Prognostic significance of miR-122 expression after curative resection in patients with hepatocellular carcinoma

Downregulation of MicroRNA-122 (miR-122) and its association with cancer progression have been reported in hepatocellular carcinoma (HCC) cell line models and a limited number of HCC samples. Recently, restoration of miR-122 expression by direct delivery of miR-122 yielded promising results in HCCs. However, the prognostic effect of miR-122 expression in human HCC samples is not fully understood. We investigated the expression level of miR-122 by quantitative real-time polymerase chain reaction in 289 curatively resected HCC samples and 20 normal liver samples and evaluated the prognostic effect of miR-122 expression. The relative quantification value of miR-122 was much lower in HCC samples than in normal liver tissues. During a median 119 months of follow-up for survival, the low miR-122 expression group showed shorter recurrence-free survival (RFS) (p = 0.033) and intrahepatic recurrence-free survival (IHRFS) (p = 0.014), and a trend of short distant metastasis-free survival (DMFS) (p = 0.149) than high expression group. On multivariate analysis, miR-122 expression was an independent prognostic factor for RFS, IHRFS and DMFS. Downregulation of miR-122 expression, frequently found in HCC samples, was an independent prognostic factor for RFS after curative resection. Emerging therapeutic approaches targeting miR-122 could be applicable in patients with miR-122 downregulated hepatocellular carcinoma.

miR-122 expression in HCC tissues and its association with clinicopathologic parameters. The assay showed a linear dynamic range of log concentration, and an efficiency of 97% ( Supplementary Fig. S1). The intra-and inter-assay coefficient of variabilities ranged from 0.3% to 1.8% and 0.8% to 1.7%, respectively.
The relative quantification (RQ) value of miR-122 was significantly lower in HCC samples than in normal liver tissues (mean ± standard deviation: 0.12 ± 0.37 vs. 1.07 ± 1.07, p < 0.001 by Mann-Whitney U test) (Fig. 1). About 90% of the HCC samples showed a value lower than the minimum value (0.24) of normal liver tissues, whereas only five HCC cases showed a value higher than the mean value (1.07) of normal liver tissues.
The RQ value of miR-122 showed a tendency to be lower in large sized (>5 cm) HCCs compared with small (≤5 cm) HCCs (mean 0.073 ± 0.100 vs. 0.150 ± 0.445) and in BCLC (Barcelona Clinic Liver Cancer) stage B,C compared with BCLC stage 0,A (mean 0.076 ± 0.098 vs. 0.159 ± 0.477), but these differences did not reach statistical significance by Mann Whitney test. The RQ value of miR-122 in HCC with other prognostic factors such as advanced AJCC T stage, intrahepatic metastasis, microvascular invasion, or Edmondson grade III or IV was slightly lower than in other groups, but failed to demonstrate statistical significance (data not shown).
The median of RQ value in 289 HCC samples was 0.036 (range, 0.000-4.205). The estimated cutoff value of miR-122 with the highest level of statistical significance related to Recurrence free survival (RFS) was 0.17 by the minimum p-value approach. This cutoff value was internally validated using 1,000 bootstrap samples generated from the study data with replacement. The estimated cutoff value of 0.17 was located between 0.007 and 0.400, which were the lower and upper 5th percentiles of the estimates, respectively. Out of 289 cases, 44 (15.2%) were classified into the high expression group and 245 (84.8%) into the low expression group. The associations between miR-122 expression and clinicopathologic parameters are summarized in Table 2. miR-122 expression was not significantly associated with any clinicopathologic factor including well-known prognostic factors of HCC. Clinical information and RQ value of miR-122 in each sample are provided in Supplementary Dataset 1.

Recurrence pattern and predictive factors for early and late intrahepatic recurrence.
During the median 120 months (range 2-193) of follow-up, recurrence was detected in 197 patients (68.2%). Among them, 82 patients (28.4%) had both intrahepatic recurrence (IHR) and distant metastasis (DM), 102 patients (35.3%) had IHR only, and 13 patients (4.5%) had DM only. Among all patients with IHR (with or without DM), early IHR developed in 125 (43.4%) patients, and late IHR was detected in 59 of the 145 patients (40.7%) that were followed for over 2 years.
