MicroRNA-140-5p inhibits hepatocellular carcinoma by directly targeting the unique isomerase Pin1 to block multiple cancer-driving pathways

Hepatocellular carcinoma (HCC) is the second leading cause of cancer related-death. As a major common regulator of numerous cancer-driving pathways and a unique therapeutic target, the prolyl isomerase Pin1 is overexpressed in a majority of HCCs, whereas the mechanism underlying Pin1 overexpression remains elusive. Here we find that miR-140-5p inhibits HCC by directly targeting Pin1 to block multiple cancer-driving pathways. Bioinformatics analysis, miRNA binding and functional assays identify that miR-140-5p directly interacts with the 3′UTR of Pin1 and inhibits Pin1 translation. Furthermore, like stable Pin1 knockdown, moderate overexpression of miR-140-5p not only eliminates Pin1, but also inhibits cells growth and metastasis. Importantly, these effects of miR-140-5p are largely rescued by reconstitution of Pin1. Moreover, miR-140-5p inhibits multiple Pin1-dependent cancer pathways and suppresses tumor growth in mice. The clinical significance of these findings has been substantiated by the demonstrations that miR-140-5p is frequently down-regulated and inversely correlated with Pin1 overexpression in HCC tissues and cell lines. Given prevalent miR-140-5p downregulation in other cancers and major impact of Pin1 overexpression on activating numerous cancer-driving pathways including global miRNA downregulation, the miR-140-5p/Pin1 axis may play a major role in tumorigenesis and offer promising therapeutic targets for HCC and other cancers.


Results
Pin1 is a direct target for miR-140-5p in HCC. Pin1 overexpression has been observed in ~70% human HCC patients 29 , but little is known about molecular mechanisms leading to Pin1 overexpression in HCC. To identify whether Pin1 is regulated by miRNAs, we used miRNA target prediction programs miRnada 43 , TargetScan 21 and PITA 44 to search for candidate miRNAs for Pin1 (Supplementary Table S1). We identified eleven candidate miRNAs with the potential targeting the 3′ UTR of Pin1, which were predicted by at least two of the three bioinformatics databases. Among these eleven candidates, miR-140-5p and miR-200s (miR-200b/200c/429) were predicted by all of the three databases (Fig. 1a) and found to be frequently downregulated in HCC 1,45 . Interestingly, in addition to a miR-200s target site, as reported 21,22 , Pin1 3′ UTR has a predicted miR-140-5p target site ( Fig. 1b and c).
Notably, the binding site for miR-140-5p in Pin1 3′ UTR is evolutionarily conserved (Fig. 1d), suggesting a potential biological significance. In order to demonstrate that miR-140-5p indeed targets Pin1, we performed a luciferase reporter assay, as described 46 . The four miRNA mimics were cotransfected with the pmirGLO vector bearing wild-type or mutated miR-140-5p binding site of Pin1 3′ UTR. We observed that as shown for the positive control miR-200b and miR-200c 21,22 , miR-140-5p inhibited the luciferase activity of the pmirGLO plasmid with wild-type 3′ UTR of Pin1 by about 50% (Fig. 1d), whereas mutation of miR-140-5p seed region in Pin1 3′ UTR counteracted the regulating effects of miR-140-5p (Fig. 1e). To confirm that miR-140-5p indeed regulates Pin1 expression, we transiently transfected four different miRNA mimics into Huh7 cells for 48 hours, followed by monitoring expression of Pin1 with cyclin D1 as a functional readout for Pin1, because Pin1 controls gene transcription and protein stability of cyclin D1 20,47 . MiR-140-5p mimic significantly decreased both Pin1 and cyclin D1 levels (Fig. 1f), as shown for miR200b and miR200c, which have been shown to regulate Pin1 expression in other cells 21,22 . However, while miR-429 inhibited the luciferase