DDX21 functions as a potential novel oncopromoter in pancreatic ductal adenocarcinoma: a comprehensive analysis of the DExD box family

Abstract

Background

Pancreatic ductal adenocarcinoma (PDAC) is a highly lethal tumor with an ill-defined pathogenesis. DExD box (DDX) family genes are widely distributed and involved in various RNA metabolism and cellular biogenesis; their dysregulation is associated with aberrant cellular processes and malignancies. However, the prognostic significance and expression patterns of the DDX family in PDAC are not fully understood. The present study aimed to explore the clinical value of DDX genes in PDAC.

Methods

Differentially expressed DDX genes were identified. DDX genes related to prognostic signatures were further investigated using LASSO Cox regression analysis. DDX21 protein expression was analyzed using the UALCAN and human protein atlas (HPA) online tools and confirmed in 40 paired PDAC and normal tissues through Tissue Microarrays (TMA). The independent prognostic significance of DDX21 in PDAC was determined through the construction of nomogram models and calibration curves. The functional roles of DDX21 were investigated using gene ontology (GO), Kyoto Encyclopedia of Genes and Genomes (KEGG), and Gene Set Enrichment Analysis (GSEA). Cell proliferation, invasion, and migration were assessed using Cell Counting Kit-8, colony formation, Transwell, and wound healing assays.

Results

Upregulation of genes related to prognostic signatures (DDX10, DDX21, DDX60, and DDX60L) was significantly associated with poor prognosis of patients with PDAC based on survival and recurrence time. Considering the expression profile and prognostic values of the signature-related genes, DDX21 was finally selected for further exploration. DDX21 was overexpressed significantly at both the mRNA and protein levels in PDAC compared to normal pancreatic tissues. DDX21 expression, pathological stage, and residual tumor were significant independent prognostic indicators in PDAC. Moreover, functional enrichment analysis revealed that Genes co-expressed with DDX21 are predominantly involved in RNA metabolism, helicase activity, ribosome biogenesis, cell cycle, and various cancer-related pathways, such as PI3K/Akt signaling pathway and TGF-β signaling pathway. Furthermore, in vitro experiments confirmed that the knockdown of DDX21 significantly reduced MIA PaCa-2 cell viability, proliferation, migration, and invasion.

Conclusions

Four signature-related genes could relatively precisely predict the prognosis of patients with PDAC. Specifically, DDX21 upregulation may signal an unfavorable prognosis by negatively affecting the biological properties of PDAC cells. DDX21 may be considered as a candidate therapeutic target in PDAC.

Supplementary Information

The online version contains supplementary material available at 10.1007/s12672-024-01204-9.

Introduction

Pancreatic cancer (PC) is a deadly malignancy with an extremely poor prognosis [1–4]. According to the GLOBOCAN report, with 511,000 new cases and 467,000 fatalities globally in 2022, pancreatic cancer ranks sixth in terms of cancer-related mortality and 12th in terms of cancer incidence in both sexes [5]. The incidence of PDAC is gradually increasing. Pancreatic ductal adenocarcinoma (PDAC) accounts for approximately 90% of primary PC cases and causes serious threats to global human health [3, 6]. Owing to late-stage diagnosis, resistance to multiple anticancer drugs, and a high rate of tumor recurrence or metastasis, the 5-year overall survival (OS) rate of PDAC is still below 10% [3, 6]. Despite rapid advances in surgical resection, radical surgical resection remains the sole curative option; however, various treatment options are also available, including immunotherapy based on checkpoint inhibitors such as the PD-L1/PD-1/CTLA-4 blockers, targeted therapy such as anti-cancer vaccine (GVAX vaccination) and engineered T cells [for example chimeric antigen receptor T cells (CAR-Ts)], and other systemic treatments that can benefit some patients [3, 4, 7–9]. Moreover, although numerous potential biomarkers, molecular mechanisms, and therapeutic targets for PDAC have been identified [6, 10, 11], significant breakthroughs are still lacking in its early screening, prevention, prediction, diagnosis, and treatment [8, 12–14]. Therefore, further uncovering the groundbreaking molecular markers, biological properties, and pathogenesis of PDAC will help improve the overall management of PDAC.

