Circulating miRNAs (microRNAs) are emerging as promising biomarkers for several pathological conditions, and the aim of this study was to investigate the feasibility of using serum miRNAs as biomarkers for liver pathologies. Real-time qPCR (quantitative PCR)-based TaqMan MicroRNA arrays were first employed to profile miRNAs in serum pools from patients with HCC (hepatocellular carcinoma) or LC (liver cirrhosis) and from healthy controls. Five miRNAs (i.e. miR-885-5p, miR-574-3p, miR-224, miR-215 and miR-146a) that were up-regulated in the HCC and LC serum pools were selected and further quantified using real-time qPCR in patients with HCC, LC, CHB (chronic hepatitis B) or GC (gastric cancer) and in normal controls. The present study revealed that more than 110 miRNA species in the serum samples and wide distribution ranges of serum miRNAs were observed. The levels of miR-885-5p were significantly higher in sera from patients with HCC, LC and CHB than in healthy controls or GC patients. miR-885-5p yielded an AUC [the area under the ROC (receiver operating characteristic) curve] of 0.904 [95% CI (confidence interval), 0.837–0.951, P<0.0001) with 90.53% sensitivity and 79.17% specificity in discriminating liver pathologies from healthy controls, using a cut off value of 1.06 (normalized). No correlations between increased miR-885-5p and liver function parameters [AFP (α-fetoprotein), ALT (alanine aminotransferase), AST (aspartate aminotransferase) and GGT (γ-glutamyl transpeptidase)] were observed in patients with liver pathologies. In summary, miR-885-5p is significantly elevated in the sera of patients with liver pathologies, and our data suggest that serum miRNAs could serve as novel complementary biomarkers for the detection and assessment of liver pathologies.
INTRODUCTION
Liver diseases including CHB (chronic hepatitis B), LC (liver cirrhosis) and HCC (hepatocellular carcinoma) are public health concerns and major medical challenges in the People's Republic of China [1,2]. Clinically, a panel of serological biochemical markers including aminotransferase [ALT (alanine aminotransferase)/AST (aspartate aminotransferase)] and AFP (α-fetoprotein) have been used to monitor liver pathologies for several decades, but they have limited sensitivity and specificity particularly with regard to the insidious progress of HCC [3,4].
miRNAs (microRNAs) are small non-coding RNAs of 22–25 nucleotides that play important roles in regulating gene expression by binding to and regulating the activity of their target mRNAs post-transcriptionally, contributing to diverse cellular processes such as proliferation, differentiation, apoptosis and carcinogenesis [5]. Expression of miRNAs is consistently deregulated in some physiopathological conditions including cancer [6–9]. miRNAs in plasma/serum are emerging as a new class of potential markers for minimally invasive diagnosis and monitoring of patients with several diseases [10–12]. Serum miR-21 has been reported to be elevated in lymphoma patients [13], plasma miR-92a can be used to identify patients with colorectal cancer [14] and miR-1 and miR-208 could serve as indicators of acute myocardial infarction [15,16].
The regulatory roles of miRNAs in the development of liver-related pathologies including HCC, CHB and chronic hepatitis C have been elucidated [17–20], and circulating miRNAs including miR-122 and miR-192 have been reported to have diagnostic value for toxin-induced hepatocellular damage in mice [21]. However, the potential clinical value of circulating miRNAs for diagnosing and managing liver pathologies has not been evaluated. The present study investigates the hypothesis that differentially expressed miRNAs circulate in patients with liver pathologies. We first profiled serum miRNA expression in patients with HCC and LC. Next, five differently expressed miRNAs were selected and further evaluated in independent patient cohorts. Our study indicates that the various levels of specific miRNAs, especially miR-885-5p, could serve as markers for the preclinical detection and clinical assessment of patients with such diseases.
MATERIALS AND METHODS
Study design
The present study was performed in three phases: (i) global serum miRNA profiling using a TaqMan Human MicroRNA Array, in combination with Megaplex RT and Megaplex preamplification techniques; (ii) analysis and preliminary evaluation of candidate miRNAs; and (iii) independent validation of selected miRNAs. Each phase was conducted using independent groups of participants.
Participants
To screen miRNAs more efficiently as potential markers for liver-related diseases during the circulating miRNA profiling phase, we first generated seven pools of serum aliquots derived from individuals in the case and control groups. In brief, serum samples from 15 pathologically confirmed HCC patients with moderate histological grade and various degrees of liver cirrhosis (light to moderate) were randomly divided into three groups to generate three serum pools, and serum samples from ten patients clinically diagnosed as moderate to decompensated LC were similarly used to generate two pools. Each serum pool consisted of equal volumes of serum from all patients within the group. In addition, serum samples from ten age-matched healthy individuals were obtained to generate two NC (normal control) pools, so a total of seven serum pools were generated for miRNA profiling.
To quantify individual miRNAs, independent sets of serum samples were obtained between November 2008 and January 2010 from patients with liver pathologies [CHB, LC, HCC, ICC (intrahepatic cholangiocellular carcinoma) and FNH (focal nodular hyperplasia)] who were admitted to or hospitalized in the Department of Hepatobiliary Surgery or the Department of Gastroenterology in the Chinese PLA General Hospital (Beijing).
Collectively, none of the enrolled HCC, ICC and FNH patients had received any adjuvant therapy or other treatment before blood sampling, none of the CHB and LC patients had received treatment within 3 weeks before blood sampling and most of the LC patients were diagnosed as HBV (hepatitis B virus)-related liver cirrhosis. Additionally, age-matched healthy individuals were recruited as NCs, and serum samples from patients with newly diagnosed GC (gastric cancer) were used as disease controls. The baseline characteristics of the recruited subjects are displayed in Supplementary Table S1 (available at http://www.clinsci.org/cs/120/cs1200183add.htm) and see Table 2.
