KDM5B promotes self‐renewal of hepatocellular carcinoma cells through the microRNA‐448–mediated YTHDF3/ITGA6 axis

Abstract Histone methylation plays important roles in mediating the onset and progression of various cancers, and lysine‐specific demethylase 5B (KDM5B), as a histone demethylase, is reported to be an oncogene in hepatocellular carcinoma (HCC). However, the mechanism underlying its tumorigenesis remains undefined. Hence, we explored the regulatory role of KDM5B in HCC cells, aiming to identify novel therapeutic targets for HCC. Gene Expression Omnibus database and StarBase were used to predict important regulatory pathways related to HCC. Then, the expression of KDM5B and microRNA‐448 (miR‐448) in HCC tissues was detected by RT‐qPCR and Western blot analysis. The correlation between KDM5B and miR‐448 expression was analysed by Pearson's correlation coefficient and ChIP experiments, and the targeting of YTH N6‐methyladenosine RNA binding protein 3 (YTHDF3) by miR‐448 was examined by luciferase assay. Additionally, the effect of KDM5B on the proliferation, migration, invasion and apoptosis as well as tumorigenicity of transfected cells was assessed using ectopic expression and depletion experiments. KDM5B was highly expressed in HCC cells and was inversely related to miR‐448 expression. KDM5B demethylated H3K4me3 on the miR‐448 promoter and thereby inhibited the expression of miR‐448, which in turn targeted YTHDF3 and integrin subunit alpha 6 (ITGA6) to promote the malignant phenotype of HCC. Moreover, KDM5B accelerated HCC progression in nude mice via the miR‐448/YTHDF3/ITGA6 axis. Our study uncovered that KDM5B regulates the YTHDF3/ITGA6 axis by inhibiting the expression of miR‐448 to promote the occurrence of HCC.


| INTRODUC TI ON
Hepatocellular carcinoma (HCC) is the leading type (~90%) of primary liver cancer, having an incidence of about 850 000 cases every year and causing many deaths worldwide. 1,2 Chronic HBV/HCV infection, alcohol, cigarette smoking, fatty liver and diabetes are all risk factors for HCC. 3 Most HCC patients are diagnosed at an advanced stage, and only 30% of HCC cases can be treated by resection, leading to dismal prognosis. 4 In addition to supporting early diagnosis, the advent of novel molecular biomarkers could improve the survival rates of the patients with advanced HCC. 5 For instance, MYB protooncogene like 2 and lysine-specific demethylase 5B (KDM5B) regulate hub genes obtained via the Gene Expression Profiling Interactive Analysis (GEPIA) database that are related to the poor prognosis of patients with HCC. 6 However, the underlying molecular mechanisms of the HCC progression still have not been fully elucidated, 7 which drew our attention to explore novel therapies targeting HCC-specific molecular disorders for patients with HCC. 8 KDM5B is a histone demethylase contained in the JmjC domain, which is capable of demethylating tri-and dimethyl modifications of H3 lysine 4. 9 It is overexpressed in multiple cancers, including stomach cancer, glioma and breast cancer. 10 KDM5B has emerged as a cancerogenic factor in HCC, as demonstrated by promoted HCC cell proliferation and colony formation. 11 In addition, microRNA-448 (miR-448) has been reported to have low expression in HCC samples and to delay epithelial-mesenchymal transition and invasion in HCC cells by down-regulating rho-associated kinase 2 (ROCK2). 12 Accumulating evidence indicates that KDM5B knockdown could increase miR-448 expression to impede the growth of papillary thyroid cancer cells. 13 Furthermore, a bioinformatic analysis in our present study suggests that miR-448 may target YTH N6-methyladenosine RNA binding protein 3 (YTHDF3). YTHDF3 is known as a member of the readers of RNA methylation (m6A) family, and it is reported to be highly expressed in HCC cells. 14,15 Hence, we speculated that the demethylase KDM5B may inhibit the expression of miR-448 to mediate YTHDF3, thus contributing to the occurrence of HCC. In support of this hypothesis, we transfected HCC cells with a series of mimics, inhibitors and short hairpin RNAs (shRNAs) to analyse in detail the potential molecular mechanisms regarding the KDM5B/ miR-448/YTHDF3 axis in HCC development.