Asp-Glu-x-Asp/His (DExD/H)-box helicase family contains more than six subfamilies, such as the DExD-box (DDX) and DExH-box families. DDX helicases contain nine conserved motifs (I, Ia, Ib, II, III, IV, V, VI and Q), and are named after the composition of conserved amino acids in their motif II [15]. Each motif plays a crucial role in regulating cellular processes. Motif III is involved in ATPase activity. Three motifs (Ia, Ib, and IV) participate in RNA binding [16]. DDX helicases are widely distributed from prokaryotes to eukaryotes [15, 17]. These multifunctional RNA helicases participate in various aspects of RNA metabolism, ribosome biogenesis, gene expression, different posttranslational modifications, signal transduction, and more [15, 17–19]. The biological process of RNA linking DNA with proteins is the foundation of all cellular activities. Therefore, dysfunctional or abnormally expressed DDX genes can lead to aberrant cellular processes and even cause human diseases, such as human neurological disorders, genetic diseases, and malignancies [20–22]. The carcinogenic roles of DDX helicases have been studied across various human tumors. DDX helicase 1 (DDX1) and DDX56 were overexpressed and closely associated with poor outcomes in hepatocellular carcinoma (HCC) [23, 24]. DDX21 was identified as a valuable prognostic biomarker in various human malignancies based on a pan-cancer analysis [21]. The proliferative capacity of lung cancer cells was suppressed after DDX10 or DDX21 knockdown [21, 25]. DDX21, DDX39B, and DDX54 were found to be overexpressed in colorectal cancer (CRC) tissues and to promote proliferation and metastasis of CRC cells which showed their cancerous roles [26–28]. A previous study reported that DDX60L was highly expressed in PDAC tissues and cells [29]. DDX60L knockdown suppressed cell invasion, and induced cell apoptosis, but not proliferation of PDAC cells [29]. A recent study reported that elevated expression of DDX60 indicated poor prognosis of PDAC and strongly correlated with the expression of tumor immune-related molecules [30]. These findings suggest that multiple DDX genes could be candidate oncogenes in PDAC as well as other malignancies. However, the expression patterns, functional properties, and prognostic roles of most members of the DDX family in PDAC remain unclear.

Over the past two decades, the rapid development of various high-throughput technologies has generated large volumes of cancer-related data in genomics, epigenomics, transcriptomics, proteomics, and metabonomics. These multi-omic data sets and analyses have yielded various molecules or biomarkers related to the diagnosis and prognosis of PDAC at different biological levels, and have predicted the functional properties and molecular mechanisms of candidate molecules and biomarkers [31–33]. The easily accessible and publicly available multi-omics data warrant further integration and analysis to promote personalized medicine for PDAC in the future [34, 35]. However, multi-omics also face several challenges and shortcomings, such as a lack of interpretability, heterogeneity, and the complexity of multi-modal data [34]. Therefore, the results from multi-omics must be verified both in vitro and in vivo.

In this study, the expression hallmarks of DDX family genes and their clinical significance in PDAC were systematically analyzed. Subsequently, DDX21 was identified as the most promising candidate gene in PDAC and warrants further investigation. Consequently, DDX21 protein expression was further validated in PDAC and adjacent normal pancreatic tissues. The prognostic significance and functional properties of DDX21 in PDAC were uncovered through bioinformatics analysis. The biological properties of DDX21 were initially investigated in pancreatic cancer cells.

