The human hepatoma cell lines MHCC97H, Huh7, PLC, MHCCLM3, Hep3B and human normal liver cells LO2 were from Liver Cancer Institute, Zhongshan Hospital, Fudan University. Cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM High Glucose Pyruvate, Gibco) with 10% fetal bovine serum (FBS, Sigma) and 1% antibiotics in an incubator with 5% CO2 at 37°C. For hypoxic treatment, cells were cultured in a hypoxia incubator with 1% O2, 5% CO2, and 94% N2 for 48h, then were harvested. All cells used in this study have tested negative for mycoplasma.

Five-week-old female BALB/c nude mice housed in a suitable environment according to the protocols approved by the Animal Ethics Committee of Zhongshan Hospital. 3×10⁶ Huh7 cells with USP2-AS1 overexpression or normal control after treated with hypoxia (1% O₂) for 48h were subcutaneously injected into either side of the flank of each female nude mouse (n=4 mice per group). All mice were sacrificed after 4 weeks for tumor weight and volume measurement.

To explore the anti-tumor effect of lenvatinib combined with USP2-AS1 knockdown, the mice were randomly divided into two groups (n=4 mice per group), 3×10⁶ Huh7 cells with USP2-AS1 knockdown or Normal Contrast (NC) after treated with hypoxia (1% O₂) for 48h were subcutaneously injected into either side of the flank of each female nude mouse. When the tumor volume exceeded 50 mm³, lenvatinib solvent diluted with PBS (22.6mg/kg per time for each nude mouse according to the manufacturer’s instructions) or PBS (as Normal Contrast) was administered by oral gavage, five times a week. Two weeks after the injection of lenvatinib, the mice were sacrificed for tumor weight and volume measurement.

The cells were plated in a six-well plate and washed twice with PBS, then SDS-PAGE Sample Loading Buffer,1X 200uL (purchased from Shanghai Beyotime Biotechnology Company) was directly added to lyse the cells for protein extraction. The samples were then heated at 95°C for 10 minutes and then loaded onto a sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) to separate proteins based on their different molecular weights. The gel was then transferred to a nitrocellulose membrane. The membrane was incubated with the corresponding primary and secondary antibodies, and the signals were visualized by ECL (Enhanced Chemiluminescence). The film was scanned by the ImageStudio system, photographed, and preserved for follow-up analysis.

The RNA was extracted from cells or tissues with RNAeasy™ Animal RNA Isolation Kit with Spin Column (purchased from Shanghai Beyotime Biotechnology Company, China) according to the manufacturer’s protocol. Following the manufacturer’s instructions, Hifair^®II 1st Strand cDNA Synthesis SuperMix (YEASEN Biotechnology, Shanghai, China) PCR premix was used for qRT-PCR. The gene expression level relative to GAPDH was calculated by the 2^-ΔΔCT method. Each experiment was performed in triplicates. The sequences of primers used in qRT-PCR are listed in Supplementary Table 2 .

USP2-AS1 overexpression and knockdown lentiviral expression vectors are provided by Tsingke Biotechnology Co., Ltd (Beijing, China). Transfection was performed with corresponding transfection reagents when the cell density reached 30%-50%. 6-8h after transfection, the medium was changed, and puromycin was used for 48h for selection. After 2 weeks, a stably transfected cell line was obtained. YBX1-siRNAs and negative control siRNAs were provided by Ribobio Biotechnology Co., Ltd (Guangzhou, China). Viral transfections were carried out using the corresponding viral transfection reagents, Polyplus jetPRIME transfection reagents were used for all siRNAs transfection.

Cell proliferation assays were carried out using the BeyoClick™ EdU Cell Proliferation Kit with Alexa Fluor 555(EdU-555, Beyotime, China), according to the manufacturer’s protocols. In brief, 5000 cells were cultured in a 96-well plate. After the adherence of the cells, the 2 × Edu culture medium was added and incubated for 2 hours. After fixing, washing, and permeating, the configured Click Additive Solution was added and incubated at room temperature in the dark for 30 minutes. The nuclei were stained with Hoechst 33342 and incubated at room temperature in the dark for 10 minutes. Finally, the cells were observed and photographed using a fluorescence microscope.

A total of 2000 cells were inoculated in a six-well plate. After the cells adhered to the wall, they were cultured in the anoxic incubator of 1% O₂, and the culture medium was changed twice a week. Cells were harvested after 15 days. Each experiment was performed with three replicates.

