Human pancreatic cancer cells (including AsPC-1, BxPC-3, CFPAC-1, PANC-1 and SW1990) and a normal human pancreatic duct epithelial cell line (HPDE6-C7) were commercially obtained from National Model and Characteristic Laboratory Cell Resource Bank, Chinese Academy of Sciences (Shanghai, China) and Bnbio Biotechnology Research Institute (Beijing, China). IMDM/DMEM/RPMI-1640 (Gibco, Grand Island, NY, United States) medium containing 10% FBS (C04001-500, VivaCell, Shanghai, China) and 1% penicillin-streptomycin mix solution (SV30010, HyClone, Logan, UT, United States) were applied to culture cells in 37°C conventional incubator with 5% CO₂. Gemcitabine (GEM) and deoxyelephantopin (DET) were purchased from Gloria Pharmaceuticals (Harbin, China) and BioBioPha (Kunming, China), and dissolved in cell culture grade saline and dimethyl sulfoxide (DMSO) to prepare stock solutions, the specific procedures have been validated in our previous studies (Ji et al., 2020). The DET formula is C₁₉H₂₀O₆, molecular weight is 344.36, and CAS number is 29307-03-7. The GEM-resistant PC cell line (BxPC-3/GR and CFPAC-1/GR) used in present study was induced through continuous different concentration of GEM stimulation (Zhou et al., 2020). Finally, the BxPC-3/GR and CFPAC-1/GR cell lines that could persist in the concentration of 1 μM (for BxPC-3/GR) or 500 nM (for CFPAC-1/GR) GEM were obtained. To ensure better cells status, a fresh tube of frozen cells was resuscitated at an interval of 3 months and the mycoplasma detection (CA1080, Solarbio, Beijing, China) was conducted.

A total of 74 pairs of PC tissue samples, matched paracancerous tissue samples and clinicopathological data were jointly acquired from the Second and the Forth Affiliated Hospital of Harbin Medical University. Each tissues were divided into two parts, one part was directly cryopreserved in liquid nitrogen and the other part was immersed into 4% paraformaldehyde solution. The written informed consent was obtained from all patients involved in our present study. Additionally, the operation also obtained approval from the Ethics Board of Harbin Medical University and followed the basic principles of the Declaration of Helsinki.

Cell Counting Kit-8 (CCK-8) and 5-ethynyl-2' -deoxyuridine (EdU) assays were carried out to evaluate the cell viability. For CCK-8 assay, 3 × 10³ cells were inoculated into 96-well plates and cultured for different time points. After that, 10 μL of CCK-8 reagent (GK10001, Glpbio, Montclair, CA, United States) was added to each well and continued to incubate for 2.5 h. Finally, the specific absorbance value at 450 nm was obtained with a multimode microplate analyzer (SCR_024560, Tecan, Mannedorf, Switzerland) and the cell viability was calculated. EdU assay was conducted using EdU-555 Cell Proliferation Kit (C0075S, Beyotime, Shanghai, China). The cells were incubated with 10 μM EdU working solution for 2 h and then fixed with paraformaldehyde solution. After staining the nucleus with Hoechst 33342, the EdU positive labeling cells were analyzed under a fluorescent microscope (SCR_000011, Leica, Wetzlar, Germany).

In order to further examine the ability of cell proliferation, clone forming approaches were performed. 800 cells were seeded into 6-well plated and continuously cultured for 10-12 days under conventional conditions. Next, the visible cell colonies which contained at least 50 cells in each well were fixed and stained. Finally, the colonies were photographed under a microscope and the clone formation rates between different groups were assessed.

The mobility of PC cells was primarily determined using scratch wound healing assay. In brief, cells on logarithmic phase were inoculated into 6-well plate, and a 200 μL sterile micropipette tip and experimental ruler were utilized to scratch longitudinally straight cell-free areas. After rinsing away floating cells with PBS buffer, the cells were cultured under FBS-free medium, and the percentage of regional confluence from different points was compared using inverted microscope.

