LncRNA CRNDE decreases the sensitivity of sorafenib in HCC cells via promoting ATG4B-mediated autophagy

Background: Autophagy has been conrmed to be closely related to the growth and drug resistance of cancer cells, and autophagy related 4B (ATG4B) performs a crucial role in the process of autophagy. lncRNA CRNDE (colorectal neoplasia differentially expressed) promotes the growth of hepatocellular carcinoma (HCC) cells, but it is unclear whether the tumor-promoting effect of CRNDE is associated with the regulation of ATG4B and autophagy. Methods: The expression of CRNDE and ATG4B in HCC tissues and cells was detected by quantitative real-time PCR (qPCR) and Western blot. The effect of CRNDE on autophagy of HCC cells was examined by uorescent assay, GFP-LC3 analysis and Western blot. The relationship among CRNDE, miR-543 and ATG4B was claried using bioinformatic analysis, qPCR, Western blot and dual-luciferase reporter assay. The effect of CRNDE on sensitivity of sorafenib in HCC cells was evaluated in vitro and in vivo by CCK-8 assay, Western blot, Hoechst staining, ow cytometry, IHC staining and TUNEL assay. Results: CRNDE induced autophagy via upregulating ATG4B in HCC cells. Mechanistically, CRNDE enhanced the stability of ATG4B mRNA by suppressing miR-543, leading to the increase of ATG4B expression and autophagy in HCC cells. Moreover, sorafenib dramatically elevated CRNDE and ATG4B as well as autophagy. Knockdown of CRNDE signicantly sensitized HCC cells to sorafenib in vitro and in vivo . Conclusion: The study has revealed the presence of “CRNDE/ATG4B/autophagy” pathway which alleviates the sensitivity of sorafenib in HCC cells, suggesting that this pathway may serve as a novel target to improve the killing effect of sorafenib in HCC treatment.


Background
Hepatocellular carcinoma (HCC) is one of the most common malignant tumor in human digestive system with high incidence and low survival rate [1,2]. The main strategies for HCC treatment with proven survival bene t include surgical resection, ablation, transplantation, chemotherapy, and so on [3]. In recent years, therapeutic reagents are increasingly used to treat HCC, but the e cacy tends to be limited due to acquired drug resistance [4,5]. Autophagy, a lysosome-mediated conserved process which degrades and recycles cellular proteins and organelles in response to starvation or cellular stresses, plays an essential part in maintaining intracellular homeostasis. Mounting studies have veri ed that autophagy attenuates the sensitivity of therapeutic drugs such as sorafenib, 5-uorouracil (5-FU) and oxaliplatin, ultimately protecting cancer cells from death [6][7][8]. Therefore, it is of great signi cance to understand the speci c regulating mechanisms of autophagy in drug resistance and interfere with protective autophagy in order to improve the drug sensitivity in HCC.
Autophagy related 4B (ATG4B), a member of the autophagin protein family, performs a critical role in regulation of autophagy in mammalian cells. In the process of autophagosome formation, C-terminal arginine residue of the cytoplasmic light chain 3 (LC3) is proteolytically cleaved to generate LC3-I, which subsequently conjugates with phosphatidylethanolamine and converts to membrane-bound LC3-II [9]. The cysteine protease ATG4 is essential for activation of LC3 precursor and delipidation of LC3-II. Among the four isoforms of ATG4 (4A, 4B, 4C and 4D), ATG4B exhibits the dominant proteolysis activity in autophagic homeostasis [10]. Recent studies have shown that ATG4B promotes the proliferation and progression of various tumor cells, and enhances the drug resistance of HCC cells through protective autophagy [11][12][13][14]. However, it is not well known about the molecular mechanisms underlying the regulation of ATG4B in HCC cells.
Long noncoding RNAs (lncRNAs), a subgroup of non-encoding transcripts longer than 200 nucleotides in length, perform functions of regulating gene expression directly or indirectly at several levels such as transcriptional, post transcriptional, translational and epigenetic regulation [15]. An increasing number of lncRNAs have been reported to play a vital role in the carcinogenesis and development of HCC [16,17]. As a new member of lncRNA family, CRNDE (colorectal neoplasia differentially expressed) is initially identi ed as the speci cally upregulated lncRNA in human colorectal cancer, and afterwards con rmed to be highly expressed in many other human malignancies including HCC [18][19][20]. CRNDE promotes the growth and invasion of HCC cells by regulating multiple signalling pathways, such as PI3K/Akt, BCAT1, MAPK and Wnt/β-catenin signalling pathways [21][22][23][24], suggesting that CRNDE may act as an oncogenic lncRNA which could be served as a potential target for HCC therapy [25]. Nevertheless, whether CRNDE is associated with ATG4B and autophagy, and the roles of CRNDE in HCC cells in response to therapeutic agents remain unclear.
In the present study, we for the rst time demonstrated that CRNDE induced autophagy of HCC cells via upregulating ATG4B. Mechanistic research showed that CRNDE enhanced the stability of ATG4B mRNA by suppressing miR-543, thus elevating ATG4B and autophagy in HCC cells. Furthermore, sorafenib promotes CRNDE and ATG4B as well as autophagy, while inhibiting the "CRNDE/ATG4B/autophagy" pathway sensitized HCC cells to sorafenib. The in vivo experiments in nude mice showed that knockdown of CRNDE dramatically strengthened the anti-HCC effect of sorafenib. Together, these ndings suggest that the pathway "CRNDE/ATG4B/autophagy" may be a promising target to increase the sensitivity of sorafenib against HCC.

