Tumor cell lines

The cell lines used in this study were human hepatic (Huh-7, HepG2 and PLC/PRF/5), ovarian (PA-1, TOV-21G, SK-OV-3, BG-1, NIH:OVCAR-3, and ES-2), renal (786-O and ACHN), bladder (T24 and TSGH-8301), breast (MCF-7), pulmonary (A549), and colon (SW480) cancer cell lines. Huh-7 was obtained from JCRB. HepG2, PLC/PRF/5, PA-1, TOV-21G, NIH:OVCAR-3, ES-2, MCF-7, SW480, T24 and TSGH-8301 were obtained from BCRC. A549 was obtained from ATCC. SK-OV-3 and BG-1 were a kind gift from Prof. Tzu-Hao Wang[23]. 786-O and ACHN were a kind gift from Prof. Yeong-Shiau Pu[24]. A549, NIH:OVCAR-3, SK-OV-3, and BG-1 were maintained in DMEM/F12. T24, TSGH-8301, Huh-7, HepG2, MCF-7 and SW480 were maintained in DMEM. 786-O and ACHN were maintained in RPMI1640. PLC/PRF/5 and PA-1 were maintained in MEM. ES-2 was maintained in McCoy’s 5a. TOV-21G was maintained in MCDB105 and M199 (1:1). All media were supplemented with 10% fetal bovine serum, 100U/ml penicillin and 100mg/ml streptomycin (Gibco, Grand Island, NY,USA) under 5% CO2 at 37°C.

The cDNA libraries and paired tumor and non-tumor tissue samples

The fetal cDNA libraries purchased from BioChain (Hayward, CA, USA) were compared with normal human adult cDNA libraries purchased from Clontech (Becton-Dickenson, Franklin Lakes, NJ). Paired tumor and non-tumor tissue from various organs was collected from patients admitted for surgery in National Cheng Kung University Hospital, Tainan, Taiwan.

Magnetic cell sorting to enrich EpCAM+ cells

EpCAM-positive PLC/PRF/5, Huh-7 and HepG2 cells were isolated by magnetic cell sorting (MCS) using monodispersed magnetizable particles according to the manufacturer’s instructions (CELLection^TM Pan Mouse IgG kit) using anti-EPCAM antibody (Epitomics, Burlingame, CA). EpCAM⁺ and EpCAM^- PLC/PRF/5, Huh-7 and HepG2 cells were lysed for western blot assays.

Retroviral Infection

We used a retrovial system to overexpress Lin28B in the HCC cell lines. pMSCVpuro (BD Clontech), pMSCV-LIN28B and pSUPERretro vectors were co-transfected into GP2-293T package cells with VSV-G plasmids using the calcium phosphate method for 48 h. The HepG2 cell was seeded in 1×10⁶ cells per well in a 6-cm dish and incubated overnight under 5% CO₂ at 37 °C. Retroviral supernatant was added with 8 ng/ml of polybrene (Sigma, St Louis, MO, USA), and used to infect HepG2 cell. Pooled HepG2 cell populations expressing either pMSCVpuro or pMSCV-LIN28B were selected with 0.7μg/mL of puromycin (Sigma-Aldrich, St Louis, MO).

shRNA lentivirus production

We used a lentiviral shRNA system to knock down Lin28B in the HCC cell lines. pLKO.1 plasmids expressing small hairpin RNA (shRNA) were purchased from the National RNAi Core Facility (Academia Sinica, Taipei, Taiwan). The lentivirus particles were obtained from the RNAi Core, the Research Center of Clinical Medicine, National Cheng Kung University Hospital. To knock down Lin28B expression, the shRNA of Lin28B (TRCN0000219860, target sequence: 5'- CATAACAGGTCTTCTTCATAT-3') was adopted. A plasmid pLKO_TRC005 was used as the negative control.

Western blotting

Collected cells were dissolved by lysis buffer (Complete Lysis M, EDTA free, Roche) on ice and then centrifuged at 10000 xg, 4°C for 20 minutes. Then, 100μg of protein was loaded and separated by 12 % SDS-PAGE, and transferred to a polyvinylidene fluoride membrane (Millipore, Billerica, MA, USA). Lower amount of protein (40μg) was loaded in the magnetic cell sorting assay due to lower total cell number collected. Primary antibodies included rabbit anti-Lin28B (Cell signaling technology), mouse anti-OCT4 (Santa Cruz Biotechnology, Santa Cruz, CA), rabbit anti-Nanog (Epitomics, Burlingame, CA), rabbit anti-SOX2 (Epitomics, Burlingame, CA), rabbit anti-EpCAM (Epitomics, Burlingame, CA) and mouse anti-β-actin (Millipore).