The migration and invasion abilities of PC cells were further detected using Transwell assays (725321, Nest, Wuxi, China) holding chambers with 8 μm aperture polyethylene terephthalate membrane. PC cells (5 × 10⁴ cells for BxPC-3 and SW1990, 4 × 10⁴ cells for CFPAC-1 and 6 × 10⁴ cells for PANC-1) resuspended in 200 μL FBS-free medium were seeded into the upper chamber module, and 600 μL complete medium were added to the lower chamber module as the putative chemoattractant. After 24 h routine culture, the cells adherent to the lower outer-field portion of the upper chamber were fixed with 4% paraformaldehyde and stained with 0.1% crystal violet aqueous solution. For invasion assay, the membrane inner side of upper chamber module was pre-enveloped with 100 μL Matrigel matrix (HY-K6001, MCE, Princeton, NJ, United States) and the remaining steps were essentially same. The numbers of cell permeating membrane were counted under inverted microscope.

Acridine Orange (AO)/Ethidium Bromide (EB) and Hoechst 33342 fluorescent staining assays were applied for apoptosis detection of PC cells. For AO/EB assay, cells were seeded in 48-well culture plate with a density of 1 × 10⁴ per well. After stimulation with different strategies, the cells were rinsed with PBS buffer, then the AO and EB reagents were added at a concentration of 100 μg/mL. After 20 min of reaction, the apoptosis rate was calculated via the ratio of green fluorescence to red fluorescence obtained by a fluorescence microscope (SCR_000011, Leica, Wetzlar, Germany). In Hoechst 33342 assay, the method of cell inoculation and stimulation was the same as above. After fixation with 4% paraformaldehyde for 15 min, 200 μL of Hoechst 33342 (40731ES, Yeasen, Shanghai, China) were dripped into culture well and incubated for another 20 min. The apoptosis was assessed by relatively strong bright blue fluorescence.

Flow cytometry assay was performed to further evaluate the cell apoptosis. In brief, after treatment with different conditions, the cells were collected with 0.25% EDTA-free trypsin. After rinsed three times with PBS and filtered with a 200-mesh strainer, the cells were resuspended in 500 μL buffer solution, then 5 μL of PI and 5 μL Annexin-Ⅴ (556547, BD Biosciences, Franklin Lakes, NJ, United States) were added. The apoptosis ratio was determined through the sum of data from Q2 and Q4 quadrants. Q1 quadrant represents the mechanically damaged cells. Q2 quadrant represents the non-viable apoptotic cells. Q3 quadrant represents the normal viable cells. Q4 quadrant represents the viable apoptotic cells.

The stem cells self-renewal was evaluated using three-dimensional spheroid formation assay. Briefly, 600 cells were seeded in 24-well ultra-low adhesion culture plate (3,473, Corning, NY, United States) with 500 μL optimal stem cell culture medium per well, and 50 μL medium was added every 2 days. After 2 weeks of routine cell incubation, the number of spheroid with a diameter ≥75 μm was counted under microscope, and the spheroid formation rate was calculated. The stem cell culture medium was composed of DMEM/F12 basic medium appended with 1×B27, 20 ng/mL bFGF, 20 ng/mL EGF and 4 μg/mL insulin.

The target genes of circTNPO3 were predicted through CircBank (http://www.circbank.cn/), ENCORI (https://rnasysu.com/encori/) and TargetScan (https://www.targetscan.org/vert_80/). The binding sites between miR-188-5p and CDCA3 were predicted by ENCORI. The expression level of CDCA3 in PC tissues and its relationship with patient survival were analyzed using TCGA (https://www.cancer.gov/ccg/research/genome-sequencing/tcga), GEPIA (http://gepia.cancer-pku.cn/) and TNMplot (https://tnmplot.com/analysis/) databases. The correlation analysis between CDCA3, TRAF2 and RelA (NF-κB-p65) was carried out based on the TIMER2.0 database (http://timer.cistrome.org/).

