC cell lines, including CCLP-1, RBE and HuCCT-1 [64]. Zhang et al. [65] showed the in vitro anticancer effect of miR-497 by suppressing its functional target, YAP. Overexpression of miR-497 inhibits the growth and survival of HCC cell lines such as HepG2 and Huh-7, whereas silencing of miR-497 has the opposite effect [65].

Chemoresistance is one of the major issues in HCC treatment [66]. Chen et al. [66] found that the level of miR-590-5p negatively correlates with YAP expression in HCC, which is resistant to adriamycin (also known as doxorubicin), the most common first-line chemotherapeutic agent for transarterial chemoembolization [66]. Furthermore, they found that miR-590-5p directly targets YAP and that dysregulation of the miR-590-5p/YAP axis leads to chemoresistance in HCC. The potential anticancer role of miR-509-3p through the suppression of YAP in cholangiocarcinoma (CCA) has been reported by Jung et al. [67]. They found that anticancer drugs such as gemcitabine show synergistic effects when treated in combination with a novel inhibitor of histone deacetylase, CG200745, in CCA cell lines and that CG200745 effectively suppresses tumor growth in xenograft mouse models of gemcitabine-resistant CCA [67]. By performing a microarray and a miRNA array, they further found that CG200745 inhibits the expression of YAP-TEAD4 signaling and increases the expression of miR-509-3p, which downregulates the level of YAP protein [67], possibly by directly targeting YAP mRNA, as in ovarian cancer cells [68].

miRNAs that target TAZ, a paralog of YAP, in HCC include miR-338-3p, miR-9-3p, and miR-125b [69–71]. miR-338-3p has also been reported to be suppressed by the hepatitis B virus (HBV)-encoded protein preS2, upregulating the expression of TAZ in HBV-associated HCC [69]. Both miR-9-3p and miR-125b are known to be downregulated in human HCC tissues and hepatic tumor cells, while the expression of TAZ is upregulated [70,71]. In particular, miR-9-3p is involved in the noninvasive proliferation of tumor cells via signaling pathways, including AKT, ERK1/2, and β-catenin [70]. miR-125b is capable of inhibiting cell invasion and migration through the regulation of TAZ expression [71]. However, the roles of miR-9-3p and miR-125b in human cancer are controversial, and they could function as either tumor suppressors or oncogenes depending on the type of cancer [70,71]. For example, miR-9-3p has been reported as a tumor suppressor in liver, breast, oral, and gastric cancers, whereas it is known as an oncogene primarily in brain cancer [70]. miR-223 plays a critical role in the progression of nonalcoholic steatohepatitis (NASH) to HCC by targeting Taz and the inflammatory gene Cxcl10 in mouse hepatocytes [72]. He et al. [72] found that chronic high-fat diet treatment in miR-223 knockout mice increases the prevalence of liver cancer by activating oncogenic and inflammatory pathways through TAZ and CXCL10 compared to wild-type mice. These findings are consistent with other reports that the level of miR-223 is substantially downregulated in human HCC [72] and that chronic inflammatory injury promotes hepatobiliary carcinogenesis [72].


**Table 1.** List of tumor suppressive microRNAs inactivating YAP/TAZ and their effect(s) on liver cancer.

Abbreviation: miR, microRNA; HCC, hepatocellular carcinoma; FLC, fibrolamellar carcinoma; NASH, nonalcoholic steatohepatitis; HBV, hepatitis B virus. † All direct targets were confirmed by luciferase reporter assay. ‡ ↑ indicates the promotion, ↓ indicates the suppression.

#### 4.1.2. MiRNAs Interacting with LATS1/2

Yang et al. [75] reported that miR-195 is one of a few miRNAs expressed differently between the drug-resistant HCC cell line BEL-7402/5-FU and its parental cell line BEL-7402. miR-195 was shown to be downregulated in HCC cells that acquired drug resistance [75]. Overexpression of miR-195 upregulates the expression of LATS2 and downregulates BCLw, an anti-apoptotic protein, which sensitizes BEL-7402/5-FU cells to anticancer drugs by suppressing proliferation and inducing apoptosis of the cells [74,75].

