CircPAK1 promotes the progression of hepatocellular carcinoma via modulation of YAP nucleus localization by interacting with 14-3-3ζ

Overall, HCC tissue samples (60 cases) were collected from HCC patients who received a primary surgery attempt between September 2015 and August 2019 at the First Affiliated Hospital of Nanjing Medical University. Patients who lacked detailed clinical pathological information and who received preoperative chemotherapy or radiotherapy were excluded. Written informed consent was obtained from all patients and the study protocol was approved by the Ethical Committee of the First Affiliated Hospital of Nanjing Medical University.

Human HCC cell lines, including Hep-G2, Focus, Hep-3B, MHCC-97 L, HCC-LM3 (LM3), YY8103, Huh-7, and L02 cell lines were purchased from the Chinese Academy of Sciences Cell Bank (CASCB, Shanghai, China). All the HCC cell lines were cultured with DMEM (Gibco, MD, USA). HUVECs were obtained from the American Type Culture Collection (ATCC, VA, USA) and cultured with RPMI-1640 (Gibco, MD, USA). All the cells were supplemented with 10% fetal bovine serum (HyClone, UT, USA) and 1% penicillin/streptomycin at 37 °C with 5% CO₂ in an incubator.

The shRNA targeting the junction site or random sequence of human circPAK1 was designed and synthesized to establish the sh-circPAK1 vector, while the lentiviral vectors containing human circPAK1 were also to construct the Lv-circPAK1 vector. The above lentiviral vectors were synthesized by GeneChem (Shanghai, China). After the transfection of lentiviral vectors into target cells, stably transduced cells were selected by puromycin and validated by qRT-PCR. The silencing RNA against YAP (si-YAP) and 14–3-3ζ (si-14-3-3ζ) were synthesized and purchased from Corues Bio (Nanjing, China). The target sequence of shRNA or siRNAs were listed in Table S1, and the full sequence of circPAK1 was listed in Table S2. We selected LM3 and Hep-3B cells to induce lenvatinib-resistant HCC cells. Lenvatinib was purchased from MCE (MedChemExpress, NJ, USA). Briefly, cells were exposed to graded drug concentrations (5 μmol/L to 25 μmol/L) step by step to induce lenvatinib resistance. After 4 months of induction, the two lenvatinib-resistant HCC cell lines were generated.

CS/si-circPAK1 nanocomplexes was designed and synthesized by Shennuoqing Biotechnology (Nanjing, China).

A total of 84 female BALB/c nude mice (4–6 weeks old) were purchased from the Model Animal Research Center of Nanjing Medical University (NJMU). The mice were randomly divided into 12 groups of 6 each. All mice were housed in a pathogen-free environment, and animal experiments were authorized by the NJMU Institutional Animal Care and Use Committee.

To evaluate the roles circPAK1 played in vivo, we first randomly chose four groups to establish a subcutaneous tumor model: Two groups of mice were subcutaneously injected with 1 × 10⁶ LM3 cells infected with sh-circPAK1 or sh-NC, the other two groups subcutaneously injected with 1 × 10⁶ Hep-3B cells transfected with Lv-circPAK1 or Lv-NC. Tumor diameters were recorded every 4 days using a caliper. The tumor volume was calculated using the following formula: volume (mm³) = 0.5 × width² × length. After 32 days, the mice were sacrificed to harvest and weigh the grafted tumors. To evaluate the roles of circPAK1 on lung metastasis, we set another 4 groups to receive tail vein injection transfected cells (LM3 and Hep-3B) to establish a lung metastasis model. The mice were sacrificed after 30 days, and the lungs were separated and stained with hematoxylin and eosin (H&E).

In addition, CS/si-circPAK1 nanocomplexes treatment and non-treatment groups were established. LM3 cells were used to generate subcutaneous xenograft tumor and lung metastasis models according to the above protocol. After 8 days, saline, si-circPAK1 (10 nmol, in vivo-grade cholesterol-conjugated RIG-I siRNA, RiboBio), or CS/si-circPAK1 nanocomplexes (10 nmol) were injected intratumorally (50 μl) or intravenously (200 μl) into the six groups, once every 4 days, respectively. Mice were treated as described above.