In order to explore the binding capacity between circTNPO3 and miR-188-5p, and between miR-188-5p and CDCA3, Dual-Luciferase Reporter Assay (E2920, Promega, Madison, WI, United States) was applied. Briefly, the wild type (WT)/mutant (MUT) of circTNPO3/CDCA3 gene fragments were designed and added to pmirGLO vector. After transfection of blank control vector, mimics negative control and miR-188-5p mimics into cells using Lipofectamine 3000, the fluorescence signal intensity was detected.

The total protein was extracted from cells under different conditioned stimulus using RIPA lysis buffer added with protease inhibitor (G2006, Servicebio, Wuhan, China). After cell spatula scraping and ultrasonic concussion, the protein lysis was collected using 4°C low temperature precooling centrifuge at 12,000×g for 15 min. To detect the translocation of NF-κB-p65 from cytoplasm to nucleus, the Nuclear and Cytoplasmic Protein Extraction Kit (P0027, Beyotime, Shanghai, China) was applied. The protein concentration was quantified using Pierce BCA Protein Assay Kit (23227, Thermo Scientific, Waltham, MA, United States). After denaturation at 100°C metal bath for 5 min, the protein was separated using 10% SDS-PAGE (G2067-50T, Servicebio, Wuhan, China) and transferred to 0.45 μm pore size PVDF blotting membrane (10600029, Cytiva, Wilmington, DE, United States). Then, the PVDF membrane was blocked with 5% (W/V) skim milk for 1 h and incubated with primary antibody overnight at 4°C. Finally, after incubation with the second antibody, the exposure of protein blot was achieved using enhanced chemiluminescence agent (MA0186, Meilunbio, Dalian, China) under biomolecular imaging system (Tanon, Shanghai, China). The primary antibodies involved in this study were listed below: rabbit anti-CDCA3 antibody (bs-7894R, 1:1,500, Bioss, Beijing, China), rabbit anti-TNPO3 antibody (T58524S, 1:2000, Abmart, Shanghai, China), rabbit anti-GRPDH antibody (ab181602, 1:8,000, Abcam, Shanghai, China), rabbit anti-cleaved-caspase 9 antibody (AF5240, 1:1,500, Affinity, Jiangsu, China), rabbit anti-cleaved-caspase 3 antibody (AF7022, 1:1,500 Affinity, Jiangsu, China), rabbit anti-TRAF2 antibody (T55841S, 1:2000, Abmart, Shanghai, China), mouse anti-p65 antibody (sc-8008, 1:1,000, Santa Cruz, Dallas, TX, United States).

After conditioned stimulus, the total RNA, cytoplasmic and nuclear RNA were extracted using RNAsimple Total RNA Kit (DP419, TIANGEN, Beijing, China) and Cytoplasmic and Nuclear RNA Purification Kit (NGB-21000, NORGEN, Ontario, Canada), respectively. The cDNA synthesis and real-time PCR were conducted using QuantiTect Reverse Transcriptase Kit (205311, QIAGEN, Redwood City, CA, United States) and Talent qPCR PreMix (FP209, TIANGEN, Beijing, China) according to the supplier’s protocol. U6 and GAPDH were selected as a reference gene, the quantitative expression of target genes was analyzed through 2^−ΔΔCt formula. The primer sequences for qRT-PCR assay were listed in Supplementary Table S1.

The circular attribute of circTNPO3 was confirmed using RNase R (LGCRNR07250, Sigma, Taufkirchen, Germany) treatment. Total RNA was extracted from PC cells and the concentration was detected by applying ultraviolet spectrophotometer. Next, a certain amount of total RNA (≤5 μg) was treated with RNase R in a ratio of 3 U/μg RNA. After co-incubation at 37°C for 20 min, the RNA was isolated and purified by the TRIzol Plus RNA Purification Kit (12183555, Thermo Scientific, Waltham, MA, United States).

