UPF1 Promotes Chemoresistance to Oxaliplatin Through Maintenance of Stemness by Increasing Phosphorylated TOP2A in Colorectal Cancer.

Background: UPF1 is proved to dysregulate in multiple tumors and inuence carcinogenesis. However, the role of UPF1 on oxaliplatin resistance in colorectal cancer (CRC) remains unknown. Methods: Firstly, we investigated the clinical relevance of UPF1 in CRC patients. Then, we explored the inuence of UPF1 on chemoresistance to oxaliplatin in vitro and in vivo. Finally, we disclosed the underlying mechanisms of oxaliplatin resistance induced by UPF1. Results: UPF1 is upregulated in CRC and overexpression of UPF1 more likely results in recurrence in CRC patients and predicts a poorer overall survival (OS). UPF1 maintains stemness in CRC cell lines and promotes chemoresistance to oxaliplatin in CRC. UPF1-induced oxaliplatin resistance can be associated with interaction with TOP2A and increasing phosphorylated TOP2A. Conclusions: UPF1 was overexpressed and predicted a poor prognosis in CRC. UPF1 enhanced the stemness and chemoresistance to oxaliplatin by interaction with TOP2A and increase of phosphorylated TOP2A in CRC, which may provide a new therapy strategy for chemoresistance to oxaliplatin in CRC patients.


Introduction
Colorectal cancer (CRC) is the third most frequent for incidence and the second most frequent for mortality worldwide [1]. While in China, CRC is the fourth most common and the fth leading cause of cancer death [2]. Surgical resection is the potential radical treatment for CRC. Other than surgery, radiochemotherapy, targeted therapy and immunotherapy can be applied to CRC patients with lymph node metastases or distant metastases [3][4][5]. However, the acquisition of drug resistance is a major hurdle for good clinical prognosis [6].
Oxaliplatin, a third-generation platinum coordination complex, can be used for treatment in multiple cancers [7][8][9]. Oxaliplatin-based combined chemotherapy is routinely applied in advanced and metastatic CRC and signi cantly increases the overall survival rate and metastases resection rate [10].
Unfortunately, acquired or intrinsic resistance brings failure to treatment. The underlying mechanisms of oxaliplatin resistance include DNA damage response and repair, inhibition of cell death, cellular transport, detoxi cation and epigenetic alteration [11]. UPF1, an mRNA surveillance factor, is a RNA-dependent ATPase and helicase for nonsense-mediated decay (NMD) of mRNAs containing premature stop codons [12]. In addition, UPF1 is proved to dysregulate and in uence carcinogenesis in multiple tumors. UPF1 is commonly mutated in pancreatic adenosquamous carcinoma (ASC) and there is little or no UPF1 expression in many ASC tumors compared to adjacent normal tissue [13]. It is revealed that UPF1 is downregulated in hepatocellular carcinoma and inhibits the tumor progression [14,15]. However, the role of UPF1 in oxaliplatin resistance in CRC remains unclear.

Materials And Methods
Patients and specimens 76 patients were pathologically diagnosed with CRC and received surgical therapy at Fudan University Shanghai Cancer Center. Patients received neoadjuvant radiochemotherapy were excluded from our study. Specimens from the 76 patients were made into tissue microarrays (TMA). Our study was carried

Lentivirus production and transfection
Core plasmid was co-transfected with psPAX2 and pMD2.G into HEK-293T cells using Hieff Trans TM Liposomal Transfection Reagent (Yeasen, Shanghai, China). 48 hours later, virus supernatant was collected. 3×10 5 cells were cultured in a 6-well plate. After incubation in virus supernatant for 48 hours, stable cell lines were selected with puromycin or ow cytometry and transfected e ciency was evaluated by immunoblot.

Flow cytometry
Apoptosis assay was conducted using Annexin V-PE/7-AAD apoptosis detection kit (Yeasen, Shanghai, China). Brie y, 3×10 5 cells were cultured in a 6-well plate with or without being treated with oxaliplatin for 48 hours. Adherent cells and cells in supernatant were collected and washed by phosphate buffer saline (PBS) twice. Cells suspended with 1× binding buffer with the addition of 5 μL Annexin V and 10 μL 7-AAD.
After incubation protecting from light for 15 minutes, apoptosis rate was detected by Flow cytometer.