Mice lung tissues were embedded in Paraffin and cut into 5 μm sections for H&E staining. While mice tumor tissues were fixed with 4% paraformaldehyde, dehydrated using ethanol solution, and then embedded in paraffinand cut into 4 μm sections, incubated with 3% H₂O₂ solution. Twenty minutes later, blocking the slice with blocking solution for 30 min, and incubating with the corresponding primary antibody at 4 °C overnight. The rest steps were consistent with previously described. Images were captured using a microscope (Leica Microsystems, Germany). The staining results were evaluated in a double-blind manner.

Total RNA was extracted from cells or frozen human tissues using RNA Quick Purification kit (YiShanbio, Shanghai, China) in accordance with the manufacturer’s instructions. Reverse transcription was performed using a cDNA kit (Vazyme Biotech, Nanjing, China) to synthesize the cDNA. All the primer sequences used in this study were listed in Table S3. qRT-PCR was performed using SYBR Green PCR kits (Vazyme Biotech, Nanjing, China) on the ABI 7900 detection system (Applied Biosystems, CA, USA). and the CT values were determined. The relative expression of circPAK1 and mRNA were calculated using the 2^-ΔΔCt method. GAPDH and U6 were used as internal standards.

Proteins were extracted by using RIPA buffer (Beyotime, Shanghai, China) containing protease inhibitors. Protein lysates were separated using 10% sodium dodecyl sulfate-polyacrylamide gels (SDS-PAGE) and subsequently transferred onto polyvinylidene fluoride membranes (Millipore, MA, USA). The membranes were subsequently blocked with fast blocking buffer (Beyotime, Shanghai, China) for 50 min and then incubated with primary antibodies at 4 °C overnight. After incubating with the respective secondary antibodies for 1 h and washing three times every 10 min with Tris-buffered saline with Tween-20 (TBST), the ECL signals were visualised by using ECL Western Blotting Kit (Millipore, MA, USA). Antibodies used in this study were listed in Table S4.

Cell Counting Kit-8 was performed as we described before [15].

For colony formation assay, cells were plated at a density of 600 cells per well into 6-well plates and cultured in complete media for approximately 10 days. After 10 days, the colonies were fixed using formaldehyde and then stained with 0.1% crystal violet (Vicmed, China). These colonies were subsequently photographed and counted visually.

The EdU proliferation assay was performed using an EdU kit (Beyotime, Shanghai, China). Briefly, cells were inoculated in 96-well plates at a density of 1 × 10⁴ cells per well, then cultured for 12 h and incubated with EdU for 2 h. The rest steps are performed according to the manufacture’s instructions. After staining for nucleic acids with DAPI (Beyotime, Shanghai, China), images were acquired by using an inverted fluorescence microscope (Nikon, Tokyo, Japan).

The invasion and migration assay were performed using trans-well chambers (Corning, USA) pre-covered or uncovered with Matrigel (BD Biosciences, USA), respectively. Briefly, 5 × 10⁴ cells were seeded onto the upper chamber, while 600ul medium with 10% FBS was provided in the lower chamber. After 24-36 h of incubation, cells on the upper membrane surface were scraped off, the invaded and migrated cells were fixed with methanol and stained with 0.1% crystal violet for 30 min. Images were acquired by using an inverted microscope.

Cells were seeded onto 6-well plates. When the confluence of cells reached more than 80%, the monolayers were scratched by a 200 μL pipette, and the cell debris was removed by washing with PBS. After that, cells were cultured with medium without FBS. Images of cell migration were captured at the same locations at 0 h, 24 h and 48 h.

Cell cycle and apoptosis analysis were performed as we described before [15].

RNA pull-down was performed using a kit (Thermo Fisher Scientific, CA, USA). The biotinylated circPAK1 probe and control probe were designed by RiboBio (Guangzhou, China), and the probe sequences were listed in Table S5. Approximately 1 × 10⁷ LM3 cells were lysed and then incubated with biotinylated probes for 4 h at 4 °C. After that, incubated with 50 μL of streptavidin-coated magnetic beads at room temperature for 1 h. Then, the beads-probe-protein complex was washed with wash buffer. The retrieved proteins were boiled in a SDS buffer and separated using SDS-PAGE, followed by staining with fast silver stain kit (Beyotime, Shanghai, China). The protein bands in circPAK1 probe group in comparison with control group were analyzed by Q Exactive mass spectrometer (Thermo Fisher Scientific, CA, USA).