The half-life of target mRNA and circRNA was monitored by administering actinomycin D (A4262, Sigma, Taufkirchen, Germany). After cells were cultured in medium containing10 μg/mL of actinomycin D for different duration, the total RNA was extracted. Finally, the half-life of RNA was analyze through reverse transcription and qRT-PCR.

The subcellular localization of circTNPO3 in PC cells was detected using Fluorescence in Situ Hybridization Kit for RNA (R0306S, Beyotime, Shanghai, China) according to the supplier’s instructions. PC cells were seed in 35 mm confocal petri dishes, fixed with 4% polyformaldehyde and permeated with proteinase K. Then the cells were co-incubated with Cy3-labeled circTNPO3 antisense probes (designed and synthesized by GenePharma, Suzhou, China) in hybridization solution overnight. After nucleus counterstained with DAPI, circTNPO3 localization was analyzed through the distribution of red fluorescence.

The expression of circTNPO3 in PC specimens was tested by digoxin (DIG)-labeled circTNPO3 probes (synthesized by GenePharma, Suzhou, China). Paraffin sections of PC tissues were deparaffinized by different concentrations of xylene and ethanol, executed antigen retrieval using proteinase K and co-incubated with circTNPO3 probes at 40°C for 12 h. The signals of circTNPO3 and nucleus in pathologic tissue slices were displayed after visualization processing by applying anti-DIG-POD antibody, DAB chromogenic reagent and hematoxylin, respectively. The expression level of circTNPO3 was quantified through staining intensity (scored: 0, none; 1, light brown; 2, brown; 3, dark brown) and percentage of positively stained cells (scored: 1, <25%; 2, 25%-50%; 3, 50%-75%, 4, 75%-100%) comprehensively.

Female BALB/c nude mice aged 4–6 weeks were purchased from Charles River Laboratories (Beijing, China) and raised in individual ventilated cages system (IVC) with the standard of specific pathogen free (SPF). After acclimation for 1 week, the lab mice were randomly divided into 3 groups of 5. In the subcutaneously xenograft tumor model, 100 μL of cell suspension containing 4 × 10⁶ conditionally stimulated cells were subcutaneously grafted into the left axilla of nude mice. Tumor growth curves were plotted using the tumor volume measured every 3 days. The tumor volume was calculated via the following formula: Tumor volume = π/6 × longest diameter × shortest diameter². After 36 days of aborative feeding, the mice were sacrificed through spinal cord transaction. The tumors were dissected, weighed, measured and split in two parts, with one soaked in 4% paraformaldehyde solution and the other preserved in liquid nitrogen immediately. To investigate the effect of circTNPO3 on metastasis of PC, the lung metastatic tumor model was constructed. 200 μL of cell suspension containing 3 × 10⁶ cells were transferred into mice lung via rapid tail vein injection using Vascular Visualization System (Yiyan, Jinan, China). After 6 weeks of meticulous care, the mice were euthanized. And the lungs were dissected and preserved in fixative solution for hematoxylin-eosin staining. The level of lung metastasis was comprehensively assessed through the number of globular nodules and the lesion of pulmonary alveoli. The above animal research has been approved by the Ethics Committee of Harbin Medical University, and the review file number is SYDW2020-058.

SPSS Statistics 28.0.1 software (https://www.ibm.com/docs/zh/spss-statistics) was used for quantitative analysis of data. The experimental data in the present research were repeated at least three times, and listed as the mean ± standard deviation (SD). The statistical diagrams were visualized using GraphPad Prism 8 (https://www.graphpad-prism.cn/). The Student’s t-test was introduced to analyze the difference between the two sets of data. Kaplan–Meier curve was applied to compare the overall survival of cohorts with different circTNPO3 expression levels. The correlations between miR-188-5p, CDCA3, TRAF2 and NF-κB-p65 were analyzed using Pearson correlation coefficient, respectively. P value of less than 0.05 was defined as statistically significant and marked with asterisk (*) or pound sign (#).

