Diagnostic and prognostic potential of serum microRNA-4651 for patients with hepatocellular carcinoma related to aflatoxin B1

Background The serum microRNAs have been reported as potential biomarkers for hepatitis virus-related hepatocellular carcinoma (HCC); however, their role in aflatoxin B1 (AFB1)-related HCC to has not yet been evaluated. Materials and Methods We conducted a case-control study, including 366 HCC cases and 662 controls without any evidence of tumors, to identify and assess diagnostic and prognostic potential of serum microRNAs for AFB1-related HCC. The sensitivity, specificity, and area under the receiver operating characteristic curve (AUC) were used to elucidate diagnostic performance, and to compare the microRNAs with α-fetoprotein (AFP) at a cutoff of 20 ng/mL (AFP20) and 400 ng/mL (AFP400). Results We found 8 differentially expressed microRNAs via the microRNA array analysis; however, only microRNA-4651 was further identified to detect AFB1-positive HCC but not AFB1-negative HCC. For AFB1-positive HCC, microRNA-4651 showed higher accuracy and sensitivity than AFP400 (AUC, 0.85 vs. 0.72; Sensitivity, 78.1% vs. 43.0%). Compared to AFP20, microRNA-4651 exhibited higher potential in identifying small-size (0.68 vs. 0.84 for AUC and 36.7% vs. 75.5% for sensitivity, respectively) and early-stage HCC (0.69 vs. 0.84 for AUC and 38.7% vs. 75.7% for sensitivity, respectively). Additionally, miR-4651 was also associated with HCC prognosis (hazard risk value, 2.67 for overall survival and 3.62 for tumor recurrence analysis). Conclusions These data suggest that serum microRNA-4651 may be a useful marker for HCC diagnosis and prognosis, especially AFB1-positive cases.


INTRODUCTION
Hepatocellular carcinoma (HCC), accounting for more than 90% of primary liver cancer, is the sixth most common malignant tumor, and the third-leading cause of cancer-related death in the world. [1] From a global perspective, the infection of hepatitis virus B (HBV) or C (HCV) and the exposure to aflatoxin B1 (AFB1) are the two main risk factors for HCC, especially in the highly epidemic areas of many Asian and African countries including China [1]. With early detection, HCC patients often have a good chance of a successful curative

Clinical Research Paper
Oncotarget 81236 www.impactjournals.com/oncotarget operation, and the 5-year overall survival (OS) rate is more than 50%. However, with late diagnosis, 5-year OS rate is reduced to less than 10%, and the dismal prognosis is largely caused by late detection of the tumors after disease progression [1][2][3]. Therefore, early detection of HCC at a curative stage offers the best chance of survival for these patients.
Currently, the early diagnosis of HCC is based on imaging examination and serological tests. Although advances in computed tomography (CT) and magnetic resonance imaging (MRI) technology have greatly improved the early screening of HCC, these procedures are costly and not suitable for daily clinical practice [3,4]. For serological tests, α-fetoprotein (AFP) is the most widely used biomarker for HCC worldwide. However, the accuracy of AFP is modest (with sensitivity of 40 -65%; and specificity of 87 -96%), and about 30% of cases of early-stage HCC are missed using AFP analysis. Additionally, serum AFP levels of benign liver diseases, including hepatitis and cirrhosis, may give false-positive results [3]. Thus, performance of this biomarker to detect disease can be inadequate.
Increasing evidence has shown a link between aberrant expression of microRNAs and HCC [5][6][7]. In particular, microRNAs are highly stable in circulation and expression patterns seem to be tissue-specific, suggesting circulating microRNAs may be potentially ideal biomarkers for early-stage HCC [7][8][9]. Recent studies have demonstrated that serum microRNAs can identify preclinical HCC related to hepatitis viruses such as HBV and HCV [10][11][12][13][14][15][16][17][18][19][20]. However, information on whether serum microRNAs are correlated with AFB1-related HCC is limited. Therefore, monitoring circulating microRNA signatures during AFB1 exposure and correlated diseases may be clinically relevant as a non-invasive diagnostic tool for early detection of HCC. In this study, we assessed the significance of the serum microRNAs in detecting HCC from individuals with positive AFB1 exposure. Our findings revealed that microRNA-4651 (miR-4651) could serve as a potential biomarker for early diagnosis of AFB1-related HCC.