Materials and methods

Data sources and processing

The PDAC RNA sequencing data (n = 179) and corresponding clinical information were retrieved from the Cancer Genome Atlas (TCGA; https://portal.gdc.cancer.gov/repository). In addition, mRNA data of normal samples (n = 332) were obtained from GTEx (https://xenabrowser.net/datapages/). DDX genes were screened from the Human Protein Atlas (HPA, https://www.proteinatlas) and the Gene Cards (https://www.genecards.org/), and literature [15, 36]. A total of 43 members of the DDX gene family (Supplementary Table S1) were screened and further validated in the NCBI database via the “Gene” module (https://www.ncbi.nlm.nih.gov/gene/?term=). Moreover, there is evidence that all the DDX genes existed at either the protein or mRNA level. Differentially expressed genes (DEGs) between PDAC and normal samples were explored by the “Limma” package in R software (version 4.2.1) and visualized as a volcano plot. We identified the DEGs using [log2FC] > 1 and the adjusted p-value < 0.05 to differentiate the PDAC and the normal groups based on the datasets from TCGA and GTEx databases. Then, the overlap between DEGs and DDX genes was analyzed using the “VennDiagram” R package.

Prognostic signature construction

Univariate Cox regression analysis was conducted using the "survival" R package to identify genes that are significantly correlated with overall survival (OS) in patients with PDAC. Subsequently, Least Absolute Shrinkage and Selector Operator (LASSO) Cox regression analysis was performed using the “glmnet” R package to construct prognostic risk score models for predicting OS. Ten-fold cross-validation was employed to determine the parameter (λ) of the model. The risk score was calculated according to the formula below: risk score = ExpGene1*Coef1 + ExpGene2*Coef2 + Expression3*Coef3 + ⋯ + ExpGenen*Coefn.

The "ExpGene" represented the expression value of a gene from the prognostic risk score model. The “Coef” indicated the regression coefficient. For the Kaplan–Meier (KM) curve, the log-rank test was used to calculate the p-value and hazard ratio (HR). Moreover, the prognostic values of each signature-related DDX gene were further explored by the “survival” R package.

Protein expression profile of signature-related DDX genes

The HPA database (https://www.proteinatlas.org/) freely provides protein expression data for thousands of genes, as determined by immunohistochemical (IHC) staining. The Clinical Proteomic Tumor Analysis Consortium (CPTAC) (https://proteomics.cancer.gov/programs/cptac), which includes protein expression data of tumors, is integrated into the UALCAN online portal (https://ualcan.path.uab.edu/). The protein levels of signature-related DDX genes were determined by “Proteomics” module in UALCAN. The IHC staining intensity and quality of DDX genes were explored by HPA. The information on pancreatic tissues, including patient ID, antibody name, and staining intensity of PDAC cells, was displayed in this study.

Validation of DDX21 at the protein level

A total of 40 patients with PDAC were enrolled in this study, and the corresponding clinical data were obtained. None of the patients had received any form of immunotherapy, radiotherapy, or chemotherapy before undergoing radical surgical procedures. All patients, aged between 18 and 75, underwent radical operations at our center. All samples were evaluated by two independently experienced pathologists. All patients were diagnosed with primary PDAC. The work in this study was approved by the Research and Ethics Committee (No. 2023-ZFYJ-049-01). All participants have provided written informed consent. This study conformed to the guidelines set by the Declaration of Helsinki.

Tissue Microarrays (TMAs) with PDAC and adjacent normal samples were constructed as previously described [37]. Subsequently, IHC staining was conducted on TMA using antibody DDX21 (ab182156, Abcam, Britain) and evaluated by two independent, experienced pathologists as previously described [37]. The DDX21 staining intensity was measured and scored as follows: negative (0), weak (1), moderate (2), and strong (3). The staining extent ranged from 0, 1, 2, 3, and 4, indicating 0–5%, 6–30%, 31–55%, 56–80%, and 81–100%. The final expression score of DDX21 was calculated as the product of DDX21 staining intensity score * DDX21 staining extent score. All pancreatic samples were finally divided into four groups according to the expression score: not detected (0), low (1–4), medium (5–8), and high (9–12).

The prognostic value of DDX21 in PDAC from the TCGA database

Univariate and multivariate Cox regression analysis for OS and disease-free survival (DFS) was conducted to explore the independent prognostic factors by “survival” package. Notably, an insufficient sample size of independent groups, divided according to clinicopathological factors, could lead to unrepresentative samples. Thus, these groups (pathologic M2 stage, pathologic stage III, pathologic stage IV, and histologic grade G4) and their samples were excluded as they had fewer than 6 samples. Nomogram models were constructed using the “rms” package based on the chosen independent prognostic factors. Additionally, the nomogram model fit was assessed by calibration curves.

