The association between KLF4 as a tumor suppressor and the prognosis of hepatocellular carcinoma after curative resection

Different patterns were used for detecting KLF4 expression in HCC

KLF4 expression in most HCC cells (PLC/PRF/5, Hep3B, Huh7, HepG2, but except SMMC- 7721) was lower than in normal liver cells (L02) not only at mRNA level but also at the protein expression level (Figure 1A and 1B). Same KLF4 expression patterns were found in the TCGA database (https://tcga-data.nci.nih.gov/tcga/), with lower expression observed in many kinds of tumors including lung adenocarcinoma (LUAD), bladder urothelial carcinoma (BLCA), and cholangiocarcinoma (CHOL) (opposing to matched para-tumorous tissues) (Figure 1D).

Figure 1. KLF4 expression in hepatocellular carcinoma (HCC) tissues and cell lines. (A) KLF4 expression was detected in the mRNA level among five HCC cell lines and one normal liver cell (L02). (B) KLF4 expression was detected in the protein level in six cell lines. The internal control was β-actin. (C) Immunohistochemical results were analyzed by chi-square test to compare the distribution of KLF4 in HCC tumors and adjacent tissues. (D) KLF4 expression was analyzed in tumor and para-tumorous tissues in TCGA tumors. (E) KLF4 expression was exhibited through characteristic photos of immunostaining in HCC tumor and para-tumorous tissues. Image scale= 100μm. P < 0.05 was considered statistically significant, * P < 0.05, ** P < 0.01, *** P < 0.001.

Immunohistochemical staining intensity in a tissue microarray (TMA) was divided into four levels, which is our scoring criterion for KLF4 expression in tissues (Figure 1E). According to the score statistics of staining degree and the staining area, high KLF4 expression rate in para-tumorous tissues was 58.92% (109 of 185: 16 of strong and 93 of moderate) while high KLF4 expression rate in HCC tissues was 23.78% (44 of 185: 4 of strong and 40 of moderate), which showed KLF4 expression in tumor tissues is lower than the paired para-tumorous tissues (*P <0.05) (Figure 1C). The results demonstrated that KLF4 may be a potential prognostic marker for those patients with HCC.

KLF4 expression is related to some clinicopathological features

Depending on the scores of immunohistochemical staining, 185 HCC patients were distributed into two groups: one group had high KLF4 expression and the other group had low KLF4 expression. High KLF4 expression group showed a statistical score of strong and moderate staining while the low KLF4 expression group showed a statistical score of weak and negative staining. Table 1 shows that low KLF4 expression was associated with many clinical phenotypes, such as vascular invasion (***P <0.001), high alpha-fetoprotein level (*P = 0.012), and advanced Barcelona Clinic Liver Cancer (BCLC) stage (**P = 0.002). There was no remarkable correlation between other clinical characteristics and KLF4 in our research.

The level of KLF4 expression was related with OS and RFS of HCC patients

The results showed a relation between the outcome of HCC and KLF4 expression when evaluating OS and RFS of patients. Figure 2A showed HCC patients with high KLF4 expression in tumor tissues displayed a favorable prognosis in OS (OS, ***P <0.001) and RFS (RFS, ***P <0.001) analyses compared to those patients with low KLF4 expression.

Figure 2. The prognostic analysis of KLF4 in HCC patients (n = 185) and subgroup analysis based on differentiation degree and BCLC stages. (A) According to KLF4 expression level in patients’ tissues with HCC, the curves described OS and RFS of patients, respectively. The OS in (B) differentiation I-II patients and (C) differentiation III-IV patients was analyzed by Kaplan- Meier method. The RFS in (D) differentiation I-II patients and (E) differentiation III-IV patients were analyzed by Kaplan- Meier method. Kaplan- Meier method was employed to analyze the OS of patients in (F) BCLC 0-A group and (G) BCLC B-C group. The RFS of patients in two subgroups (H and I) of the BCLC stage was analyzed through Kaplan- Meier analysis. P < 0.05 was considered statistically significant.