The characteristics of HCC cases and controls
According to eligibility criteria listed in Table 1, we collected 1028 serum samples from the five groups of participants (Table 2). For each group, the age, sex, race, and the time and location sample collection were wellmatched. About 40% of HCC cases were in the early stage of the BCLC system.

Diagnostic performance of serum miR-4651
According to the results presented above, we chose miR-4651 to further investigate whether serum levels could differentiate cases with AFB1-positive HCC from controls by the receiver operating characteristic (ROC) curves. Results showed that miR-4651 had the comparable area under ROC curves (AUC, 0.88; 95% confidence interval [CI], 0.86 -0.91; P < 0.001), sensitivity (77.4%) and specificity (92.7%) to differentiate AFB1-positive HCC cases from controls ( Figure 2). At the best cut-off value (9.306), miR-4651 had an AUC 0.85 (0.82-0.88) to discriminate cases with AFB1-positive HCC from controls with a sensitivity of 78.1% and specificity of 92.1% (Table  4 and Figure 3). However, miR-4651 in this setting did not detect AFB1-negative HCC cases from controls (P = 0.15, Figure 4). The diagnostic potential of serum miR-4651 and AFP for AFB1-positive HCC was further compared. The AUC for miR-4651 was not significantly different from AFP20, but was significantly greater than AFP400 (difference between areas = 0.13, P < 0.001) ( Table 4 and Figure 3).
Next, we analyzed the diagnostic performance of miR-4651 for both AFB1-positive and AFP-negative HCC ( Figure 6). In this study, 38.7% (108/279) of individuals with AFB1-positive HCC were AFP-negative. The miR-4651 demonstrated high accuracy in discriminating individuals with AFB1-positive HCC but AFP-negative from controls (AUC, 0.80; 95% CI, 0.75-0.86) ( Figure  6A). This microRNA had about 70% sensitivity and 90% specificity for detection of AFP-negative HCC. In accordance with the above-mentioned results, miR-4651 also exhibited value in identifying the AFP-negative HCC with a small size ( Figure 6B) or at an early stage ( Figure  6C).

Serum miR-4651 significantly correlated with HCC prognosis
To study the effects of the serum miR-4651 on outcome of HCC patients, we analyzed the survival including miR-7-2-3p A., miR-4651 B., miR-127-3p C., miR-192-5p D., miR-382-5p E., miR-10b-5p F., miR-532-3p G., and miR-16-5p H., were further analyzed using TaqMan-PCR method in the training stage. The relative levels of microRNA expression were calculated according to 2 -ΔCT method (see Materials and Methods). The microRNA data are shown as box plots, with horizontal lines representing the median, the bottom and the top of the boxes representing the 25th and 75th percentiles, respectively. We compared expression data between groups using the Mann-Whitney U test.
Oncotarget 81238 www.impactjournals.com/oncotarget Abbreviations: AFP20, the cutoff of 20 ng/mL AFP; AFP400, the cutoff of 400 ng/mL AFP; C miR+AFP20 , the combination of miR-4651 and AFP20; C miR+AFP400 , the combination of miR-4651 and AFP400. Oncotarget 81239 www.impactjournals.com/oncotarget follow-up information of all HCC patients. Among these subjects, 111 received the same curative resection as the initial treatment but not did other treatments before resection, according to Chinese Manage Criteria of HCC [21], and were included for final survival analysis. We first analyzed the distribution difference of the miR-4651 status among different clinic-pathological characteristics of cases (Table 7). A significant difference between different serum levels of miR-4651 among AFP (odds ratio [OR] = 11.72, 95% CI = 2.59-53.01), but not in age, gender, race, tumor sized, tumor grade, or cirrhosis was observed. Results of the Kaplan-Meier survival analysis showed that increasing serum miR-4651 level significantly correlated with shorter overall survival (OS) and recurrence-free survival (RFS) of HCC cases ( Figure 7A and 7B). From Cox regression analysis ( Figure 7A and 7B) we showed that the miR-4651 is correlated with poor prognosis of HCC (miR-4651-positive risk value, hazard ratio [HR] = 2.67 and P = 1.44 × 10 -4 for OS and 3.62 and 4.52 × 10 -3 for RFS, respectively). Taken together, these results indicated that the serum miR-4651 is independent of other clinical covariates and suggested its potential as an independent prognostic factor for HCC.