Western blotting
Total protein was extracted with NuPAGE ® LDS sample buffer (Thermo Fisher Scienti c, Waltham, USA). A certain amount of protein was separated by SDS-PAGE gel and transferred to PVDF membranes. After blockage with non-fat milk powder and incubation with primary antibodies and horseradish peroxidaseconjugated secondary antibodies, images were captured with ImageQuant™ biomolecular imager. All the antibodies used in our study were listed in additional le named Table S1.
Drug cytotoxicity assay Drug cytotoxicity was evaluated by Cell Counting Kit-8 (CCK-8) assay. Cell suspension was seeded in 96well plate and cells were treated with a gradient concentration of oxaliplatin for 48 hours after adhesion. Medium containing CCK-8 reagent (Yeasen, Shanghai, China) was added to each well of the plate. After two-hours' incubation, absorbance at 450nm was detected by microplate reader.
Clone formation assay 1000 cells were cultured in a 6-well plate and treated with oxaliplatin for 48 hours after adhesion. The medium was replaced by the fresh every three days. 2 weeks later, cells were xed with 4% paraformaldehyde for 30 minutes and stained with crystal violet for 30 minutes. Images were photographed and visible colonies were counted.
Co-immunoprecipitation (Co-IP) assays and mass spectrometry Cells in culture dish were lysed with IP buffer containing protease inhibitors and phosphatase inhibitor cocktail (Topscience, Shanghai, China). IP buffer was a mixture of 1 mM EDTA, 20 mM HEPES, 150 mM NaCl, 0.05% sodium deoxycholate and 0.05% NP-40. Lysate was spun at the speed of 12,000 rpm at 4 ℃ for 12 minutes. 60 μL was kept for input and the rest of the extract was immunoprecipitated with anti-HA beads or anti-FLAG resins at 4 °C for more than 3 hours. After washing 5 times with IP buffer, beads or resins of protein bound were lysed with 1.25× SDS loading buffer and boiled at 100 °C for 10 minutes. Bands appeared after electrophoresis and silver staining in SDS-PAGE gel received mass spectrometry analysis (Wayen Biotechnologies, Shanghai, China) for protein identi cation. Interacted protein was further con rmed by immunoblot.

Mammosphere formation assay
Mammosphere formation assay was conducted as previously described [16]. Cells were digested by trypsin and washed by PBS twice. 300 cells suspended in 200 μL mammosphere medium were seeded in 96-well plate with ultra-low attachment surface. The formulation of mammosphere medium was shown in additional le named Table S2. 10 days later, images were photographed and mammosphere were counted.

Immunocytochemistry (ICC)
Sterilized round coverslips were laid in a 24-well plate and cells were seeded on them. After 24 hours incubation, cells were xed with 4% paraformaldehyde for 30 minutes and permeabilized on ice with 0.25% Triton X-100. Then cells were blocked with BSA and incubated with primary antibodies for 2 hours and uorescent secondary antibodies protecting from light for 1 hour at room temperature. After mounting with DAPI Fluoromount mounting medium for 15 minutes, images were captured using Leica confocal system. Xenograft experiments 5×10 6 cells were subcutaneously injected into four-week-old male BALB/c nude mice. 5 days later, mice were randomly divided into groups and treated with oxaliplatin (5mg/kg) or 5% glucose solution by intraperitoneal injection, twice per week for three weeks, respectively. All mice were sacri ced and tumors were collected and weighed. Tumor volume equaled length × width^2 × 0.5 and was measured twice a week. Xenografts were saved in 4% paraformaldehyde for following experiments. Tumor weight and volume were presented as means ± standard deviation (SD). Apoptosis rates in the xenograft were determined by terminal deoxynucleotidyl transferase mediated dUTP nick end labeling (TUNEL) technology. Xenograft experiments were approved by the Committee on Animals Handling of Fudan University Shanghai Cancer Center.