RIP was performed using Magna RIP RNA Binding Protein Immunoprecipitation Kit (Millipore, MA, USA) according to the manufacturer’s protocols. Antibodies including anti-IgG (Millipore, MA, USA), anti-YAP (Proteintech, Wuhan, China) and anti-14-3-3ζ(Proteintech, Wuhan, China) were listed in Table S4. The relative interaction between circRNA and proteins were assessed by qPCR, and normalized to input.

To evaluate the binding between LATS2, YAP and 14–3-3ζ, proteins were immunoprecipitated by using a IP kit (Thermo Fisher Scientific, CA, USA) according to the manufacture’s protocols and then assessed by immunoblotting.

Cytoplasm and nucleus fractionation was performed with Nuclear and Cytoplasmic Protein Extraction Kit (Beyotime, Shanghai, China) in accordance with the manufacturer’s instructions (Beyotime, Shanghai, China). Cytoplasm and nucleus fractionations of protein and RNA were analyzed by western blot and qPCR respectively, to determine the level of YAP.

Briefly, after the fixation and permeabilization of HCC cells, Hybridization with circPAK1 probe (Sequence is shown in Table S5) (RiboBio, Guangzhou, China) was performed with a Fluorescent in Situ Hybridization Kit (RiboBio, Guangzhou, China) in accordance with the manufacturer’s instructions. Fluorescence images were acquired using a confocal laser scanning microscopy (Zeiss, Jena, Germany).

To identify differentially expressed circRNAs, circRNA sequencing was performed using 3 paired HCC tissues and adjacent normal tissues. The tissue quality control, sample preparation and circRNA sequencing were performed by Huada Gene (Shenzhen, China). To explore the influence of circPAK1 knockdown on global gene expression profiles, circPAK1-knockdown and control LM3 cells were lysed in Trizol. The RNA sequencing was analyzed by Tiangen Biotech (Beijing, China).

Exosomes were isolated from CM by differential ultracentrifugation. Briefly, cells were cultured in DMEM with 10% exosome-free FBS (Absin, Shanghai, China). When the confluence of cells reached more than 80%, CM was collected and centrifuged at 300 g for 10 min. Subsequently, cells and debris were removed through 2 × 10³ g centrifugation for 20 min and 1 × 10⁴ g centrifugation for 30 min. The resulting supernatant was pre-cleared through 0.22 μm filters (Millipore, MA, USA), followed by ultra-centrifugation at 1× 10⁵ g for 70 min. Repeat the previous step and the exosomes can be collected. For exosomal RNA extraction, exosomes were pre-treated with RNase, and an equal number of exosomes were used for RNA extraction.

Exosomes were examined by electron microscopy using negative staining and quantified using the NanoSight NS300 instrument (Malvern Instruments Ltd. UK) equipped with NTA 3.0 analytical software (Malvern Instruments Ltd. UK).

To identify circRNAs involved in HCC progression, high-throughput sequencing was performed on 3 paired tissues of HCC and adjacent tissues. Among the 351 differential expressed circRNAs (fold change≥2 and q value≤0.001)，a total of 211 were upregulated and 140 were downregulated in the HCC tissues than their paired normal tissues. As shown in Fig. 1A, from the circular RNA sequencing results, 44 differentially expressed genes (DEGs) with fold change ≥15 were identified, all higher expressed in tumor tissues than those in normal tissues. The 44 upregulated circRNAs were listed in Table S6. Usually, the levels of circRNAs are in accordance with their respective mRNA levels [16]. Therefore, 25 DEGs whose host genes are significantly upregulated in HCC tissues were selected from these 44 DEGs by using The Cancer Genome Atlas (TCGA) database. We designed specific primers targeting the junction site of these 25 DEGs, respectively, and only 11 circRNAs could be successfully amplified by qRT-PCR. Then, we examined the expression of these 11 circRNAs by qRT-PCR in 20 HCC tissues and matched adjacent normal tissues, and only 5 DEGs were statistically significant. As the P value of paired comparison of circPAK1 is the most significant, so circPAK1 was eventually selected for the subsequent study (Fig. S1A).