Serum miR-4651 positively associated with serum AAA levels
To investigate whether AFB1 exposure was associated with miR-4651 expression, we explored the correlation between serum AAA levels and serum miR-4651 levels in subjects with positive AFB1. The increasing    Oncotarget 81241 www.impactjournals.com/oncotarget serum levels of miR-4651 were found among HCC cases with higher serum AAA levels ( Figure 8A). Spearman's correlation analysis further exhibited the serum miR-4651 levels were positively related with the serum levels of AAA (r = 0.79, P < 0.01). However, this positive association between miR-4651 levels and AAA levels was  AFP20, the cutoff of 20 ng/mL AFP; AFP400, the cutoff of 400 ng/mL AFP; CmiR+AFP20, the combination of miR-4651 and AFP20; CmiR+AFP400, the combination of miR-4651 and AFP400. www.impactjournals.com/oncotarget not observed among controls with positive AFB1 ( Figure  8B and 8C). Taken together, our data suggested that miR-4651 expression might be correlated with AFB1 exposure among cases with positive AFB1. * The p value indicates the statistical significance for the differences of AUC/sensitivity/specificity as compared with miR-4651. Abbreviations: ROC, the receiver operating characteristic curve; AFP, α-fetoprotein; AUC, the area under the ROC curve; AFP20, the cutoff of 20 ng/mL AFP; AFP400, the cutoff of 400 ng/mL AFP.  . Abbreviations: AFP20, the cutoff of 20 ng/mL AFP; AFP400, the cutoff of 400 ng/mL AFP; C miR+AFP20 , the combination of miR-4651 and AFP20; C miR+AFP400 , the combination of miR-4651 and AFP400.

DISCUSSION
Increasing studies have proved that microRNAs have a crucial role in human carcinogenesis, including hepatocarcinogenesis, via acting as oncogenes or tumor suppressor genes [22][23][24][25]. Recent evidence has shown that serum microRNAs are remarkably stable and expression patterns may be tissue-specific; thus, they may be important potential candidates for non-invasive cancer testing [26]. It has been hypothesized that serum microRNAs might have diagnostic potential for HCC, including AFB1-related HCC. Therefore, in the present we conducted a case-control study to screen and analyze potential serum microRNAs for the early diagnosis of AFB1-related HCC in a high AFB1 exposure area, Guangxi area of China [27]. Our results proved that serum microRNAs have different expression between HCCs and non-HCCs, and exhibit useful diagnostic value. Supporting our findings, recent several reports have shown that serum microRNAs could detect HCC cases from controls [2, 14-16, 18, 24, 28-33]. For example, Wang et al. [18] investigated the differential profiles of circulating microRNAs expression in HBV-related small liver cancer and found six differentially expressed serum microRNAs in HCC related to HBV. Yu et al. [16] and Li et al. [32] identified miR-150 and miR-18a as potential markers for hepatitis B virus-related HCC screening. Motawi et al. [14] and Lin et al. [31] analyzed the combination value of several serum microRNAs in the early-stage diagnosis of HCC, and found microRNA classifier should be more valuable to detect HCC. Collectively, these results suggest that serum microRNAs might be important biomarkers for HCC diagnosis.
In this study, miR-4651 was particularly concerned because of its different expression between HCC patients with and without positive-AFB1 status. This microRNA, a kind of small RNA encoded by miR-4651 gene (located at chr7: 75915197-75915269), was found in breast cancer    Oncotarget 81245 www.impactjournals.com/oncotarget [34]. Until now, it has been not clear whether miR-4651 acts as a tumor suppressor or an oncogene. However, we observed that miR-4651 had higher expression in the AFB1-positive HCC cases than in non-HCC-harboring individuals, and this increasing expression was further positively associated with AAA levels among AFB1positive HCCs and poor prognosis of HCC related to AFB1. This implies that miR-4651 might act as an oncogene or have a similar role of oncogenes through decreasing detoxication and genomic DNA damage repair capacity because the levels of AAA correlate with HCC risk and prognosis, and can reflect the deficiency of detoxication and genomic DNA damage repair capacity [35]. Interestingly, miR-4651 had higher accuracy and sensitivity to identify small liver cancer and early-stage AFB1-positive HCC compared with AFP; highlighting its diagnostic value in AFP-negative tumor cases. Taken together, these results implied that miR-4651 might be a useful diagnostic biomarker for HCC induced by AFB1 exposure.
This study has several strengths. First, we completed a high-throughput screening analysis for serum microRNAs that exhibited differential levels between HCC cases with positive-AFB1 status and healthy controls featuring risk factor AFB1 exposure. Through this methodology, we not only improved the chance to identify serum biomarkers, but obtained 8 possible AFB1-related microRNAs. Second, to control effects of other carcinogenetic factors such as HBV and HCV, only HBV-and HCV-negative cases were included in this study; whereas HBV-or HCV-positive individuals were excluded. Third, the present study included both healthy controls without risk factor and controls with different risk stages, where the serum levels of microRNAs could change at the different states. Finally, a control group whose patients with HCC were not caused by AFB1 exposure was included in the present study to evaluate whether miR-4651 had the similar diagnostic potential for non-AFB1-related HCC. Our data showed that miR-4651 can distinguish individuals with AFB1-positive HCC from controls with different risk stages. However, this microRNA did not exhibit similar diagnostic value for AFB1-negative cases. This implies the practicality of miR-4651 in clinical practice.
However, there were several limitations to our study. First, potential selection bias might have occurred through the selection of hospital-based control subjects. Second, although the serum levels of miR-4651 among AFB1positive HCC cases were significantly associated with serum AAA levels, we did not investigate the association between miR-4651 and other known causes of HCC such as HBV and HCV. Finally, we did not examine additional functional analysis. Thus, more functional analyses should be performed based on large samples and a combination of biomarkers and different causes.
In summary, this study represents the first report describing the serum levels of microRNAs and their early diagnostic and prognostic roles in AFB1-related HCC. The findings provide an additional insight into serum miR-4651 as a biological determinant for identifying small-size, early-stage, and AFP-negative HCC related to AFB1. Therefore, these findings, in combination with AFP, could improve the identification of HCC, especially in high AFB1-exposure areas.