Written informed consent was obtained from each individual or their responsible guardians. This study protocol was approved by the ethics committee of the Chinese PLA General Hospital (Beijing).
Serum sample processing and RNA isolation
Peripheral blood was collected in tubes containing separating gel and clot activator and centrifuged at 3400 g for 7 min at room temperature, and the supernatants were transferred to Eppendorf tubes. The samples were centrifuged at 15000 g for 8 min to precipitate cell debris, and the supernatants were stored at −80 °C pending RNA extraction. All blood samples were processed within 24 h after they were obtained.
Total serum RNA was isolated and eluted in 100 μl of RNase-free water using a mirVana PARIS kit (#1556; Ambion) following the manufacturer's protocol for liquid samples. For miRNA profiling, 800 μl of pooled serum was used, and for individual miRNA tests, 400 μl of serum from each participant was used. All serum RNA preparations were quantified using a DU 800 spectrophotometer (Beckman Coulter) and then pretreated with RNase-free DNase I (Promega) to eliminate potential DNA contamination.
Serum miRNA profiling and data analysis
In our study, a three-step procedure was performed to profile the miRNAs in the pooled serum samples. First, for cDNA synthesis from the miRNAs, 10 ng of total RNA from the pooled serum was subjected to RT (reverse transcription) using a TaqMan® microRNA Reverse Transcription Kit (#4366596; Applied Biosystems) and Megaplex RT primers (Human Pool A, #4399966; Applied Biosystems) following the manufacturer's protocol, allowing simultaneous reverse transcription of 380 mature human miRNAs to generate an miRNA cDNA library corresponding to each serum pool. RT was performed on a Mastercycler Epgradient thermocycler (Eppendorf) with the following cycling conditions: 40 cycles of 16 °C for 2 min, 42 °C for 1 min and 50 °C for 1 s, followed by a final step of 80 °C for 5 min to inactivate the reverse transcriptase.
Thereafter, to generate enough miRNA cDNA template for the following real-time PCR, the cDNA libraries were pre-amplified using Megaplex PreAmp primer (Humam Pool A, #4399233; Applied Biosystems) and PreAmp Master Mix (#4384266; Applied Biosystems) following the manufacturer's instructions. The PreAmp primer pool used here consisted of forward primers specific for each of the 380 human miRNAs and a universal reverse primer. The preamplification cycling conditions were as follows: 95 °C for 10 min, 55 °C for 2 min, 72 °C for 2 min followed by 12 cycles of 95 °C for 30 s and 60 °C for 4 min; the samples were then held at 99.9 °C for 10 min.
After the preamplification step, the products were diluted with RNase-free water, combined with TaqMan gene expression Master Mix and then loaded into TaqMan Human MicroRNA Array A (#4398965; Applied Biosystems), which is a 384-well formatted plate and real-time PCR-based microfluidic card with embedded TaqMan primers and probes in each well for the 380 different mature human miRNAs; the Mamm U6 (mammalian U6) transcript was used as a normalization signal. qPCR (quantitative PCR) was performed according to the manufacturer's instructions. Real-time PCR was performed on an ABI PRISM 7900HT sequence detection system (Applied Biosystems) with the following cycling conditions: 50 °C for 2 min, 94.5 °C for 10 min followed by 40 cycles of 95 °C for 30 s and 59.7 °C for 1 min. The Ct (cycle threshold) was automatically given by SDS 2.3 software (Applied Biosystems) and is defined as the fractional cycle number at which the fluorescence passes the fixed threshold of 0.2. Mamm U6 embedded in the TaqMan Human MicroRNA Arrays was used as an endogenous control.
The relative expression levels of miRNAs were calculated using the comparative ΔΔCt method as described previously [22,23]. The fold changes in miRNAs were calculated by the equation 2−ΔΔCt. Cluster 3.0 software was used to perform unsupervised hierarchical clustering using Pearson's correlation metrics and average linkage methods. Java Treeview 1.1.3 was used to visualize the clustering results.
Measurement of liver function parameters
The concentrations/activities of serum aminotransferases (ALT and AST) and AFP in each serum sample were measured using a Hitachi 7600 Modular Analytic system.
TaqMan miRNA assay for individual miRNAs
Individual miRNA tests were performed on independent sets of serum samples and a two-step procedure was used. First, 5 ng of total RNA isolated from 400 μl of each serum sample was subjected to RT using an miRNA-specific primer from the TaqMan MicroRNA Reverse Transcription Kit (Applied Biosystems) as described previously [10]. Briefly, RT was conducted in a scaled-down RT reaction volume of 7.5 μl, which contained 2.08 μl of water, 0.75 μl of 10× RT buffer, 0.095 μl of RNase inhibitor, 0.075 μl of dNTPs with dTTP, 0.5 μl of multiscribe reverse transcriptase, 1.5 μl of miRNA-specific stem-loop RT primer (Applied Biosystems) and 2.5 μl (5 ng) of digested RNA preparations. RT was carried out on a Mastercycler Epgradient at 16 °C for 30 min, 42 °C for 30 min and 85 °C for 5 min.
Thereafter, real-time qPCR was performed using a TaqMan MicroRNA assay (Applied Biosystems) to quantify individual miRNAs as described previously [10]. In brief, 4.5 μl of 5:28 diluted RT product was combined with 5.0 μl of TaqMan gene expression Master Mix and 0.5 μl of Taqman miRNA assay mix. qPCR was carried out on an ABI PRISM 7300 sequence detection system at 95 °C for 5 min, followed by 40 cycles of 95 °C for 15 s and 60 °C for 1 min. All qPCR reactions were performed in triplicate and the Ct values greater than 35 from the real-time PCR assays were treated as 35. The PCR products were analysed by electrophoresis on 3.0% agarose gels to validate the specific generation of the expected product. The expression levels (2−ΔΔCt) of miRNAs were calculated as described previously [22].