Statistical analysis
Data are shown in mean ± SD. All analyses were performed by IBM SPSS 22.0 software. Quantitative variables were analyzed using Student's t-test. The Log-rank test in the Kaplan-Meier method and the Cox regression model were used to assess patients' survival outcome and prognostic factors. All the experiments were performed in triplicate. A two-tailed value of P < 0.05 was considered statistically signi cant (*P < 0.05; **P < 0.01; ***P < 0.001).

UPF1 is upregulated and predicts a poor prognosis in CRC.
To investigate the expression of UPF1 in CRC tissues, we performed bioinformatics analysis of UPF1 in mRNA level using the public RNA sequencing (RNA-seq) datasets from The Cancer Genome Atlas (TCGA) and found that UPF1 was overexpressed (P = 0.043, Fig. 1a). we also detected it by IHC staining of TMA using a retrospective cohort containing 76 patients with paired CRC tumor and normal tissues. Similarly, the results veri ed that UPF1 was dramatically upregulated in CRC tissues (P < 0.001, Fig. 1c). The representative IHC images of UPF1 staining were shown in Fig. 1b. In these patients, UPF1-positive was detected in 52 (68.4%) of the tumor tissues, whereas only 4 (5.3%) of the normal tissues was UPF1positive (Table 1, P < 0.001). To explore the clinical signi cance of UPF1 in CRC, we analyzed the correlation between the expression of UPF1 and clinicopathological characteristics of CRC patients (Table 1). A signi cant association was observed between UPF1 positive and TOP2A positive (P = 0.012, 58.3% in UPF1-negative group vs 84.6% in UPF1-positive group). High UPF1 expression (P = 0.035) and high TNM stage (P = 0.040) more likely result in recurrence in CRC patients. However, high expression of UPF1 and recurrence in CRC patients had no correlation with gender, age, tumor location, tumor size, lymphovascular invasion, perineural invasion and carcinoembryonic antigen (CEA) level. Additionally, Kaplan-Meier curves showed a strong correlation between UPF1 higher expression and a poorer overall survival (OS) (P = 0.045, Fig. 1d). UPF1positive patients also had a shorter recurrence-free survival (RFS) (P = 0.029, Fig. 1e). In addition, univariate and multivariate COX regression analysis suggested that UPF1-positive (HR = 3.719; 95% CI = 1.313 to 10.529; P = 0.013) and poor differentiation (HR = 2.927; 95% CI = 1.186 to 7.222; P = 0.020) were independent prognostic risk factors for RFS in CRC (Fig. 1f, 1 g). In summary, UPF1 is aberrantly upregulated in CRC and high expression of UPF1 predicts a poor prognosis in CRC patients.
UPF1 promotes oxaliplatin resistance in CRC in vitro.
The baseline expression of UPF1 in CRC cell lines and normal colonic epithelial cell line was detected by immunoblot. Notably, UPF1 expression in NCM460 was remarkably less compared with CRC cell lines ( Fig. 2a). We chose DLD1 to overexpress UPF1 due to its relative low baseline expression. And based on the same rule, UPF1 was knocked down in HCT116 using three small hairpin RNAs. Overexpression and knockdown were con rmed by western blotting and SH2 was the most e cient in UPF1 knockdown (Fig. 2b, 2c). HA and FLAG tag were used for co-immunoprecipitation (Co-IP) assay. Drug cytotoxicity assay showed that overexpression of UPF1 decreased oxaliplatin sensitivity in DLD1 while knockdown of UPF1 increased oxaliplatin sensitivity in HCT116 (Fig. 2d, 2e). In apoptosis assay, overexpression and knockdown of UPF1 had no in uence on apoptosis rate in DLD1 and HCT116 respectively. Whereas after treatment for 48 hours, oxaliplatin-induced apoptosis was signi cantly reduced in DLD1-UPF1 and raised in HCT116-shUPF1 (Fig. 2f-2i). Clone formation assay also a rmed that UPF1 promoted oxaliplatin resistance in CRC in vitro. The number of colonies were more in DLD1-UPF1 group and less in HCT116-shUPF1 group compared with control group with treatment of oxaliplatin for 48 hours (Fig. 2j, 2 k). However, the size of colonies remained no obvious change between different groups, which indicated that UPF1 may have no effect on cell proliferation. CCK-8 assay also showed that UPF1 did not in uence proliferation in DLD1 and HCT116 cell lines (Fig. S1d, S1e).