From the analysis of the impact of different differentiation degree or BCLC stages on OS and RFS, we can see patients with higher KLF4 expression in “differentiation I-II” subgroups had a better OS (***P <0.001, Figure 2B) and RFS (**P =0.0029, Figure 2D). However, differences in KLF4 expression of the “differentiation III-IV” subgroup failed to predict similar outcomes (Figure 2C and 2E). Then to further analyze the relationship between KLF4 and prognosis in different BCLC subgroups, results show the patients in all stages of BCLC with low KLF4 expression group had shorter OS (**P = 0.0019, Figure 2F and ***P <0.0001, Figure 2G) and RFS (*P =0 .0183, Figure 2H and Figure 2I, **P =0.0063).

Low expression of KLF4 may mean poor prognosis for HCC patients

Table 2 showed the prognostic role of KLF4 in HCC patients. In univariate analysis, tumor size and the degree of tumor differentiation were significantly related to HCC patients’ OS and RFS. The OS was also significantly related to tumor number and microvascular invasion while RFS was also significantly associated with serum HBsAg level. However, other features including age, gender, ALT, liver cirrhosis, tumor encapsulation, and serum AFP level showed no significant prognostic associations with OS or RFS. A multivariate Cox regression analysis was then conducted with all the prognostic factors of *P < 0.05 in univariate analysis. Based on the results, we concluded high KLF4 was identified as a positive prognostic factor for OS and RFS of HCC patients.

Prognostic prediction of nomograms for HCC based on KLF4 expression

Two new prognostic nomograms were built based on the results of univariate analysis to predict the OS and RFS of HCC patients, which aimed to further clarify the relationship between KLF4 expression and prognosis of patients (Figure 3A and 3B). As seen in Figure 3C–3F, the calibration curves were used to compare the value of nomograms-prediction and the actual observed value, and found the predicted results is similar to the actual results. The decision curve analyses were drawn to compare the prediction of clinical net benefits between prognostic nomograms and the BCLC staging. According to the comparison, it is found that prognostic nomograms has better prediction capability (Figure 3G–3J). To enhance the predictive accuracy of KLF4 for HCC, C-index (Harrell’s concordance index) was employed to evaluate the performance. Table 3 showed the nomograms for OS and RFS (based on KLF4 expression) owned a better postoperative prediction effect compared to the BCLC staging (***P < 0.001).

Figure 3. Nomograms and decision curves analyses were to further study the effect on the prognosis of KLF4. Pictures (A) and (D) showed the predictive analysis through prognostic nomogram. First, the plumb line between each factor and the point scale was drawn, and then got the point of each factor. The sum of the points of all factors is the total set of points. finally, a plumb line was drawn from all point scales to the probability scale to obtain the probability of OS or RFS. 3- year and 5- year OS (B and C) and 3-year and 5- year RFS (E and F) were shown by calibration curves. The X-axis represented the predicted value of OS or RFS by nomograms and the Y-axis represented actual OS or RFS. The clinical effects of different models were exhibited by decision curve analyses. The comparison of predictions between nomogram-predicted and conventional staging system for 3-year OS and RFS (G and H) and 5-year OS and RFS was shown (I and J). BCLC: Barcelona Clinic Liver Cancer staging. Dashed lines: The probability of the clinical net benefit crossing a certain threshold; the solid black horizontal line: to suppose no patients suffer the incident; the gray solid line: to assume all patients suffer the incident.