Statistical analysis
The Mann–Whitney test or Kruskal–Wallis test was performed to determine the significance of serum miRNA levels. MedCalc 9.6.4 software was used to construct ROC (receiver operating characteristic) curves and calculate the AUC (area under the ROC curve). Linear regression analysis was used to examine correlations between the levels of the miRNAs and liver function parameters. P values <0.05 were considered statistically significant. All statistical calculations were performed using SPSS software (version 11.0).
RESULTS
Serum miRNA profiling and data analysis
The first aim of the present study was to investigate whether aberrantly expressed miRNAs are present in patients with liver pathologies. To screen differentially expressed miRNAs in serum samples efficiently, we generated seven serum pools of serum aliquots derived from individuals in the NC, LC and HCC groups (two NC, two LC and three HCC pools). The circulating miRNAs in the seven serum pools were profiled separately using RT-preamp-qPCR (RT-preamplification-qPCR) as described in the Materials and methods section.
The data indicated that a total of 115 miRNAs could be detected (assays giving Ct values <33 in at least four pools were classed as detectable) (results not shown). By comparing the miRNA profiles between HCC and NC and LC and NC, two miRNA expression patterns were obtained (Table 1). This process generated a list of more than 20 up-regulated candidate miRNAs. As shown in Table 1, the average levels of 26 miRNAs in the HCC serum pools and 22 miRNAs in the LC serum pools were higher than in NC.
Fold change was calculated as described previously [14] and is presented as 2−ΔΔCt. miRNAs with mean fold change >2 in HCC serum pools (n=3) and >1.5 in LC serum pools (n=2) when compared with NC (n=2) are listed respectively. miRNAs underlined are common miRNAs with up-regulated levels in both HCC and LC when compared with NC. The five miRNAs highlighted in bold represent HCC/LC-related candidate miRNAs selected for subsequent validation.
miRNAs in serum of HCC compared with NC . | miRNAs in serum of LC compared with NC . | ||
---|---|---|---|
miRNA . | Average fold change . | miRNA . | Average fold change . |
miR-99a | 191.3 | miR-99a | 328.4 |
miR-224 | 87.1 | miR-100 | 59.5 |
miR-100 | 82.5 | miR-125b | 29.7 |
miR-122 | 16.3 | miR-128 | 14.7 |
miR-885-5p | 13.5 | miR-215 | 14.4 |
miR-125b | 8.5 | miR-224 | 14.1 |
miR-95 | 7.9 | miR-192 | 8.9 |
miR-215 | 7.8 | miR-194 | 8.2 |
miR-134 | 7.3 | miR-133a | 7.7 |
miR-194 | 6.3 | miR-210 | 7.0 |
miR-192 | 5.7 | miR-375 | 6.6 |
miR-193b | 5.7 | miR-423-5p | 5.7 |
let-7b | 5.2 | miR-185 | 5.3 |
miR-133a | 5.1 | miR-141 | 5.2 |
miR-483–5p | 5.0 | miR-122 | 5.2 |
miR-886–3p | 4.7 | miR-19b | 4.3 |
miR-574-3p | 4.6 | miR-885-5p | 2.7 |
let-7d | 3.7 | miR-16 | 2.7 |
miR-146a | 3.4 | miR-574-3p | 1.9 |
miR-18a | 3.1 | miR-30b | 1.8 |
miR-141 | 2.9 | miR-146a | 1.6 |
miR-486-5p | 2.8 | miR-26a | 1.5 |
miR-222 | 2.7 | ||
miR-181a | 2.5 | ||
miR-92a | 2.2 | ||
miR-16 | 2.0 |
miRNAs in serum of HCC compared with NC . | miRNAs in serum of LC compared with NC . | ||
---|---|---|---|
miRNA . | Average fold change . | miRNA . | Average fold change . |
miR-99a | 191.3 | miR-99a | 328.4 |
miR-224 | 87.1 | miR-100 | 59.5 |
miR-100 | 82.5 | miR-125b | 29.7 |
miR-122 | 16.3 | miR-128 | 14.7 |
miR-885-5p | 13.5 | miR-215 | 14.4 |
miR-125b | 8.5 | miR-224 | 14.1 |
miR-95 | 7.9 | miR-192 | 8.9 |
miR-215 | 7.8 | miR-194 | 8.2 |
miR-134 | 7.3 | miR-133a | 7.7 |
miR-194 | 6.3 | miR-210 | 7.0 |
miR-192 | 5.7 | miR-375 | 6.6 |
miR-193b | 5.7 | miR-423-5p | 5.7 |
let-7b | 5.2 | miR-185 | 5.3 |
miR-133a | 5.1 | miR-141 | 5.2 |
miR-483–5p | 5.0 | miR-122 | 5.2 |
miR-886–3p | 4.7 | miR-19b | 4.3 |
miR-574-3p | 4.6 | miR-885-5p | 2.7 |
let-7d | 3.7 | miR-16 | 2.7 |
miR-146a | 3.4 | miR-574-3p | 1.9 |
miR-18a | 3.1 | miR-30b | 1.8 |
miR-141 | 2.9 | miR-146a | 1.6 |
miR-486-5p | 2.8 | miR-26a | 1.5 |
miR-222 | 2.7 | ||
miR-181a | 2.5 | ||
miR-92a | 2.2 | ||
miR-16 | 2.0 |
To investigate the relative abundances of the miRNAs detected, they were normalized in each serum pool (NC, LC and HCC) to Mamm U6 in the corresponding pool. As shown in Figure 1, miRNAs such as miR-223, miR-16 and miR-146a were present in relatively high abundance in the serum samples, while miR-99a, miR-224, miR-192, miR-128 and miR-100 were present at significantly lower abundance compared with the corresponding Mamm U6 (28.1 in NC, 29.4 in LC and 30.0 in HCC pool, respectively). In addition, the abundances of different miRNA species in a specified serum pool varied widely. For example, the relative abundances (normalized to Mamm U6) of miR-223 and miR-100 in the NC serum pool were 295.09 and 0.00026, respectively.