| Tissue sample collection
Twelve pairs of fresh HCC tissues and corresponding normal tissues were obtained from patients with HCC. The fresh tissues were frozen by immersion in liquid nitrogen for later mRNA analysis. Enrolled patients had no previous radiotherapy or chemotherapy history and were followed up until death or the end of the study, which ranged from 3 to 59 months. All samples used in this study underwent histopathological examination.

| Cell culture and transfection
Hep3B, SMMC7721 and HEK293T cells were purchased from American Type Culture Collection (ATCC) and cultured with Roswell Park Memorial Institute (RPMI)-1640 medium (Gibco Company) containing 10% foetal bovine serum (FBS) (Gibco), 10 μg/mL streptomycin and 100 U/mL penicillin in an incubator (Thermo Fisher Scientific Inc) at 37°C with 5% CO 2 . HCC cells in the logarithmic growth phase were seeded into 6-well culture plates at a density of 4 × 10 5 cells/well. When reaching 80%-90% confluence, the cells were transfected with miR-448 mimic and miR-448 inhibitor as well as their corresponding negative controls (NCs) (mimic NC and inhibitor NC) according to the instructions of the Lipofectamine 2000 reagents (11668-019; Invitrogen). The sequences were designed by and plasmids purchased from Shanghai GenePharma Co, Ltd.
HEK293T cells were transfected with KDM5B overexpression vector (oe-KDM5B), YTHDF3 overexpression vector (oe-YTHDF3), ITGA6 overexpression vector (oe-ITGA6) or NC of overexpression vector (oe-NC) produced by Shanghai GenePharma Co, Ltd., through the packaging virus and the target vector, and the supernatant was collected after 48 hours of culture. Exponentially replicating viruses in the supernatant were collected, and the cells were infected with oe-NC, oe-KDM5B, oe-YTHDF3, oe-ITGA6 and oe-KDM5B + sh-ITGA6. Then, the cells in the logarithmic growth phase were detached with trypsin and isolated to obtain suspension of 5 × 10 4 cells/mL, which was seeded in a 6-well plate (2 mL per well) for incubation at 37°C overnight.

| Reverse transcription quantitative polymerase chain reaction (RT-qPCR)
Total RNA was extracted using TRIzol reagents (15596026; Invitrogen) and reversely transcribed into complementary DNA (cDNA) according to the instructions of PrimeScript RT reagent Kit (RR047A; Takara Bio Inc). RT-qPCR was performed on the synthesized cDNA using the Fast SYBR Green PCR kit (Applied Biosystems Inc) and an ABI PRISM 7300 RT-qPCR system (Applied Biosystems). The relative expression of mRNA or miRNA was normalized to glyceraldehyde-3-phosphate dehydrogenase (GAPDH) or U6, respectively, and was calculated using the 2 −ΔΔCt method.
The primer design is shown in Table 2.

| Western blot analysis
Cells from each culture group were detached by trypsin, collected, and lysed with an enhanced radioimmunoprecipitation assay lysis buffer (Wuhan Boster Biological Technology Co., Ltd.) containing a protease inhibitor. After measurement of protein concentration with a bicinchoninic acid (BCA) quantitation kit (Boster Biological Technology), samples of protein were separated by 10% sodium dodecyl sulphate-polyacrylamide gel electrophoresis, and electrotransferred onto a polyvinylidene fluoride membrane. The membrane was blocked in 5% bovine serum albumin (BSA) at room temperature for 1 hour and added with diluted primary antibodies (antibody to KDM5B, ab181089; Abcam; antibody to YTHDF3, Cat # 25537-1-AP; Proteintech; antibody to ITGA6, Cat # 3750; CST) for incubation overnight at 4°C. The following day, the membrane was washed 3 times with Tris-buffered saline Tween-20, re-probed with horseradish peroxidase (HRP)-labelled secondary antibody of goat anti-rabbit for 1 hour at room temperature and developed with enhanced chemiluminescence working solution (EMD Millipore).
Finally, ImageJ analysis software was used to quantify the grey levels of each band in the Western blot image normalized to GAPDH.

| Dual-luciferase reporter gene assay
The dual-luciferase reporter gene vectors for 3′ untranslated region (UTR) of YTHDF3 and mutant plasmid of YTHDF3 with mutations in the binding sites with miR-448 were constructed:  Promega) was used to detect luciferase activity. A total of 100 μL Firefly luciferase working fluid and 100 μL Renilla luciferase working fluid were added into each cell sample with the Renilla luciferase as the internal standard, and the ratio of Firefly luciferase activity to Renilla luciferase activity represented the relative luciferase activity.