Study design and participants
We conducted a hospital-based case-control study of AFB1-related HCC in the Guangxi area. To validate biomarkers, the present study consists of three control groups and two case groups (Table 1): healthy controls without positive AFB1 (Group 1), healthy controls with positive AFB1 (Group 2), controls with positive AFB1 and liver diseases exception tumors (Group 3), patients with positive AFB1 and diagnosed HCC (Group 4), and cases with diagnosed HCC but without positive AFB1 (Group 5). All newly diagnosed HCC patients in hospitals affiliated with Youjiang Medical College for Nationalities and Guangxi Medical University from January 2006 to December 2010 were utilized. All controls were recruited from the general health checkup center at the same hospitals during the same period for comparison. To control the effects of confounders, controls were individually matched to controls based on gender, ethnicity (Han, Zhuang), age ( ± 5 years), time when sampled, and hospital locations. All controls were surveyed to ascertain their willingness to participate in the study and to provide preliminary demographic data. In this study, a total of 366 cases and 662 controls, representing 92% of eligible cases and 95% of eligible controls were interviewed and included the final analysis.
In this report, all participants were defined as Group 1 (n = 338), 2 (n = 292), 3 (n = 32), 4 (n = 279), or 5 (n = 87) by medical doctors, according to eligibility criteria listed in Table 1. All HCC cases (including AFB1negative and -positive HCC cases) were confirmed histopathologically according to the American Association for the Study of Liver Diseases (AASLD) guidelines. Tumour stage was defined according to the Barcelona Clinic Liver Cancer (BCLC) staging system, and tumours with BCLC stage 0 and stage A were defined as early-stage HCC. Liver cirrhosis was evaluated by biopsy or hepatic ultrasound technology with CT or MRI. All controls underwent serial monitoring of AFP, ultrasonography, and hepatitis virus biomarkers, and there was not any evidence of HBV, HCV, or tumors. The controls were www.impactjournals.com/oncotarget screened at baseline; whereas cases were screened at primary diagnostic time. Written informed consent was obtained for all participants and 5 mL of peripheral while blood was collected along with demographic data. This study was approved by the Institutional Ethics Committee of Youjiang Medical University for Nationalities, and was carried out in accordance with the approved guidelines (No. AYJM20090112).

Serum preparation
For serum preparation, 5 mL peripheral whole blood was collected into drying tubes, and then was centrifuged at 4 °C, 3000 r.p.m. for 10 min, which was followed by an additional centrifugation at 12,000 r.p.m. for 15 min to completely remove all remaining cells. The serum samples were aliquoted and stored at -80 °C until analysis.