activity and Pin1 protein levels as the same as miR-140-5p, miR-140-5p decreased Pin1 protein levels slightly more obvious than miR-429 ( Fig. 1d and f). To further support this finding, we generated HCC cell lines Huh7 cells and PLC/PRF/5 cells stably expressing miR-140-5p using lentiviral infection. The expression of miR-140-5p, which was confirmed by RT-PCR, led to significant reduction of Pin1 protein expression without significant effects on Pin1 mRNA level (Fig. 1g), suggesting that miR-140-5p mainly regulates Pin1 at the translational level. Similarly, moderate overexpression of  (top panel) and miR-140-5p reduces Pin1 expression (bottom panel), as assayed by a luciferase reporter. The seed sequences of miR-140-5p within the Pin1 3′ UTR are evolutionarily highly conserved across mammals as marked red. Capitalized letters are the conserved binding sites that directly interact with miR-140-5p. Dual-luciferase assay showed that miR-140-5p and miR-200s reduce luciferase activity by about 50%. (e) MiR-140-5p targets wild-type Pin1 3′ UTR, but not its mutant. Luciferase constructs bearing a Pin1 3′ UTR (WT) or Pin1 3′ UTR containing mutated binding sequences of miR-140-5p (Mut) were cotransfected with miR-140-5p. Results showed that miR-140-5p reduces luciferase activity by 50%, but that was abolished when miR-140-5p binding sequences on Pin1 3′ UTR was mutated. (f) MiR-140-5p downregulates Pin1 and cyclin D1 expression, as detected by Western blot analysis. β -actin served as loading control. (g) MiR-140-5p downregulates Pin1 at the translational level. Huh7 cells were infected with lentiviruses expressing miR-140-5p followed by selection with indicated concentration of puromycin. The expression of miR-140-5p and Pin1 protein were significantly increased and decreased, respectively, while Pin1 mRNA level had no significant change. In all panels, bar graphs represent mean ± SEM of three independent experiments. The statistical significance of all tests was accepted for P < 0.05. miR-140-5p also downregulated Pin1 expression in PLC/PRF/5 cells (Supplementary Fig. S1a). Taken together, these results support that Pin1 is a miR-140-5p downstream target.

Moderate overexpression of miR-140-5p potently inhibits cell growth, colony formation, migration and invasion of human HCC cells. Pin1 overexpression in human HCC is an independent factor
for poor prognosis, correlating with larger tumor size and increased portal vein invasion in HCC 48 . Given the downregulation of Pin1 by miR-140-5p in HCC cells, we wondered whether such regulation has any functional impact on HCC cells. Moderate Pin1 overexpression in non-transformed human mammary epithelial HMLE cells induces the epithelial-mesenchymal transition (EMT) 21 , which was linked to invasion and metastasis of cancer 49 . We hypothesize that miR-140-5p-mediated Pin1 downregulation might cause the mesenchymal-epithelial transition (MET) in HCCs. To test this possibility, we stably infected Huh7 HCC cells with lentiviruses expressing miR-140-5p, followed by examining its effect on cell phenotypes. MiR-140-5p overexpression not only resulted in a significant reduction in Pin1 protein levels, as compared with vector controls, but also induced epithelial-like morphological feature in Huh7 cells (Fig. 2a). To confirm that miR-140-5p-expressing cells have undergone the MET, we detected the expression of epithelial and mesenchymal markers using Western blot analysis. Indeed, miR-140-5p overexpression drastically upregulated protein levels of the epithelial marker E-cadherin, but downregulated expression of the mesenchymal markers MMP9 and vimentin ( Fig. 2a and Supplementary Fig. S2a). These results are consistent with the notion that miR-140-5p overexpression causes the MET.