Functional enrichment analysis of DDX21

Gene ontology (GO) (https://geneontology.org/) enrichment (including biological process, BP; cellular component, CC; molecular function, MF) and Kyoto Encyclopedia of Genes and Genomes (KEGG) (https://www.genome.jp/kegg/) analysis of DDX21 and its co-expression genes were executed and visualized using the “ClusterProfiler”, “ggplot2”, and “ggraph” R packages.

To further predict the potential functions of DDX21 involved in PDAC, Gene Set Enrichment Analysis (GSEA) (https://www.gsea-msigdb.org/gsea/index.jsp) was conducted and visualized by the “ClusterProfiler” and “ggplot2” packages. The gene set of c2.cp.all.v2022.1.Hs.symbols.gmt served as the reference. The false discovery rate (FDR) and normalized enrichment score (NES) were obtained to evaluate statistical differences. FDR < 0.25 and p-value < 0.05 were established as cut-off criteria.

The dependency map (DepMap)

The DepMap online portal (https://depmap.org/portal/) is a visualization tool for the Cancer Cell Line Encyclopedia (CCLE) database. It can be used to discover factors related to cancer vulnerabilities and provide access to key analyses of cancer dependencies. In this study, the DepMap database (https://depmap.org/portal/) was utilized to determine gene expression and knockdown data for several PDAC cell lines. Fourteen PDAC cell lines were analyzed to explore DDX21 knockdown data, and 28 PDAC cell lines were studied to assess DDX21 expression data. The dependency score was adopted to display the changes in cell proliferation after gene knockdown. A dependency score of DDX21 less than − 0.5 indicates that DDX21 is essential in PDAC cell lines. The smaller the dependency score value, the stronger the dependency of PDAC cells for DDX21.

Cell culture and transfection

MIA PaCa-2 cells were maintained in Dulbecco’s modified Eagle’s medium (DMEM) with 10% fetal bovine serum (Gibco) at 37 °C and 5% CO₂. The DDX21 gene sequence was obtained from the National Center for Biotechnology Information database. Three small interfering RNA (siRNA) against DDX21 (siDDX21) were designed according to the DDX21 sequence by Gene Pharma (Shanghai, China). The sequence primers (from 5ʹ to -3ʹ) are shown in Table 1. To knock down the DDX21 gene, PDAC cells were transfected with siDDX21 using LipoSmart RNAiMax Transfection Reagent (130202, Genetic Bio, Shanghai, China). Each experiment was replicated a minimum of three times. In vitro experiments were categorized into three groups: blank, negative control (NC), and experimental (siDDX21).

Table 1.

The sequences of small interfering RNAs (siRNA) of DDX21 (siDDX21)

Name	Sequence (5ʹ-3ʹ)	Molecular weight	Base
siDDX21-1
Sense	CCUGAGGUUGAUUUGGUUAUATT	14,555.86	23
Antisense	UAUAACCAAAUCAACCUCAGGTT
siDDX21-2
Sense	CCCAUAUCUGAAGAAACUAUUTT	14,540.85	23
Antisense	AAUAGUUUCUUCAGAUAUGGGTT
SiDDX21-3
Sense	GCGGAGUUUCAGUAAAGCAUUTT	14,570.88	23
Antisense	AAUGCUUUACUGAAACUCCGCTT