UPF1 promotes oxaliplatin resistance in CRC in vivo
Nude mice were divided into 2 groups and injected with HCT116-shNC and HCT116-shUPF1 cell lines respectively. One half of each group was treated with 5% glucose solution (GS) and the other was treated with Oxaliplatin (5 mg/kg). The tumor xenografts were shown in Fig. 3a. In the groups treated with GS, there is no signi cant difference between HCT116-shNC and HCT116-shUPF1 groups. Whereas, xenografts in HCT116-shNC group grew faster and were much larger and heavier after treatment with oxaliplatin (Fig. 3b, 3c). In line, these ndings were further con rmed by TUNEL assay (Fig. 3d). Apoptosis rate in tissues remained no change in HCT116-shUPF1 but after being treated with oxaliplatin, apoptosis rate remarkably raised compared with control. xenografts assay in vivo revealed that UPF1 promotes oxaliplatin resistance in CRC.
UPF1 interacts with TOP2A and increases the level of phosphorylated TOP2A.
The protein in DLD1-UPF1 and DLD1-Ctrl cell lines with FLAG tag was lysed by IP buffer. After immunoprecipitation with anti-FLAG resin, protein extracts were separated in 10% SDS-PAGE gel. some signature bands of DLD1-UPF1 cell line emerged after silver staining (Fig. 4a). SDS-PAGE gel containing signature bands was used for mass spectrometry. 167 kind of proteins were identi ed which were listed in additional le named Table S3. TOP2A, TOP1, CCAR2 and XRCC6 were selected as candidate proteins interacted with UPF1 according to pathway analysis (Fig. S1f) and biological function. The interaction between UPF1 and TOP2A was proved in co-IP assay. The results showed co-IP of TOP2A with UPF1-FLAG using an anti-FLAG antibody. Similarly, UPF1 was also precipitated using an anti-HA antibody in HEK-293T cell line transfected into TOP2A-HA (Fig. 4b, 4c). In DLD1 cell line transfected into UPF1-FLAG, the results of ICC also indicated the colocalization of UPF1 and TOP2A (Fig. 4e). We further found that expression of total TOP2A remained no signi cant change but phosphorylated TOP2A at the site of Ser 1106 was increased after upregulation of UPF1. Correspondingly, silencing UPF1 had no in uence on total TOP2A expression but attenuated phosphorylated TOP2A at the same site (Fig. 4d).
To identify the binding region between UPF1 and TOP2A, we generated the truncated mutants of UPF1 and TOP2A to perform co-IP assays. The full length of TOP2A contains 1531 amino acids and six truncated mutants were generated (Fig. 5a, 5b). Co-IP assay revealed that Toprim domain may be the binding region with UPF1 (Fig. 5c). The full length of UPF1 contains 1129 amino acids and ve truncated mutants were generated (Fig. 5d, 5e). The results demonstrated that the zinc nger domain of UPF1 may interact with TOP2A (Fig. 5f).
To verify the function of TOP2A in UPF1-induced oxaliplatin resistance, TOP2A was silenced using three small hairpin RNAs. Sh1 was selected for following experiments due to its biggest e ciency (Fig. 6a). Of note, drug cytotoxicity assay demonstrated that UPF1-induced resistant cells regained the sensitivity to oxaliplatin after knockdown of TOP2A (Fig. 6c). As shown in clone formation assay, DLD1-Ctrl, DLD1-UPF1 and DLD1-UPF1-shTOP2A were treated by oxaliplatin for 48 hours. We can see that UPF1 overexpression signi cantly increase the number of colonies while the effect was abolished by silencing the TOP2A gene in the cells. The size of colonies had no difference among different groups (Fig. 6d, 6e). The results in apoptosis assay came to the same conclusion (Fig. 6f, 6 g). In vivo, xenografts in DLD1-UPF1 group were much larger and heavier under treatment with oxaliplatin. The trend of chemoresistance was brought to a halt with TOP2A knockdown (Fig. 6h). In same, TUNEL assay testi ed that inhibition of apoptosis in tissue induced by UPF1 was deactivated by silencing TOP2A (Fig. 6i). Altogether, these data strongly suggested that TOP2A played an essential role in the UPF1-induced chemoresistance to oxaliplatin in CRC.
UPF1 maintains stemness of CRC cell lines.
Cancer stem cells (CSCs) are considered as a subpopulation of cells in the tumor characterized by an increased ability to self-renew and seed secondary tumors [18]. The ability was de ned as CSC-like properties or stemness. Studies have identi ed an essential role of CSCs in the drug resistance in colorectal cancer progression [19,20]. Mammosphere formation assay can be used to de ne the functional CSC-like properties in vitro and larger size and bigger number of mammosphere indicate enhanced stemness. In our study, the ability of mammosphere formation was enhanced after upregulation of UPF1 in DLD1 and was attenuated after knockdown of UPF1 in HCT116 (Fig. 7a, 7b).

