TCGA PAAD transcriptome FPKM data were downloaded from the GDC Data Portal¹. Clinical data, such as age, sex, clinical stage, and survival time, were also downloaded. The 377 E3 ligase genes were acquired from the online database².

Gene expression in PC and control samples was compared using the LIMMA R package. A fold change ≥ 2 or ≤ 0.5 and p < 0.05 were set as the cutoff values for EG screening. The heatmap R package was used to draw heatmaps. The Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis, Gene Ontology (GO) enrichment analysis, and protein–protein interaction enrichment analysis for DEGs were also performed using the EnrichR (Kuleshov et al., 2016) website tool³.

The genomic alteration (GA) information of DEGs was acquired using cBioPortal⁴. Two PAAD datasets were included in this study, the TCGA Pan-cancer atlas and the UTSW study (Witkiewicz et al., 2015). All 184 and 109 samples from the two datasets were included.

The PANC-1 and ASPC-1 human PC cell lines were obtained from the American Type Culture Collection (United States) and cultured at 37°C with 95% air and 5% CO₂. PANC-1 and ASPC-1 cells were cultured in Dulbecco modified Eagle medium (Gibco, Germany) supplemented with 10% fetal calf serum (Germany), 2 mM L-glutamine, 105 U/L penicillin, and 100 mg/L streptomycin. Short hairpin RNAs specifically targeting RNF223 and Kruppel-like factor 4 (KLF4) were designed and synthesized by Generay Biotech (Shanghai) Co., Ltd., RNF223 shRNAs were as follows: shRNA1 (5′–3′): GCACAGCAGCCACTGGAAGTC, shRNA2 (5′–3′): GCGAAAGGAGCCTGGCATCTC, and shRNA3 (5′–3′): GGAGCCTGGCATCTCTGAGGA.

KLF4 shRNAs were as follows: shRNA1 (5′–3′): GCTCC ATTACCAAGAGCTCAT, shRNA2 (5′–3′): CCAGCCAGAAA GCACTACAAT, and shRNA3 (5′–3′): GCCTTACACATGAAG AGGCAT.

The cell proliferation reagent WST-8 (Roche, Germany) was measured. Cell growth: 10 μL of cell counting kit-8 (CCK8) was added to each well at the time of harvest after plating cells in 96-well microtiter plates (Corning, NY) at 1.0 × 10³/well, according to the manufacturer’s instructions. Cellular viability was determined by measuring the absorbance of the converted dye at 450 nm 2 h after adding CCK8.

PANC-1 and ASPC-1 cells were seeded into 6-well plates and incubated for 24 h, and a linear wound was created by dragging a 100-μL pipette tip through the monolayer prior to transfection. Cellular debris was removed by gentle washes with culture medium, following which transfection was immediately performed, and the cells were allowed to migrate for an additional 48 h. The healing process was dynamically imaged after the wound was introduced using a microscope (Olympus 600 Autobiochemical Analyzer, Tokyo, Japan). Migration distance was measured using images (five fields) taken at each indicated time point. The gap size was analyzed using Image Pro Plus 6.0 software. The residual gap between the migrating cells from the opposing wound edge is expressed as a percentage of the initial gap size.

Total RNA was extracted from ASPC-1 and PANC-1 cells using TRIzol^® RNA Isolation Reagent (Invitrogen, Carlsbad, CA) according to the manufacturer’s instructions. Reverse transcription was performed using the PrimeScript^TM RT reagent kit (Takara, Dalian, China). All mRNA levels were normalized to the housekeeping gene GAPDH. The following RNF223 and KLF4 primers were used in this study: RNF223: 5′-TGATGCTCTTCTGTGTGGCA-3′ (F) and 5′-TTATCAGTCAG AGGCCCGAG-3′ (R). KLF4: 5′-CCCACATGAAGCGACTTC CC-3′ (F) and 5′-CAGGTCCAGGAGATCGTTGAA-3′(R). The GAPDH primers were as follows: 5′-TGACTTCAACAGC GACACCCA-3′ (F) and 5′- CACCCTGTTGCTGTAGCCAAA-3′ (R). All samples were treated under the same conditions and analyzed by quantitative reverse transcriptase–polymerase chain reaction (qRT-PCR) using SYBR Premix ExTaq^TM (Takara, Dalian China) according to the manufacturer’s protocol.

Univariate Cox regression analysis was performed to identify prognosis-associated E3 ligase genes. Genes with a hazard ratio (HR) < 1 were considered favorable for OS, whereas HR > 1 presented unfavorable for OS. Genes with p < 0.05 were considered significant markers. The Kaplan–Meier survival plot was constructed using the “survplot” R package. Pan-cancer survival analysis was conducted using the KM plotter online tool (Nagy et al., 2021)⁵.