Confirmation of DDX21 knockdown efficiency

DDX21 knockdown efficiency was evaluated at both the mRNA and protein levels using RNA extraction and real-time polymerase chain reaction (RT-PCR) and Western blot (WB) analyses, as previously described [38]. After transfection, 2 × 10⁵ MIA PaCa-2 cells in Blank, NC, and siDDX21 groups were incubated for 48 at 37 °C in a 5% CO₂ incubator. TRIzol reagent (15596018, Thermo Fisher, United States) was used to harvest total RNA from the cell lines. RT-PCR was conducted using RT-PCR kit (11732088, Thermo Fisher, United States) based on the manufacturer’s instructions. The sequence primers (from 5ʹ to -3ʹ) used for RT-PCR are displayed in Table 2. Similarly, 2 × 10⁵ MIA PaCa-2 cells in Blank, NC, and siDDX21 groups were incubated for 72 h. Subsequently, the BCA Protein Assay Kit (P0009, Beyotime Biotechnology, Shanghai, China) was used to isolate and quantify total protein. WB was performed based on traditional WB assay method [39]. β-Actin (1:2000 dilution, ab8227, Abcam) was selected as the internal control. DDX21 (1:2000 dilution, ab181870, Abcam) was the target protein.

Table 2.

Primer sequences of housekeeping gene (β-Tubulin) and target gene (DDX21) used in this study

Gene name	Primer sequence (5ʹ-3ʹ)
β-Tubulin
Forward	TCCATGAAGGAGGTCGATGA
Reverse	CAGACGGCTGTCTTGACATT
DDX21
Forward	AGTGGAGCAAAAAGCGGGAA
Reverse	GCACGGAATCCAAAAGCCTG

Cell counting kit-8 (CCK8) assay and colony formation assay

Cell viability was assessed using a CCK8 assay (10 μl/well; C0037, Beyotime Biotechnology, Shanghai, China). 3 × 10³ MIA PaCa-2 cells were seeded in a 96-well plate. Subsequently, the plate was incubated for 0, 24, 48, and 72 h at 37 °C in a 5% CO₂ incubator. Next, the CCK8 solution was added to each well. Lastly, the absorbance was measured using a microplate reader at 450 nm.

MIA PaCa-2 cells were seeded at a density of 1 × 10³ per well in six-well plates and incubated for 10 days at 37 °C in a 5% CO₂ incubator. Subsequently, the cells were fixed using 4% paraformaldehyde and stained with 0.1% crystal violet. Colonies were measured using ImageJ software (National Institutes of Health, United States).

Wound healing assay

MIA PaCa-2 cells were seeded at 2 × 10⁵ per well in a six-well plate. The cell layer was scratched using a P100 pipette tip. Images were captured at time points of 0, 24, and 48 h. Images of the wounds were pictured and determined by ImageJ software. The cell migration rate was calculated by the scratch area as follows: Cell migration rate = (wound area at 0 h − wound area at 48 h)/(wound area at 0 h).

Transwell migration and invasion assays

Transwell migration and invasion assays were performed in Transwell Chambers (3783, Corning Costar, China). The upper surface of the chamber with 8 um pores was precoated without (migration) or with (invasion assay) Matrigel. MIA PaCa-2 cells were transferred into the upper chamber at a density of 5 × 10⁴ cells in 200 μl per well. Following a 48-h incubation, migrated or invasive cells in the lower chamber were fixed using 4% paraformaldehyde and stained with 0.1% crystal violet. The cells were then imaged using an inverted fluorescence microscope and quantified using ImageJ.

Statistical analysis

Most of the statistical analyses were conducted using R software and its associated packages (version 4.2.1). The Student’s t-test was used to perform two-group comparisons, and a one-way ANOVA test was conducted for multi-group analysis. Differences were assessed using a two-sided test, and a p-value of < 0.05 was considered statistically significant.

Results

Screening of differentially expressed DDX genes

A differential expression analysis was performed to screen the differentially expressed DDX genes, considering all of the genes. Overall, the results indicated that 8451 genes were up-regulated while 606 down-regulated genes were recorded based on the screening criteria of |log2FC|> 1 and adjusted p-value < 0.05 (Fig. 1A). Subsequently, 17 DDX genes were identified as DEGs after intersecting the list of all of the DEGs and that of DDX genes (Fig. 1B). All of the 17 DDX genes were overexpressed in PDAC tissues (Fig. 1C).

Fig. 1.