Discussion
CRC maintains a high rate of incidence and mortality worldwide, as well as that in China [1,2]. For CRC patients in stage , resection is enough and yet combined chemotherapy should be applied in patients in high-risk stage and stage / . Stage CRC exhibits a recurrence rate of about 20% after radical resection [24]. For patients with stage disease, recurrence rate can be as high as more than 50% [25]. Chemotherapy resistance is a major cause of recurrence and poor prognosis in CRC patients [26].
Oxaliplatin, irinotecan and uoropyrimidine are three backbones of rst-line systemic treatments for CRC. As a single agent, oxaliplatin has limited e cacy applied as treatment for CRC patients [27]. Combination of oxaliplatin and uorouracil is promising in treatment for the disease due to its highly synergistic effect [28]. Oxaliplatin-based combined chemotherapy is routinely applied in advanced and metastatic CRC and signi cantly increases the overall survival rate and metastases resection rate [10]. Unfortunately, drug resistance brings failure to treatment, which becomes a pressing problem in clinical treatment in CRC patients.
In our study, we used wild type CRC cell line HCT116 (HCT116-wt) to select over times with a gradual increasing concentration of oxaliplatin for acquired oxaliplatin-resistant cell line (HCT116-L). RNA-seq (Genminix Informatics, Shanghai, China) was used for transcriptome-wide analysis of differential gene expression in HCT116-wt and HCT116-L cell lines (Fig. S2). And we found that UPF1 was highly expressed in HCT116-L cell line.
It is reported that UPF1 can play a role in regulated mRNA and protein decay. Feng Q et al. reported that UPF1 acted as an E3 ubiquitin ligase to repress human skeletal muscle differentiation [29]. In addition, UPF1 is proved to dysregulate in multiple tumors and in uence carcinogenesis [13][14][15]. However, there is lack of detailed study of UPF1 in CRC. It is reported that UPF1-mediated NMD might play a part in the selection of target gene mutations with a functional role in MSI-H carcinogenesis [30]. In line, Ada Collura et al. ascertained that inhibition of NMD in vivo using amlexanox reduced MSI tumor growth [31]. Currently, the function of UPF1 and potential mechanisms remains unclear in CRC.
Firstly, we demonstrated that UPF1 was aberrantly overexpressed in CRC tissues in mRNA and protein levels by bioinformatics analysis and IHC staining respectively. And high UPF1 expression more likely resulted in recurrence and predicted a worse OS in CRC patients. In addition, COX regression analysis suggested that UPF1-positive (HR = 3.719; 95% CI = 1.313 to 10.529; P = 0.013) was an independent prognostic risk factor for RFS in CRC.
Herein, how UPF1 acts as a oncogene in CRC needed deep study. CCK-8, clone formation assays and ow cytometry indicated that UPF1 may have no effect on cell proliferation and apoptosis. While assays in vitro showed that overexpression of UPF1 weakened sensitivity to oxaliplatin in DLD1 while knockdown of UPF1 reinforced sensitivity to oxaliplatin in HCT116. In vivo, we reached the same conclusion that UPF1 promotes oxaliplatin resistance in CRC.
Cancer stem cells (CSCs) are de ned by their functional properties and could be able to self-renew and propagate the tumor [32]. CSC-like properties or stemness have an essential role in the drug resistance in CRC [19,20]. Mammosphere formation assay can be used to de ne the functional stemness in vitro [16]. In our study, the stemness was progressed in UPF1-upregulation cell line and was impeded in UPF1knockdown cell line. CSCs are characterized by speci c markers, such as EpCAM. Flow cytometry indicated that the percentage of EpCAM-positive cells was diminished after silencing UPF1 and increased after upregulating UPF1.