The results of univariate and multivariate analyses of clinicopathologic parameters and early or late IHR are summarized in Tables 3 and 4 Impact of miR-122 expression on the recurrence-free survival of HCC patients. Patients with low miR-122 expression showed shorter IHRFS and RFS rates than those with high miR-122 expression (p = 0.014 and p = 0.033 by log-rank test, respectively) and a trend toward a shorter DMFS rate (p = 0.149 by log-rank test) (Fig. 2). Results of univariate analysis of probable prognostic factors for intrahepatic RFS, DMFS, and RFS are summarized in Table 5. In multivariate analyses, significant prognostic factors were intrahepatic metastasis, multicentric occurrence, ALBI grade, and miR-122 expression for intrahepatic RFS; Edmondson grade, intrahepatic metastasis, and miR-122 expression for DMFS; and intrahepatic metastasis, ALBI grade, and MiR-122 expression for RFS (Table 6). In addition to intrahepatic metastasis, miR-122 expression was an independent prognostic factor for IHRFS (

Discussion
In this study, we demonstrated that miR-122 expression was downregulated in 289 HCC samples compared with normal liver tissues and was an independent predictive factor for late IHR as well as an independent prognostic factor for RFS regardless of recurrence pattern (intrahepatic recurrence or distant metastasis) in a large cohort of HCC patients with long-term follow-up.
www.nature.com/scientificreports www.nature.com/scientificreports/ There have been numerous reports on the role of miR-122 as a tumor suppressor in hepatocarcinogenesis, performed at various confidence levels. Downregulation of miR-122 in HCCs has been reported in several studies [13][14][15][16][17][18][19][20] , leading to further in vitro and in vivo studies. Tsai et al. showed that miR-122 was significantly downregulated in T3 HCCs with intrahepatic metastasis compared with T1 HCCs or adjacent normal liver tissue, and that restoration of miR-122 reduced in vitro migration, invasion, and anchorage-independent growth as well as tumorigenesis  In their studies, mice with germline knockout or liver-specific knockout of miR-122 developed spontaneous HCCs in addition to steatohepatitis and fibrosis, and restoration of miR-122 reduced tumor development 10,22 . However, the clinical effect of miR-122 expression in HCC patients, especially on patient survival, has not been fully elucidated. Coulouarn et al. reported that patients with low miR-122 expression showed shorter overall survival times than those with high miR-122 expression (30.3 ± 8.0 months vs. 83.7 ± 10.3 months, p < 0.001) in 64 HCC patients 26 . They also showed that miR-122 expression was low in HCCs with poor differentiation, high proliferation, low apoptotic index, and large tumor size 26 . In a recent meta-analysis by Zhang et al. incorporating of 4 studies containing 328 patients, low miR-122 expression in tissue or fine needle aspiration sample was significantly associated with unfavorable overall survival and progression free survival. In validation using TCGA dataset, low miR-122 expression was significantly associated with OS and marginally associated with PFS 30 . Also, significance of serum miR-122 expression in HCC have been reported. Qiao et al. showed that serum miR-122 expression was significantly lower in HBV-related HCC patients than in benign liver disease group or control group 31 . A few recent reports suggested serum exosomal miR-122 as a predictive biomarker in transarterial chemoembolization-treated HCC patients or diagnostic biomarkers for HCC 32,33 .
In this study, we showed that miR-122 expression was downregulated in 289 HCC samples compared with normal liver tissues. About 90% of the HCC samples showed a value lower than the minimum expression value of normal liver tissues. Also, we demonstrated that miR-122 expression was an independent prognostic factor for RFS in a large cohort of 289 HCC patients with long-term follow-up. These results are consistent with results from previous studies and provide clinical evidence suggesting miR-122 as a potential therapeutic tool targeting HCC. Interestingly, survival curves are clearly distinguished after 2 years from surgery, as depicted in survival curve of Fig. 2, and low expression of miR-122 was associated with frequent late intrahepatic recurrence, which could represent a de novo HCC, not metastasis from primary tumor. Based on these results, we can suggest that patients with HCC having low expression of miR-122 may require intensive follow-up even after 2 years from surgery. These results are somewhat different from previous studies showing association between low expression of miR-122 and aggressive tumor behavior which represents early intrahepatic recurrence 22,26,34 . This study includes HCCs with curative resection, which had no known metastasis at the time of surgery. We observed that miR-122 expression is slightly lower in HCCs with factors associated with tumor aggressiveness, such as larger size (>5 cm), intrahepatic metastasis, higher Edmondson grade, or microvascular invasion, but it did not reach the statistical significance. The effect could be less obvious because of the relatively uniform study population. Meanwhile, miR-122 down-regulation is known to be involved in the progression of liver fibrosis and emergence of de novo HCC 10 . It can be inferred that de-novo HCCs by down-regulation of miR-122 could have low expression of miR-122. There have been only 4 studies regarding prognostic effect of miR-122 expression in HCC tissue samples, and the largest patient number was 144 [23][24][25][26] . Further study is necessary for validating the prognostic effect of miR-122 expression.