Clinical specimens
Human HCC specimens and the adjacent noncancerous tissues were collected from Southwest Hospital of Army Medical University (Third Military Medical University) (Chongqing, China).
TCGA(The Cancer Genome Atlas) data HCC RNA-Seq data and the related clinical data were from TCGA database (http://cancergenome.nih.gov/).

Cell culture
Human HCC cell lines PLC, HepG2 and Hep3B were purchased from the American Type Culture Collection (Manassas, VA, USA). Huh7, SMMC7721 cell lines and hepatic cell line THLE-3 were from the Cell Bank of Chinese Academy of Sciences (Shanghai, China). The cells were cultured in high-glucose DMEM (Gibco, Carlsbad, CA, USA) containing 10% FBS at 37 ℃ in a humid incubator with 5% CO 2 .
Quantitative real-time PCR (qPCR) Total RNA from HCC cells or tissues was extracted using the total RNA extraction kit (BioFlux, Hangzhou, China) according to the manufacturer's instructions. For nuclear/cytoplasmic separation assay, cytoplasmic and nuclear RNA were separately isolated using PARIS kit (Invitrogen, Carlsbad, CA, USA) according to the manufacturer's protocol. The RNA was reversely transcribed to rst-strand cDNA using PrimeScript RT reagent kit (Takara, Dalian, China). qPCR was performed with SYBR qPCR master mix (Takara), taking GAPDH as the internal reference. Reverse transcription and qPCR of miRNAs were conducted with All-in-One miRNA qRT-PCR detection kit (GeneCopoeia, Guangzhou, China), taking U6 RNA as the internal reference. The primers were listed in Supplementary Table 1.

Western blot
Total proteins from HCC cells or tissues were extracted with RIPA lysis buffer (Beyotime, Shanghai, China) and the concentrations were detected using BCA protein assay kit (Beyotime). Then Western blot was performed as previously described [6]. The primary antibodies anti-ATG4B and anti-GAPDH were from Proteintech (Chicago, IL, USA), anti-PARP and anti-SQSTM1/p62 were from Cell Signaling Technology (CST, Beverly, MA, USA), and anti-LC3 was from Sigma-Aldrich (St Louis, MO, USA).

Construction of plasmids
The DNA fragments encoding the wild type and the mutant human CRNDE containing mutations of the predicted miR-543 binding site (1008 CTTTATTGGATTGAATGAATGTTT 1031, the underlined nucleotides were mutated) were chemically synthesized by Sangon Biotech (Shanghai, China), and separately inserted into pcDNA3.1(+) (pcDNA3.1) expression vector (Invitrogen, Carlsbad, CA, USA) after digestion with EcoR (Takara) and BamH (Takara). The reconstructed plasmids were named as pcDNA-CRNDE and pcDNA-CRNDE-mut, respectively. The DNA fragments encoding the wild type and the mutant 3 -UTR of human ATG4B mRNA containing mutations of the putative miR-543 binding site (1565 TGTCAGACACAGACATGAATTTCT 1588, the underlined nucleotides were mutated) were separately synthesized by Sangon Biotech, digested with Sac (Takara) and Sal (Takara), and then cloned into pmir-GLO reporter vector (Thermo Scienti c, Waltham, MA, USA). The reconstructed plasmids were named as pmir-ATG4B and pmir-ATG4B-mut, respectively. The overexpression plasmid pCMV-ATG4B was bought from Lab Cell Biotechnology (Chongqing, China).