Expression of 115 miRNAs in representative serum pools from the NC, LC and HCC groups
miRNAs were quantified using RT-preamp-qPCR as described in the Materials and methods section. The relative abundance of each miRNA is presented as 2−ΔCt, where ΔCt was calculated by subtracting the Ct values of Mamm U6 (28.1 in NC, 29.4 in LC and 30.0 in HCC pool, respectively) from the Ct of each miRNA detected within the corresponding groups. The levels of various miRNAs are shown (normalized) in representative NC, LC and HCC serum pools.
miRNAs were quantified using RT-preamp-qPCR as described in the Materials and methods section. The relative abundance of each miRNA is presented as 2−ΔCt, where ΔCt was calculated by subtracting the Ct values of Mamm U6 (28.1 in NC, 29.4 in LC and 30.0 in HCC pool, respectively) from the Ct of each miRNA detected within the corresponding groups. The levels of various miRNAs are shown (normalized) in representative NC, LC and HCC serum pools.
Furthermore, unsupervised hierarchical clustering analysis demonstrated that different serum pools had different miRNA spectra (Figure 2A), and a small panel of miRNAs including miR-146a, miR-224, miR-574-3p and miR-885-5p were clustered together because of their similarly up-regulated expression in HCC serum pools compared with NC serum pools (Figure 2B).
Serum miRNAs profiled using real-time qPCR-based arrays
(A) miRNA expression profiles in two NC, two LC and three HCC serum pools. Assays with Ct values <33 in at least four pools were classed as detectable, and a total of 115 miRNAs were detected. The expression levels of miRNAs were normalized to Mamm U6 and data are presented as 2−ΔCt. miRNA expression was mean-centred, and a heat map was generated using Pearson's correlation metrics and average linkage methods of Cluster 3.0 software. Samples are shown in columns and miRNAs in rows. (B) miR-146a, miR-128, miR-224, miR-574-3p, miR-134 and miR-885-5p were clustered together. The colour of each cell represents the expression level of the corresponding miRNA in the corresponding sample. The higher intensities of yellow relate to higher expression levels, and the increasing intensities of blue reflect lower expression of miRNAs.
(A) miRNA expression profiles in two NC, two LC and three HCC serum pools. Assays with Ct values <33 in at least four pools were classed as detectable, and a total of 115 miRNAs were detected. The expression levels of miRNAs were normalized to Mamm U6 and data are presented as 2−ΔCt. miRNA expression was mean-centred, and a heat map was generated using Pearson's correlation metrics and average linkage methods of Cluster 3.0 software. Samples are shown in columns and miRNAs in rows. (B) miR-146a, miR-128, miR-224, miR-574-3p, miR-134 and miR-885-5p were clustered together. The colour of each cell represents the expression level of the corresponding miRNA in the corresponding sample. The higher intensities of yellow relate to higher expression levels, and the increasing intensities of blue reflect lower expression of miRNAs.
Candidate miRNAs selection and preliminary evaluation
In this phase, preliminary tests of candidate miRNAs were carried out on a second set of serum samples including NC (n=16) and patients with LC (n=12) or HCC (n=20) (Supplementary Table S1).
Initially, the expression of U6 snRNA (small nuclear RNA) and miR-16 (both proposed as the most commonly used internal controls in the literature [14,24–26]) in sera from NC, LC and HCC was determined using real-time qPCR in order to select suitable internal controls. Both U6 snRNA and miR-16 were specifically and consistently detected in all sera (Supplementary Figures 1A and 1B available at http://www.clinsci.org/cs/120/cs1200183add.htm), with an average raw Ct of 32.8 for U6 snRNA and 24.7 for miR-16. No significant differences in raw Ct values of U6 snRNA were detected among the three groups (P=0.128, Kruskal–Wallis test), but there was variability among the raw Ct values within the respective groups. miR-16 was significantly higher in HCC sera than in NC and LC (P=0.014, Kruskal–Wallis test). Therefore U6 snRNA was selected as the internal normalization control for the subsequent miRNA qPCR.
Additionally, miRNAs such as miR-99a, miR-100 and miR-125b were excluded from further analysis because their expression levels were significantly lower than Mamm U6 in global miRNA profiling (Figure 1). As a pilot study, the potentially increased expressions of a subset of miRNAs (i.e. miR-146a, miR-215, miR-224, miR-574-3p and miR-885-5p) were assessed using a TaqMan microRNA assay (Figure 2 and Table 1). Both miR-885-5p and miR-146a were consistently and specifically detected in all 48 serum samples (i.e. all with average Ct values <35) (Supplementary Figure 1C). However, regarding mean Ct <35 as a positive signal, the detection rates for miR-574-3p and miR-224 were 71% and 23%, respectively.