KLF4 inhibits HCC cells proliferation

Based on the dysregulation of KLF4 expression in HCC patients, we further studied the effects of KLF4 on the biological behaviors of HCC cells. PLC/PRF/5 cells were transfected with pCDHCMV-MCS-EF1-Puro-KLF4 to select stable KLF4 overexpressed cells (OE-KLF4), while SMMC-7721 cells were transfected with pLKO.1-shRNA to select stable KLF4 silenced cells (sh-KLF4). The overexpression control (OE-NC) and shKLF4 control (sh-NC) were transfected with empty plasmids, respectively. The efficiency of overexpression or knockdown of KLF4 in PLC/PRF/5 and SMMC- 7721 cells was detected by Western blot (Figure 4A). We applied the CCK8 assay to detect the cell proliferation rate. As shown in Figure 4B, the KLF4 overexpression could significantly reduce the cell proliferation activity in PLC/PRF/5 cells. While, KLF4 knockdown significantly improved the cell proliferation activity in SMMC- 7721 cells. The results confirmed that KLF4 inhibits cell proliferation of HCC cells in vitro.

Figure 4. KLF4 inhibits the proliferation and migration ability of HCC cells. (A) Knockdown of KLF4 in SMMC-7721 and overexpression of KLF4 in PLC/PRF/5 was verified by western blot, and β-actin was used as an internal control in western blot assays. (B) CCK8 assay was implemented to detect the proliferation rate of steadily transfected SMMC-7721 and PLC/PRF/5. (C) Cell wound scratch assay and (D) transwell assay was executed to evaluate the migration rate of steadily transfected SMMC-7721 and PLC/PRF/5. Student’s t-test was used in line charts and bar charts. P < 0.05 was considered statistically significant. * P < 0.05, ** P < 0.01, *** P < 0.001.

KLF4 inhibits HCC cells migration

Cell wound scratch and transwell assays were employed to explore the effects of KLF4 on cell motility. Compared to the control group, the cell wound scratch assay revealed that over-expressed KLF4 inhibited cell migration in PLC/PRF/5 cells, and knocked-down KLF4 promoted cell migration in SMMC-7721 cells, as shown in Figure 4C. Also, transwell assay showed that over-expressed KLF4 inhibited cell migration in PLC/PRF/5 cells and knocked-down KLF4 promoted cell migration in SMMC-7721 cells in comparison with the respective control group (Figure 4D). Thus, we concluded that KLF4 could inhibit HCC cell migration in vitro.

KLF4 inhibits the invasion of HCC cells

Transwell assay was used to investigate the effects on invasion by changing KLF4 expression. We found that up-regulated KLF4 expression in PLC/PRF/5 cells inhibited cell invasion, and decreased KLF4 expression in SMMC-7721 cells promoted cell invasion. These results illustrated that KLF4 inhibits the invasion of HCC cells in vitro (Figure 5A). Judging from the cellular morphology of SMMC-7721 cells with shKLF4, reduced KLF4 expression leads to the morphological transformation from the epithelial phenotype into the mesenchymal phenotype (Figure 5B). Western blot and qPCR assays showed that KLF4 overexpression in PLC/PRF/5 cells could effectively upregulated the expression of E-cadherin. While KLF4 knockdown could significantly downregulate the expression of N-cadherin and Vimentin (Figure 5C). These results revealed that KLF4 exhibits the capacity to block the activation of EMT pathway.

Figure 5. KLF4 inhibits the invasion of HCC cells. (A) Transwell assay was determined to detect the effect of KLF4 on cell invasion in PLC/PRF/5 and SMMC-7721 cells. Student’s t-test was used in bar charts. (B) Changes in cell morphology in SMMC-7721 with shKLF4. (C) Western blot and qPCR were used to detect EMT-associated protein expression in steadily transfected PLC/PRF/5 and SMMC-7721 cells. P < 0.05 was considered statistically significant. * P < 0.05, ** P < 0.01, *** P < 0.001.