Some miRNAs play a tumor suppressive role by interacting with other epigenetic regulators [76]. For example, Wu et al. [73] found that miR-29c-3p promotes DNA demethylation of the LATS1 gene through the direct inhibition of DNA methyltransferase 3B (DNMT3B), resulting in upregulated LATS1 expression (and thus increased Hippo kinase activity) to suppress oncogenic YAP activation. Consistent with the findings that the expression of DNMT3B is increased in many malignancies [73], Wu et al. [73] observed that the level of DNMT3B is higher in HCC tumor tissues than in adjacent nontumor tissues, while miR-29c-3p and LATS1 are expressed at lower levels in HCC tumors than in nontumors. Moreover, patients with either high DNMT3B, low LATS1 or low miR-29c-3p/LATS1 with high DNMT3B have worse outcomes [73].

#### *4.2. MiRNAs as Oncogenes*

#### 4.2.1. MiRNAs Targeting MST1 and LATS1/2

Cheng et al. [77] showed that miR-3910 is highly expressed in human HCC tissues and various HCC cell lines compared to its level in nontumor liver tissues and normal liver cells. They found that miR-3910 directly inhibits the expression of MST1, promoting the growth and migration of HCC cells in vitro and tumor formation in vivo through the activation of oncogenic YAP (Table 2).

Other studies have shown that miRNAs, including miR-103 [78] and miR-650 [79], directly bind to and inhibit LATS2 mRNA, leading to an increase in YAP, which promotes EMT, metastasis, and invasion of cancer cells, particularly in the liver. In addition, miR-1307-3p is induced by MEIS2, a homeobox protein that promotes HCC development and downregulates LATS1 [80]. Thus, the oncogenic function of MEIS2 is accomplished by the miR-1307-3p/LATS1 axis promoting YAP nuclear translocation in addition to its association with Wnt/β-catenin signaling [80]. Interestingly, a recent study reported that miR-15b in extracellular vesicles (EVs) derived from macrophages after exposure to arsenite, a

carcinogen, is transferred to HCC cells and inactivates Hippo signaling by directly targeting LATS1 to promote the proliferation, migration, and invasion of HCC cells [81]. This study suggested the role of miR-15b as a messenger between tumor-associated macrophages and tumor cells in the progression of HCC and that targeting miR-15b could be a strategy for the treatment of liver cancer.

#### 4.2.2. MiRNAs Interacting with YAP/TAZ

Fibrosis is known as a precancerous condition that raises the risk for cancers, including liver cancer [82], and ECM stiffness increases in liver fibrosis and cirrhosis [83]. The mechanosensitive miRNAs of the miR-130/301 family have been shown to regulate fibrosisrelated pathways, and their expression depends on the activation of YAP/TAZ, which is promoted by and enforces ECM stiffening [84]. miR-130a also plays a central role in positive feedback regulation of YAP expression by directly inhibiting VGLL4, an antagonist of YAP for TEAD binding, after its expression is induced by YAP [85]. Consequently, endogenous miR-130a strengthens the YAP-TEAD complex, which is a major player in hepatic carcinogenesis [85].

In addition, Hu et al. [86] reported that miR-665 is highly expressed in HCC, suppresses LATS1 activity and enhances activated YAP by negatively regulating the tyrosine phosphatase receptor type B (PTPRB) gene, the protein of which phosphorylates the LATS1 Hippo kinase. As a result, overexpression of miR-665 markedly enhances EMT, cell cycle progression, migration, and invasion of HepG2 cells in vitro [86]. Furthermore, Hu et al. [86] demonstrated that miR-665 promotes tumor growth and metastasis in HepG2 xenograft mouse models in vivo. Recently, Lu et al. [87] reported that miR-1254 is upregulated in human HCC tissues compared with adjacent nontumor tissues and various HCC cell lines, including Hep3B and Huh-7. They also demonstrated that miR-1254 promotes the proliferation, migration, and invasion of HCC cells in vitro and enhances tumor size, vascular invasion, histological grade of HCC assessed by the Edmondson-Steiner scoring system, and lung metastasis in xenografts in vivo [87]. Notably, they found that paired box gene 5 (PAX5) is directly regulated by miR-1254 and that miR-1254 inhibits the phosphorylation of LATS1 and YAP, which seems to be dependent on the availability of PAX5 [87]. Consequently, overexpression of miR-1254 promotes HCC progression [87].

