The human liver cancer cell lines LM3, HepG2, PLC/PRF/5, Li-7 and Huh7 and the murine HCC cell line Hepa1-6 were provided by the Hepatobiliary Institute of Nanjing University, while the normal hepatocyte cell line MIHA was provided by the Shanghai Institute of Cell Biology. Unless otherwise specified, all the experiments were conducted using the Huh7 cell line. Short tandem repeat profiling was conducted to authenticate all the cell lines used in the present study. The cells were cultured in DMEM or RPMI-1640 supplemented with 10% FBS (all from Gibco; Thermo Fisher Scientific, Inc.), 100 U/ml penicillin and 100 mg/ml streptomycin, and maintained at 37°C and 5% CO₂ in a humidified incubator. When the cell density reached 60–70%, the cells were treated with HDAC3 inhibitors, RGFP966 (cat. no. S7229; Selleck Chemicals) or Trichostatin A (TSA; cat. no. S1045; Selleck Chemicals).

HCC tissues and the corresponding adjacent normal tissues were provided by the Affiliated Drum Tower Hospital of Nanjing University Medical School (Nanjing, China) with written informed consent from the patients concerned. The present study was approved by the local ethics committee. Western blot (WB) and immunohistochemistry (IHC) assays were used to evaluate the protein levels in the HCC tissues, as described previously (19). The following primary antibodies were purchased from Abcam: anti-KLB (WB, 1;1,000 dilution; IHC, 1:200 dilution; cat. no. ab106794) and anti-β-catenin (WB, 1:1,000 dilution; IHC, 1:400 dilution; cat. no. ab223075).

Transcript levels of KLB in different cancer tissues and their corresponding normal tissues were analyzed by the ONCOMINE (https://www.oncomine.org/) and Tumor IMmune Estimation Resource (TIMER) (https://cistrome.shinyapps.io/timer/) databases. The transcriptomic data of KLB in normal and HCC samples were obtained from the Cancer Genome Atlas (TCGA) database (https://www.cancer.gov/tcga) and its expression in HCC samples was further determined based on tumor grade, individual cancer stage, TP53 mutation status and nodal metastasis status using the UALCAN database. The Kaplan-Meier plotter (https://kmplot.com/) was used to analyze the prognostic value of the KLB gene in patients with HCC. Gene set evaluation analysis (GSVA) (https://www.aclbi.com/static/index.html#/) of TCGA data was used to determine the potential biological function of KLB.

siRNAs were used to alter the expression of KLB and HDAC3 in the Huh7 cell line. The siRNAs with the highest silencing efficiency were screened and used for gene function research. After screening, the Huh7 cells (60–70% confluency in a 6-well plate) were treated with KLB- or HDAC3-siRNAs (si-KLB1, 5′-CGCUAUAGGAAUACAAUGUTT-3′; si-KLB2, 5′-GCUUCAAGCAAUAAGGUUATT-3′; and si-HDAC3, 5′-CACAAAUACGGAAAUUACUTT-3′) and the corresponding control-siRNAs (TSINGKE Biological Technology), which were encapsulated by the INTERFERin reagent (Polyplus-transfection SA). The sequence of the control siRNA was not provided by the supplier.

To inhibit KLB expression, the Huh7 cell line was infected with lentiviral vectors carrying short hairpin RNA against KLB (sh-KLB; target sequence, 5′-GCTTCAAGCAATAAGGTTA-3′). To overexpress FGF21, the huh7 cell line was transfected with lentiviral vectors carrying the FGF21 gene (GeneChem; sequences provided in supplementary data). Empty lentiviral vectors were used as internal controls.

Proteins were extracted from the HCC tissues and cell samples using RIPA buffer (KeyGEN BioTECH) with phenylmethylsulfonyl fluoride, according to the manufacturer's instructions, and the WB and co-IP assays were conducted as described previously (20). Anti-KLB (WB, 1:1,000 dilution; cat. no. ab106794), anti-E-cadherin (WB, 1:2,000; cat. no. ab40772), anti-claudin-1 (WB, 1:2,000; cat. no. ab211737), anti-MMP9 (WB, 1:2,000; cat. no. ab76003), anti-β-catenin (WB, 1:1,000; cat. no. ab223075), anti-laminB1 (WB, 1:5,000; cat. no. ab108536), anti-β-actin (WB, 1:1,000; cat. no. ab8226) and anti-GAPDH (WB, 1:10,000; cat. no. ab181602) were purchased from Abcam; anti-vimentin (WB, 1:20,000; cat. no. 60330-1-Ig) and anti-HDAC3 (WB, 1:1,000; cat. no. 10255-1-AP) were purchased from ProteinTech; and anti-acetylated lysine (WB, 1:1,000; cat. no. 9441) was purchased from Cell Signaling Technology, Inc. The bicinchoninic acid assay was conducted to determine the protein concentration. Each experiment was replicated thrice. The results of the WB assay were processed by Image J software (National Institutes of Health, version 1.8.0).