Sphere formation assay

Sphere formation assay was performed as previously described[25]. Briefly, HepG2 cells were seeded on uncoated 6-well culture plates (BD Labware, Bedford, MA) in DMEM/F-12 serum-free medium (caisson) contained 1% MEM NEAA, 1X N2, 20 ng/ml EGF, 10 ng/ml bFGF, 100 μg/ml penicillin G, and 100 U/ml streptomycin (Invitrogen, Grand Island, NY). After 9 day culture, wells were examined under an inverted microscope at x20 magnification, and the number of spheres of >50 μm in diameter were counted under a light microscope.

Patients, samples and clinical data

One hundred and nineteen patients who had primary HCC underwent hepatectomy at National Cheng Kung University Hospital from January 2006 through December 2011, were included. Pre-surgery whole blood samples were sent for peripheral blood mononuclear cell collection. Patients with a previous diagnosis of cancer, distant metastasis before hepatectomy, positive surgical margins, a diagnosis of combined hepatocellur-cholangiocarcinoma (CHC), or poor sample RNA quality were excluded. Ten patients were excluded due to a positive surgical margin or a diagnosis of CHC and 13 patients were excluded due to poor sample RNA quality. The remaining 96 patients were included for further study (Figure S1). The follow-up interval was every 3 months. Recurrence of HCC was documented upon typical findings of computed tomography or magnetic resonance imaging with or without raised serum AFP level or pathological confirmation. Recurrence-free survival (RFS) was defined as time from surgery to the first occurrence of either local or distant recurrence. Disease-specific survival (DSS) was defined as time from surgery to HCC-related death. Subjects were censored at the last follow-up appointment or at death without recurrence. The patient profiles of the 96 HCC patients were shown in Table S2. The median duration of follow-up was 19.7 months (range, 0.1-41.5 months). Forty patients (41.7%) had recurrent HCC (median duration until recurrence, 7 months; range, 1.5-31.6 months), including local recurrence in 32 patients, metastasis in 5 patients, and both local recurrence and metastasis in 3 patients. Ten patients (10.4%) died of HCC (median survival, 13.8 months; range, 1.6-21.9 months). For comparison, 60 individuals without HCC (non-HCC group) were also included: 31 healthy individuals without a liver disease and 29 patients with viral hepatitis, including 8 patients with cirrhosis (16 HBV and 13 HCV). The mean age of the healthy individuals without liver disease was 44.1 years (range, 20-75 years; 10 men and 21 women), and of the patients with hepatitis was 49.4 years (range, 24-67 years; 20 men and 9 women).

Informed consent in writing was obtained from each patient and the study protocol conformed to the ethical guidelines of the 1975 Declaration of Helsinki as reflected in a priori approval by the Human Experiment and Ethics Committee of National Cheng Kung University Hospital in Tainan, Taiwan.

Peripheral blood sample preparation

Whole blood samples were collected in 10-ml pyrogen-free tubes containing 0.12 ml of K₃EDTA 15% (BD Vacutainer K₃EDTA; Becton Dickinson, Franklin Lakes, NJ) and then layered on equal volumes of Ficoll-Hypaque density gradient (Histopaque-1077; Sigma-Aldrich, Taufkirchen, Germany). Peripheral blood mononuclear cells were recovered from the interphase. The cell pellets were washed and collected for subsequent RNA extraction.

RNA extraction

Total RNA of primary tissue and cell lines were extracted using the Trizol reagent (Life Technologies, Carlsbad, CA). The RNA of blood samples was extracted using a kit (QIAamp RNA Blood Mini Kit; Qiagen) according to the manufacturer’s protocols.

Reverse transcription polymerase chain reaction (RT-PCR) and Reverse transcription quantitative real-time PCR (RT-qPCR)