Based on previous studies supporting a tumor suppressive effect of miR-122 in HCCs, restoration of miR-122 expression represents an interesting strategy for the treatment of HCC. Direct delivery of miRNA mimic oligonucleotides into tumors by viral vectors such as the adeno-associated virus 8 (AAV8) serotype or LNP-DP1, a cationic lipid-based nanoparticle, has been tested with promising results 22,[27][28][29] . It is noteworthy that the first miRNA mimic, MRX34, a liposomal miRNA 34 mimic, is now undergoing testing in a multicenter phase I study in a variety of cancers including HCCs (ClinicalTrials.gov.identifier: NCT01829971A) 35 . In addition, miR-122 has therapeutic applications in the aspect of sensitization of chemotherapy. It has been reported that restoration of miR-122 sensitizes HCCs to chemotherapeutic agents such as sorafenib, doxorubicin, vincristine, and cisplatin [36][37][38][39] . A combination of miR-122 restoration and chemotherapy could be another therapeutic strategy against HCC.  www.nature.com/scientificreports www.nature.com/scientificreports/ This study has some limitations. First, this study was retrospectively performed in a single institution and included patients who had undergone curative resection for primary tumor, thus possibly introducing unavoidable selection bias. Consequently, our results should be interpreted cautiously, and further study with a large population and prospective design is needed. Second, the estimated cutoff value of miR-122 expression was not validated in an independent cohort. We redeemed this by internal validation using 1,000 bootstrap samples generated from the study data with replacement, but further external validation would be required.   Not included in multivariable analysis due to multicollinearity with tumor size. www.nature.com/scientificreports www.nature.com/scientificreports/ Despite these limitations, our study provides valuable information confirming the prognostic significance of miR-122 expression in the largest cohort of HCC patients with long-term follow-up. Our data showed that miR-122 is an independent predictive factor for late IHR and an independent prognostic factor for RFS regardless of recurrence pattern, either intrahepatic recurrence or distant metastasis.
In conclusion, miR-122 expression is frequently downregulated in HCCs, and miR-122 expression is an independent prognostic factor for recurrence-free survival after curative resection. Emerging therapeutic approaches targeting miR-122 could be applicable in patients with miR-122-downregulated hepatocellular carcinoma.

Material and Methods
Patients and samples. Initially, a total of 291 patients who underwent curative resection for primary HCC at Samsung Medical Center, Seoul, Korea, from July 2000 to May 2006 were enrolled. This cohort is the same as the previous study of our group 40 , and this study has been performed independently of previous study. Samples from 2 patients had poor RNA quality and experiments were failed in these samples. Finally, analyses were performed in remaining 289 patients. Curative resection was defined as complete resection of all tumor nodules without involvement of microscopic resection margins and no detected residual tumor on computed tomography scans at 1 month after surgery. Twenty normal liver samples, which were confirmed biochemically and histologically and mostly obtained from patients with liver resection due to metastatic colorectal cancer, were selected as normal controls. The Institutional Review Board of Samsung Medical Center approved this study (2016-11-098), and waived informed consent. All research was performed in accordance with relevant guidelines/regulations. By reviewing the medical records, clinical parameters such as age, gender, date of surgery, serum α-fetoprotein (AFP), and serum albumin were collected. HCC was diagnosed by basically histological feature by HE slide which shows hepatocytic differentiation, occasionally with aid of immunohistochemical staining of Hepatocyte antigen, AFP, or cytokeratin 19 when differentiation is poor. Histopathologic features of HCCs, including tumor  www.nature.com/scientificreports www.nature.com/scientificreports/ differentiation, microvascular invasion, major portal vein invasion, intrahepatic metastasis, multicentric occurrence, and non-tumor liver pathology, were reviewed by two liver pathologists (S.Y.H. and C.-K.P.). Tumor differentiation was determined according to the criteria of Edmondson and Steiner 41 . Intrahepatic metastasis and multicentric occurrence were distinguished according to the criteria of the Liver Cancer Study Group of Japan 42 . Patients were staged according to the AJCC staging system 43 and BCLC staging classification 44 .