Cell migration and invasion assays were carried out using the Boyden chamber (BD Bioscience) with an aperture of 8um with or without Matrigel, and the experiments were performed according to the manufacturer’s protocols. The cells were washed with PBS and digested into a single-cell suspension. 1 × 10⁵ cells (suspended in 200ul medium containing 10%FBS) were added to the upper chamber and 600ul containing 30% FBS was added to the lower chamber and cultured in the hypoxia incubator for an appropriate time. The cells were fixed with 4% paraformaldehyde, stained with crystal violet, and washed by PBS. The cells on the surface of the upper chamber were observed and photographed using a microscope. The experiments were repeated three times.

RIP was performed by the MagnaRIPTM RNA-Binding Protein Immunoprecipitation Kit (Merck Millipore) according to the manufacturer’s protocols. 2 × 10⁷ cells were lysed by Complete RIP Lysis Buffer, centrifuged and the supernatant was incubated with magnetic beads conjugated to antibody (Five micrograms per analysis) against Y-box binding protein 1 (YBX1) at 4°C overnight. After the protein was digested by protease K, RNA was purified by phenol-chloroform. Finally, the qRT-PCR test was performed with specific primers for quantitative analysis of lncRNA USP2-AS1 and HIF1α mRNA.

LncRNA pulldown assay was performed by the Pierce™ Magnetic RNA-Protein PullDown Kit from Thermo according to the manufacturer’s protocols. Approximately 50 pmol of the RNA 3’ end desthiobiotinylation labeled lncRNA USP2-AS1 full-length or its truncated probes (synthesized by BersinBio, Guangzhou, China) were incubated with 50ul of streptavidin magnetic beads for 30 minutes at room temperature. The beads were then washed by using an equal volume of 20 mM Tris (pH 7.5). 100ul 1× RNA-protein binding buffer (Containing 100ug total protein) was added to the tube containing streptavidin magnetic beads and gently mixed and incubated at room temperature for 60 minutes. After washing the beads 1-2 times with an equal volume of 1x wash buffer (100ul), the beads were mixed with 50ul Elution Buffer and incubated at 37°C for 30 minutes. Supernatants were collected for silver staining or WB analysis.

HCC tissue microarrays were purchased from Shanghai Outdo Biotech Company (T16-339 TMA1, T16-339 TMA2, T16-T16-TMA3) for the FISH experiment. The clinical information was statistically analyzed based on the high and low USP2-AS1 positive cell rates. A total of 208 pairs of non-duplicate liver cancer paired samples and 216 liver cancer tissue samples (8 samples were repeated) were collected on the three tissue microarrays, of which the third tissue microarray had a small number of duplicate samples. These samples were included in the analysis of USP2-AS1 expression in normal liver and HCC tissues. In the survival and prognostic analysis, we excluded the duplicate samples. Two liver cancer samples used in the USP2-AS1 FISH experiment and 36 paired samples of HCC used in the qRT-PCR experiment came from Zhongshan Hospital Fudan University (approved by the Institutional Review Board (IRB) at Zhongshan Hospital Fudan University).

LncRNA USP2-AS1 in situ hybridization kit was provided by Wuhan Servicebio technology Co., Ltd. Briefly, the dehydrated and fixed tissue microarray was dripped with a probe specific for lncRNA USP2-AS1 and hybridized overnight at 55°C in a thermostat. After hybridization, tissue microarray was washed by 2 × SSC at 37°C for 10 minutes, 1 × SSC at 37°C for 2 × 5 minutes and 0.5 × SSC for 10 minutes at room temperature, or formamide was added to wash nonspecific hybrids. Then the microarray was blocked with BSA and incubated with anti-DIG-488 at 37°C for 50 minutes, washed 4 times (5 minutes per wash). Finally, DAPI dye was added and incubated for 8 minutes, protected from light. After being washed, anti-fluorescence quenching tablets were added to seal. The rate of USP2-AS1 positive cells was counted using a fluorescence microscope.