Transfection assay
The siRNAs targeting human CRNDE, ATG4B and control siRNA were synthesized by Lab Cell Biotechnology. Mimics and inhibitor of miRNAs, and the corresponding negative controls were purchased from GeneCopoeia. The cells were transfected using Lipofectamine 3000 (Invitrogen) according to the manufacturer's instructions. The sequences of siRNA were presented in Supplementary Table 2, and the sequences of miRNA mimics and inhibitor were listed in Supplementary Table 3.

Fluorescent assay of autophagy
Autophagy in live cells was analyzed using the Cyto-ID autophagy detection kit (Enzo, NY, USA) with proprietary probes speci cally staining autophagosomes according to the manufacturer's protocol.
Brie y, the cells were harvested, resuspended in 500 μL of freshly diluted Cyto-ID green detection reagent and incubated at room temperature for 30 min in dark. Then uorescence intensity was analyzed using a ow cytometer (Beckman, CA, USA).

RNA FISH ( uorescence in situ hybridization)
The Cy3-labelled probes for RNA FISH of CRNDE, 18S RNA and U6 RNA were synthesized by RiboBio (Guangzhou, China). RNA FISH was performed using the uorescent in situ hybridization kit (RiboBio) according to the manufacturer's instructions. DAPI was used for counterstaining of nuclei. The cells were photographed under a uorescence microscope (OlympusIX81, Tokyo, Japan).

RIP (RNA immunoprecipitation) assay
RIP assay was performed using the EZ-Magna RIP kit (Millipore/Merck, Darmstadt, German) according to the manufacturer's protocol. Brie y, cells were lysed and incubated with antibody-coated beads at 4 ℃ overnight. Subsequently, the co-immunoprecipitates were treated with proteinase K at 55 ℃ for 30 min.
Dual-luciferase reporter assay HCC cells were seeded into 24-well plates and co-transfected with the indicated luciferase reporter plasmids and miRNA mimics, siRNA, expression plasmid, or the corresponding control using Lipofectamine 3000 (Invitrogen) for 24 h. Then the cells were lysed, and the reporter assay was performed using the dual-luciferase reporter assay kit (GeneCopoeia) according to the manufacturer's instructions.

Cell viability assay
The cell viability assay was performed using cell counting kit-8 (CCK-8; Beyotime) according to the manufacture's protocol. Brie y, the cells were seeded into 48-well plates, followed by different treatments. 20 μL CCK-8 reagent was added to each well and incubated at 37 ℃ for 1 h. Then the OD values at 450 nm were examined using a microplate reader (Molecular Devices, Sunnyvale, CA, USA).

Cell apoptosis detection
Cell apoptosis was detected using the following methods: (1) The cells were xed with 4% paraformaldehyde for 10 min and stained with Hoechst 33258 (Beyotime) for 10 min in dark. Then the cells were photographed under a uorescence microscope (OlympusIX81) and the apoptotic cells were characterized by the nuclear morphology changes. (2) The cells were harvested and stained with Annexin-/FITC and PI (BD, California, USA) at room temperature for 15 min in dark. The apoptotic cells were analyzed under a ow cytometer (Beckman).

Animal experiments
Male nude mice (six weeks old) were obtained from Beijing Huafukang Bioscience (Beijing, China). Lentivirus system was used to construct the HepG2 cell lines with (or without) stable knockdown of CRNDE (LV-shCRNDE or LV-NC) as previously described [6]. 5×10 6 cells in 0.1 mL PBS (for each mouse) were subcutaneously injected into the right ank of each mouse (n = 10 per group). One week after inoculation, the mice in each group were randomized into two subgroups (n=5 per subgroup) and given daily administration of sorafenib (Selleckchem, Houston, USA, 30 mg/kg) or vehicle control by gavage. The size of xenograft tumors was measured every three days and the volume was calculated with the following formula: volume = width 2 × length × 1/2. Twenty-ve days after inoculation, the mice were sacri ced and the tumors were excised for photographing and weighting. The cell apoptosis in the xenograft tumors was determined using TUNEL staining, and the levels of corresponding RNAs and proteins were analysed by immunohistochemical (IHC) staining, qPCR and Western blot, respectively.