2 μg RNA was reversely transcribed using SuperScript II (Invitrogen, Grand Island, NY). The semiquantitative RT-PCR primer sequences were shown in Table S3. β-actin was used as a internal control gene in RT-PCR. The cDNAs was subjected to RT-qPCR using a LightCycler system (Roche, Mannheim, Germany). For RT-qPCR in various fetal and normal adult organs and blood samples, the levels of Lin28B were normalized to GAPDH which has been used as a reference gene in detecting circulating tumor cells [19,26]. For liver and HCC tissue samples, the levels of Lin28B were normalized to tyrosine 3-monooxygenase/tryptophan 5-monooxygenase activation protein, zeta polypeptide (PLA) instead of GAPDH because there was a wide range of variation in the expression levels of GAPDH among the liver and HCC tissue samples. PLA had a medium expression level in liver tissue [27] and its expression levels were more stable among HCC tissue samples. Both RT-qPCR primers and probes were purchased (Applied Biosystems; sequences not shown) or synthesized as designed by LightCycler Probe Design Software 2.0. Both the fluorescent TaqMan probes and the designed amplification primer sequences are listed in Table S4. Each standard curve was calculated with serial dilutions (10⁰-10⁶ copies) of plasmid templates. The threshold cycle (Ct) of samples was converted into the copy number of the mRNA using standard curves derived from serial dilutions of constructs of Lin28B and GAPDH, respectively. Nucleic acid extracted from HepG2 cells was used as a positive control. A negative control in which template total RNA was replaced by sterile water was included. The dynamic range of the RT-qPCR for Lin28B was from 10 to 10⁶ copies of plasmid templates (Figure S2). A Ct value less than 38 and greater than zero indicates positive expression of Lin28B gene, whereas a Ct value equal to or greater than 38 or no Ct value indicates negative expression of Lin28B gene.

Statistical analysis

The difference in tumorsphere formation between Lin28B-expressing and -knockdown cell lines was tested for significance using student's t-test. Lin28B mRNA expression in peripheral blood was compared between groups using Wilcoxon rank sum test and was correlated with clinicopathological indicators using chi-square test or Fisher’s exact test. Survival curves and probabilities were estimated using the Kaplan-Meier method, and the log-rank test was used to assess the significance of difference between groups. A leave-one-out cross validation was used to determine a cutoff value associated with survival. Univariate and multivariate Cox proportional hazards regression model was used to determine the significance of different prognostic factors. The multivariate analyses were adjusted simultaneously for age, sex, type of viral infection, cirrhosis, tumor grade, status of multifocal tumors, satellite nodules, vascular invasion, tumor size, American Joint Committee on Cancer (AJCC 7th edition) stage, Barcelona-Clinic Liver Cancer (BCLC) stage and serum AFP. Statistical significance was set at P < 0.05. The analysis of peripheral blood samples from 96 patients will have 80% power to detect a hazard ratio of 2.2 between Lin28B (+) and Lin28B (-) HCC patients. The proportional hazards assumptions were checked by the martingale and deviance diagnostic plots and no significant deviation from the assumptions of the proportional hazard regression model existed.

The oncofetal characteristics of Lin28B

To test whether Lin28B is an oncofetal gene, its expression was examined in a panel of human total RNA (Master Panel II, BD Clontech) using RT-PCR (Figure 1A). Lin28B was expressed in fetal brain, fetal liver, normal adult placenta, testis, brain, and spinal cord. It was not expressed in other adult tissues. The change of expression from fetal to adult tissues was quantitated by RT-qPCR using total RNA selected from Human Fetal Normal Tissue (BioChain) and Master Panel II (BD Clontech) tissue panels (Figure 1B). Lin28B was expressed in the fetal organs, but was significantly down-regulated in their adult counterparts, except for the brain.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 1. The oncofetal expression profile of Lin28B.

A, RT-PCR showed that Lin28B was expressed in fetal brain and liver and in adult testis, brain, spinal cord and placenta. It was not detected in other adult tissues. B, RT-qPCR showed that Lin28B was markedly downregulated from fetal (closed bar) to adult (open bar) tissues except for the brain. C, RT-qPCR was performed using pairs of HCC (closed bar) and non-tumor (open bar) liver tissues. Overexpression of Lin28B (> 100×) was observed in 8 HCC samples (53.3%). NC, negative control; PC, positive control.

https://doi.org/10.1371/journal.pone.0080053.g001

For screening possible Lin28B expression in different tumor types, RT-PCR was performed in various human cancer cell lines. Lin28B could be detected in the ovarian, hepatic, and colorectal cancer cell lines (Figure S3). To examine its expression in tumors, RT-PCR was performed on five cases of paired tumor and non-tumor tissues from 5 types of common cancers (Figure S4). Lin28B was overexpressed in some of the tumors from breast (1/5), uterus (1/5), lung (4/5), liver (1/5), and ovary (1/5), but was expressed at very low levels in the non-tumor tissue. RT-qPCR for Lin28B expression was further performed on 15 HCC tumor and non-tumor tissue pairs (Figure 1C). Over-expression of Lin28B (> 100×) was observed in 8 tumor tissue samples (53.3%). These results confirmed the oncofetal nature of Lin28B. Two non-tumor liver tissue samples (13%) showed low level of Lin28B expression. Microscopically, one of these two samples showed cirrhosis with focal large cell change and the other showed chronic hepatitis with mild interface activity. The histological findings showed no obvious difference from those of the 13 non-tumor liver tissue samples without Lin28B expression, in which seven cases showed cirrhosis and 6 cases showed chronic hepatitis with interface activity. Four cases showed large cell change and one case showed small cell change.