| Agarose colony formation assay
HCC cells of each group in the logarithmic growth phase were suspended in the culture medium and 1 mL 0.5% agarose was added to a 6-well plate and solidified at room temperature to prepare the bottom gel. Then, 500 μL cell suspension containing 5000 cells was mixed with 500 μL 0.5% agarose to prepare the upper gel, which was placed on the bottom and allowed to solidify at room temperature.
After coagulation, the sample was added into a 2 mL medium and cultured at 37°C with 5% CO 2 for 3 weeks. Cells were fixed with ethanol, stained with 0.1% crystal violet (C0004; Banmanbio Co, Ltd.) and then imaged with a microscope. The number of clones was counted with ImageJ software.

| Scratch test
Horizontal lines were drawn with a ruler and a marker at intervals of 0.5-1 cm on the bottom surface of a 6-well plate, with at least five lines passing through each well. Cells were added to the 6-well plate at a density of approximately 5 × 10 5 cells/well and were incubated overnight in a medium containing 10% FBS. A sterile 10 μL pipette was used to make scratches perpendicular to the horizontal lines.
The length of the wounds was measured under an optical microscope at 0 and 24 hours incubation, and images were collected under an inverted microscope to observe the cell migration in each group.

| Transwell assay
The apical chamber surface of the bottom membrane of the Transwell chamber was coated with Matrigel (BD Biosciences) which was polymerized into a gel at 37°C for 30 minutes, with hydration of the basal membrane before use. The cells were cultured in serum-free medium for 12 hours, followed by collection and resuspension of cells in serum-free medium (1 × 10 5 /mL). The lower chamber was added with 10% FBS, and 100 μL cell suspension was added into the Transwell chamber at 37°C for 24 hours. The cells that had not invaded the Matrigel membrane surface were gently removed with a cotton swab, and the remaining cells were fixed with 100% methanol and stained with 1% toluidine blue (Sigma-Aldrich Chemical Company). Stained invading cells in five randomly selected areas were manually counted under an inverted light microscope (Carl Zeiss).

| Chromatin immunoprecipitation assay (ChIP)
ChIP assay was performed using a ChIP kit (Millipore). After reaching 70%-80% confluence, cells collected from each group were added with 1% formaldehyde and fixed at room temperature for 10 minutes to induce DNA-protein cross-linking. Then, the cells were subjected to ultrasonic treatment to produce fragments of appropriate size. The fragments were centrifuged at 6540 g at 4°C with the supernatant collected into three tubes which were added with positive control antibody to RNA polymerase II, as well as NC antibodies to human immunoglobulin G and KDM5B (ab181089; Abcam) or H3K4me3 (ab12209; Abcam) for incubation at 4°C overnight.
Subsequently, Protein Agarose/Sepharose was used to precipitate the endogenous DNA-protein complex, with the supernatant discarded after centrifugation. The non-specific complexes were washed, and de-cross-linking was performed following incubation at 65°C overnight. Afterwards, phenol/chloroform was added to purify and recover DNA fragments. Finally, RT-qPCR was performed to examine the expression of miR-448.

| Xenograft tumour in nude mice
Thirty female BALB/c nude mice aged 3-4 weeks were purchased from Beijing Institute of Pharmacology, Chinese Academy of Medical Sciences, fed in a specific pathogen-free animal laboratory in separate cages with humidity of 60%-65% and a temperature of 22-25°C, and provided with free access to food and water under a 12-hour light/dark cycle. The experiment was started one week after adaptive feeding, before which the health status of mice was observed. The mice were randomly arranged into six groups according to the bodyweight with 5 mice in each group. HCC cells (5 × 10 6 cells/mouse) that had been transfected with oe-NC + sh-NC, oe-KDM5B + sh-NC, or oe-KDM5B + sh-ITGA6, or treated with NS, GSK-467 (a selective inhibitor of KDM5B) + oe-NC or GSK-467 + oe-ITGA6 were then subcutaneously implanted into the back of mice. The status of nude mice was monitored following the procedure, and tumour formation and growth were evaluated weekly.
After 4 weeks, the mice were killed under deep anaesthesia, tumours were removed, and their weight and volume were measured after back tumour was stripped, and the expression of miR-448 and ITGA6 was measured using RT-qPCR.