Using U6 snRNA as the normalization control, we demonstrated that miR-885-5p was significantly higher in HCC and LC serum samples than in NC (P<0.0001), with a 6.5-fold increase in HCC and an 8.8-fold in LC (Figure 3A). In addition, increased amounts of serum miR-146a and miR-224 were observed in the HCC and LC groups (Figures 3B and 3C). However, no significant differences in the levels of miR-574-3p were observed among the NC, LC and HCC groups (Figure 3D), and the low level of miR-215 made it difficult to quantify its abundance in the sera accurately using real-time qPCR (most mean Ct values >35).
Comparisons of levels of serum miRNAs in NC, HCC and LC groups
Serum abundances of miR-885-5p (A), miR-146a (B), miR-224 (C) and miR-574-3p (D) in an independent set of serum samples from NC (n=16), HCC (n=20) and LC (n=12) were quantified using real-time qPCR. Each qPCR was carried out in triplicate in 96-well plates. Expression levels of selected miRNAs were normalized to U6 snRNA and are presented as fold changes (2−ΔΔCt) above NC. Data are shown as the mean fold changes compared with the mean value in NC (±S.E.M.). A Mann–Whitney or Kruskal–Wallis test was used to determine statistical significance. ***P<0.0001, **P<0.001, *P<0.01.
Serum abundances of miR-885-5p (A), miR-146a (B), miR-224 (C) and miR-574-3p (D) in an independent set of serum samples from NC (n=16), HCC (n=20) and LC (n=12) were quantified using real-time qPCR. Each qPCR was carried out in triplicate in 96-well plates. Expression levels of selected miRNAs were normalized to U6 snRNA and are presented as fold changes (2−ΔΔCt) above NC. Data are shown as the mean fold changes compared with the mean value in NC (±S.E.M.). A Mann–Whitney or Kruskal–Wallis test was used to determine statistical significance. ***P<0.0001, **P<0.001, *P<0.01.
Validation of miR-885-5p as a potential liver pathology-associated marker
Considering that the level of miR-885-5p was consistently increased in serum samples from the HCC and LC groups (Figure 3A), the abundance of this miRNA was further measured in another independent cohort of serum samples consisting of LC (n=26), HCC (n=46), CHB (n=23), ICC (n=9), FNH (n=6) and NC (n=24). In addition, 17 patients with GC were included as disease controls to investigate the specificity of miR-885-5p as a liver pathology-associated molecule. Information regarding the disease control individuals in terms of demographics, biochemical testing, virological investigation and histological grade is summarized in Table 2.
Ages are given as means (S.D.), and AFP, ALT, AST and TB are given as medians (range). No-HCC, patients with liver diseases including nine with ICC and six with FNH. In the LC group, 18 patients had decompensated liver function. HBsAg, hepatitis B surface antigen; ND, not detected; TB, total bilirubin.
Group . | HCC . | Non-HCC . | LC . | CHB . | GC . | NC . |
---|---|---|---|---|---|---|
Individuals (n) | 46 | 15 | 26 | 23 | 17 | 24 |
Gender (n) | ||||||
Male | 39 | 12 | 23 | 17 | 15 | 17 |
Female | 7 | 3 | 3 | 6 | 2 | 7 |
Age (years) | 54.2 (9.2) | 45.7 (8.6) | 61.2 (10.2) | 41.3 (8.9) | 58 (7.5) | 50.6 (5.6) |
AFP (ng/ml) | 38.8 (1.62–20000) | 2.97 (1.1–422.3) | 4.22 (0.61–365.9) | 3.6 (1.22–23.7) | 1.81 (1.02–6.87) | 2.6 (1.23–6.38) |
Abnormal (AFP >20 ng/ml) (n) | 24 | 3 | 3 | 2 | 0 | 0 |
Normal (AFP<20 ng/ml) (n) | 22 | 12 | 19 | 11 | 12 | 24 |
ND (n) | 0 | 0 | 4 | 10 | 5 | 0 |
ALT (units/l) | 34.1 (17.8–421.9) | 29.8 (9.1–281.8) | 28.9 (14.5–101.9) | 27.4 (5.8–137.2) | 13.8 (7.6–50.1) | 17.2 (13.1–32.6) |
AST (units/l) | 31.1 (9.1–281.8) | 26.3 (9.4–540.5) | 33.7 (14.7–85.1) | 25.1 (10.2–296) | 15.8 (10.6–24.1) | 15.1 (8.2–29.6) |
GGT (units/l) | 61.2 (11.5–908.0) | 62.3 (15.7–1310) | 59.3 (24.8–199.3) | 26.1 (12.3–118.5) | 18.6 (7.6–135.2) | ND |
TB (μmol/l) | 14.2 (6.0–227.3) | 11.4 (5.4–377.2) | 19.6 (7.5–47.7) | 10.3 (2.8–32.8) | 9.6 (2.3–56.9) | ND |
HbsAg (n) | ||||||
Positive | 33 | 2 | 20 | 23 | 0 | 0 |
Negative | 13 | 12 | 5 | 0 | 16 | 24 |
Not known | 0 | 1 | 1 | 0 | 1 | 0 |
Cirrhosis (n) | ||||||
Normal | 0 | 13 | 0 | Unknown | – | – |
Low | 17 | 2 | 0 | Unknown | – | – |
Moderate | 29 | 0 | 12 | Unknown | – | – |
High | 0 | 0 | 14 | Unknown | – | – |
Histological grade (n) | ||||||
Good | 6 | – | – | – | – | – |
Moderate | 32 | – | – | – | – | – |
Poor | 8 | – | – | – | – | – |
Group . | HCC . | Non-HCC . | LC . | CHB . | GC . | NC . |
---|---|---|---|---|---|---|
Individuals (n) | 46 | 15 | 26 | 23 | 17 | 24 |
Gender (n) | ||||||
Male | 39 | 12 | 23 | 17 | 15 | 17 |
Female | 7 | 3 | 3 | 6 | 2 | 7 |