Elevated expression of Lin28B in EpCAM-enriched stem cell-like population

EpCAM is a surface marker that was reported to be able to enrich the hepatic stem cell-like population [28]. Therefore, to investigate whether Lin28B is expressed in cancer stem cells, we performed magnetic cell sorting (MCS) to separate HCC cell lines into EpCAM⁺ and EpCAM^- populations. The protein level of Lin28B was increased in EpCAM⁺ PLC/PRF/5 cells and Huh-7 cells (Figure 2A, upper panel), along with increased levels of stem cell markers, SOX2, Nanog and OCT4 in PLC/PRF/5 cells and SOX2 and OCT4 in Huh-7 cells (Figure 2A, lower panel). In contrast, EpCAM-capture did not enrich the stem cell-like population in HepG2 cells, since there was no difference of SOX2, Nanog, or OCT4 expression between EpCAM⁺ and EpCAM^- HepG2 cells (Figure 2A, lower panel). Lin28B also showed no difference between these two populations (Figure 2A, upper panel). Taken together, our findings supported the notion that Lin28B is expressed in cancer stem cell-like subpopulations.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 2. Expression of Lin28B associated with stemness markers in HCC cells.

Western blot analyses showed that Lin28B was overexpressed in captured EpCAM⁺ PLC/PRF/5 cells and Huh-7 cells but not in EpCAM⁺ HepG2 cells (A, upper panel). EpCAM⁺ PLC/PRF/5 cells demonstrated elevated levels of SOX2, Nanog and OCT4; EpCAM⁺ Huh-7 cells, SOX2 and OCT4; but EpCAM⁺ HepG2 cells, none (A, lower panel). Lin28B-pMSCV HepG2cells showed increased levels of OCT4, Nanog, SOX2 and EpCAM in western blot (B) and formed more spheres than vector control cells (P=0.032) (C). sh-Lin28B HepG2 cells showed decreased levels of OCT4, Nanog, SOX2 and EpCAM in western blot (D) and tended to form fewer spheres than vector control cells (P = 0.059) (E). sh-Lin28B PLC/PRF/5 cells and Huh-7 cells also decrease expression of stem cell markers (F). Lower amount of protein (40μg) was loaded in the MCS assay than that (100μg) loaded in the Lin28B-overexpression assay due to lower total cell number collected in MCS assay. Different viral systems were used in the experiments: retrovirus in Lin28B over-expression assay and lentivirus in Lin28B knock-down assay. Cell growth was slower when infected with lentivirus. MCS, magnetic cell sorting. Scale bar in C and E, 50 μm.

https://doi.org/10.1371/journal.pone.0080053.g002

Increased in vitro stemness with Lin28B overexpression

Because EpCAM could not enrich the Lin28B-expressing stem cell-like population in HepG2 cells, we established a Lin28B-overexpression HepG2 stable pool by retrovirus infection to investigate the effect of Lin28B on stem cell-like phenotypes. Representative stem cell markers were analyzed by Western blotting. In comparison to the vector control, Lin28B overexpression upregulated the stem cell markers: SOX2, Nanog, and EpCAM. A mild increase of OCT4 was also observed (Figure 2B).

To determine the effect of Lin28B on tumor sphere-formation, both Lin28B-expressing and control cells were cultured in suspension to generate spheres as an indicator of a cancer stem-like property in vitro [25,29,30]. As shown in Figure 2C, Lin28B-expressing HepG2 cells showed an increased tumorsphere forming ability compared with the vector control cells (P=0.032).

On the contrary, knocking down Lin28B in HepG2 cell line downregulated the expression of OCT4, Nanog, SOX2 and EpCAM (Figure 2D) and tended to reduce the tumorsphere formation (P=0.059) (Figure 2E). Furthermore, knocking down Lin28B in PLC/PRF/5 cells and Huh-7 cells also downregulated the expression of the stem cell markers (Figure 2F).