| Statistical analysis
The SPSS 21.0 version (IBM Corp.) was used for statistical analysis. Measurement data were presented as mean ± standard deviation. The data of cancer tissues and corresponding normal tissues were analysed by paired t test, and the data of the other two groups were analysed by unpaired t tests. Besides, the data among multiple groups were analysed by one-way analysis of variance (ANOVA) and Tukey's post hoc test, the data among groups at different time were analysed by repeated measures ANOVA followed by Bonferroni's post hoc test. Pearson's correlation coefficient was used to assess the correlation of indicators. Statistical significance was assumed when P < .05.

| KDM5B deletion inhibited self-renewal of HCC cells
A total of 3482, 3594 and 2525 differentially expressed genes were obtained after differential analysis of the microarray data sets GSE45267, GSE62232 and GSE11 7361, respectively, by R language. GEPIA analysis also obtained 6,987 LIHC differentially expressed genes in the TCGA database. In addition, 318 human transcription factors were obtained from Cistrome, and 4 important transcription factors (SOX4, SMARCC1, KDM5B and JUNB) in HCC were identified by taking the intersection of differentially expressed genes and transcription factors ( Figure 1A). A previous study has shown that KDM5B is related to HCC and shows a significantly increased expression in HCC tissue. 11 By extracting the expression data of the microarray data set GSE11 7361, we further confirmed that KDM5B is highly expressed in HCC ( Figure 1B).
GEPIA analysis suggested that its high expression predicted significantly lower survival of HCC patients ( Figure 1C)

| KDM5B inhibited miR-448 by demethylating H3K4me3
KDM5B is a transcription repressor, which contains histone demethylase active sites. 16 Based on the demethylase activity of KDM5B, we searched the ChIPBase database for targets that might be regulated by KDM5B. The results showed a potential KDM5B binding site at 763 bp upstream of miR-448 ( Figure 2A). It has been reported that miR-448 inhibits HCC by inhibiting the self-renewal of HCC cells. 12 We measured the expression of KDM5B and miR-448 using RT-qPCR in 12 HCC tissues and 12 corresponding normal tissues and analysed their correlation. The results showed that miR-448 had low expression in HCC tissues ( Figure 2B) and was negatively correlated with the KDM5B expression ( Figure 2C).
Next, we examined the expression of miR-448 in Hep3B (sh-NC) and Hep3B (sh-KDM5B) cell lines using RT-qPCR. The results indicated that miR-448 was distinctly up-regulated in the sh-KDM5Btreated Hep3B cell line ( Figure 2D). In order to investigate whether the demethylase activity of KDM5B directly inhibited the expression of miR-448, we first performed ChIP analysis of the miR-448 promoter with an antibody to KDM5B. RT-qPCR results suggested that KDM5B knockdown significantly reduced the binding region content of the miR-448 promoter. Then, we performed ChIP analysis of the miR-448 promoter with an antibody against H3K4me3.
We observed that the miR-448 promoter binding region was significantly enriched in the sh-KDM5B-treated Hep3B cells ( Figure 2E).