Age (years) | 54.2 (9.2) | 45.7 (8.6) | 61.2 (10.2) | 41.3 (8.9) | 58 (7.5) | 50.6 (5.6) |
AFP (ng/ml) | 38.8 (1.62–20000) | 2.97 (1.1–422.3) | 4.22 (0.61–365.9) | 3.6 (1.22–23.7) | 1.81 (1.02–6.87) | 2.6 (1.23–6.38) |
Abnormal (AFP >20 ng/ml) (n) | 24 | 3 | 3 | 2 | 0 | 0 |
Normal (AFP<20 ng/ml) (n) | 22 | 12 | 19 | 11 | 12 | 24 |
ND (n) | 0 | 0 | 4 | 10 | 5 | 0 |
ALT (units/l) | 34.1 (17.8–421.9) | 29.8 (9.1–281.8) | 28.9 (14.5–101.9) | 27.4 (5.8–137.2) | 13.8 (7.6–50.1) | 17.2 (13.1–32.6) |
AST (units/l) | 31.1 (9.1–281.8) | 26.3 (9.4–540.5) | 33.7 (14.7–85.1) | 25.1 (10.2–296) | 15.8 (10.6–24.1) | 15.1 (8.2–29.6) |
GGT (units/l) | 61.2 (11.5–908.0) | 62.3 (15.7–1310) | 59.3 (24.8–199.3) | 26.1 (12.3–118.5) | 18.6 (7.6–135.2) | ND |
TB (μmol/l) | 14.2 (6.0–227.3) | 11.4 (5.4–377.2) | 19.6 (7.5–47.7) | 10.3 (2.8–32.8) | 9.6 (2.3–56.9) | ND |
HbsAg (n) | ||||||
Positive | 33 | 2 | 20 | 23 | 0 | 0 |
Negative | 13 | 12 | 5 | 0 | 16 | 24 |
Not known | 0 | 1 | 1 | 0 | 1 | 0 |
Cirrhosis (n) | ||||||
Normal | 0 | 13 | 0 | Unknown | – | – |
Low | 17 | 2 | 0 | Unknown | – | – |
Moderate | 29 | 0 | 12 | Unknown | – | – |
High | 0 | 0 | 14 | Unknown | – | – |
Histological grade (n) | ||||||
Good | 6 | – | – | – | – | – |
Moderate | 32 | – | – | – | – | – |
Poor | 8 | – | – | – | – | – |
The data demonstrated that patients with HCC, LC or CHB had significantly (P<0.0001) higher serum levels of miR-885-5p than normal controls and GC patients (Figure 4A and Supplementary Table S2 at http://www.clinsci.org/cs/120/cs1200183add.htm). There was no statistically significant difference in the serum levels of miR-885-5p between normal controls and patients with GC (P=0.19, Mann–Whitney test). Furthermore, ROC analysis indicated that using a cut off value of 1.06 (2−ΔCt in comparison with U6 snRNA), miR-885-5p yielded an AUC of 0.904 (95% CI: 0.837–0.951, P<0.0001) that could be used to discriminate patients with liver pathologies (LC, n=26; HCC, n=46; and CHB, n=23) from normal controls (n=24), with specificity and sensitivity of 90.5 and 79.2%, respectively (Figure 4B). For comparison, ALT was used to distinguish patients from healthy controls and yielded an AUC of 0.742 (Figure 4C). In addition, miR-885-5p levels in the sera of patients with ICC and FNH were also increased (Figure 4D).
Validation of miR-885-5p in an independent set of serum samples
(A) Box plots of serum levels (2−ΔCt) of miR-885-5p in NC (n=24), HCC (n=46), LC (n=26), CHB (n=23) and GC (n=17). Expression levels of miRNAs were normalized to U6 snRNA (Log10 scale on y-axis). The upper and lower borders of the boxes indicate the 75th and 25th percentiles, respectively. The whisker caps indicate the 90th and 10th percentiles. The horizontal lines in the boxes indicate the median values. Outliers are illustrated as circles. A Mann–Whitney or Kruskal–Wallis test was used to determine statistical significance. (B) ROC curve analysis of serum miR-885-5p to differentiate patients with liver pathologies (n=95) from NC (n=24). miR-885-5p yielded an AUC of 0.904 with 93.8% sensitivity and 75.0% specificity in discriminating liver pathologies. (C) ROC curve analysis indicated that ALT yielded an AUC of 0.742 in distinguishing patients with liver pathologies (n=95) from NC. (D) Scatter plots indicate that the expression of miR-885-5p in serum samples from patients with ICC and FNH is greater than in controls (Log10 scale on y-axis). The horizontal lines across the data points indicate the median values.
(A) Box plots of serum levels (2−ΔCt) of miR-885-5p in NC (n=24), HCC (n=46), LC (n=26), CHB (n=23) and GC (n=17). Expression levels of miRNAs were normalized to U6 snRNA (Log10 scale on y-axis). The upper and lower borders of the boxes indicate the 75th and 25th percentiles, respectively. The whisker caps indicate the 90th and 10th percentiles. The horizontal lines in the boxes indicate the median values. Outliers are illustrated as circles. A Mann–Whitney or Kruskal–Wallis test was used to determine statistical significance. (B) ROC curve analysis of serum miR-885-5p to differentiate patients with liver pathologies (n=95) from NC (n=24). miR-885-5p yielded an AUC of 0.904 with 93.8% sensitivity and 75.0% specificity in discriminating liver pathologies. (C) ROC curve analysis indicated that ALT yielded an AUC of 0.742 in distinguishing patients with liver pathologies (n=95) from NC. (D) Scatter plots indicate that the expression of miR-885-5p in serum samples from patients with ICC and FNH is greater than in controls (Log10 scale on y-axis). The horizontal lines across the data points indicate the median values.