Data-independent acquisition quantitative proteomics was conducted according to a previous study (Zhang et al., 2021). Briefly, cells underwent protein extraction and trypsin digestion into peptides, and then a spectral library was generated and quantified. Ten fractions were collected, and each fraction was dried in a vacuum concentrator. The fractions were redissolved in 0.1% formic acid and analyzed using nanospray liquid chromatography with tandem mass spectrometry (LC-MS/MS) on an Orbitrap Fusion Lumos Tribrid (Thermo Fisher Scientific, MA, United States) coupled to a Waters nanoACQUITY UPLC System (Waters, MA, United States). The mass spectrometer was run in DDA mode and automatically switched between MS and MS/MS modes. The DDA data were processed and analyzed using Spectronaut X (Biognosys, Schlieren, Switzerland) with default settings to generate an initial target list. A false discovery rate cutoff at the precursor and protein levels was applied at 1%. Finally, proteins were identified and quantified.

The ChIP sequencing peaks of RNF223 transcription factors were identified using the online tool ChIP-Atlas⁶. PC, including cell lines, datasets in bigwig format were downloaded and imported into the IGV browser. Transcription factors with binding peaks within the distance of ≤ 1 kb from the transcription start sites were considered candidate peaks.

A luciferase reporter assay was performed according to a standard protocol as previously described (Gong et al., 2016). Briefly, ASPC-1 cells (3 × 10⁴ cells/well) were seeded into 24-well plates in triplicate and allowed to attach for 24 h. The sequences (−28∼−54) of the wild-type (WT) and mutant ASPC-1 cells were WT: TATACCCTATGGGCCAAGGGTGTGGC and mutant (MUT): CGCCTTTACTGGGAACCGGGAGCAAA. The indicated plasmids and 1.5 ng pRL-TK Renilla plasmid were transfected using Lipofectamine 3000 Reagent (Thermo Fisher Scientific, Waltham, MA, United States, cat. no. L3000008). Forty-eight hours posttransfection, luciferase and Renilla signals were assessed using a Dual Luciferase Reporter Assay Kit (Promega, cat. no. E1980) according to the manufacturer’s instructions as previously described (Hahn et al., 2002).

The data are presented as means ± SD. SPSS 18.0 software was used to perform statistical analyses, and graphs were generated using GraphPad Prism 8.0 (GraphPad Software, San Diego, CA, United States). Differences were examined using Student t-test or one-way analysis of variance. p < 0.05 was considered statistically significant.

To examine the expression differences in the E3 ligases in PC, 377 genes were applied for statistical analysis. By applying the cutoff criteria of fold change ≥ 2 or ≤ 0.5 and p < 0.05, 27 DEGs were acquired, including RNF166, MARCHF1, TRIM10, TRIM7, and so on (Figure 1A), and their expression in cancer and normal samples is shown in Figure 1B. Then, we examined their GAs in PC studies. As shown in Figure 1C, MEX3A exhibited the highest alteration frequency of 11%, followed by CBLC (10%), RNF43 (9%), MECOM (8%), TRIM58 (8%), and RNF39 (8%), mostly comprising amplification events. In another study (UTSW) by Witkiewicz et al. (2015), GA events displayed a comparatively different pattern. As shown in Figure 1D, RNF223 (26%), TRIM46 (24%), MEX3A (18%), CBLC (16%), TRIM58 (13%), and RNF43 (9%) showed the highest alteration frequencies. Notably, high occurrence of amplification events in MEX3A, CBLC, and MECOM and truncating mutations in RNF43 were observed in both datasets. In addition, RNF223 showed a distinct GA pattern in the two datasets, with only 2.2% in TCGA and 26% GA events (amplification and deep deletion events combined) in UTSW.

Next, we conducted univariate survival analysis using the 27 DEGs and clinicopathological parameters of the TCGA dataset. As shown in Figure 2A, eight DEGs (MECOM, CBLC, MARCHF4, RNF166, TRIM46, LONRF3, RNF39, and RNF223) and two clinical parameters (pathological N stage and T stage) showed prognostic significance. The Kaplan–Meier survival plot of the eight DEGs is shown in Figure 2B. Of the eight genes, five genes, CBLC (HR = 1.7, p = 0.015), LONRF3 (HR = 2.1, p = 0.00092), RNF39 (HR = 1.6, p = 0.02), MECOM (HR = 1.8, p = 0.0059), and RNF223 (HR = 1.9, p = 0.0034), exhibited unfavorable prognostic significance, whereas TRIM46 (HR = 0.65, p = 0.046), MARCHF4 (HR = 0.63, p = 0.029), and RNF166 (HR = 0.65, p = 0.046) conveyed favorable survival significance. Finally, using the pan-cancer dataset, overall survival (OS) and relapse-free survival (RFS) analyses were conducted using the online tool Kaplan–Meier plotter. Consistent with the results in Figure 3, all eight genes showed prognostic significance, including the RFS data. Taken together, we acquired the expression and GA data of 27 DEGs, among which eight genes exhibited potential performance as prognostic indicators in PC.