Meanwhile, we conducted ChIP experiments in Hep3B cells treated
with inhibitors targeting the activity of the KDM5B catalytic region.
Results showed that the miR-448 promoter region was also signifi-
Previous studies have shown that YTHDF3 and ITGA6 are both highly expressed in HCC tissues. 15,17 We predicted through microRNA analysis that the 3′UTR of YTHDF3 may contain a site that binds to miR-448, and determined that YTHDF3 and ITGA6 were highly expressed in HCC by StarBase analysis ( Figure 3B,C), showing significant positive correlation ( Figure 3D). MEM prediction also showed significant coexpression of YTHDF3 and ITGA6 ( Figure 3E). Based on the above predictions, we speculated that miR-448 could target YTHDF3 and inhibit the YTHDF3/ITGA6 axis, thereby inhibiting the occurrence of HCC.
As shown in Figure 3F, the 3′UTR of YTHDF3 contains a possible binding site to miR-448. In order to verify this, we performed a luciferase assay. For this, we cloned WT and MUT 3′UTR of YTHDF3 into the pmirGLO plasmid, and cotransfected miR-448 mimic or NC mimic with YTHDF3-WT or YTHDF3-MUT into HEK293T cells. The results showed that the luciferase activity was significantly reduced after cotransfection with miR-488 and YTHDF3-WT but was not affected when cotransfected with YTHDF3-MUT, suggesting that miR-448 could indeed bind to YTHDF3 ( Figure 3G). In order to further verify the inhibitory effect of miR-448 on the YTHDF3/ITGA6 axis, we observed changes in the expression of YTHDF3 and ITGA6 upon overexpressing/ inhibiting miR-448 expression. RT-qPCR results indicated that overexpressed miR-448 could significantly inhibit the mRNA level of YTHDF3 and ITGA6. When miR-448 was inhibited, YTHDF3 and ITGA6 mRNAs were elevated to varying degrees ( Figure 3H), which was confirmed by Western blot analysis ( Figure 3I). Next, we cotransfected sh-KDM5B and miR-448 inhibitor into cells and found that the altered YTHDF3 and ITGA6 expression caused by the sh-KDM5B treatment alone was rescued after cotransfection of sh-KDM5B and miR-448 inhibitor ( Figure 3J,K), demonstrating that KDM5B affected downstream molecular pathways by regulating miR-448. Therefore, miR-448 targets the YTHDF3 and to suppress the YTHDF3/ITGA6 axis.

| miR-448 inhibited self-renewal of HCC cells by inhibiting YTHDF3/ITGA6 axis
A previous study has shown that miR-448 could inhibit the selfrenewal of HCC cells and thus suppress the occurrence of HCC. 18 It was also confirmed in the above-mentioned experiments that miR-448 could target YTHDF3 to regulate the function of ITGA6. In order to The data were measurement data, which were presented as mean ± standard deviation. The data in two groups were compared using unpaired t test, and the data in multiple groups were compared using one-way ANOVA and Tukey's post hoc test. *P < .05 compared with treatment of sh-NC

| KDM5B inhibitor GSK-467 depressed hepatocarcinogenesis in vitro and in vivo
We The data were measurement data, which were presented as mean ± standard deviation. The data of tumour tissues and corresponding normal tissues were compared using unpaired t test, and the data in the other two groups were compared using unpaired t test. Pearson's correlation coefficient was used to assess the correlation of indicators. *P < .05 when compared with treatment of sh-NC Results showed that nude mice treated with GSK-467 had smaller tumours and slower tumour growth rate, which was annulled by oe-ITGA6 ( Figure 6E,F). RT-qPCR results showed that miR-448 expression in mice treated with GSK-467 was increased, whereas F I G U R E 3 miR-448 targets YTHDF3/ITGA6 axis. A, Venn diagram predicting downstream genes of miR-448 based on TargetScan, StarBase, mirDIP, miRDB and microRNA. B, Box plot of YTHDF3 expression after obtained by StarBase, wherein the left red boxes indicate its expression in HCC samples, and the right purple boxes represent its expression in normal samples. C, Box plot of ITGA6 expression obtained by StarBase, wherein the left red boxes represent its expression in HCC samples, and the right purple boxes represent its expression in normal samples. D, Expression correlation diagram between YTHDF3 and ITGA6 in HCC obtained by StarBase. E, MEM analysis displaying a significant co-expression relationship between YTHDF3 and ITGA6. F, Binding sites between miR-448 and YTHDF3. G, Luciferase activity after cotransfection of miR-448 mimic and pmirGLO-YTHDF3-WT or pmirGLO-YTHDF3-MUT relative to Renilla luciferase activity. H, RT-qPCR determination of mRNA level of YTHDF3 and ITGA6 in HCC cells after miR-448 mimic and miR-448 inhibitor treatment. I, Western blot analysis of the protein level of YTHDF3 and ITGA6 in HCC cells normalized to GAPDH after miR-448 mimic and miR-448 inhibitor treatment. J, Detection of YTHDF3 and ITGA6 mRNA expression in HCC cells after cotransfection of sh-KDM5B and miR-448 inhibitor by RT-qPCR. K, Detection of YTHDF3 and ITGA6 protein expression in HCC cells after cotransfection of sh-KDM5B and miR-448 inhibitor by Western blot analysis. The data were measurement data, which were presented as mean ± standard deviation. The data in two groups were compared using unpaired t test, and the data in multiple groups were compared using one-way ANOVA and Tukey's post hoc test. *P < .05 when compared with treatment of sh-NC F I G U R E 4 miR-448 inhibits self-renewal of HCC cells by suppressing the YTHDF3/ITGA6 axis. A, RT-qPCR determination of YTHDF3 and ITGA6 expression in Hep3B cells after oe-YTHDF3 and oe-ITGA6 treatment. B, Formation of cell spheroids after oe-YTHDF3, oe-ITGA6 and miR-448 mimic treatment. C, Colony-forming ability assessed by the colony formation assay after oe-YTHDF3, oe-ITGA6 and miR-448 mimic treatment. D, Migratory capacity assessed by the scratch test after oe-YTHDF3, oe-ITGA6 and miR-448 mimic treatment. E, Invasive ability assessed by the Transwell assay after oe-YTHDF3, oe-ITGA6 and miR-448 mimic treatment. The data were measurement data, which were presented as mean ± standard deviation. The data in two groups were compared using unpaired t test, and the data in multiple groups were compared using one-way ANOVA and Tukey's post hoc test. *P < .05 when compared with treatment of oe-NC. & P < .05 when compared with treatment of oe-NC/oe-YTHDF3/oe-ITGA6 + mimic NC