TGF-β signaling is closely associated with YAP/TAZ signaling, and both signaling pathways are known to involve many cellular processes, including proliferation, differentiation and tumor formation and progression [88,89]. Hong et al. [90] reported that miR-21-3p is enriched in human HCC tissues and exerts oncogenic effects on hepatocarcinogenesis and HCC progression by targeting SMAD7, one of the negative regulators of the TGF-β signaling pathway [90], to upregulate oncogenic YAP.



Abbreviation: miR, microRNA; MΦ, macrophage; HCC, hepatocellular carcinoma; NASH, nonalcoholic steatohepatitis. † All direct targets were confirmed by luciferase reporter assay. ‡ ↑ indicates the promotion, ↓ indicates the suppression.

#### **5. Potential for Clinical Use of MiRNAs Interacting with the Hippo-YAP/TAZ Signaling Pathway in Liver Cancer**

Liver cancer is often asymptomatic in the early stages and is usually diagnosed after the advent of metastasis and advanced stages, limiting the chances for surgical treatment with an optimistic prognosis [17]. Although liver biopsy is currently the gold standard for diagnosing liver cancer, it can cause severe complications due to its invasiveness and is restricted to some subset of patients [91]. Various blood-based tests are used clinically, but the currently available biomarkers show insufficient specificity and sensitivity [92]. Hence, there is an unmet need for biomarkers that can diagnose liver cancer early and differentiate the stage of liver cancer. Over the past several years, many studies have focused on identifying circulating miRNAs that have the potential to be used as biomarkers of liver cancer, since they can be detected in cell culture media and different biological fluids, such as serum, plasma, saliva, tears, urine, and breast milk [93–95]. Unlike cellular miRNAs or other RNAs that are degraded within a few seconds, circulating miRNAs are relatively stable, viable for a long time and resistant to endogenous RNase activity in the extracellular environment [96]. Chen et al. [97] reported that circulating miRNAs remain stable under harsh conditions, such as boiling, high or low pH, prolonged storage time, and multiple freeze–thaw cycles. Furthermore, circulating miRNAs in serum maintain their expression patterns after incubation for 24 h at room temperature and are still detectable in serum after a maximum of 10 freeze–thaw cycles [97].

However, no reports have determined whether the miRNAs involved in hepatic carcinogenesis that regulate Hippo-YAP/TAZ signaling can be released into the circulation under either normal or pathological conditions. However, previous findings that some of the miRNAs discussed above (e.g., miR-15b [81,98], miR-9-3p [70], miR-223 [72]) are also detected in the serum of liver cancer patients give rise to the intriguing question of whether they could predict the risk or progression of liver cancer. In lung cancer, for example, upregulation of miR-328-3p targeting NF2 to inactivate Hippo kinase activities is promoted by hypoxic bone marrow mesenchymal stem cells, which deliver EVs containing miR-328-3p to the lung tissue [99]. miR-328-3p is detected not only in cancer tissues but also in the serum of lung cancer patients, suggesting the potential of miR-328-3p as both a biomarker and a therapeutic target of lung cancer [99]. Moreover, miRNAs can monitor tissue responses to therapeutic interventions [100]. Bie et al. [101] have shown that the miRNA expression profiles changes in BEL-7402 HCC cells after treatment with baicalein, an anticancer drug. Notably, the putative target genes for the differentially expressed miRNAs after baicalein treatment are enriched in pathways involved in cell proliferation, including the Hippo signaling pathway [101], indicating that the miRNAs interacting with Hippo-YAP/TAZ signaling represent the status of liver cancer.