CircPAK1 is derived from chromosome 11: 77085372–77,103,586 and consists of 4 adjacent exons in the PAK1 gene, sanger sequencing confirmed the junction site of circPAK1 (Fig. 1B). To further identify the ring structure, convergent and divergent primers were designed, and the PCR product was determined by agarose gel electrophoresis (Fig. 1C). Additionally, the RNase R digestion and actinomycin D RNA stability assays confirmed the stability of circPAK1 compared with PAK1 (Fig. 1E, F). The FISH and qRT-PCR results suggested that the localization of circPAK1 was mostly in cytoplasm (Fig. 1G, H).

The levels of circPAK1 in 60 pairs of HCC tissues and adjacent non-tumor tissues were determined by qRT-PCR. The results showed that circPAK1 was significantly overexpressed in HCC tissues than normal liver tissues (Fig. 2A, B). Besides, circPAK1 was also overexpressed in HCC cell lines compared with normal hepatocytes (L02) (Fig. 2C). For the subsequent study, we chose LM3 and Focus cells to establish circPAK1 stable knockdown cell lines by infection with shRNA specially targeting the junction sites of circPAK1, while Hep-3B to infect with a lentiviral vector to establish circPAK1 stable overexpression cell lines. As shown in Fig. 2D, stable circPAK1-knockdown and circPAK1-overexpression cell lines were successfully established. To evaluate the clinical value of circPAK1, the 60 HCC patients were divided into two groups based on the median expression of circPAK1 (Table 1). Our data suggested that higher expression of circPAK1 was more likely to develop larger tumor size, higher risks of LN metastasis, advanced TNM stage and microvascular invasion (MVI) (all P values<0.05); however, no statistical significance for age, gender, AFP, PIVKA-II, HBsAg status, tumor multiplicity, differentiation and vascular invasion was observed. Importantly, patients in the high expression group showed poor overall survival and disease-free survival than the patients in the low expression group (Fig. 2E). Taken together, these results demonstrate that circPAK1 is abundant and upregulated in HCC tissues and cell lines, and it could serve as a valuable molecular biomarker for HCC.

To explore the effect of circPAK1 on cell proliferation, CCK-8, EdU and colony formation assays were performed, and the results showed that circPAK1 knockdown significantly attenuated the growth of LM3 and Focus cells, while converse results were observed in circPAK1 overexpression cell lines (Fig. 3A, B and C). To further investigate the effect of circPAK1 on HCC cell migration and invasion, wound-healing and transwell assays were performed. The results revealed that circPAK1 inhibition could significantly reduce the migration and invasion ability of LM3 and Focus cells, while circPAK1 overexpression markedly increased these abilities (Fig. 3D, E). Taken together, these results indicated that circPAK1 enhances HCC cell proliferation and motility.

We also explored whether circPAK1 could regulate the cell cycle and apoptosis. We found that the intervention of circPAK1 did not affect the cell cycle (Fig. S1B). However, it is worth noting that circPAK1 knockdown promoted cell apoptosis in both LM3 and Focus cells (Fig. 4A). These results suggest that the promotion ability of circPAK1 on HCC cell growth may depend on inhibiting tumor apoptosis.

Tumor cells can promote angiogenesis by modifying the tumor microenvironment, contributing to early progression, invasion, postoperative recurrence, and metastasis of HCC [17, 18]. Therefore, the ability to promote angiogenesis is also one of the important indicators to measure tumor invasiveness. Then, we performed an angiogenesis experiment to study whether circPAK1 has a promoting effect on tumor angiogenesis.

HUVEC tube formation, migration and invasion were evaluated using the conditioned medium (CM) from LM3 (circPAK1 knockdown) and Hep-3B (circPAK1 overexpression) cells. As shown in Fig. 4B, CM from circPAK1 knockdown cells significantly inhibit the migration and invasion of HUVEC cells. In contrast, the opposite results were found in circPAK1 overexpression cells. As envisioned, the tube formation assay also showed that the number of branch sites was decreased in HUVECs treated with CM from circPAK1 knockdown cells compared with that from control cells. The above abilities of HUVECs were enhanced after incubation with CM from circPAK1 overexpression cells. Taken together, these findings suggest that circPAK1 could induce angiogenesis of HCC in vitro.