Identification of differentially expressed DDX genes among PDAC tissues and normal tissues. A Volcano plot showing differentially expressed genes (DEGs) between PDAC and normal samples from the TCGA and GTEx databases. B Venn Diagram showing the intersection of DEGs between the list of DDX genes and that of DEGs. C Box plots showing the expression profile of the 17 DDX genes differentially expressed in PDAC. Wilcoxon rank sum test was performed. *** p < 0.001

Construction of the prognostic signature

To pinpoint the survival-related DDX genes, Univariate Cox regression analysis was performed based on the 17 DDX genes. The results showed that four DDX genes (DDX10, DDX21, DDX60, and DDX60L) among the 17 were identified as significantly associated with OS of patients (p-value < 0.05).

Next, the risk score signature was developed based on the formula:

risk score = ExpGene1*Coef1 + ExpGene2*Coef2 + Expression3*Coef3 + ⋯ + ExpGenen*Coefn.

Based on the survival DDX genes, the risk score for OS was as follows:

Risk score = 0.3252 * DDX10 + 0.053 * DDX21 + 0.1011 * DDX60 + 0.2848 * DDX60L (Fig. 2A and B).

Fig. 2.

Construction of the prognostic signature for overall survival (OS) and disease-free survival (DFS). A Lasso coefficient profiles of the signature genes obtained from the LASSO feature selection analysis. B The best cut-off variables were calculated based on the likelihood of deviance. C The distribution of risk scores in patients ranked according to their risk score. D Distribution of patients according to their survival time and survival status. E The KM plot depicts the survival of patients according to the risk model. F–I Association of high expression of DDX10, DDX21, DDX60, and DDX60L with OS. J–M Association of high expression of DDX10, DDX21, DDX60, and DDX60L with DFS. N The 1-, 2-, and 3-year ROC curves indicate higher accuracy of the risk model in predicting clinical features

The distributions of risk score as well as the status of patients with PDAC, were displayed in Fig. 2C and D. Besides, the high-risk group of patients with PDAC indicated a poorer prognostic value than patients with a lower risk score [Fig. 2E for OS, Log-rank P = 3.73e−06, HR (High groups) = 2.749].

Furthermore, the prognostic properties of the aforementioned four signature genes were further evaluated by KM plot. High expression of DDX10 (HR = 1.61 (1.07–2.43), p = 0.024), DDX21 (HR = 1.77 (1.17–2.67), p = 0.007), DDX60 (HR = 1.98(1.30–3.01), p = 0.001), and DDX60L (HR = 2.51(1.64–3.83), p < 0.001) indicated worse OS in patients with PDAC (Fig. 2F–I). The expression of DDX10 (HR = 1.83(1.24–2.70), p = 0.002), DDX21 (HR = 1.87(1.26–2.76), p = 0.002), and DDX 60L (HR = 1.62 (1.10–2.38), p = 0.015) was significantly associated with short DFS (Fig. 2J–L). However, DDX60 expression was not correlated with DFS (Fig. 2L, HR = 1.46 (0.99–2.15), p = 0.055).

The accuracy of this risk score model was further verified by the ROC curve. The 1-year [AUC = 0.663, 95% CI (0.584–0.742)], 2-year [AUC = 0.705, 95%CI (0.621–0.79)], and 3-year [AUC = 0.708, 95% CI (0.594–0.822)] ROC curves for the model based on OS (Fig. 2N), further demonstrating that this model displayed great reliability. These results suggested that DDX10, DDX21, DDX60, and DDX60L are potentially robust prognostic signature-related genes.