The protein-coding ability of USP2-AS1 has displayed by PhyloCSF (PhyloCSF in the UCSC Genome Browser (https://genome.ucsc.edu/). We use the lncATLAS (https://lncatlas.crg.eu/) to predict the location of USP2-AS1 in HCC cells. The binding site of USP2-AS1 with YBX1 was predicted through the online website catRAPID (http://service.tartaglialab.com/update_submission/380192/4af202c73a). We also used the RNAfold Webserver (http://rna.tbi.univie.ac.at/cgi-bin/RNAWebSuite/RNAfold.cgi) to predict the secondary structure of USP2-AS1. TCGA-LIHC (The Cancer Genome Atlas-Liver hepatocellular carcinoma) was used for verifying the overexpression of lncRNA USP2-AS1 in liver cancer and the correlation between YBX1 and HIF1A, YBX1 and USP2-AS1.

All experiments in this study were performed at least three times. Data were expressed as mean± SD. Differential comparisons of data between two groups were performed using paired or unpaired two-tailed t-tests. Statistical analysis was performed using GraphPad Prism 8.0.2 and R 4.0.5. The results were considered statistically significant when p < 0.05.

To find hypoxia-associated lncRNAs with potential promoting effect on HCC, RNA sequencing analysis was performed using HCC tissues and adjacent normal tissues from 4 patients and HCC cell line Huh7 treated with hypoxia for 48h ( Figures 1A, B ). A total of 2722 lncRNAs were significantly up-regulated in HCC tissues, and 101 lncRNAs were up-regulated in hypoxia-treated Huh7 cells, the intersection between the two categories included 25 lncRNAs ( Figure 1B ).

Given that lncRNA USP2-AS1 (ubiquitin specific peptidase 2 antisense RNA 1) has the most pronounced fold change in the two categories, so we focused on the role of USP2-AS1 in HCC in our following studies. USP2-AS1 is a long non-coding RNA with 3798 nucleotides in length (USP2-AS1-208, ENST00000659432.1). Similar to other lncRNAs, we found USP2-AS1 has almost no protein-coding ability by PhyloCSF ( Figure S1A ). We also found that USP2-AS1 is likely entirely located in HCC cytoplasm by lncATLAS. ( Figure S1B ), which is confirmed by Fluorescent in situ hybridization (FISH) assay ( Figure S1C ).

To explore the relationship between USP2-AS1 and liver cancer, we analyzed the expression of USP2-AS1 in 36 pairs of HCC and adjacent non-tumor specimens by RT-qPCR. As expected, the expression of USP2-AS1 in HCC was significantly higher than in para-cancer tissues ( Figure 1C ). It was also confirmed in the FISH experiment using three tissue chips of human liver cancer ( Figures 1D, E ), which is consistent with the TCGA-LIHC (The Cancer Genome Atlas, Liver Hepatocellular Carcinoma) data ( Figure S1D ).

USP2-AS1 was differentially upregulated in hepatoma cell lines (Huh7, MHCC97H, HCCLM3, PLC) compared with normal human hepatocyte LO2 under both normoxic and hypoxic conditions ( Figures S2A, B ). We also found that the expression of USP2-AS1 was significantly up-regulated in these cell lines following 48h hypoxia ( Figure S2C ), which was consistent with our RNA sequencing results ( Figure 1A ). By analyzing the positive cell rate of USP2-AS1 in human liver cancer tissue microarrays and determining the cutoff value of USP2-AS1 by X-Tile software, we observed that patients with low expression of USP2-AS1 had better prognosis while patients with high expression of USP2-AS1 had shorter OS (Overall Survival) and DFS (Disease-Free Survival) time ( Figure 1F ).

Notably, clinical information of the tissue microarrays showed that USP2-AS1 remarkably correlated with tumor size and the TNM stage of HCC ( Table 1 ). Importantly, ROC curve analysis showed lncRNA USP2-AS1 had a high AUC (Area Under Curve, 0.824) for HCC diagnosis ( Figure S2D ). The above results indicate that hypoxia-related lncRNA USP2-AS1 is up-regulated in HCC. USP2-AS1 positively correlated with worse outcomes in HCC patients and may be a potential prognostic tool.

To investigate the role of USP2-AS1 in HCC, we constructed stable transfectants of MHCC97H and Huh7 cell lines overexpressing USP2-AS1 and knocked down USP2-AS1 in Huh7 cell lines by lentiviral transfection ( Figure 6A ). Under normoxic conditions, overexpression or knockdown of USP2-AS1 did not affect the proliferation, migration, and invasion ability of HCC (data not given). However, our previous RNA sequencing and FISH experiments have confirmed a significant difference in the expression of USP2-AS1 between liver cancer and para-cancer tissues ( Figures 1C–E ). And given that hypoxia treatment can increase USP2-AS1 expression in HCC cell lines ( Figure S2C ), we hypothesized USP2-AS1 promotes the progression of tumors by affecting tumor hypoxic metabolism.