Based on the chromosome localization and splicing schematic (Figure 1A) of has_circ_0001741 (circTNPO3), the divergent primers crossing the junction (cyclization of exons 2-4 of a TNPO3 mRNA transcript) were designed. As shown in Figure 1B, the Sanger sequencing verified the back-splicing site of circTNPO3. In order to illustrate the cyclization characteristic of circTNPO3, the primers for amplifying circular and linear TNPO3 were synthesized. As shown in Figure 1C, the data of agarose gel electrophoresis implied that circTNPO3 could only be amplified from cDNA of BxPC-3 and PANC-1 cells using divergent primers, not from gDNA. By comparison, the linear TNPO3 could be amplified from both cDNA and gDNA using convergent primers. Next, the expression level of circTNPO3 in PC cell lines was analyzed using qRT-PCR. As shown in Figure 1D, the expression of circTNPO3 was significantly increased in PC cells (AsPC-1, BxPC-3, CFPAC-1, PANC-1 and SW1990) compared with that in normal human pancreatic ductal epithelial (HPDE6-C7) cells. Similar status were verified in a cohort of PC tissues and corresponding paracancer tissues collected from 74 patients, indicating that the expression of circTNPO3 was higher in tumor tissues than that in the normal tissues (Figure 1E). This differential trend was further validated in ISH assay (Figure 1F). Next, the 74 patients were classified into lower-expression (n = 31) and higher-expression (n = 43) subgroups through setting the median expression of circTNPO3 as a criterion. The Kaplan-Meier survival curve analysis suggested that patients with lower circTNPO3 expression obtained better overall survival times than those with higher circTNPO3 expression (Figure 1G).

To explore the role of circTNPO3 in PC cells, function loss and gain strategies were carried out by transfecting circTNPO3-siRNAs, circTNPO3-shRNAs or circTNPO3-overexpresion lentiviruses. Firstly, circTNPO3 was downregulated by si-circTNPO3-1/2 in BxPC-3 and PANC-1 cells with comparatively higher expression level. The knockdown efficiency was confirmed using qRT-PCR, suggesting that the expression of circTNPO3 was indeed downregulated in si-circTNPO3-1/2 group compared with that in si-NC group (Figure 2A). Moreover, the unique targeting property of si-circTNPO-1/2 was confirmed by immunoblotting assay, showing that the expression of linear gene TNPO3 was not intervened (Supplementary Figure S1A). The CCK-8 and colony formation assays showed that the cell viability and proliferation of BxPC-3 and PANC-1 were significantly inhibited in si-circTNPO3-1/2 group (Figures 2B, C). The EdU assay exhibited the similar results (Figure 2D). However, although the circTNPO3 was efficiently knockdown in HPDE6-C7 cells, significant changes in cell viability were not found (Supplementary Figures S1B,C). This result indicated that circTNPO3 might act as a tumor specific gene in PC. The wound healing and Transwell assays found that the migration and invasion abilities of PC cells were suppressed in circTNPO3 silencing group compared with that in control group (Figures 2E–G). Furthermore, the tendency of apoptotic cells was assessed through canonical apoptosis-associated makers. As shown in Figures 3A,B, the silence of circTNPO3 promoted the conversion of pro-apoptotic protein caspase 9 and downstream caspase 3 to activated form. Meanwhile, the elevation in the expression of cleaved-caspase 9 and cleaved-caspase 3 was detected by immunoblotting assay (Figure 3C). The AO/EB, Hoechst 33342 and flow cytometry assays also showed that the apoptosis rate was higher in si-circTNPO3-1/2 groups, implying that the effect of circTNPO3 on the proliferation of PC cells might be partly dependent on the regulation of apoptosis (Figures 3D–F).