Next, a multivariate analysis was conducted to analyze the prognostic significance of the 27 DEGs and the OS rate, and as a result, only RNF223 (among the eight prognostic genes) displayed prognostic significance, indicating that RNF223 may serve as an independent prognostic marker in PC (Figure 4A). Then, we compared the expression of RNF223 in groups of clinical parameters and as shown in Figure 4B, RNF223 in males (gender), high alcohol consumption, ductal/lobular neoplasms (disease type), pathological stage N0, tumor stage IIa, and vital status (dead) groups exhibited significantly higher expression compared to the other groups.

Next, we focused on RNF223 and explored the function of RNF223 silencing on PC cell line phenotypes (Figure 5A). After RNF223-targeting shRNA transfection, the RNF223 silencing efficiency was examined using qRT-PCR, and because shRNA2 exerted consistently higher (>50%) knockdown efficiency, we used shRNA2 as the subsequent shRNA (shRNF223) (Figure 5B). Then, CCK8 and wound healing assays were applied to study the impact of RNF223 silencing on the proliferation and migration capacity of ASPC-1 and PANC-1 cells. As shown in Figure 5C, RNF223 knockdown significantly reduced the cell number in both cell lines, indicating that RNF223 may promote PC growth. In addition, the wounds of shRNF223-transfected ASPC-1 and PANC-1 cell lines demonstrated reduced migration distance compared to the control group, indicating that RNF223 knockdown decreases the migration ability in both cell lines (Figure 5D). In summary, the above results revealed that RNF223 may represent an independent prognostic marker that promotes PC growth and migration.

Then, the downstream mechanism of RNF223 in PC was investigated using quantitative proteomics in RNF223 knockdown and control ASPC-1 cells. After protein quantification, all 885 DEPs were acquired (Figure 6A), and their expression is shown in Figure 6B. Then, the functions of these DEPs were annotated, and their enriched pathways and functions were identified using KEGG pathways and GO databases. Based on the enrichment score (−log10 p-value), the top 15 enriched pathways and GO biological processes (BPs), cellular components (CCs), and molecular functions (MFs) are shown in Figures 6C–F. The most enriched pathways were oxidative phosphorylation, and other items, such as regulation of cytoskeleton, pathways in cancer, metabolism pathways, and HIG1α signaling pathways, were also enriched. For GO-BPs, metabolism-related BPs were also enriched, such as cellular metabolic, primary metabolic, and nitrogen compound metabolic processes. In addition, the cell cycle process was also significantly enriched. For the GO-CC result, catalytic complex was the top enriched component, supporting the pathway and BPs of enrichment of metabolism-related factors. As expected, the GO-MM category identified enriched protein binding as the most significant item, consistent with the biochemical role of RNF223 as an E3 ligase. Finally, a functional network was created of RNF223 targets, and genes in the DNA synthesis and transforming growth factor β signaling pathways are shown. In summary, we identified potential protein targets and metabolism-related pathways of RNF223 in PC (Figure 6G).

Finally, the mechanism underlying the elevated expression of RNF223 was explored. As the genetic alteration frequency of RNF223 was not remarkably high in PC, we speculated that transcription factors may contribute to this process. First, we downloaded PC ChIP sequencing data from the recently published online tool ChIP-Atlas. As shown in Figure 7A, KLF4 exhibited a prominent peak in the promoter region of RNF223 DNA. In addition, coexpression analysis of KLF4 with RNF223 revealed a strong coefficient (R = 0.51) in TCGA PAAD datasets (Figure 7B). Then, to validate the role of KLF4 on the mRNA expression of RNF223, qRT-PCR was performed to examine the expression of RNF223 in the KLF4 knockdown cell line ASPC-1. As shown in Figure 7C, RNF223 exhibited significantly decreased expression in KLF4-silenced cells, indicating that KLF4 may upregulate RNF223 expression in ASPC-1 cells. Finally, a luciferase assay was conducted to validate the above results. As shown in Figure 7D, KLF4 silencing remarkably decreased luciferase intensity in the RNF223 WT group, whereas no significant difference was observed in the RNF223 MUT group. Altogether, these data show that KLF4 contributes to the increased expression of RNF233 in PC.