After being treated with hypoxia for 48h, overexpression of USP2-AS1 dramatically increased HCC proliferation and clone formation in vitro ( Figures 2A, B ). Conversely, USP2-AS1 knockdown significantly inhibited the proliferation and clone formation ability of HCC ( Figures 2A, B ). We observed that the migration and invasion of HCC cells decreased after knockdown of USP2-AS1, while the reverse was observed with USP2-AS1 overexpression ( Figure 2C ). These results indicate that USP2-AS1 significantly enhances the proliferation, migration, and invasion of liver tumor cells in vitro under hypoxia.

To understand the mechanism of USP2-AS1 in HCC progression under hypoxia, we performed RNA pulldown assays on Huh7 cells cultured in hypoxia for 48 hours ( Figures 3A and S3A, B ). A total of 75 proteins were identified as potential molecules that may bind to USP2-AS1 by using mass spectrometry ( Table S1 ). We conducted GO enrichment analysis as well as KEGG pathway enrichment analysis of these 75 protein molecules. GO enrichment results showed that MF (Molecular Function) principally included protein binding and RNA binding, BP (Biological Process) mainly comprised of mRNA Splicing, cell-cell adhesion, and translation, and CC (Cellular Component) was primarily in the cytoplasm and nucleus ( Figures 3B, C ). Among these protein molecules, we noticed that YBX1(Y-box binding protein 1) is highly likely to bind to USP2-AS1 with the highest protein spectrum score ( Figure 3D ). The binding of YBX1 to USP2-AS1 was confirmed by YBX1-RIP assay in the MHCC97H cell line ( Figure 3E ), and in the Huh7 cell line ( Figure 3F ).

To investigate how USP2-AS1 binds to YBX1, we predicted the binding site of USP2-AS1 with YBX1 through the online website catRAPID It indicated the segment 500nt-1000nt of USP2-AS1 was likely to bind to YBX1 ( Figure S2E ). We also found USP2-AS1 contains 6 fragments with specific secondary structures by the RNAfold Webserver ( Figure 4A ). We hypothesized that these secondary structures (A-F fragments) are the binding sites of USP2-AS1 with YBX1. We designed six truncated probes (corresponding to A-F fragments, respectively) of USP2-AS1, such as A (1-790nt), B (791-1370nt), C (1371-1960nt), D (1961-2450nt), E (2451-3330nt) and F (3331-3798nt) for RNA pulldown assays ( Figures 4B and S3B, C ). We demonstrated that fragment A and C are necessary to mediate the binding of USP2-AS1 and YBX1 ( Figure 4C ). These results show USP2-AS1 physically interacts with YBX1 by the A and C fragments under hypoxia.

Hypoxia-inducible factors (HIFs) are critical transcriptional activators mediating tumor hypoxia metabolism, and their imbalance plays an important role in tumor development (27), so we explored whether the expression of USP2-AS1 affects the levels of HIF1α under hypoxia. However, the overexpression of USP2-AS1 did not affect the levels of HIF1α mRNA in Huh7 or MHCC97H cells, and the knockdown of USP2-AS1 also did not affect the HIF1α mRNA in Huh7 ( Figure 4D ). Interestingly, we found that overexpression of USP2-AS1 increased HIF1α protein levels under hypoxia while USP2-AS1 knockdown reduced the protein levels of HIF1α compared with the control group ( Figures 4E and S4A ). However, no changes in HIF2α mRNA and protein levels were observed after knockdown or overexpression of USP2-AS1 (data not shown), indicating that the effect of USP2-AS1 on HIF1α protein levels is specific.

Previous studies have indicated that YBX1 promotes the translation of HIF1α protein by directly binding to the 5’ UTR of HIF1α mRNA to mediate the metastasis of osteosarcoma (28). At the same time, it indicated that the expression of YBX1 and HIF1α showed a significant positive correlation in the TCGA-LIHC ( Figure S4B ), which is consistent with previous studies. In this study, we confirmed USP2-AS1 binds to YBX1, but whether USP2-AS1 changed the protein levels of HIF1α by affecting the transcription or translation level of YBX1 remains unclear. We found a positive correlation of expression between YBX1 and USP2-AS1 in TCGA-LIHC ( Figure S4C ), so we hypothesized USP2-AS1 may increase HIF1α protein levels by affecting the expression of YBX1. Unfortunately, We further demonstrated that overexpression or knockdown of USP2-AS1 failed to cause significant changes in YBX1 mRNA or protein levels ( Figures 4F, G and S4D ). These results suggested that USP2-AS1 did alter the protein levels of HIF1α under hypoxia, but not affected the transcriptional or translation levels of YBX1.