In vivo experiments were conducted to estimate the effect of circTNPO3 on PC growth and metastasis. The data from subcutaneous graft tumor model suggested that the tumor growth velocity and final volume were significantly reduced in sh-circTNPO3-1/2 groups as compared with that in the sh-NC group (Figures 4A–C). And the above results were supported by the decreased expression of proliferation-related proteins, such as Ki-67 and PCNA (Figure 4D). Additionally, the lower number of lung metastasis nodules, more complete alveolar structure and increased expression of EMT-related marker E-cadherin showed that silence of circTNPO3 also attenuated the metastatic ability of PC (Figure 4E).

In order to further investigate the regulatory role of circTNPO3, the circTNPO3 was overexpressed in SW1990 and CFPAC-1 cells with comparatively lower expression level. In vitro experiments data showed that circTNPO3 overexpression promoted PC cells proliferation, migration and invasion (Figure 5). The phased research data confirmed that circTNPO3 might act as an oncogene in PC.

To explore the effect of circTNPO3 on GEM chemosensitivity of PC, the GEM resistant cell lines BxPC-3/GR and CFPAC-1/GR were established. As shown in Figure 6A, the expression of circTNPO3 was further analyzed in chemoresistant PC cells compared with that in the non-chemoresistant parental PC cells. And, the similar function loss or gain experiments were carried out in GEM resistant cells and non-resistant parental cells (Figures 6B,C). The recent oncology perspectives indicated that the immortality and chemoresistance of tumor cells might be closely associated with cancer stem cells (Sarabia-Sánchez et al., 2024). So the spheroid formation assay which assessed the stem cells self-renewal was performed. As shown in Figure 6D, the spheroid formation rate of GEM resistant cells was suppressed in circTNPO3 silencing group compared with that in the sh-NC group. On the contrary, the spheroid formation rate of non-resistant cells was significantly increased in circTNPO3 overexpression group. Furthermore, the silence of circTNPO3 partially improved GEM chemosensitivity in BxPC-3/GR and CFPAC-1/GR cells, mainly consisting of the diminished cell viability and soaring apoptosis rate induced by GEM cytotoxicity (Figures 6E, G). However, circTNPO3 overexpression stimulated chemoresistance characteristics in parental non-resistant BxPC-3 and CFPAC-1 cells (Figures 6F, H). The available data from our study supported the concept that circTNPO3 not only acted as an oncogene but also a GEM resistance-related factor in PC.