Relationships between serum miR-885-5p and liver pathology parameters
To examine whether elevated serum miR-885-5p was correlated with traditional liver disease markers such as ALT and AST, linear regression analysis was used to investigate the relationships between miR-885-5p and AFP, ALT, AST and GGT in patients with liver pathologies. There was no correlation between serum miR-885-5p and ALT in 95 patients with liver pathologies (LC, n=26, HCC, n=46, CHB, n=23) (Supplementary Figure 2A available at http://www.clinsci.org/cs/120/cs1200183add.htm). Similar results were obtained for miR-885-5p compared with AST and miR-885-5p compared with GGT (results not shown). There was an increase in AFP concentrations in 46 patients with pathologically confirmed HCC (closed circles), but the levels of serum miR-885-5p (open circles) did not increase (Supplementary Figure 2B).
Multiple factors including age, gender, miR-885-5p, AFP, ALT, AST and GGT in patients with liver pathologies were used to establish a logistic regression model. This model revealed that miR-885-5p was a potential marker for the detection of liver pathologies (P<0.0001), and the odds ratio for an individual with miR-885-5p >1.12 being associated with liver pathologies was 16.3 (95% CI: 6.43–25.37).
DISCUSSION
Recent findings have exposed circulating miRNAs as potential biomarkers for several disease conditions including human cancer [10–16,20–21,24–27]. To our knowledge, this report is the first to investigate the feasibility of using serum miRNAs as indicators of liver pathologies. We found that individuals with HCC, LC and CHB had significantly elevated levels of serum miR-885-5p. miR-885-5p yielded an AUC of 0.904 (95% CI: 0.837–0.951, P<0.0001) with 90.53% sensitivity and 79.17% specificity in terms of discriminating patients with liver pathologies from healthy controls and, more importantly, had a better diagnostic performance than ALT (74.2%). The odds ratio for individuals with miR-885-5p >1.012 having liver pathologies was 16.3 (95% CI: 6.43–25.37). However, there was no correlation between circulating miR-885-5p and AFP, ALT, AST and GGT, and the level of serum miR-885-5p did not increase despite an increase of AFP in the HCC group. There was no significant difference in serum levels of miR-885-5p between normal controls and patients with GC. These results suggest that serum miR-885-5p could potentially be an independent and complementary liver pathology-associated biomarker.
Clinically, new biomarkers are necessary to improve the diagnosis and management of patients with liver pathologies, particularly insidious liver impairment and early HCC [1–2,28]. Recent advances in genomics and proteomics have accelerated the discovery of many potential markers for clinical use. The present study comprises a preliminary investigation of the expression of circulating miRNAs in clinical serum samples of patients with liver pathologies. Using RT-preamp-qPCR [12,29,30], more than 110 miRNA species were detected in clinical samples. In agreement with the findings of Mitchell et al. [10], miRNAs such as miR-223 and miR-16 were present in the circulation with relatively high abundance (Figure 1). In addition, a wide dynamic range of distribution of serum miRNA species was observed; for example, the relative abundances of miR-223 and miR-100 were 295.09 and 0.00026 in the NC serum pool, respectively (Figure 1), which may reflect the different origins and metabolic characteristics of different circulating miRNA species in vivo. Furthermore, miRNAs in the pooled serum samples from patients with HCC and LC were differentially expressed. As shown in Figure 1, which shows the relative abundance of each miRNA in three representative serum pools, similar trends were observed when the miRNAs were ranged according to their relative abundances, e.g. with consistently relatively high copy numbers of miR-16 and relatively low abundance of miR-100 in each pool. Additionally, some miRNAs showed up-regulated levels in both the HCC and LC (Table 1) serum pools, which may reflect the fact that, in most cases, HCC develops in association with liver cirrhosis. Furthermore, unsupervised hierarchical clustering analysis revealed that different serum pools had different miRNA spectra (Figure 2A), which besides some possible technical variations in miRNAs profiling (discussed below) might indicate that the circulating miRNAs in serum pools derived from different patients reveal different metabolic characteristics. Additionally, Figure 2(B) shows that miR-146a, miR-224, miR-574-3p and miR-885-5p were clustered together because of their similarly up-regulated expression in the HCC serum pools (Figure 2B).
As a pilot study, five miRNAs, miR-885-5p, miR-574-3p, miR-224, miR-215 and miR-146a, that are up-regulated in the HCC and LC serum pools (Table 1) were selected and further quantified using real-time PCR in independent sets of patients with HCC, LC, CHB, ICC and FNH. Our case control study confirmed that serum miR-885-5p is consistently and significantly up-regulated in individuals with HCC, LC and CHB (Figure 4A). The median values of miR-885-5p in limited ICC and FNH cases were also elevated (Figure 4D). Nevertheless, although four other miRNAs (miR-574-3p, miR-224, miR-215 and miR-146a) were increased in the global miRNAs profiling experiment (Table 1), we failed to validate their elevation in the second groups of HCC and LC patients (Figures 3B–3D). A probable reason for the discrepancies is that although the TaqMan microRNA array is promising for simultaneously measuring miRNAs expression in a high-throughput manner [31,32], and the introduction of an miRNA cDNA preamplification step significantly reduces the amount of RNA needed [30], potential preamplification bias [30] may be responsible for some variation in the Ct values and may thus diminish the reproducibility with which miRNAs of low abundance such as miR-574-3p, miR-224 and miR-215 can be quantified in the serum.