Afterwards, the underlying mechanism of UPF1-induced oxaliplatin resistance remained to be revealed. Mass spectrometry analysis identi ed 166 kind of proteins that may interact with UPF1. Enrichment analysis exhibited related pathways and TOP2A was unearthed to be involved in platinum-resistant pathway. TOP2A, DNA topoisomerase, an enzyme that controls and alters the topologic states of DNA during transcription. Phosphorylation of serine 1106 in the catalytic domain of TOP2A regulates enzymatic activity and drug sensitivity [33,34]. TOP2A is upregulated and could induce tumor development and progression in multiple tumors [35][36][37][38] and is proved important therapeutic targets of anticancer agents [39]. Etoposide, as an inhibitor of TOP2A, is a kind of cell cycle speci c antitumor drug and applied in multiple tumors. In CRC tissues, TOP2A was proved to be upregulated in mRNA and protein levels (Fig. S1a-S1c). In the IHC staining of TMA, a signi cant association was observed between UPF1 positive and TOP2A positive. TOP1 could modulate colorectal cancer response to irinotecan [40].
CCAR2 overexpression decreased the chemosensitivity to oxaliplatin in CRC [41]. XRCC6 as a DNA repair gene may participate in platinum resistance by modulating the DNA repair capacity [42]. TOP2A, TOP1, CCAR2 and XRCC6 were selected as candidates according to pathway analysis and biological function.
Forward and reverse co-IP assay proved the interaction of UPF1 with TOP2A, instead of TOP1, CCAR2 and XRCC6. In DLD1 transfected into UPF1-FLAG, the results of ICC assays also indicated the colocalization of UPF1 and TOP2A. To identify the binding region between UPF1 and TOP2A, we generated the truncated mutants of UPF1 and TOP2A to perform co-IP assays. Toprim in TOP2A is a catalytic domain involved in DNA strand breakage and rejoining and may be the binding region with UPF1. The zinc ngers vary widely in structure, as well as in function. And this domain of UPF1 may interact with TOP2A. We further found that expression of total TOP2A remained no signi cant change after upregulation of UPF1. While the interaction with UPF1 increased phosphorylated TOP2A at the site of Ser 1106 thus may enhance the enzyme activity of TOP2A. Of note, UPF1-induced resistant cells regained the sensitivity to oxaliplatin after knockdown of TOP2A. We can see that UPF1 overexpression signi cantly reduced cytotoxicity induced by oxaliplatin while the resistance was abolished by silencing the TOP2A gene in the cells. Notably, UPF1-induced CSC-like properties was abated by silencing the expression of TOP2A. Altogether, these data strongly suggested that TOP2A played an essential role in the UPF1-induced chemoresistance of CRC cells in response to oxaliplatin in vitro and in vivo and TOP2A could be a therapeutic target in UPF1-overexpressed CRC patients.
There are some limitations in this study that could be addressed in future research. We have not gured out how interaction between UPF1 and TOP2A increases the level of phosphorylated TOP2A. And we wonder whether combination with the drugs targeted for TOP2A could make synergistic effect in patients with highly expressed UPF1.

Conclusions
UPF1 was overexpressed and predicted a poor prognosis in CRC. UPF1 promoted the stemness and chemoresistance to oxaliplatin by interaction with TOP2A and increasing phosphorylated TOP2A, which may provide a new therapy strategy for chemoresistance to oxaliplatin in CRC patients. Availability of data and materials The data used to support the ndings of this study are included within the article and the supplementary materials.

Competing interests
The authors declare that they have no competing interests.

Funding
The nancial support of our study was from the Science and Technology Commission of Shanghai Municipality (19511121202, 20DZ1100100).

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.