Protein expression profiles of four signature genes

In order to verify the protein expression of the prognostic genes in the PDAC, the protein expression data of these genes were acquired from the CPTAC and HPA databases. As shown in Fig. 3A1–5, DDX10 protein expression showed no significant difference (p > 0.05) between PDAC and normal samples. On the contrary, DDX21, DDX60, and DDX60L proteins were significantly (p < 0.05) upregulated in PDAC samples compared to normal specimens in the CPTAC database (Fig. 3B1, C1, and D1). In addition, in the HPA database, DDX21 (Fig. 3B2–5) and DDX60 (Fig. 3C2–5) protein expression levels in PDAC tissues were either high or medium, while the expression levels in normal tissues were moderate (medium), which was consistent with the protein expression profiles of DDX21 and DDX60 observed in the CPTAC database. Moreover, based on the HPA database, the DDX60L protein expression was high in normal tissues and moderate in PDAC tissues (Fig. 3D2–5), which was inconsistent with its protein expression profile observed in the CPTAC database. Additional analysis indicated that DDX10 was mostly expressed in type 1 ductal cells, endothelial cells, macrophages, fibroblasts, and T cells in the pancreas of patients (Fig S1A), and its expression was mostly localized in the cytoplasmic/membranous or the cytoplasmic/membranous and nuclear compartments (Fig S1B). The expression of DDX21 was detected in the type 1 ductal cells, endothelial cells, and macrophages in the pancreas of patients (Fig S1C) and was mostly localized in the nuclear compartments (Fig S1D). The expressions of DDX60 (Fig S1E) and DDX60L (Fig S1G)were generally recorded in type 1 ductal cells, endothelial cells, macrophages, fibroblasts, and T cells in the pancreas of patients and were mostly localized in the cytoplasmic/membranous compartment, followed by the nuclear compartments (Fig S1F and Fig S1H).

Fig. 3.

Protein expression patterns of four signature-related genes sourced from the CPTAC and HPA databases. A₁, B₁, C₁, and D₁ The comprehensive protein profiles of DDX10, DDX21, DDX60, and DDX60L in both PDAC and normal tissues, as sourced from the CPTAC database. A₂, B₂, C₂, and D₂ The protein expression levels of DDX10, DDX21, DDX60, and DDX60L in normal tissues, as sourced from the HPA database. A_3–4, B_3–4, C_3–4, and D_3–4 The protein expression levels of DDX10, DDX21, DDX60, and DDX60L in PDAC tissues. A₅, B₅, C₅, and D₅ The IHC staining intensities of DDX10, DDX21, DDX60, and DDX60L in pancreatic tissues are depicted in the bar charts

Validation of DDX21 protein expression

Based on the above results, we eliminated DDX60 since its expression was not correlated with DFS. In addition, since the expression of DDX60L was not consistent among both databases and because there was no significant difference in the expression levels of DDX10 among the PDAC and normal samples, these proteins were also eliminated in subsequent validation analysis. Thus, only DDX21 was ultimately identified as a particularly promising biomarker worthy of further investigation. To validate the expression of DDX21 in tumor samples, immunohistochemistry analysis was performed. The results indicated that DDX21 staining was predominantly localized to the nuclei of tumor cells within PDAC tissues (Fig. 4). Among the normal pancreatic samples, 21 (52.5%) showed weak staining (Fig. 4A1–3), 18 (45.0%) exhibited moderate staining (Fig. 4B1–3), and only one (2.5%) showed high staining. Of the 40 paired PDAC samples, none showed weak staining, nine (22.5%) exhibited moderate staining (Fig. 4C1–3), while 31 (77.5%) showed high staining (Fig. 4D1–3). DDX21 protein levels were higher than that in normal tissues. These results further confirmed the importance of DDX21 in PDAC.

Fig. 4.

Verification of DDX21 protein expression across 40 paired samples of PDAC and normal tissues. A_1–3, B_1–3 Weak and medium expression levels of DDX21 were observed in representative normal tissues, respectively. C_1–3, D_1–3 Medium and high expression levels of DDX21 were observed in representative PDAC tissues, respectively