To further support this opinion, we carried out cell proliferation assay, colony formation assay and transwell assay for testing the migration and invasion of Huh7 cells. Strikingly, we found that miR-140-5p overexpression in Huh7 cells led to a significant decrease in the capacity of cell growth (Fig. 2b), migration and invasion (Fig. 2c), and colony formation (Fig. 2d). If these effects are specific due to miR-140-5p-mediated Pin1 downregulation, we expressed a Pin1 mutant construct that is resistant to miR-140-5p due to lack of the Pin1 3′ UTR would rescue the phenotypes. Indeed, miR-140-5p-resistant Pin1 significantly rescued the expression of vimentin and cyclin D1, as a functional readout for Pin1 (Fig. 2a), as well as rescued the ability of miR-140-5p overexpression to inhibit the growth capability ( Fig. 2b and d) and to reduce the migratory and invasive capacity ( Fig. 2c) of Huh7 cells. To further demonstrate the function of miR-140-5p in HCC, we carried out cell proliferation assay and transwell assay in PLC/PRF/5 cells. Notably, we found that miR-140-5p overexpression in PLC/PRF/5 cells significantly reduced the expression of vimentin and cyclin D1 ( Supplementary Fig. S2b), and decreased cell growth ( Supplementary Fig. S2c) and migration ( Supplementary Fig. S2d), which were significantly rescued by expression of miR-140-5p-resistant Pin1, consistent with the above results in Huh7 cells (Fig. 2a,b and c). The lack of a full rescue is likely due to that miR-140-5p has multiple targets such as FGF9 and TGFβ R1 1 , as shown for most other miRNAs 50 . Since the original hypothesis of this study is that Pin1 is a downstream target of miR-140-5p, we would expect that Pin1 knockdown might not further enhance the tumour suppressive functions of miR-140-5p in Huh7 cells as detected by transwell migration assay. Indeed, combination of Pin1 knockdown and miR-140-5p overexpression did not significantly enhance the migratory capacity of miR-140-5p in Huh7 cells ( Supplementary Fig. S2e). Taken together, these data demonstrate that miR-140-5p not only down-regulates Pin1, but also inhibits cell growth, colony formation, migration and invasion of human HCC cells in vitro.
Effect of Pin1 knockdown on HCC cells. If the above functional effects of miR-140-5p on cell growth, colony formation, migration and invasion of human HCC cells is due to miR-140-5p-mediated Pin1 downregulation, we would expect that Pin1 genetic knockdown would produce the similar phenotypes. To address this possibility, we stably silenced Pin1 function by infecting Huh7 cells with lentiviruses expressing a validated shRNA of Pin1 followed by examining its effect on cell phenotypes, as described 32 . Indeed, stable knockdown of Pin1 in Huh7 cells induced the epithelial-like morphological feature ( Supplementary Fig. S3a), drastically upregulated expression of the epithelial marker E-cadherin and downregulated expression of the mesenchymal marker vimentin ( Supplementary Fig. S3a). Furthermore, Pin1 knockdown also significantly inhibited cell proliferation, migratory and invasive behavior of HCC cells, as detected by MTT assay ( Supplementary Fig. S3b), wound healing assay ( Supplementary Fig. S3c) and transwell assay ( Supplementary Fig. S3d). Moreover, Pin1 knockdown also markedly inhibited cell colony formation of HCC cells ( Supplementary Fig. S3e). Notably, some of these Pin1 knockdown phenotypes are similar to those observed independently as described 32 . Thus, these results indicate that both Pin1 shRNA and miR-140-5p lead to similar phenotypes, inhibiting cell growth, colony formation, migration and invasion of HCC cells in vitro.

MiR-140-5p inhibits HCC tumor growth by targeting Pin1 in vivo.