Laboratory tests
Fasting venous blood samples were collected from all patients for routine workup, including complete blood picture, liver function tests, prothrombin concentration and prothrombin international normalized ratio, AFP, and HBV and HCV markers using commercially available assays. In this report, AFP20 and AFP400 were defined as the cutoffs of 20 ng/mL and 400 ng/mL AFP, respectively.

AFB1 status analysis
Because the serum AAA is a stable AFB1 exposure biomarker [36], the levels of AAA were used to evaluate AFB1 exposure levels of all subjects. AAA levels in the serum were tested using the comparative enzyme-linked immunosorbent assay as previously published [37]. According to our previous reports with respect to AFB1 exposure [37], value of > 2.18 ln fmol/mg was considered as positive-AFB1 status.

Serum microRNAs expression profiling analysis
In this report, we screened and optimized the serum microRNAs via two methods: RT 2 Profiler PCR Array and TaqMan-PCR analysis. For the PCR Array analysis, we randomly selected six HCC cases with positive AFB1 from Group 4 and six age-, sex-, race-, and sample collected time and location-matched controls with positive AFB1 exposure but without any evidence of liver tumors from Group 2. The sera were collected from these individuals and sent to Shanghai Oe-Bio-Tech Medical Company (Shanghai, China) for microRNA detection. Briefly, total RNA from 1 mL serum was extracted with the PAXgene ® Blood RNA Kit (cat#762174, Qiagen, Dusseldorf, Germany). RNA concentration and purity were evaluated on a NanoDrop spectrophotometer and RNA integrity (RIN) was evaluated on a BioAnalyzer 2100; the results showed good RNA quality (RIN value 7.2). The relative cDNA was synthesized by RT 2 First Strand Kit (cat#330401, Qiagen). After that, the amounts of human microRNAs in the serum samples were tested via the real-time PCR (on an Applied Biosystems 7900HT Real-Time PCR System) using RT 2 Profiler PCR Arrays (cat# PAHS-028ZF-2, Qiagen) in combination with RT 2 SYBR Green Mastermixes (cat# 330500, Qiagen). The cycle threshold (CT) values were processed and subsequently analyzed using the ΔΔ CT method by the PCR Array Data Analysis Web portal (www.SABiosciences. com/pcrarraydataanalysis.php). Briefly, the relative expression level of candidate microRNAs to endogenous control cel-miR-39-3p was calculated as 2 -ΔCT , where ΔCT = CT microRNA -CT cel-miR-39-3p . To determine fold change in gene expression, the normalized expression of the candidate microRNAs in the case group is divided by the normalized expression of the same microRNAs in the control group, also called 2 -ΔΔCT , where ΔΔCT = CT (case group) -CT (control group) . In this phase, a total of 8 candidate microRNAs were chosen for TaqMan-PCR analysis in all participants according to the following criteria: >2-fold change between AFB1-positive HCC cases and controls with positive AFB1, and coefficient of variation (CV) CT < 0.05, and high expression (CT average < 29 cycles) in AFB1-positive HCCs.

TaqMan-PCR assay for candidate microRNAs in the training and validation study
The serum levels of 8 candidate microRNAs were tested using quantitative reverse transcription-PCR with TaqMan probe described in our previous reports [5,6]. Briefly, total RNA was extracted from 400 µL serum with 0.2 nM of cel-miR-67 using PureLink ® RNA Mini Kit (cat#12183018A, Ambion, USA), and corresponding first-strand cDNAs were synthesized using High Capacity cDNA Reverse Transcription Kit (cat# 4368814, Invitrogen Grand Island, NY) and TaqMan MicroRNA Reverse Transcription Kit (cat#4366596, Applied Biosystems, Carlsbad, CA). After that, TaqMan-PCR analysis was performed using standard protocols on a Bio-Rad iCycler CFX Detection System. The serum levels of candidate microRNAs were assessed using TaqMan microRNA assays (cat#4427975, Applied Biosystems) with cel-miR-67 as the endogenous control. PCR reactions were run in a 5-µL final volume containing 1 TaqMAN Universal Master Mix II (cat#4440041, Applied Biosystems), 1 × TaqMan microRNA probes and primers (cat#4427975, Applied Biosystems), and about 15 ng of cDNA. Cycling conditions were 95 °C for 30 s, and 50 cycles of 95 C for 15 s and 60 C for 1 min. All reactions were conducted in triplicate and controls were