IHC staining and TUNEL assay
The xenograft tumor tissues were xed with 4% paraformaldehyde, and then embedded in para n and sectioned. Next, the tumor sections were immunostained using histostain-plus kit (Zhongshan Biotechnologies) according to the manufacturer's instructions, counterstained with hematoxylin and photographed under a microscope (OlympusIX81). For apoptosis analysis of the xenograft tumors, TUNEL assays were conducted according to the manufacturer's protocol (Zhongshan Biotechnologies) and the tumor sections were visualized using a laser confocal microscope (Carl Zeiss AG).

Statistical analysis
The data were presented as means ± SD. Two-tailed unpaired t-test was used for the comparisons between two independent groups and one-way analysis of variance (ANOVA) was for the comparisons among three or more groups. Pearson's test was employed for the analysis of the correlation between CRNDE and ATG4B levels. P<0.05 was considered statistically signi cant.

CRNDE upregulates ATG4B in HCC cells
To preliminarily explore the relationship between CRNDE and ATG4B, TCGA database was used to analyze the levels of CRNDE and ATG4B in HCC tissues. The results showed that the levels of CRNDE and ATG4B mRNA in HCC tissues were signi cantly higher than those in noncancerous liver tissues (Fig. 1a,   b). The correlation analysis revealed that the level of ATG4B was positively correlated to that of CRNDE in the HCC tissues (Fig. 1c). Furthermore, both CRNDE and ATG4B mRNA levels were associated with a shorter overall survival of HCC patients ( Supplementary Fig. S1a, b). The results of qPCR (Fig. 1d, e) and Western blot (Fig. 1f) showed that the levels of CRNDE as well as ATG4B mRNA and protein in HCC tissues were higher than those in the corresponding adjacent noncancerous liver tissues. In addition, the levels of CRNDE and ATG4B in 5 HCC cell lines (SMMC-7721, HepG2, Hep3B, Huh7 and PLC) were also higher than those in relatively normal hepatic cell line THLE-3 ( Supplementary Fig. S2a, b). To investigate whether ATG4B could be regulated by CRNDE, CRNDE was overexpressed and silenced in HCC cells ( Supplementary Fig. S3a, b). Overexpression of CRNDE markedly increased ATG4B at both mRNA and protein levels (Fig. 1g, h), whereas silence of CRNDE dramatically reduced the mRNA and protein levels of ATG4B in HCC cells (Fig. 1i, j). Taken together, these results indicate that CRNDE elevates ATG4B in HCC cells.

CRNDE promotes ATG4B mRNA stability
In order to inquiry the mechanism by which CRNDE elevates ATG4B mRNA and protein levels in HCC cells, RNA FISH and nuclear/cytoplasmic separation assay were used to detect the subcellular localization of CRNDE. As shown in Fig. 3a, b, the majority of CRNDE distributed in cytoplasm and the minority located in nucleus. It has been reported that lncRNAs could directly bind to the corresponding proteins to exert their functions [27], thus RIP was performed to verify whether CRNDE could directly bind to ATG4B protein. However, the result showed that CRNDE was not speci cally enriched in the coimmunoprecipitated RNAs with ATG4B (Fig. 3c). Subsequently, we examined the effect of CRNDE on the mRNA stability of ATG4B in HCC cells. In the presence of the transcription inhibitor actinomycin D (Act D), the level of ATG4B mRNA in CRNDE-overexpressed cells was signi cantly higher than that in the control cells (Fig. 3d). Conversely, the level of ATG4B mRNA in CRNDE-knockdown cells was markedly lower than that in the corresponding control (Fig. 3e). These ndings reveal that CRNDE upregulates ATG4B through increasing its mRNA stability in HCC cells.