To demonstrate that miR-140-5p could inhibit HCC tumor growth by targeting Pin1 in vivo, we subcutaneously injected stable miR-140-5p-or miR-NC (negative control)-expressing Huh7 cells into the either flank of the same nude mice, followed by monitoring tumor growth for 8 weeks after implantation. The size of subcutaneous tumors originated from miR-140-5p-transduced Huh7 cells was dramatically smaller than that of miR-NC (negative control)-transduced cells, as revealed by tumor growth curves, photographic illustration of final tumors or their weights (Fig. 4a,b and c). Moreover, miR-140-5p significantly decreased the abundance of Pin1 and its functional readout cyclin D1 in tumors as compared to controls (Fig. 4d), consistent with the above in vitro results (Figs 1f and 3c). Taken together, our data show that miR-140-5p has potent anti-tumor activity against HCC through targeting Pin1 to block multiple cancer pathways in vitro and in vivo. MiR-140-5p is downregulated and correlated with Pin1 overexpression in HCC cell lines and human HCC tissues. To demonstrate the clinical significance of miR-140-5p-mediated Pin1 regulation in HCC, we first determined expression levels of miR-140-5p and Pin1 protein by qRT-PCR and Western blot assay, respectively, in normal human liver cell line LO2 and 7 human HCC cell lines HepG2, Huh7, Hep3B, PLC/ PRF/5, MHCC-97H, HCCLM3, SMMC-7721. The normal human liver cell line LO2 had lower Pin1 levels but higher miR-140-5p levels, whereas the HCC cell lines examined had higher Pin1 levels but lower miR-140-5p levels (Fig. 5a,c and d). Statistical analysis revealed Pin1 protein levels were inversely correlated with miR-140-5p levels (R = − 0.6621 and P value = 0.0263) in normal and HCC cells (Fig. 5e). Next, we extended our experiments to the clinical samples obtained from HCC patients. By comparing HCC tumor tissues and paired adjacent non-cancerous liver tissues (ANLTs) of 25 HCC patients, we found that Pin1 was overexpressed in about 64%  (Fig. 5b and f), which was consistent with the previous findings 29 . We also found that miR-140-5p was markedly reduced in HCC tissue as compared with the paired ANLTs of 25 HCC patients (Fig. 5g), as shown by a separate study 1 . More importantly, we found a significantly inverse correlation between expression levels of Pin1 and miR-140-5p (R = − 0.5183 and P value = 0.0080) in human HCC samples (Fig. 5h), as expected from our above results in vitro and in vivo. These results indicate that loss of miR-140-5p is a major factor contributing to Pin1 overexpression in HCC, further supporting the significance of miR-140-5p-mediated regulation of Pin1 in HCC.

Discussion
It has been shown that Pin1 is markedly overexpressed in a wide range of human cancers, including breast cancer 55 , prostate cancer 56 , glioblastoma 57 and HCC 26,29,58 . Pin1 overexpression promotes tumorigenesis by activating multiple cancer-driving pathways at the same time. High Pin1 expression has been shown to be an independent factor for cancer development and poor prognosis and is along with larger tumor size and increased portal vein invasion of HCC 48 . However, while mechanisms regulating Pin1 expression have been studied in other cancers 28 , molecular mechanisms leading to Pin1 overexpression in HCC and its impact on the malignancy of HCC are still unclear. This question is important given that HCC is one of the most lethal cancers with few therapeutic options. The present study uncovers for the first time that miR-140-5p directly targets Pin1 to block multiple cancer-driving pathways and exert potent antitumor activity in vitro and in vivo.
Recently, it has been shown that deregulation of miRNAs contributes to the tumorigenesis and progression of tumor 41,59,60 , and that miRNAs carry out their function by targeting multiple genes 50 . Notably, Takata et al. have shown that miR-140 acts as a liver tumor suppressor by controlling NF-κ B activity by directly targeting Dnmt1, and demonstrated that miR-140 is related to hepatocarcinogenesis 61 . Hao Yang et al. found that miR-140-5p is strikingly down-regulated in HCC and that ectopic miR-140-5p expression inhibits the capacity of HCC tumor growth and metastasis by targeting TGFβ R1 and FGF9 1 .