To assess the in vivo effect of circPAK1 on HCC growth and metastasis, we then generated tumor xenograft and lung metastasis models. The xenograft tumor model showed that no matter the tumor size or the tumor weight was significantly decreased in the sh-circPAK1 group, while increasing in the Lv-circPAK1 group (Fig. 5A). The results of IHC staining also showed that the level of Ki-67 was lower in the sh-circPAK1 group and higher in the Lv-circPAK1 group (Fig. 5B). Additionally, the lung metastasis model showed a similar trend as the results of xenograft model, for there were less metastasis foci in the sh-circPAK1 group, and more in the Lv-circPAK1 group (Fig. 5C). These results indicate that circPAK1 could promote tumor growth and metastasis of HCC.

Based on the new understanding of tumor pathogenesis, gene targeting therapy in cancer has attracted a wide range of attention and proved to be a rational approach in cancer therapy. Chitosan (CS)-based materials have a long history as drug delivery vehicles due to their characteristics of non-toxicity, biodegradability, high drug carrying capability, high plasma membrane permeability, including transport across mucosal membranes, the pH-dependent release of therapeutic agents, multi-functionality, and suitable circulation time [19‐21]. We established three nanocomplexes in different proportions based on the weight ratios of CS and si-circPAK1. The loading efficiency of chitosan to si-circPAK1 was more than 99% under these three ratios (Fig. 5D). Therefore, to ensure the loading efficiency, we choose the largest chitosan utilization ratio-CS/si-circPAK1 of 50. The TEM (Fig. 5E) and particle size potentiometer (Fig. S2A) revealed that the particle size of CS/si-circPAK1 nanocomplexes was on an average of 101.9 nm. The zeta potential was − 21.4 and + 31.6 mV (Fig. S2B). By simulating the temperature and pH of human body to detect the release of siRNA from CS, we found that siRNA can be released persistently and efficiently (Fig. S2C).

To study the effect of CS/si-circPAK1 nanocomplexes on the proliferation and metastasis of HCC in vivo, we generated several groups of subcutaneous xenograft tumor and lung metastasis models. Figure 5F shows the injection scheme with saline, si-circPAK1 or CS/si-circPAK1 nanocomplexes into mice. We noticed that no matter the tumor size or the weight in the CS/si-circPAK1 nanocomplexes treated group was much smaller and lighter than the saline-treated group, even better than the si-circPAK1 group (Fig. 5G). IHC staining of tumors from CS/si-circPAK1 nanocomplexes treated group also revealed a significant decrease of Ki67 (Fig. S2D). Furthermore, the lung metastasis models from CS/si-circPAK1 nanocomplexes treated group showed a better inhibition effect on lung metastasis lesions (Fig. 5H). Collectively, these findings suggest that circPAK1 could promote the in vivo growth and metastasis of HCC, and the CS/si-circPAK1 nanocomplexes could effectively inhibit these abilities.

To explore the molecular mechanisms of circPAK1 involved in HCC development, we performed RNA-sequencing. By listing the top 10 of pathway enrichment, a strong correlation between the Hippo signaling pathway and the downstream of circPAK1 is shown (Fig. 6A). The hippo signaling pathway is highly conserved in evolution, and it can regulate organ size by regulating cell proliferation and apoptosis [22, 23]. The dysregulation of Hippo signaling pathway will cause uncontrolled proliferation of cells, thus leading to the overgrowth of tissues and organs [24]. Several studies have pointed out that the inactivation of Hippo signaling pathway was closely correlated with the progression of HCC [25‐27].

Combined with the above tumor-promoting phenotype of circPAK1 and the result of RNA-seq, we, therefore, explored the effect of circPAK1 on Hippo signaling pathway. We first performed a western blot to analyze the changes in proteins related to Hippo signaling pathway. We found that neither the knockdown nor the overexpression of circPAK1 did not affect the level of LATS1/2 or its phosphorylation level (Fig. 6B); however, p-YAP was increased in the circPAK1 knockdown group, while decreased in the circPAK1 overexpression group. Cytoplasmic retention of YAP plays a vital role in the Hippo pathway-mediated control of cell proliferation and apoptosis. It is well established that non-phosphorylated YAP can be transported into the nucleus and interact with TEAD1–4 to activate multiple downstream target genes, which is very important to enhance tumor proliferation and apoptosis inhibition [28‐30]. However, YAP (p-YAP) phosphorylation by p-LATS1/2 will be sequestrated in the cytoplasm and eventually degraded [31].