LncRNAs can inhibit or promote the translation of HIF1α by binding to YBX1 protein without changing the expression level of YBX1 protein itself (29–31). We investigated whether USP2-AS1 could enhance the effect of YBX1 on the translation of HIF1α protein by binding to YBX1. RIP experimental results indicated HIF1α mRNA did bind to YBX1 protein ( Figures 3E, F ), which was consistent with the previous outcomes. We observed that under hypoxic conditions, overexpression of USP2-AS1 increased the binding efficiency of YBX1 to USP2-AS1 and significantly increased the effective binding of YBX1 to HIF1α mRNA ( Figures 5A, B ). The knockdown of USP2-AS1 had the opposite effect on the binding level of YBX1 and HIF1α mRNA ( Figure 5C ).

We used small interference RNA to change the expression of YBX1 in USP2-AS1 overexpression or knockdown cell lines ( Figure 5D ). The effect of overexpression or knockdown of USP2-AS1 on the translation level of HIF1α protein under hypoxia was significantly abrogated after the knockdown of YBX1 ( Figure 5D ). Cell proliferation, migration, and invasion experiments also showed that knockdown of YBX1 eliminated the effect of USP2-AS1 on the growth of HCC cells ( Figures S5A, B , S6A, B ). The results suggest that USP2-AS1 can increase the protein levels of HIF1α by enhancing YBX1 protein binding to HIF1α mRNA under hypoxia, ultimately promoting the growth and development of HCC.

To further explore the role of USP2-AS1 acting on the growth of HCC in vivo, we constructed tumor xenograft model in mice by using USP2-AS1 overexpression or knockdown Huh7 cell line. The tumor xenograft model showed that overexpression of USP2-AS1 promoted the growth of tumor while USP2-AS1 knockdown inhibited the proliferation of HCC in vivo ( Figures 6B–E ). These results suggest USP2-AS1 overexpression or knockdown can affect the growth of tumor in our mice tumor xenograft model.

Hypoxia often mediates the resistance of solid tumors to chemotherapeutic drugs. Lenvatinib is a multikinase inhibitor like sorafenib, which was developed in Japan as a multitargeted kinase inhibitor for the treatment of advanced-stage HCC in 2018. Lenvatinib significantly improved the overall survival and progression-free survival time of patients with advanced HCC compared with sorafenib treatment (32). Sorafenib treatment is also limited by drug resistance. Under hypoxic conditions, transcription factor HIF1α can mediate the transcriptional activation of some genes related to glucose uptake, metabolism, and cell proliferation, such as MAPK1, VEGF, MDR1 (ABCB1), GLUT1, to facilitate the adaptive response of liver tumor cells to hypoxia, thereby increasing the therapeutic resistance of HCC to sorafinib (14). Therefore, the therapeutic effect of sorafenib is increased in liver cancer when it is combined with HIF1α inhibitors (33, 34). However, the mechanisms of HCC resistance to lenvatinib are poorly understood.

Our previous study showed that USP2-AS1 increased the protein levels of HIF1α in an anoxic environment, RT-qPCR also showed that the expression of target genes downstream of HIF1α was also significantly down-regulated in USP2-AS1 knockdown Huh7 cell line ( Figure 6F ), suggesting that USP2-AS1 knockdown may enhance the efficacy of lenvatinib. Therefore, we further explored whether USP2-AS1 knockdown enhanced the therapeutic effect of lenvatinib on HCC in mice tumor xenograft model. We observed that the tumor volume and tumor weights of mice treated with lenvatinib were significantly lower than those of the control group in the HCC xenograft model ( Figures 6D, E ). Importantly, compared with control groups, the tumors’ weight and volume were the smallest in the group treated with lenvatinib and USP2-AS1 knockdown ( Figures 6D, E ), indicating that USP2-AS1 knockdown can enhance the therapeutic effect of lenvatinib in HCC in our mice tumor xenograft model.