In the present study, our bioinformatics analysis and miRNA binding and functional assays showed that miR-140-5p directly interacted with a conserved target sequence in the 3′ UTR of Pin1 mRNA and inhibited Pin1 translation in human HCC cells. Furthermore, like Pin1 knockdown, miR-140-5p exerted potent anticancer activity against HCC by targeting Pin1 to block multiple Pin1-dependent cancer pathways simultaneously, thereby suppressing the migratory, invasive and proliferous capacity of HCC cells. Importantly, these effects of miR-140-5p were significantly, but not fully rescued by reconstitution of Pin1 resistant to miR-140-5p. These results are consistent with the previous findings that miR-140-5p has other targets such as TGFβ R1 and FGF9 1 . Interestingly, some downstream targets of TGFβ R1 and FGF9 such as TGF-β and the ERK/MAPK signaling pathways 62,63 are regulated by Pin1 28 . For example, in our study, miR-140-5p suppressed the expression of a few endogenous ERK/ MAPK pathway-related proteins (such as ERK and pERK) (Fig. 4b) by ablating Pin1 and thereby inhibited the activity of ERK/MAPK signaling pathway. Therefore, Pin1 is likely important for miR-140-5p to affect HCC growth and metastasis by targeting FGF9 and TGFβ R1. Importantly, we found that miR-140-5p also inhibits many other Pin1 substrate oncoproteins. Moreover, miR-140-5p decreased tumor growth of HCC by ablating Pin1 and its downstream target cyclin D1 in vivo. Finally, expression of miR-140-5p was markedly reduced and correlated with Pin1 overexpression in human HCC tissues and cell lines compared to their respective normal controls. These results together indicate that loss of miR-140-5p contributes to tumor growth and metastasis of HCC, at least in part, through the up-regulation of Pin1 and its multiple cancer-driving pathways simultaneously.
Interestingly, in addition to HCC 1 , miR-140-5p has been shown to be gradually lost along with tumor progression and related to metastatic disease in ovarian cancer 64 , colorectal cancer 65 , lung cancer 66 , tongue cancer 67 . Notably, Pin1 has been shown to overexpress in most human cancer and activate at least 43 oncoproteins and inactivate at least 20 tumor suppressors 28 . Our results suggest that the down-regulation of miR-140-5p might also promote tumor growth and metastasis of many other cancers by directly increasing Pin1 expression to activate multiple cancer-driving pathways. Notably, Pin1 overexpression has been shown to reduce pre-miRNA export from the nucleus, resulting in global downregulation of miRNAs in liver cancer 33 . Therefore, the deregulation of the miR-140-5p/Pin1 interaction may represent an aggressive molecular lesion in cancer development by resulting in global reduction of miRNA and activation of numerous cancer-driving pathways, offering novel therapeutic targets not only for HCC, but also for other cancers. (c-e) Relative Pin1 protein levels (normalized to GAPDH) and miR-140-5p expression levels (normalized to U6) were performed by qRT-PCR in eight liver cell lines. The expression of miR-140-5p was inversely correlated with Pin1 protein expression in HCC cell lines. (f-h) Pin1 protein and miR-140-5p levels were detected in 25 pairs of surgical specimens from HCC patients. We observed that Pin1 protein levels were significantly upregulated in HCC tissues as compared to that in ANLTs (P < 0.01), on the contrary, miR-140-5p levels were significantly downregulated in HCC tissues as compared to that in ANLTs (P < 0.01). The expression of miR-140-5p was inversely related to Pin1 expression in HCC tissues.
Clinical specimens. All studies involving human subjects were approved by the Institutional Review Board at Mengchao Hepatobiliary Hospital of Fujian Medical University, and performed in accordance with the relevant guidelines and regulations. Informed consent was obtained from all patients at Mengchao Hepatobiliary Hospital of Fujian Medical University before we carried out our studies. Matched fresh HCC tissues and adjacent non-cancerous liver tissues (ANLTs) of 25 HCC patients were collected at the time of hepatic resection and immediately frozen in liquid nitrogen and then stored at − 80 °C for RNA and Protein isolation 1 . All HCC tissues were further affirmed by pathological examination. The clinicopathological characteristics of these fresh samples were provided in the Supplementary Table S2.