miR-543 targets and represses ATG4B
Previous research shows that cytoplasmic lncRNA can function as a competing endogenous RNA (ceRNA) by sequestrating miRNA, thereby regulate the expression of speci c mRNA targets [28,29]. As CRNDE mainly locates in cytoplasm and in uences the stability of ATG4B mRNA, we hypothesized that miRNAs might participate in the CRNDE-mediated regulation of ATG4B in HCC cells. Thus, the miRNAs potentially binding to both CRNDE and ATG4B were predicted using online bioinformatic tools StarBase (http://starbase.sysu.edu.cn/) and TargetScan (http://www.targetscan.org/vert_72/). Eight candidate miRNAs were picked out from the putative miRNAs and identi ed whether CRNDE was involved in their regulation. As shown in Fig. 4a, overexpression of CRNDE decreased the levels of miR-543, miR-126-5p and miR-384 in HepG2 and Hep3B cells. In contrast, knockdown of CRNDE increased the expression of these three miRNAs (Fig. 4b). Next, we determined whether these three miRNAs could affect the expression of ATG4B. As shown in Fig. 4c, only miR-543 mimics could dramatically repress ATG4B expression, while miR-543 inhibitor elevated both the mRNA and protein levels of ATG4B (Fig. 4d, e). Bioinformatic analysis showed that miR-543 might bind to CRNDE and target 3 -UTR of ATG4B mRNA, and the potential binding sites were shown in Fig. 4f. Furthermore, luciferase reporter assays revealed that miR-543 mimics obviously reduced the luciferase activity of pmir-ATG4B, but not pmir-ATG4B-mut (in which the binding site of miR-543 in 3 -UTR of ATG4B mRNA was mutated) (Fig. 4g). Collectively, these results indicated that miR-543 directly targets 3 -UTR of ATG4B mRNA and inhibits ATG4B.
The "CRNDE/ATG4B/autophagy" pathway alleviates the sensitivity of sorafenib in HCC cells As shown in Supplementary Fig. S6a, b, overexpression of CRNDE promoted the viability of HCC cells while silence of CRNDE suppressed the cell viability, which was consistent with the previous reports [23,30,31]. Since protective autophagy tends to weaken the anti-tumor effect of therapeutic reagents, we further investigated whether the CRNDE-triggered autophagy in uenced the response of HCC cells to sorafenib. Interestingly, we found that sorafenib dramatically upregulated CRNDE (Fig. 6a) and ATG4B (Fig. 6b, c), increased LC3-II while decreased SQSTM1/p62 (Fig. 6c) in HCC cells. Additionally, knockdown of CRNDE or ATG4B, or inhibition of autophagy with CQ strengthened the suppression effect of sorafenib on cell survival (Fig. 6d), and increased sorafenib-induced PARP cleavage (Fig. 6e), apoptotic bodies (Fig.  6f) and apoptotic cells (Fig. 6g). Furthermore, overexpression of ATG4B markedly attenuated the reduction of cell survival caused by the co-treatment with si-CRNDE and sorafenib (Fig. 6h). The decrease of cell survival caused by si-CRNDE (or si-ATG4B) combined with sorafenib was abolished by autophagy inducer rapamycin (RAPA) (Fig. 6i), but not affected by autophagy inhibitor CQ (Fig. 6j) co-treatment. Nevertheless, inhibition of autophagy with CQ signi cantly alleviated CRNDE (or ATG4B)-increased cell survival in the presence of sorafenib (Fig. 6k). Altogether, these results prove that sorafenib activates the "CRNDE/ATG4B/autophagy" pathway, which alleviates the sensitivity of sorafenib in HCC cells.

Knockdown of CRNDE enhances the anti-HCC effect of sorafenib in vivo
To validate whether the above phenomenon is present in vivo, the xenograft tumor models with (or without) stable knockdown of CRNDE were established in nude mice ( Supplementary Fig. S7a, b). Knockdown of CRNDE remarkably strengthened the sorafenib-mediated inhibition of xenograft tumor growth (Fig. 7a-c) and decrease of the cell proliferation marker Ki67 (Fig. 7 d), while increased sorafenibtriggered cell apoptosis (Fig. 7 e) and PARP cleavage (Fig. 7g) in HCC xenograft tumors. Sorafenib signi cantly elevated the levels of CRNDE (Fig. 7f) and ATG4B (Fig. 7d, f and g) in the xenograft tumors. Furthermore, silence of CRNDE obviously attenuated the sorafenib-induced ATG4B (Fig. 7d, f and g) and LC3-II (Fig. 7g), whereas abolished the sorafenib-mediated reduction of SQSTM1/p62 ( Fig. 7d and g). These results demonstrate that knockdown of CRNDE strengthens the anti-HCC effect of sorafenib in vivo.