Total RNA was extracted from the HCC cells using TRIzol™ (Thermo Fisher Scientific, Inc.) and subjected to RT-qPCR analysis using the ABI PRISM 7500 Real-Time PCR System (Applied Biosystems; Thermo Fisher Scientific, Inc.), as described previously (20). β-Actin was used as the internal reference gene and the 2^−ΔΔCq method was used to evaluate the relative mRNA expression (21). The PCR primers were designed by Realgene (Nanjing, China) and are listed in Table I.

Cell viability, EdU and apoptosis assays were conducted using a cell counting kit-8 (cat. no. C0037), the BeyoClick™ EdU kit (cat. no. C0075S) and an Annexin V-PE kit (C1065S; Beyotime Institute of Biotechnology), respectively, following the manufacturer's instructions. For the colony-formation assay, si-KLB-treated cells and negative control cells (n=500 each) were incubated in a 12-well plate and cultured for 2 weeks. Thereafter, the cells were fixed with methanol for 30 min and stained with 0.1% crystal violet for 30 min at room temperature. Finally, the colonies containing ≥50 cells were counted.

Wound-healing, cell migration and cell invasion assays were conducted as described previously (22). For the cell migration and invasion assays, migrated or invaded cells were counted using a microscope in six random fields. For the wound-healing assay, cell migration was calculated as the percentage of wound closure.

The IF assay was conducted as described previously (20). In brief, treated cells were seeded in 24-well plates and incubated for 24 h. Thereafter, the cells were fixed using 4% paraformaldehyde and washed thrice with PBS. The cells were then treated with anti-vimentin antibodies (IF, 1:500 dilution; cat. no. 60330-1-Ig, ProteinTech), followed by incubation with goat anti-mouse IgG H&L (Alexa Fluor^® 488; IF, 1:500 dilution; cat. no. ab150113; Abcam). The cells were then observed under a fluorescent microscope (Leica Microsystems GmbH) and the images were obtained.

5×10⁶ Huh7 cells were injected into male BALB/c nude mice (GemPharmatech; age, 3–4 weeks; weight, 16–20 g; n=6 per group) through the tail vein to generate the mouse lung metastasis model. The mice were housed with 5 mice per cage on a 12-h light/dark cycle, controlled humidity (50–60%) and temperature (24–26°C) and had free access to food and water (23). After 6 weeks, the mice were anesthetized and euthanized by intraperitoneal injection of pentobarbital sodium 150–200 mg/kg. The lungs were excised, imaged and embedded in paraffin. Tumor metastases were analyzed by hematoxylin and eosin staining according to standard procedures. The Ethics Committee of the Affiliated Drum Tower Hospital of Nanjing University Medical School (Nanjing, China) approved the animal experiments.

GraphPad Prism v8.02 (Dotmatics) was used to conduct data analyses and to calculate the P-value. Student t-test and one-way analysis of variance (ANOVA) with Dunnett's post-hoc test were used for comparisons between two or multiple groups, respectively. P<0.05 was considered to indicate a statistically significant difference.