Detection of Lin28B mRNA in the peripheral blood cells of HCC patients

RT-qPCR was applied to examine the expression of Lin28B in the peripheral blood circulating cells. The reaction was linear from 10 to 10⁶ copies of purified plasmid templates (Figure S2). The expression of Lin28B could be detected when one HepG2 cell was pooled with 10⁷ leukocytes in about 3 ml of whole blood (Figure S5). On this base, Lin28B mRNA was detected in 3 cases (5%) of non-HCC controls (1 in healthy group and 2 in hepatitis group) and in 32 cases (33.3%) of HCC group (Figure 3A). The range of Ct was from 30.57 to 37.94 for the positive samples.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 3. RT-qPCR quantification of the expression of Lin28B in peripheral blood monoculear cells.

Lin28B was expressed in 3 cases (5%) in the non-HCC controls (1 in healthy group and 2 in hepatitis group) and in 32 cases (33.3%) in the HCC group. (Bar: mean; Black dot: undetectable) (A). Patients with recurrent HCC had significantly higher expression levels of Lin28B than patients without recurrent HCC. (P<0.001). (Bar: mean; Black dot: undetectable) (B). Kaplan-Meier analysis showed that Lin28B was significantly associated with decreased recurrence-free survival (P<0.001) (C). Ratio of Lin28B/GAPDH mRNA higher than 10^-3 tended to associate with disease-specific survival (P=0.094) (D). Lin28B was significantly associated with decreased recurrence-free survival in AJCC stage I-II patients (P=0.003) (E) but not in AJCC stage IIIA-IVA patients (P=0.419) (F). Lin28B was significantly associated with decreased recurrence-free survival in AJCC stage I (P=0.030) (G) and stage II (P=0.030) patients (H).

https://doi.org/10.1371/journal.pone.0080053.g003

Patients with recurrent HCC had significantly higher expression levels of Lin28B than those without recurrence (P<0.001) (Figure 3B). The expression of Lin28B in circulating cells was significantly associated with non-cirrhotic liver (P=0.021), high tumor grade (P=0.046), large tumor size (P=0.005), high AJCC stage (P=0.044) and BCLC stage (P=0.017) (Table 1). The level of circulating Lin28B had a significant association with high BCLC stage (stage B to C diseases versus A1 to A4 diseases) (P=0.022) (Figure S6A) and had a borderline significance in associated with high AJCC stage (stage IIIC to IVA diseases versus I to IIIB diseases) (P=0.066) (Figure S6B). Kaplan-Meier analysis showed that circulating Lin28B was significantly associated with decreased RFS (P<0.001) (Figure 3C). The expression of Lin28B in circulating cells was not significantly associated with decreased DSS (P=0.140). By using a leave-one-out cross validation method, ratio of Lin28B/GAPDH mRNA higher than 10^-3 (n=15) tended to be associated with decreased DSS (P=0.094) (Figure 3D).

Univariate analysis revealed that tumor grade (P=0.047), vascular invasion (P=0.038), AJCC stage (P<0.001), BCLC stage (P=0.030) and circulating Lin28B (P=0.001) were significantly associated with decreased RFS (Table 2). In the multivariate model, cirrhosis (P=0.043), AJCC stage (P=0.001) and circulating Lin28B (P=0.047) were independent variables associated with decreased RFS (Table 2). Tumor grade (P=0.035), AJCC stage (P<0.001), serum AFP (P=0.014) and circulating Lin28B (P=0.001) were significantly associated with early recurrence less than one year in the univariate model (Table 3). AJCC stage (P=0.004) and circulating Lin28B (P=0.045) were significantly associated with early recurrence less than one year in the multivariate model (Table 3). As for DSS, satellite nodule (P=0.047) and AJCC stage (P=0.046) were independent variables in the multivariate analysis (Table S5). However, Lin28B/GAPDH mRNA higher than 10^-3 was not an independent parameter in DSS.

If further stratified by AJCC stage, circulating Lin28B significantly correlated with RFS in earlier stages (stage I and stage II) (P=0.003) (Figure 3E) although it did not reach statistical significance in patients with more advanced stages (P=0.419) (Figure 3F). Circulating Lin28B still significantly correlated with RFS in stage I and stage II patients, separately (P=0.030 and P=0.030, respectively) (Figure 3G and 3H). In multivariate analysis, circulating Lin28B (P=0.006) and AJCC stage (P=0.023) were independent variables associated with decreased RFS in earlier stages (Table S6).