| D ISCUSS I ON
HCC is a major cause of the cancer-induced deaths and poses an increasing burden to many areas around the globe, especially in countries located in Africa and Asia where medical resources are limited. 3 As HCC patients are often diagnosed at the advanced stage, the prognosis is dismal, 8 which urges scientists to explore novel therapeutic targets to improve the survival of patients with HCC. The histone demethylase KDM5B is highly expressed in several cancers including lung, stomach, breast and hepatic cancers, 11 making it a potential therapeutic target for HCC. Hence, we experimentally confirmed the possible oncogenic function of KDM5B in HCC and found that KDM5B regulates the YTHDF3/ITGA6 axis by inhibiting the expression of miR-448 to promote the occurrence of HCC (Figure 7).
Initially, we determined that silencing of KDM5B hindered the self-renewal, invasion, migration and proliferation of HCC cells. The demethylase KDM5B is known as a potential therapeutic target for cancer treatment, based on its action as a key regulator in important biological processes of cancers such as tumorigenesis, progression and antibiotic resistance, and based on its ability to block the activation of H3K4me3. 19 KDM5B was found previously to be highly expressed in HCC tissues, and its up-regulation could accelerate the growth of Hep3B cells and serve as a prognostic marker for HCC. 11 Moreover, poor expression of KDM5B correlates with better outcome in HCC. 20 By arresting the cell cycle at the G1/S phase through up-regulation of p15 and p27, KDM5B inhibits HCC cell proliferation both in vivo and in vitro. 9 In addition, a prior study elaborated that KDM5B depletion caused reduction in expression of self-renewalrelated genes in trophoblast stem cells, which was partially coincided with our results. 21 Also, another work uncovered that KDM5B silencing contributed to repression of gastric cancer cell proliferation, migration and invasion in vitro. 22 These works partially supported our results about the oncogenic role of KDM5B in HCC.
In addition, we found that miR-448 expression had low expres-  It has been well-established that miRNAs can interact with the 3′UTR of specific target mRNAs and consequently inhibit their expression. 24 In this study, the biological prediction website and luciferase reporter assay results identified that miR-448 bound to the 3′UTR of YTHDF3 mRNA and could negatively regulate its expression. Furthermore, our findings indicated a positive correlation be-  18,29 and that activation of the AMPK pathway inhibits the malignant phenotypes of lung cancer. 30 Therefore, we will focus in future studies on the role of different signalling pathways involved in the promotive effect of KDM5B on HCC through miR-448-mediated effects on the YTHDF3/ITGA6 axis, aiming to identify new therapeutic targets for treating this disease.

This study was supported by Hainan Provincial Basic and Applied
Basic Research Program (Natural Science Field) High-level Talent Project in 2019 (No. 2019RC373).

CO N FLI C T S O F I NTE R E S T
The authors declare no conflicts of interest.
F I G U R E 7 Molecular mechanism of demethylase KDM5B in the occurrence of HCC. KDM5B inhibits miR-448 expression, and miR-448 inhibits YTHDF3/ITGA6 axis, eventually leading to accelerated self-renewal of HCC cells

DATA AVA I L A B I L I T Y S TAT E M E N T
The authors confirm that the data supporting the findings of this study are available within the article.