KLB, a vital component of the endocrine FGFR complex, is involved in tumor progression in several cancers. To obtain KLB expression profiles in human cancers, pan-cancer KLB expression data were retrieved from the ONCOMINE and TIMER databases. As presented in Fig. 1A and B, KLB expression was lower in a majority of cancer tissues, including breast cancer, colorectal cancer, lung cancer and pancreatic cancer, compared with their corresponding normal tissues. However, KLB expression was markedly increased in HCC tissues compared to the corresponding normal tissues (Fig. 1B). Differential-expression analysis of the TCGA-liver hepatocellular carcinoma data using paired and unpaired t-tests revealed that KLB expression was significantly increased in the HCC tissues compared with the normal tissues (Fig. 1C and D). Immunoblotting and IHC staining were conducted to assess the protein levels of KLB in 10 human HCC tissues and their corresponding adjacent non-cancerous tissues, to further confirm KLB expression in HCC. The results revealed that KLB protein levels were markedly higher in the HCC tissues (Fig. 1E), which is consistent with previously published results (11). In addition, KLB expression was also evaluated in various HCC cell lines (LM3, HepG2, PLC/PRF/5, Li-7 and Huh 7) and a normal hepatocyte cell line (MIHA), and the results suggested that the mRNA and protein levels of KLB were significantly upregulated in Huh7, HepG2 and Li-7 compared with MIHA (Fig. 1F and G). Furthermore, KLB expression was observed to gradually increase with tumor grade (Fig. 1H). Altogether, these findings confirmed that KLB expression is high in HCC tissues and cell lines.

To determine the clinical significance of increased KLB expression in HCC, a survival analysis was performed using Kaplan-Meier plotter. It was indicated that, although KLB is elevated in HCC, its high expression was associated with a favorable prognosis in non-alcohol-consuming patients with HCC (Fig. 1I), thus implying that KLB may serve as a protective modulator in HCC. These results suggest that high expression of KLB may be considered a prognostic biomarker for a specific population of patients with HCC. Further survival analysis of KLB gene expression in non-alcohol-consuming HCC patients with and without vascular invasion revealed that KLB was a specific prognostic marker for patients with HCC without vascular invasion (Fig. S1A and B). To further determine the expression of KLB in HCC development, KLB expression was analyzed according to tumor grade, disease stage, TP53 mutation status and lymph node metastasis status. As presented in Fig. 1J, KLB expression increased continuously with HCC development, suggesting that KLB has an important role in HCC development and metastasis.

To elucidate the role of KLB in HCC progression, a KLB-knockdown Huh7 cell line was generated (Fig. S1C). Consistent with the results of a previous study (11), the EdU, apoptosis, CCK-8 and colony-formation assays revealed that KLB knockdown reduced cell proliferation and induced cell apoptosis in Huh7 cells (Fig. 2A-D). These findings demonstrate the involvement of KLB in the pathogenesis of HCC and highlight the need to investigate its role in enhancing the FGF19-FGFR4 signaling pathway. However, wound-healing migration, Transwell migration and Matrigel invasion assays in Huh7 and Hepa1-6 cell lines revealed that KLB knockdown enhanced HCC cell motility (Fig. 2E-H). This may partly explain why high expression of KLB is associated with a favorable prognosis in patients with HCC without vascular invasion.

GSVA was performed to further explore the underlying biological function of KLB in HCC and the results demonstrated that KLB is closely associated with β-catenin signaling (Fig. 3A). In addition, it was found that KLB knockdown significantly inhibited the mRNA and protein expression of several critical genes in the β-catenin signaling pathway, including β-catenin, C-MYC, MMP7, transcription factor (TCF)-7 and phosphorylated glycogen synthase kinase 3β-S⁹ in Huh7 cells; however, the expression of nuclear β-catenin was not obvious (Fig. 3B and C). These results further verified the accuracy of the enrichment analysis, thus suggesting that KLB is possibly involved in the β-catenin-mediated signaling activation in HCC. IHC staining of the HCC tissues and their corresponding normal tissues revealed a high correlation between KLB and β-catenin gene (CTNNB1) expression, consistent with the results of the gene expression profiling interactive analysis (Fig. 3D and E).