Discussion
In the present study, we have revealed a novel signalling pathway "CRNDE/ATG4B/autophagy", in which CRNDE upregulates ATG4B via inhibiting miR-543, leading to the enhancement of ATG4B mRNA stability and autophagy in HCC cells. Additionally, sorafenib can activate this pathway, which weakens the sensitivity of sorafenib. These ndings demonstrate a promoting effect of CRNDE on regulation of autophagy and sorafenib resistance of HCC cells.
CRNDE displays a tissue-specific expression pattern, in which it has little or no expression in certain normal tissues such as liver. Several lines of evidence support the important role of CRNDE as a mediator of oncogenesis [19,25]. Here, we validated that CRNDE was aberrantly upregulated in HCC cell tissues and lines, which was consistent with prior studies [23,30,31]. A recent study has shown that silence of CRNDE promotes autophagy and cell viability in neurons exposed to hypoxic-ischemic [32]. However, our study indicated that overexpression of CRNDE enhanced autophagy and cell viability in HCC cells. The discrepancy probably lies in the different pathological circumstances, which deserves indepth investigation to better understand the relationship between CRNDE and autophagy.
Numerous studies have highlighted the regulation of ATG4B by enzymes, compounds, miRNAs, siRNAs, transcription factors and their binding proteins, and so on [11,[33][34][35]. Recently, two lncRNAs have been reported to promote autophagy and chemoresistancein colorectal cancer through miR-34a/ATG4B pathway [36,37]. Nevertheless, it is still unclear whether CRNDE is associated with the regulation of ATG4B in hepatocellular carcinoma. In this study, we for the first time verified that CRNDE at least partially accounted for ATG4B upregulation and the subsequent autophagy induction in HCC cells, which disclosed a novel signalling pathway "CRNDE/ATG4B/autophagy". Additionally, as elevated CRNDE expression is typically found in a variety of malignancies, more studies are required to explore if the CRNDE-triggered upregulation of ATG4B also exists functionally in other correlated neoplastic diseases.
RNA stability could be in uenced by multiple factors such as RNA binding proteins, RNases and miRNAs [33]. Recent studies have shown that CRNDE affects the expression and/or activity of some miRNAs as a sponge [23,24,30], while several miRNAs have impact on the stability of ATG4B mRNA in various cancer cells [33,38,39]. Here, miR-543 was screened out to be a candidate which might be involved in the CRNDE-mediated regulation of ATG4B, and subsequently proved to be necessary for "CRNDE/ATG4B/autophagy" signalling axis. Interestingly, compared to wild type CRNDE, mutation of CRNDE in the binding site of miR-543 led to an attenuated upregulation of ATG4B and LC3-II while a weakened reduction of SQSTM1/p62, suggesting that CRNDE could modulate ATG4B and the subsequent autophagy through other ways besides regulating miR-543. Additionally, the action mechanisms of miR-543 are also complicated. For instance, miR-543/TRPM7 axis participates in cervical cancer progression through PI3K/AKT and p38/MARK pathways [40]. Another study has shown that miR-543/Angpt2 axis is involved in osteosarcoma metastasis and angiogenesis mediated by CTGF [41]. Hence the other miR-543-associated signalling cascades as mentioned above might be affected by the CRNDE-caused change of miR-543, which needs further investigation to clarify the comprehensive intracellular signal networks relevant to CRNDE and miR-543.
Sorafenib, an oral multi-kinase inhibitor, has dual anti-tumor effects on both tumor cell proliferation and angiogenesis [42]. To date, sorafenib is regarded as the rst approved and most widely-used systemic drug in the treatment of advanced HCC, but the therapeutic effect is less than satisfactory largely due to acquired drug resistance [42,43]. Mounting evidences have shown that protective autophagy weakens the lethality of sorafenib to HCC cells [44,45]. Mechanistically, sorafenib can induce autophagy in HCC cells through multiple ways, such as promoting IRE1 signals [44], activating Akt pathway [46], supressing mTORC1 [47], or inhibiting STAT3/Mcl-1/Beclin 1 axis [48], and so on. However, the detailed relationship between autophagy and sorafenib has not yet been clearly clari ed. Although accumulating studies have reported the signi cant role of CRNDE in HCC, there is currently no evidence whether CRNDE affects the sensitivity of sorafenib in HCC cells. In the present study, we elucidated that sorafenibactivated "CRNDE/ATG4B/autophagy" pathway contributed to the reduced sensitivity of sorafenib in HCC cells, which is a novel role of CRNDE to induce sorafenib resistance. Nevertheless, it remains unknown how sorafenib elevates CRNDE level. Therefore, more studies are warranted regarding the important mechanisms by which sorafenib in uences CRNDE expression in HCC cells.