During the follow-up period, patients underwent computed tomography (CT) with measurement of serum AFP every 2 or 3 months postoperatively. Patients with suspicious imaging findings and/or continuously elevated AFP levels were further evaluated with PET-CT and/or MRI. The median follow-up period was 119.8 months (range 14-151.4 months) for survival. Recurrence was generally diagnosed using radiologic examinations with no histologic confirmation. The site of first recurrence was classified as intrahepatic recurrence (IHR) or distant metastases (DM). IHR is defined as recurrence occurring anywhere inside the entire liver. Intrahepatic HCC recurrence within the first two years after curative resection is mainly due to intrahepatic metastasis, whereas late recurrence usually results from multicentric disease 45 . Using 2 years as a cutoff, intrahepatic tumor recurrence was classified as either early or late 46 . All other sites of recurrence were defined as DM. The duration of IHR-free survival (IHRFS), DM-free survival (DMFS), and recurrence-free survival (RFS) was calculated from the date of surgical resection to the date of each event or the last day of follow-up.  www.nature.com/scientificreports www.nature.com/scientificreports/ PRISM 7500HT Fast Real-time PCR system (Applied Biosystems). The threshold cycle (Ct), which is the fractional cycle number at which the amount of amplified target reaches a fixed threshold, was determined. Relative changes in gene expression were measured using the 2−ΔΔCt (ΔΔCt = ΔCttarget gene−ΔCtGAPDH) method. Each reaction was carried out in triplicate technical replicates. Ct values were calculated for each replicate and averaged.

RNA extraction and quantitative Reverse-Transcriptase Polymerase Chain Reaction (qRT-PCR
To generate standard curves for genotype identification assays, 10-fold serial dilution containing a HCC190 sample was analyzed in three independent runs by RT-qPCR. In order to take into account the diagnostic context of the sample matrix, each dilution was analyzed in three independent runs by RT-qPCR. The log dilution series of was tested in triplicates in each run. Standard curves for each assay were generated by plotting threshold cycle (Ct) values per three replicates per standard dilution versus the log concentration. Amplification efficiency, intra-assay and inter-assay variation were determined using 10-fold dilutions of one sample (No. 178). An efficiency of 1 corresponds to 100% amplification efficiency. The coefficient of determination (R 2 ) was assessed and considered to be suitable when it was not lower than 0.98 in a single run.
Statistical analysis. The cut-off value of miR-122 expression for RFS was determined using the log-rank test and the minimum p-value approach. The estimated cutoff value was internally validated using 1,000 bootstrap samples generated from the study data with replacement. P-values < 0.05 were considered statistically significant in two-tailed tests.
The correlation between miR-122 expression and other clinicopathologic variables was evaluated using the chi-square test or Fisher's exact test. Univariate analysis was performed using the Cox proportional hazard model or the time-dependent Cox model according to satisfaction of the proportional hazard assumption for each variable. The proportional hazard assumption was confirmed using the correlation between partial residuals from the estimated Cox proportional hazard model and time to the event. Variables with P < 0.2 in univariate analysis were included in the multivariate analysis. Multivariate analysis was conducted using a time-dependent Cox model because some variables included in the multivariate analysis failed to adhere to the proportional hazard assumption. Multicollinearity was identified by a Variance Inflation Factor (VIF) >10.0. Some of those variables were excluded from the multivariate analysis. Statistical analyses were performed using SAS version 9.4 (SAS Institute, Cary, NC) and SPSS software (SPSS Inc., Chicago, IL, USA). Variable risk was expressed as a hazard ratio (HR) with corresponding 95% confidence interval (CI).