Aberrant FGF19-FGFR4-KLB signaling is a metastatic driver for HCC (10), while the function of FGF21-KLB signaling in HCC metastasis remains to be clarified. Therefore, to explore the role of FGF21-KLB signaling in regulating HCC metastasis, Huh7 cells were transduced with lentivirus carrying the FGF21 gene to induce stable expression of FGF21 (Fig. S1D). Several studies demonstrated that β-catenin expression increased, while E-cadherin expression decreased in diabetic FGF21-knockout mice (24). In the present study, it was found that enhanced FGF21 expression inhibited the transcription of AXIN2, CTNNB1, C-MYC and TCF7. However, KLB knockdown in FGF21-overexpressing cells (Fig. S1E and F) caused increased expression of CTNNB1 and TCF7 (Fig. 4A and B), suggesting that FGF21 is an upstream mediator of the β-catenin signaling pathway and that its regulation is dependent on KLB expression. Thereafter, it was examined whether FGF21-KLB induced HCC metastasis via the EMT and it was found that FGF21 overexpression enhanced the protein level of E-cadherin (essential for the establishment of stable adherent junctions), which is downregulated during EMT (Fig. 4D). Subsequently, the mRNA levels of E-cadherin and several EMT inducers, including N-cadherin, Slug and zinc finger E-box binding homeobox 1 (ZEB1), were measured in both FGF21-overexpressing and sh-KLB-treated FGF21-overexpressing cells (Fig. 4C). Compared with FGF21-overexpressing cells, sh-KLB-treated FGF21-overexpressing cells had reduced mRNA levels of E-cadherin and increased mRNA levels of Slug and ZEB1 (key transcriptional repressors of E-cadherin). Furthermore, compared with the FGF21-overexpressing cells, claudin-1 [a repressor of E-cadherin (25)] was increased, while zonula occludens (ZO)−1 (a member of the ZO proteins that constitute tight junctions) was decreased in the sh-KLB-treated FGF21-overexpressing cells (Fig. 4E). Meanwhile, IF staining revealed an upregulation of vimentin expression in the sh-KLB-treated FGF21-overexpressing cells compared with the FGF21-overexpressing cells (Fig. 4F). Furthermore, the wound-healing migration, Transwell migration and Matrigel invasion assays revealed that the sh-KLB-treated FGF21-overexpressing cells had an enhanced migration and invasion ability compared with the FGF21-overexpressing cells (Fig. 4G,H). To confirm the effects of FGF21-KLB on the tumorigenicity of HCC in vivo, a lung metastasis model was established. Consistent with previous results, sh-KLB-treated FGF21-overexpressing cells had an enhanced in vivo metastatic ability compared with the FGF21-overexpressing cells (Fig. 4I and J). Taken together, these results suggest that FGF21 suppresses HCC cell metastasis through the β-catenin signaling pathway, while KLB knockdown promotes HCC cell motility under FGF21 overexpression.

The methylation level of the KLB promoter decreased gradually with an increase in tumor grade and disease stage and was associated with TP53 mutation status and lymph node metastasis status (Fig. S1H). It was hypothesized that the epigenetic modification of KLB may contribute to its high expression in HCC. A previous study demonstrated that HDAC3 catalyzes the acetylation of Klotho, a paralog of KLB (26); however, its role in KLB modification has not yet been determined. Therefore, to determine the regulatory effects of epigenetic changes on KLB expression, HCC cells were treated with TSA (a broad-spectrum HDAC inhibitor) and RGFP966 (an HDAC3-specific inhibitor). The results demonstrated that KLB expression decreased in a dose-dependent manner in TSA and RGFP966-treated HCC cells (Fig. 5A and B). Furthermore, it was found that the KLB promoter was indeed acetylated and that its acetylation was significantly increased after treatment with TSA and RGFP966, suggesting that KLB is acetylated in HCC (Fig. 5C and D).

Thereafter, a co-IP assay was performed to determine whether HDAC3 is able to regulate the acetylation status of KLB. The si-HDAC3-transfected Huh7 cells (Fig. S1G) were subjected to IP analysis using anti-KLB, followed by WB analysis with anti-acetylated-lysine antibodies, to detect the levels of acetylated KLB. The results revealed that KLB acetylation was significantly increased in the si-HDAC3-treated group compared with the untreated group (Fig. 5E). Subsequently, to determine the functional significance of HDAC3-mediated acetylation of KLB, the effects of decreased HDAC3 on KLB-knockdown and KLB-overexpressing cells were analyzed. As presented in Fig. 5F, treatment with either si-KLB or HDAC3 inhibitor increased the expression of vimentin and MMP9, while their combined application resulted in increased vimentin and MMP9 expression. However, these effects were reversed by KLB overexpression. Furthermore, the HDAC3 inhibitor induced a significant decrease in ZO-1 expression and an increase in vimentin expression, which was able to be inhibited by FGF21 overexpression (Fig. 5G). In addition, the wound-healing experiments demonstrated that the addition of HDAC3 inhibitor significantly enhanced HCC cell migration under FGF21 overexpression (Fig. 5H). Based on the above results, it was hypothesized that HDAC3 inhibitor inhibits KLB expression by aberrant acetylated modification, resulting in the blockade of FGF21-KLB signaling, thereby increasing the expression of EMT-inducing genes.