Patient selected and follow-up

We completed a statistical analysis of registered information from 185 HCC patients in Zhongshan Hospital, Fudan University. All patients had a hepatectomy from July 2010 to July 2011. Patients' screening criteria were as follows: patients did not receive any systemic or local treatments nor experience extrahepatic metastases before the operation. All patients who underwent primary and therapeutic hepatectomy had a definite postoperative pathological diagnosis of HCC. These patients who all showed absolute follow-up data and clinicopathological features had no infection or inflammatory response other than viral hepatitis. The patients underwent a series of examinations such as alanine aminotransferase (ALT), liver cirrhosis, hepatitis B virus surface antigens (HBsAg), alpha-fetoprotein (AFP), tumor encapsulation, tumor number, tumor size, vascular invasion, differentiated degree, and Barcelona Clinic Liver Cancer staging (BCLC) within 3 days before the operation. All patients were followed up and received blood routine examination, blood biochemical examination, serological examination of tumor markers, abdominal ultrasound, and chest imaging examination. The follow-up time was defined as every three months for the first five years after surgery and once a year for the time after. The deadline for data collection was July 2018 or the time of death in deceased patients. The study was conducted with the patients’ knowledge and permitted by the hospital ethics committee. The academic research was conducted under ethical standards.

Tissue samples and immunohistochemistry

The human HCC tissues microarray was made of the tumor tissues and para-carcinoma tissues collected from the 185 HCC patients after curative resection. We then sorted out and analyzed the corresponding information of tissue microarray. Tissue microarrays were constructed with surgical specimens fixed by formalin and embedded by paraffin. Anti-KLF4 antibody (ab215036; Abcam) was used for incubating sections. Immunohistochemical analysis was performed according to the previous method [38]. Then, immunohistochemical results were assessed by two independent pathologists who did not have access to patients’ information. The staining degree of KLF4 was scored semi-quantitatively as 3 for strong, 2 for moderate, 1 for weak, 0 for negative. The score of the staining area was calculated by the percentage of stained positive cells and was scored as 4 for >75%, 3 for 51%- 75%, 2 for 26%- 50%,1 for 5%- 25%, 0 for <5%. The final results described each specimen through multiplying the staining degree score by the staining area score. The expression level of KLF4 in each specimen was evaluated according to the score value with 9- 12 as a strong level, 6- 8 as a moderate level, 1- 4 as a weak level, 0 as a negative level.

Cell culture and transfection

There are six types of cells (five types of liver cancer cells: SMMC- 7721, HepG2, Hep3B, Huh7, and PLC/PRF/5, and a normal liver cell (L02)) were included in the experiments and all cell identification reports were provided in supplementary materials. All these cells were acquired from the cell bank of the Chinese Academy of Sciences (Shanghai, China). All the cells were cultured with Dulbecco's Modified Eagle Medium (DMEM basic; Gibco). The concentration of fetal bovine serum (FBS; Gibco) in DMEM was 10%. These cells were then cultured at cell incubator (Thermo; America) with 5% CO2 at 37°C. Lentiviral particles loaded the KLF4 overexpression plasmid (OE-KLF4) to infect PLC/PRF/5 cells, empty vectors (OE-NC) were used as overexpression control. SMMC- 7721 cells were infected with lentiviral particles, loading shRNA-KLF4 plasmid (sh-KLF4), and empty vector (sh-NC) as shRNA control.

RNA extraction and quantitative reverse transcription PCR (qRT-PCR)