Conclusion
In summary, our study presents the rst evidence that CRNDE upregulates ATG4B via suppressing miR-543 and subsequently enhances ATG4B mRNA stability, which ultimately promotes autophagy in HCC cells. Moreover, sorafenib activates the "CRNDE/ATG4B/autophagy" pathway, and knockdown of CRNDE sensitizes HCC cells to sorafenib. These ndings illustrate a key role of CRNDE in autophagy regulation and sorafenib resistance of HCC cells, and targeting the "CRNDE/ATG4B/autophagy" pathway may serve as a promising strategy to increase the sensitivity of sorafenib in HCC cells.

Declarations
Ethics approval and consent to participate

Consent for publication
All authors have agreed to publish this manuscript.

Availability of data and material
All data in our study are available upon request.

Competing interests
The authors declare that they have no competing interests.   4miR-543 targets and inhibits ATG4B expression. a HepG2 and Hep3B cells were transfected with pcDNA-CRNDE or pcDNA3.1. 24 h after transfection, the expressions of the indicated miRNAs were detected by qPCR. b 24 h after transfecting HepG2 and Hep3B cells with si-CRNDE or si-NC, the levels of miR-543, miR-126-5p and miR-384 were examined by qPCR. c HepG2 and Hep3B cells were transfected with mimics of the three candidate miRNAs separately. 24 h after transfection, the level of ATG4B protein was measured by Western blot. d, e 24 h after transfecting HepG2 and Hep3B cells with miR-543 mimics, miR-NC, miR-543 inhibitor or inhibitor-NC, levels of ATG4B mRNA and protein were measured by qPCR and Western blot, respectively. f Prediction of the potential miR-543 binding site in CRNDE and the putative miR-543 target site in 3 -UTR of ATG4B mRNA using StarBase and TargetScan, respectively. g Reporter plasmids (pmir-ATG4B, pmir-ATG4B-mut or pmir-GLO) were transfected into HepG2 cells in the presence of miR-543 mimics or miR-NC. 24 h after transfection, dual-luciferase reporter assay was performed. The activity of re y luciferase was normalized against that of rennilla luciferase. ns: no signi cance; *P<0.05; **P<0.01; ***P<0.001. Hep3B cells were co-transfected with pcDNA-CRNDE (or pcDNA3.1) and miR-543 mimics (or miR-NC). 24 h after co-transfection, mRNA and protein levels of ATG4B were determined by qPCR and Western blot, respectively. c 18 h after transfection with pcDNA-CRNDE and miR-543 mimics (or miR-NC), HepG2 and Hep3B cells were treated with 5 µg/mL actinomycin D (Act D) for 0, 2, 4 and 6 h. Then the mRNA level of ATG4B was quanti ed using qPCR. d HepG2 cells were co-transfected with pcDNA-CRNDE (or pcDNA3.1) and pmir-ATG4B. 24 h after transfection, dual-luciferase reporter assay was performed. The activity of re y luciferase was normalized against that of rennilla luciferase. e HepG2 cells were co-transfected with si-CRNDE (or si-NC) and pmir-ATG4B, and the luciferase activity was assayed as described in d. f HepG2 cells were co-transfected with pcDNA-CRNDE (or pcDNA3.1) and miR-543 mimics (or miR-NC) in the presence of pmir-ATG4B, and the luciferase activity was detected as in d. g-i HepG2 and Hep3B cells were separately transfected with pcDNA-CRNDE, pcDNA-CRNDE-mut, or pcDNA3.1. 24 h after transfection, the levels of miR-543 and ATG4B mRNA were analyzed by qPCR, and the level of ATG4B protein was examined by Western blot. ns: no signi cance; *P<0.05; **P<0.01; ***P<0.001.

Figure 6
The CRNDE/ATG4B/autophagy axis attenuates the sensitivity of sorafenib in HCC cells. a-c HepG2 and Hep3B cells were treated with various concentrations of sorafenib for 24 h. Then the levels of CRNDE and ATG4B mRNA were determined by qPCR, and the protein levels of SQSTM1/p62, LC3 and ATG4B were analyzed by Western blot. d HepG2 and Hep3B cells were transfected with si-CRNDE, si-ATG4B or si-NC, or pretreated with 20 μM CQ or vehicle control (0.1% DMSO) for 2 h. 12 h later, the cells were treated with