Based on the above perceptions, we speculate that circPAK1 may promote the nucleus transport of YAP and then affect the proliferation, invasion and metastasis of HCC. Subsequently, we conducted the cytoplasm and nucleus fractionation assay to observe the effect of circPAK1 on the localization of YAP. As shown in Fig. 6C, overexpression of circPAK1 promoted the nucleus localization of YAP, while reducing the p-YAP (ser-127) level. The opposite results were observed under the condition of circPAK1 knockdown. Meanwhile, immunofluorescence also confirmed this observation (Fig. 6D). These findings suggest that circPAK1 could guide the nucleus transportation of YAP. Besides, the main downstream target gene of YAP, CTGF and CYR61, were downregulated when circPAK1 knockdown, but upregulated when circPAK1 overexpression (Fig. 6E).

We further explored whether the tumor-promoting effect of circPAK1 is dependent on YAP by treating the circPAK1 overexpression Hep-3B cells with YAP siRNA. Fig. S3A shows the efficiency of YAP knockdown. As envisioned, the tumor promoting effect of circPAK1 on HCC was significantly weakened (Fig. S3B-F). Taken together, our data revealed that circPAK1 enhances HCC progression by accelerating YAP nucleus transport, which leads to the inactivation of Hippo signaling pathway.

As a transcription coactivator, the nucleus transportation of YAP will directly determine whether it can interact with TEAD1–4 to activate multiple downstream target genes [28‐30]. There have been many studies on downstream signals of YAP after its nucleus implantation, but little is known about by what mechanism YAP is regulated before its transportation to the nucleus.

Serving as miRNA sponge is one of the most common biological functions of circRNA, while binding with argonaute 2 (AGO2) RNA is the vital basis for circRNA to act as competitive endogenous RNA (ceRNA). However, RIP-qPCR assay showed that circPAK1 couldn’t be enriched in AGO2 antibody complexes compared with anti-IgG, so acting as ceRNA of circPAK1 was excluded (Fig. S4A).

It has been reported that circRNA can facilitate YAP nucleus transport by binding YAP protein directly, thus promoting the metastasis of colorectal cancer [32]. To verify whether circPAK1 can directly bind to YAP, we also performed RIP-qPCR and found that the result was not consistent with our hypothesis (Fig. 7A). So, how does circPAK1 affect the cellular spatial localization of YAP? We speculate that another mediator protein mediates the association between circPAK1 and YAP. Next, we performed an RNA-pulldown assay and mass spectrometry analysis to identify potential circPAK1-interacting proteins in LM3 cells. Surprisingly, 14–3-3ζ protein was identified as the main protein using silver staining (Fig. 7B) and liquid chromatography mass spectrometry (Fig. S4B). WB (Fig. 7C) and RIP were also performed and confirmed this specific binding (Fig. 7D).

Nucleus, import/export of YAP, is strictly regulated by 14–3-3ζ, as evidenced by the fact that 14–3-3ζ assembles with YAP, isolates it in the cytoplasm and prevents it from further signal amplification [33]. Therefore, the direct or indirect change of 14–3-3ζ level will affect the spatial localization of YAP in a cell. It has been proved that 14–3-3ζ can recruit p-LATS and YAP to form a complex, which can induce the phosphorylation of YAP and lead to its cytoplasmic retention, and 14–3-3ζ is the key regulatory factor of this complex (Fig. 7E) [34]. In addition, 14–3-3ζ can be dissociated from non-phosphorylated YAP under the condition of hypoxia, thus leading to the nucleus transport of YAP [35]. These studies showed that the cytoplasmic fixation of 14–3-3ζ on YAP is based on the non-phosphorylation of YAP. We hypothesized that circPAK1 could directly or indirectly affect the level of 14–3-3ζ by interacting with 14–3-3ζ, thus affecting the recruitment of P-LATs and YAP, then further inhibiting the phosphorylation of YAP, and finally, exerting the tumor promoting effect by facilitating the nucleus transport of YAP. Interestingly, we found that circPAK1 did not affect the 14–3-3ζlevel, suggesting that the level of 14–3-3ζmay be affected indirectly (Fig. 7F). To approach this, we first performed immunoprecipitation assay in circPAK1-knockdown LM3 cells. The existence of 14–3-3ζ- p-LATS-YAP protein complex was confirmed, and the interaction between p-LATS and YAP in the circPAK1-knockdown group was significantly increased compared with the sh-NC group, while significantly decreased in the circPAK1-overexpression group (Fig. 7G, Fig. S4D). Next, we silenced 14–3-3ζ in circPAK1 stable knockdown cell lines to explore whether the interaction between p-LATS and YAP can still be influenced by the change of circPAK1 level. The silencing efficiency of si-14-3-3ζ was shown in Fig. S4C. Surprisingly, we found that the increased interaction between p-LATS and YAP was significantly reversed after 14–3-3ζ knockdown (Fig. 7H). The immunofluorescence also showed that the nucleus location of YAP was restored after the knockdown of 14–3-3ζ (Fig. S4E). Taken together, the promotion of YAP nucleus transport mediated by circPAK1 is through binding with 14–3-3ζ, which weakens the recruitment of p-LATs and YAP, and then reduces the phosphorylation of YAP by p-LATs.