Firstly, PBS was used to wash the collected HCC cells. Then, total RNA was extracted with the TRIzol reagent (Invitrogen, USA). Next, 1 μg of total RNA was used for reverse transcription with Reverse Transcription Reagent Kit (Takara, Japan). The mRNA expression of KLF4 was detected with TB Green qPCR Mix (Takara, Japan) on the basis of reagent supplying protocols. The primers were used for detection as follows: KLF4-forward: 5′-CAAGTCCCGCCGCTCCATTA-3′; KLF4-reverse: 5′-CCATCCACAGCCGTCCCAGT-3′; β-actin-forward: 5′-GGACCTGACTGACTACCTCAT-3′; β-actin-reverse: 5′-CGTAGCACAGCTTCTCCTTAAT-3′; E-cadherin-forward: 5′-CGAGAGCTACACGTTCACGG-3′; E-cadherin-reverse: 5′-GGGTGTCGAGGGAAAAATAGG-3′; N-cadherin-forward: 5′-TGCGGTACAGTGTAACTGGG-3′; N-cadherin-reverse: 5′-GAAACCGGGCTATCTGCTCG-3′; Vimentin-forward: 5′-CGGGAGAAATTGCAGGAGGA-3′; Vimentin-reverse: 5′-AAGGTCAAGACGTGCCAGAG-3′. The conditions of the experimental reaction were: 1 cycle for 30 seconds (95 °C),40 cycles of denaturation for 5 seconds (95 °C), and annealing for 34 seconds (60°C). Finally, the gene expression differences were analyzed according to the experimental results. Three repetitions were performed in the experiment.

Western blot

The protein expression level of KLF4 in the HCC cell lines was detected by Western blot. RIPA lysis buffer was employed to extract total protein of HCC cells, and then 10% SDS- PAGE were used to separate protein samples. The PVDF membrane (IPVH 00010; Millipore) with a pore size of 0.45μm was used to transfer the cell protein. After 5% defatted milk blocking for 90 minutes at room temperature, anti-KLF4 antibody (ab215036; Abcam) and anti-β-actin antibody (#13E5; CST) were used for incubating the membranes at 4°C overnight. On the second day, TBST (1% Tween diluted in TBS) was used to wash the membranes, then the membranes were incubated with diluted secondary antibodies (AP132P; EMD Millipore) at room temperature for 90 minutes. Finally, the membranes were scanned with equipment (Image Quant LAS 4000; Sweden).

Cell proliferation assay

We employed CCK8 assay (http:www.dojingdo.cn) to detect the cell proliferation activity and viability of the selected SMMC- 7721 and PLC/PRF/5 cells. Counted cells were cultured with 96-well plates at 37 °C. The absorbance at 450nm was observed at different time points with each well injected into the 10ul CCK8 kit so that we can compare the rate of cell proliferation by the value of absorbance. Each well has three identical wells.

Cell migration and invasion assays

For cell wound scratch experiment, selected cells and cells of the control group were cultured in 6-well plates. Each well was covered with cells and scratched with white tips on the next day. Then pictures were taken at different time points. Cell migration and invasion was examined by transwell assays. Matrigel matrix basement membrane (#354234; Corning) was added into the chambers (8.0μm pores; Corning) before culturing cells in transwell invasion assay. For transwell migration and invasion assays, a certain amount of medium with serum was added to 24-well plate, then chambers were placed on the plate. The counted cells were cultured with serum-free DMEM in chambers for a few hours, where they can migrate through the holes planting on the membranes of the chambers. Finally, the cells on the membranes were fixed with 4% paraformaldehyde and the fixed cells stained with crystal violet. The average number of cells in the five fields was regarded as the number of migrating cells.

Statistical analysis

Using the TCGA database, we estimated the differential expression of KLF4 between tumor tissues and para-tumorous tissues in different types of cancers with the method of the Wilcoxon test. The results were shown in box plots. The relationship between clinicopathologic features and KLF4 was evaluated by student’s t-test and Pearson chi-square test. The OS and RFS were assessed by the Log-rank test and Kaplan Meier survival analysis. For other data, they were evaluated by means ± standard deviation (SD), and SPSS 19.0 was used to complete the analysis. According to the distribution of data, student’s t-test or Mann-Whitney U test was used for the comparison of differences between groups. The Cox proportional risk regression model was employed to perform univariate and multivariate analyses. The property of a nomogram was estimated according to three ways: the decision curve analysis (DCA), calibration curve and Concordance index (C- index). Only when *P < 0.05, results reached statistical significance.