The lentiviral plasmid for expression of miR-140-5p and NC (Negative control) were LV-3 (pGLVH1/ GFP + Puro) purchased from Genepharma. For overexpression of Pin1, the CDS were subcloned into the pBybe lentiviral vector, and for stable knockdown of Pin1, the shRNA construct was subcloned into the pLKO.1 lentiviral vector, as described previously 21 . The production of lentiviruses as well as the infection of target cells was on the basis of previous description 68  Luciferase Reporter Assays. For the reporter assays, HEK293T cells (5 × 10 4 ) were seeded in 24-well plate with 30% confluence, twelve hours later, the cells were cotransfected with 50 ng pmirGLO-Pin1-3′ -UTR luciferase plasmid and 50 nM miRNA mimics using TurboFect Transfection Reagent (Thermo Scientific). After 48 hours of transfection, the HEK293T cells were harvested for detection using the Dual-Luciferase Reporter Assay system (Promega) according to the specifications of the manufacturer. The Renilla luciferase activities were used to normalize the transfection efficiency.  Assays (Invitrogen) in a reaction containing 10 ng of total cDNA following the manufacturer's protocol and the specific primers for miR-140-5p was designed by Applied Biosystems. The relative expression of miRNA-140-5p in tumor tissues against ANLTs and among different cell lines was obtained by comparing the cycle threshold (Ct) values which were normalized by the internal control RNU6B (U6 snRNA). Each sample was detected in triplicates.

Cell Proliferation, Cell Cycle, Cell Colony Formation assays. Cell proliferation was determined by
counting the number of cells or using 3-(4, 5-dimethylthiazol-2-yl)-2, 5-diphenyltetrazolium bromide (MTT) assays. For cell colony formation assays, Huh7 cells (4 × 10 3 ) were seeded in each well of 6-well plates and cultured for 2 weeks at 5% CO 2 37 °C. The numbers of colonies per well were counted after staining with crystal violet. All studies were conducted with 3 replicates.
In Vitro Wound Healing, Migration and Invasion Assays. Cells (4 × 10 5 ) were seeded into 12-well plate coated with Collagen. Until the cells reached 95% confluence, wound healing assays were carried out with a sterile 10 μ L pipette tip to straightly scratch 2 lines through the confluent monolayer. The cells were cultured with fresh medium for another 48 h, and then photographed for calculating wound closure. The percent of wound closure was calculated using Image J software.
For migration assay, cells (5 × 10 4 ) in serum-free medium were seeded into the upper chamber without matrigel (Merck Millipore), and 48 hours later, the cells in the upper chamber were removed. The number of cells that adhering to the lower membrane was counted after staining with 0.1% crystal violet. For the invasion assay, cells (5 × 10 4 ) in serum-free medium were seeded into the upper chamber of the insert with matrigel (BD Biosciences), and 48 hours later, the cells in the upper chamber were removed. The number of cells that adhering to the lower membrane was counted after staining with 0.1% crystal violet. All studies were conducted with 3 replicates.
HCC Mouse Model. The hepatocellular carcinoma nude mice model was constructed as previously described 69 . Briefly, cells (3 × 10 6 ) were subcutaneously injected into the bilateral upper flank regions of the nude mice (3-4 weeks of age, female, BALB/c). Two month later, the mice were sacrificed when their tumor volume reached 1000 mm 3 , and the subcutaneous tumors were removed for calculating the size as follows: tumor volume (mm3) = (L × W 2 )/2, where L = long axis and W = short axis 1 . BALB/c nude mice were housed in laminar flow cabinets with free access to food and water in Laboratory Animal Center of Fujian Medical University. All of animal experiments were performed in accordance with the animal protocols and regulations approved by FJMU Experimental Animal Ethics Committee of Fujian Medical University. Statistical Analysis. Statistical analysis was performed using SPSS (v. 22.0, Chicago, IL). All data are presented as the mean ± SEM, and then two-tailed Student's t test or analysis of variance test (ANOVA) was used to compare unpaired samples. Paired Student's t test was used to compare paired groups of samples. The statistical significance of all tests was accepted for P < 0.05, P < 0.01 and P < 0.001.