Small molecules, such as pazopanib, dasatinib, and statins, which are under investigation in clinical trials or used currently for the treatment of liver cancer, have been reported to activate the Hippo signaling pathway, thereby reducing cancer cell viability and sensitizing tumor cells to chemotherapeutics [102,103]. Verteporfin, a YAP inhibitor, disrupts the YAP-TEAD interaction by promoting the degradation of YAP [104]. Thus, an anticancer approach using verteporfin has been suggested for liver cancers with YAP overexpression and chemoresistance [105,106]. Nevertheless, adverse side effects of these small molecules have been reported, and toxicity, short life, and unintended outcomes also limit the use of these inhibitors [107]. miRNAs may become the alternative since miRNA-based therapy is effective and biologically safe [108,109]. The efficacy of mimics of tumor suppressive miRNAs or inhibitors of oncogenic miRNAs in the prevention and treatment of liver cancer has been evaluated in preclinical models of HCC. To improve the efficiency of miRNA delivery and targeting to specific organs, nanoparticles such as liposomes have been used as delivery vehicles [110,111]. The first-in-human phase I study of miRNA-based therapy was recently completed and used a liposomal miR-34 mimic (known as MRX34, Mirna Therapeutics, Inc.) in solid tumors, including HCC [112]. Although MRX34 was shown to regulate its target genes dose-dependently, unexpected immune-related adverse events occurred in a small subset of patients, the reason for which needs to be elucidated [112]. Hence, further studies are strongly encouraged to understand the mechanism of action of miRNA-based cancer therapeutics to develop more therapeutic candidates that can be used in clinical trials.

#### **6. Conclusions**

Genetic ablation of Hippo signaling and overactivation of YAP cause liver cancer in mice [41,113], but the Hippo-YAP/TAZ signaling pathway is dysregulated in human liver cancer primarily by molecular events other than mutations [114]. Recently, it has been demonstrated that alternative RNA splicing of Hippo signaling regulators, including NF2 and CSNK1D, is important for their activities, and certain exon skipping of their mRNAs promotes hepatocyte proliferation and loss of mature hepatocyte functions, suggesting a novel post-transcriptional regulation of the Hippo-YAP/TAZ signaling pathway [115,116]. As well-known post-transcriptional regulators, several miRNAs have been proposed as regulators of the expression of Hippo-YAP/TAZ signaling components, and dysregulation of those miRNAs can lead to hepatic tumorigenesis [117].

In this review, we summarized the miRNAs that contribute to the development and progression of liver cancer by directly binding to the mRNAs of Hippo-YAP/TAZ signaling components or indirectly through interactions with related signaling pathways. Accumulating evidence indicates that these miRNAs could be used as biomarkers for the early detection, prognosis and monitoring of liver cancer and therapeutic targets against liver carcinogenesis. However, most of the studies have been conducted using liver cancer cell lines and xenograft models or on a relatively small number of human liver cancer specimens. Therefore, an increasing number of studies are required to investigate the functions of miRNAs in liver cancers with various etiologies and at different stages during carcinogenesis using relevant models of liver cancer and large-scale cohorts.

In conclusion, miRNAs that interact with the Hippo-YAP/TAZ signaling pathway are promising therapeutic targets for liver cancer. They play pivotal roles in hepatic tumorigenesis by affecting oncogenic transformation, proliferation and migration of tumor cells and modulating the cancer microenvironment.

**Author Contributions:** Conceptualization, J.H.; investigation, N.-H.L., S.J.K. and J.H.; writing original draft preparation, N.-H.L. and S.J.K.; writing—review and editing, J.H.; supervision, J.H. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by National Research Foundation of Korea, grant number 2020R1F1A1073695 (J.H.) and 2021R1C1C1003904 (J.H.).

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Not applicable.

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

#### **References**