Studies have suggested that the regulatory mechanism of circRNA is associated with unique subcellular localization (Zhao et al., 2022). As shown in RNA fluorescence in situ hybridization assays based on specific probes and analysis of RNA nucleoplasmic distribution (Figures 7A,B), circTNPO3 was mainly pinpointed in the cytoplasm, implying that circTNPO3 might be involved in gene expression control at the post-transcriptional level. The stability of circTNPO3 was also analyzed to further verify its unique ring configuration. The total RNAs from BxPC-3/GR and CFPAC-1/GR were extracted after treated with RNase R, an exoribonuclease that could induce the RNA degradation from 3′initial to 5′terminal, but not cleave circRNA. As shown in Figure 7C, circTNPO3 exhibited significant resistance to RNase R compared with linear TNPO3 mRNA. The similar data was obtained after the cells treated with actinomycin D, a recognized transcription inhibitor, implying that circTNPO3 held a longer half-time than linear TNPO3 mRNA (Figure 7D). To explore the potential of circTNPO3 to absorb miRNAs through ceRNA mechanism, the Ago2-RIP assay was carried out. As shown in Figure 7E, compared with the blank control group with immunoglobulin-G (IgG) antibody, the enrichment of circTNPO3 was significantly higher in the Ago2 immunoprecipitation group, but was interrupted by circTNPO3 silencing. After comprehensive analysis of bioinformatics databases, including circBank, ENCORI and TargetScan, four target miRNAs, including miR-188-5p, miR-199a-5p, miR-199b-5p and miR-552-3p that might bind to circTNPO3, were predicted (Figure 7F). The qRT-PCR combined RNA pulldown assay based on biotin-labeled circTNPO3 probes showed that compared to the other three miRNAs, only miR-188-5p could be simultaneously enriched in both BxPC-3/GR and CFPAC-1/GR cells (Figure 7G). To determine the direct interaction between circTNPO3 and miR-188-5p, the dual-luciferase reporter assay instructed with wt (wild type) and mut (mutant type) circTNPO3 sequences was performed. Experiment data in Figure 7H found that compared with mimic negative control, the fluorescence signal intensity was decreased by miR-188-5p mimics in circTNPO3-wt group, but no significant difference was observed in circTNPO3-mut reporter gene plasmid. Combined with the expression of circTNPO3 in PC cell lines, the expression of miR-188-5p was shown to be negatively related with circTNPO3, and with the emergence of chemoresistance, its expression level was further reduced (Figure 7I). Conversely, silencing circTNPO3 could significantly upregulate miR-188-5p expression in BxPC-3/GR and CFPAC-1/GR cells, indicating that circTNPO3 might regulate the function of miR-188-5p by direct combination (Figure 7J). Moreover, transfection with miR-188-5p inhibitor partially reversed the inhibition of proliferation and migration, and alleviated the apoptosis caused by circTNPO3 silencing in BxPC-3/GR and CFPAC-1/GR cells (Figures 7K–M).

In order to unveil the mechanism of circTNPO3 regulating PC malignant phenotype through miR-188-5p, the TargetScan bioinformatics database for predicting downstream target was applied. As shown in Figure 8A, CDCA3 might be the potential downstream target gene of miR-188-5p. Moreover, to detect the binding potential between miR-188-5p and CDCA3, cells were co-transfected with the miR-188-5p mimics designed with the reporter vector containing the 3′-UTR of CDCA3. The results indicated that compared with mimic negative group, the luciferase activity of the reporter vector containing the wt 3′-UTR of CDCA3 was significantly suppressed by miR-188-5p mimics. On the contrary, fluorescence inhibition induced by miR-188-5p was blocked in mutant 3′-UTR of CDCA3 group. This result was verified by RNA pull down assay, compared with the blank control group with oligo probe, the enrichment of CDCA3 mRNA was obviously higher in the experimental group with miR-188-5p probe (Figure 8B). Immunoblotting and qRT-PCR assay found that CDCA3 mRNA and CDCA3 protein were overexpressed in PC cell lines compared with that in HPDE6-C7 cell line, and were further increased in chemoresistant cell lines compared with that in parental non-resistant cell lines (Figure 8C). Combining the clinicopathologic data from TCGA and TNMplot, and the IHC analysis from 74 patients, it was showed that the expression level of CDCA3 was aberrantly elevated in PC tumor tissues compared with that in the adjacent peritumoral tissues, and was negatively associated with patient survival (Figures 8D–H). Additionally, Pearson correlation analysis based on the clinicopathological data of 74 patients showed that the expression of CDCA3 and miR-188-5p was negatively correlated in PC (Figure 8I). To further clarify that circTNPO3 regulated CDCA3 through absorbing miR-188-5p, the expression level of CDCA3 was detected after co-transfection with sh-circTNPO3-1 and miR-188-5p inhibitor in BxPC-3/GR and CFPAC-1/GR cells. As shown in Figures 8J,K, transfection with sh-circTNPO3-1 attenuated CDCA3 expression, but this trend was rescued by miR-188-5p inhibitor.