Recently, advances in circulating miRNAs research [10–16] have generated the concept that tissue or organ-specific intracellular miRNAs may be released and/or leaked into the circulation during processes accompanying cell death and/or apoptosis owing to cell turnover, cellular destruction or pathological injury. It has been suggested that comprehensive investigations aimed at elucidating the associations between circulating miRNAs and various physiopathological conditions may provide new opportunities to use plasma/serum miRNAs as indicators in a clinical setting [10,11]. In view of their relatively low complexity and excellent stability and the availability of high throughput approaches such as ‘deep sequencing’, it is possible to use a panel of circulating miRNAs as markers to complement and/or improve disease monitoring and management [10,33].
Furthermore, our ‘proof-of-principle’ observation supports the idea that ‘normal’ compositions and levels of circulating miRNAs can be deregulated under certain disease conditions, and the finding that miR-885-5p is up-regulated in the sera of patients with liver pathologies could reflect a metabolic imbalance of this miRNA in vivo. Additionally, other miRNAs such as miR-122 and miR-192 were elevated in patients with both HCC and LC (Table 1). Interestingly, miR-122, a liver-associated miRNA, was recently shown to be increased significantly in the circulation of toxicant-treated mice [21]. Therefore the fact that miR-122 is elevated in the sera of patients with liver pathologies merits further investigation. However, it should be noted that the conclusion that serum miR-885-5p is a potential marker for liver pathologies is based on results from a relatively small sample, and we could not analyse the relationship between the elevated levels of serum miR-885-5p and the severity and aetiology of liver pathology. Therefore it might be interesting to enlarge this study by prospective follow-up and investigate the possible correlation between aberrant circulating miR-885-5p and the progression of hepatic cirrhosis/fibrosis. In addition, given that the metabolism and function of circulating miRNAs remain unknown, it is unclear whether serum miR-885-5p is liver-restricted/derived or originates from immune cells involved in the antiviral responses that accompany the development of liver diseases. In a newly published paper [34], miR-885-5p was found to be a relatively liver tissue-enriched miRNA (Supplementary Figure S3 available at http://www.clinsci.org/cs/120/cs1200183add.htm) and our bioinformatics analysis indicates that ABCA1 (ATP-binding cassette subfamily A member 1), which functions as the gatekeeper of CRT (cholesterol reverse transport) and is a key modulator in phosphatide metabolism, is a potential target of miR-885-5p (Supplementary Figure S4 and Supplementary Table S3 available at http://www.clinsci.org/cs/120/cs1200183add.htm). That might mean that this miRNA has some important cellular function in the liver. Further investigation into the expression of miR-885-5p in a variety of clinical liver biopsies and related cell lines might help to elucidate its biological role in the progression of chronic liver diseases. Additionally, although our present observation that the circulating miR-885-5p is significantly up-regulated in patients with HCC, LC and CHB, future studies should include stratification analysis of a large sample to eliminate sample selection biases and to clarify the pathological role of circulating miR-885-5p at different stages of HCC and its correlation with the degree of hepatic chronic inflammation and the severity of liver impairment. Furthermore, as mentioned earlier, since no correlation was found between circulating miR-885-5p and traditional liver parameters such as ALT, it will be interesting to investigate whether overexpressed circulating miR-885-5p correlates with other liver function parameters and hepatic histopathological indicators such as platelets, serum albumin and Scheuer grading system in patients with liver pathologies.
In summary, miR-885-5p was significantly elevated in sera from individuals with liver pathologies. The detection method was sensitive and could be utilized for minimally invasive and complementary detection of liver pathologies in a clinical setting. Although the clinical significance of these findings needs further investigation, miRNAs could serve as novel screening and diagnostic tools for detecting individuals with liver pathologies in routine clinical practice.
FUNDING
This work was supported by the National Project of Scientific and Technical Supporting Program [grant number 2009BAI86B05] and National Science and Technology Infrastructure Program [grant number 2006FY230300] funded by the Ministry of Science and Technology of China.
We thank Yueping Hua (Orbital Instruments, Beijing, People's Republic of China) and Yuantao Zhang (AB, Beijing, People's Republic of China) for their technical support and suggestions. We acknowledge stimulating discussions with Y.M. Lo, Joseph J.Y. Sung and Hongchuan Jin (Prince of Wales Hospital, Hong Kong) and Muneesh Tewari (Fred Hutchinson Cancer Research Centre, Seattle, WA, U.S.A.) concerning the strategy for quantifying circulating miRNAs.
Abbreviations
- AFP
α-fetoprotein
- ALT
alanine aminotransferase
- AST
aspartate aminotransferase
- CHB
chronic hepatitis B
- FNH
focal nodular hyperplasia
- GC
gastric cancer
- GGT
γ-glutamyl transpeptidase
- HCC
hepatocellular carcinoma
- ICC
intrahepatic cholangiocellular carcinoma
- LC
liver cirrhosis
- Mamm
U6, mammalian U6
- miRNAs
microRNAs
- NC
normal control
- qPCR
quantitative PCR
- ROC
receiver operating characteristic
- AUC
the area under the ROC curve
- RT
reverse transcription
- RT-preamp-qPCR
RT-preamplification-qPCR
- snRNA
small nuclear RNA
AUTHOR CONTRIBUTION
Junhao Gui, Yaping Tian and Xinyu Wen designed the study. Junhao Gui, Wenhui Zhang, Pengjun Zhang, Wei Run and Liyuan Tian recruited the patients and provided the samples. Jing Gao and Xingwang Jia determined the liver function parameters, which was supervised by Yanhong Gao. Junhao Gui and Wenhui Zhang collected and assembled the data. Junhao Gui, Wenhui Zhang and Yaping Tian performed data analysis and interpretation; and Junhao Gui and Yaping Tian wrote the manuscript.