To examine whether circPAK1 is involved in mediating HCC lenvatinib resistance, we generated two lenvatinib-resistant HCC cell lines (LM3-LR and Hep-3B-LR). The IC50 of the two lenvatinib-resistant cell lines and their parental cell lines (LM3-P and Hep-3B-P) were shown in Fig. S4F. Next, we performed a qRT-PCR analysis and confirmed that circPAK1 was consistently increased in the two lenvatinib-resistant cell lines than in their parental cell lines (LM3-P and Hep-3B-P) (Fig. 8A). To explore the potential correlations between circPAK1 and lenvatinib resistance in HCC, we first silenced circPAK1 in LM3-LR and Hep-3B-LR cells and confirmed the knockdown efficiency (Fig. 8B). We found that the depletion of circPAK1 remarkably increased the lenvatinib sensitivity of the two lenvatinib-resistant HCC cell lines as determined by cell viability assays (Fig. 8C). Collectively, circPAK1 is critical for maintaining lenvatinib resistance.

It has been reported that circRNAs are abundant in exosomes and can be transferred from cell to cell via exosomes [36‐38]. Other reports demonstrated that exosomes could mediate drug resistance by exosomes [12, 13, 39, 40]. Therefore, the subsequent research was mainly focused on whether circPAK1 could confer lenvatinib resistance by integrating into exosomes. We then isolated exosomes from the CM of LM3-LR and Hep-3B-LR cells and their parental cell lines. Exosomes isolated from the CM of LM3-LR and Hep-3B-LR cells were identified by TEM, NTA and WB. The representative micrograph and video were taken by TEM; NTA analysis revealed that the average diameter of exosomes was 100 nm (Fig. 8D). The typical positive exosome biomarkers (Calnexin) and negative exosome biomarkers (Alix, Tsg101 and CD9) were detected by WB (Fig. 8E). In addition, qRT-PCR indicated that exosomes isolated from LM3-LR and Hep-3B-LR CM contained more circPAK1 than those from their parental cell lines, suggesting that exosomes may have the ability to transmit circPAK1 (Fig. 8F).

We used three prolonged stages to explore whether exosome could mediate the transfer of circPAK1 disseminates lenvatinib resistance. Firstly, we labeled exosomes isolated from LM3-LR and Hep-3B-LR cells with PKH67 followed by incubation with their parental cells, respectively, for 48 h. As shown in Fig. 8G, a strong green signal was observed in LM3-P and Hep-3B-P cells, indicating that the exosomes were taken up by recipient parental cells. Next, we determined the level of circPAK1 in recipient parental cells and found that after incubation with exosomes isolated from the lenvatinib-resistant cells, higher levels of circPAK1 were determined, but exosomes from circPAK1 knockdown lenvatinib-resistant cells failed to raise the level of circPAK1 in recipient cells (Fig. 8H). The results of these two stages indicated that exosomes could transport circPAK1 from lenvatinib-resistant cell lines to their parental cell lines. Lastly, we explored whether exosome-transferred circPAK1 could induce the resistance of recipient cells to lenvatinib. Surprisingly, when we treated the parental cells with exosomes isolated from their lenvatinib-resistant cell lines, their sensitivity to lenvatinib was significantly decreased (Fig. 8I). Taken together, our data demonstrated that circPAK1 could be transported by exosomes from lenvatinib-resistant HCC cells to recipient parental cells and confer the resistant phenotype to recipient cells.