In our previous research, it was found that the tumorigenesis and GEM chemoresistance of PC might be associated with the abnormal activation of NF-κB signaling pathway (Ji et al., 2020). Through analyzing bioinformatics database, TRAF2 that was related with NF-κB signaling pathway and had protein interaction with CDCA3 was identified as the key mediator (Supplementary Figure S1D). Meanwhile, as shown in Figure 8L, Supplementarys Figures S1D–F, Pearson correlation analysis indicated that the expression level of CDCA3 and RELA (scilicet NF-κB-p65) was positively correlated with TRAF2. In order to explore the regulatory effect of CDCA3 on the expression level of TRAF2 and translocation of p65 from cytoplasm to nucleus, sh-CDCA3-1/2 used to silence CDCA3 was synthesized. After transfection with sh-CDCA3-1/2 in BxPC-3/GR and CFPAC-1/GR cells, the knockdown efficiency was detected by immunoblotting assay, finding that sh-CDCA3-2 might be a more efficient sequence (Figures 8M,N). Nucleoplasmic separation combined with immunoblotting experiments showed that compared with parental non-resistant BxPC-3 cells, the expression of CDCA3 and TRAF2 was higher in chemoresistant BxPC-3/GR cells, and the translocation of p65 from cytoplasm to nucleus was more significant (Figure 8O). Contrastively, sh-CDCA3-2 transfection obviously downregulated CDCA3 and TRAF2 expression, and attenuated p65 nuclear translocation tendency (Figure 8O). Meanwhile, as shown in Figure 8P, immunofluorescence assay more intuitively demonstrated the translocation of p65 (stained using red fluorescence) from cytoplasm to nucleus was stirred up by chemoresistance and was suspended by CDCA3 knockdown in BxPC-3/GR cells.

More importantly, in order to further clarify that circTNPO3 ultimately regulated the malignant biological characteristics and chemoresistance of PC through NF-κB signaling pathway, the overexpression or silence of circTNPO3 was performed in the parental non-resistant BxPC-3 cells and resistant BxPC-3/GR cells. As shown in Supplementary Figure S1G, circTNPO3 overexpression promoted the translocation of NF-κB-p65 from cytoplasm to nucleus in BxPC-3 cells. By contrast, silencing circTNPO3 attenuated the activation of the NF-κB in BxPC-3/GR cells (Supplementary Figure S1H).

Deoxyelephantopin (DET), a kind of small molecule compound, was extracted from the Chinese herbal medicine E. scaber L.. In our previous research, DET had been proven to possess noticeable antitumor effect in PC (Ji et al., 2022). In order to explore the potential of circTNPO3 as a therapeutic target and further reveal the antitumor mechanism of DET in PC, DET was applied for the subsequent experiments. The specific molecular structure of DET was listed in Supplementary Figure S2A. The qRT-PCR results found that the expression of circTNPO3 was significantly downregulated after DET treatment at a concentration of 40 μM for 24 h (Supplementary Figure S2B). In CCK-8 and EdU staining assays, DET decreased the proliferation ability of CFPAC-1 cells compared with that in the control group, conversely, overexpression of circTNPO3 partially attenuated the cytotoxicity of DET (Supplementary Figures S2C,D). Similar results were further confirmed in Transwell assays, finding that the migration and invasion abilities of CFPAC-1 cells were suppressed by DET treatment, but reversed by circTNPO3 overexpression (Supplementary Figure S2E). In order to further investigate the feasibility of DET in the treatment of PC, xenograft tumor model was conducted by subcutaneous injecting normal or transfected CFPAC-1 cells. As shown in Supplementary Figures S2F–I, DET exhibited inhibitory effects on the growth of PC (n = 5), however, overexpression of circTNPO3 attenuated the cytotoxicity of DET. Furthermore, the carcinogenic effect of circTNPO3 on PC was also confirmed in vivo experiment indirectly. Our current studies suggested that the disincentive efficacy of DET on the malignant phenotype of PC was partially depended on downregulating the expression of oncogene circTNPO3. Meanwhile, the potential of circTNPO3 as a novel therapeutic target for PC could also be inferred.