Relationship between clinicopathological characteristics and prognosis in PDAC

To explore the association of DDX21 expression and the clinicopathological characteristics in the prognosis of PDAC, univariate Cox regression analysis was performed. The results revealed that pathologic stage [p = 0.032, HR 2.342(1.074–5.109)], DDX21 expression [p = 0.007, HR 1.765(1.166–2.672)], residual tumor [p = 0.018, HR 2.161(1.713–2.678)], histologic grade [p = 0.047, HR 1.560(1.007–2.418)], pathologic T stage [p = 0.026, HR 2.058(1.091–3.882)], N stage [p = 0.004, HR 2.161(1.287–3.627)] and radiation therapy [p = 0.012, HR 0.507(0.298–0.863)] were significant predictors of OS (Fig. 5A). Similarly, pathologic stage [p = 0.007, HR 2.942(1.342–6.451)], DDX21 expression [p = 0.002, HR 1.865(1.26–2. 761)], residual tumor [p < 0.001, HR 2.140(1.402–3.267)], histologic grade [p = 0.015, HR 1.677(1.108–2.542)], pathologic T stage [p = 0.004, HR 2.431(1.318–4.485)], and N stage [p = 0.014, HR 1.746(1.121–2.720)] were significant predictors of DFS in patients with PDAC (Fig. 6A). Multivariate analysis displayed that pathologic stage [p = 0.048, HR 2.012(1.087–4.516)], DDX21 expression [p = 0.046, HR 1.62(1.009–2.602)], and residual tumor [p = 0.047, HR 1.627(1.007–2.631)] were significant independent predictors for OS (Fig. 5A). Similarly, pathologic stage [p = 0.043, HR 2.03(1.054–4.902)], DDX21 expression [p = 0.019, HR 1.733(1.093–2.748)], and residual tumor [p = 0.004, HR 1.963(1.239–3.111)] were significant independent predictors for DFS (Fig. 6A). In clinical practice, though postoperative radiotherapy for pancreatic cancer is not a standardized treatment plan, nor is it the first choice, our results showed that radiation therapy was also a significant independent predictor for OS based on the existing data in the TCGA database (Fig. 5), which indicated the necessity of including this variable in the monitoring OS of PDAC patients. Subsequently, a nomogram model was developed using these independent prognostic factors to predict 1-, 2-, and 3-year OS and DFS outcomes (Figs. 5B, 6B). In addition, calibration analysis was performed to assess the predictive accuracy of the nomogram models and demonstrated reliable predictions (Figs. 5C, 6C).

Fig. 5.

Analysis of independent prognostic factors for OS and construction of a predictive nomogram model based on the TCGA database. A Univariate and multivariate Cox regression analysis of the association of clinical features and DDX21 expression with OS. B A nomogram model incorporating DDX21 expression and other independent prognostic factors to predict the likelihood of 1-, 2-, and 3-year OS. C The efficacy of the nomogram model for OS was validated using s ROC curve

Fig. 6.

Analysis of independent prognostic factors for DFS and the construction of a predictive nomogram model based on the TCGA database. A Univariate and multivariate Cox regression analysis of the association of clinical features and DDX21 expression with DFS. B A nomogram incorporating DDX21 expression and other independent prognostic factors to predict the likelihood of 1-, 2-, and 3-year DFS. C ROC curve analysis for the validation of the efficacy of the nomogram model for DFS

Functional enrichment analysis

To investigate the biological classification of DDX21 co-expressed genes, we performed GO and KEGG analyses. The top 10 positive and negative co-expressed genes associated with DDX21 are depicted in a heat map (Fig. 7A and B). The GO terms, including BP, CC, and MF, are shown in Fig. 7C. DDX21 and its co-expressed genes were mainly enriched in the BPs of ribonucleoprotein complex biogenesis (P.adj < 0.05), chromosome segregation (P.adj < 0.05), ribosome biogenesis (P.adj < 0.05), and RNA metabolic process (P.adj < 0.05). As for CCs terms, the genes showed enrichment in chromosomal regions (P.adj < 0.05), spindles (P.adj < 0.05), condensed chromosomes (P.adj < 0.05), and centromeric regions (P.adj < 0.05). MF terms indicated ATP hydrolysis activity (P.adj < 0.05), catalytic activity (P.adj < 0.05), RNA-binding activity (P.adj < 0.05), cadherin binding (P.adj < 0.05), and helicase activity (P.adj < 0.05). KEGG pathway analysis revealed that DDX21 and its co-expressed genes were primarily involved in the cell cycle (P.adj < 0.05), ribosomal biogenesis in eukaryotes (P.adj < 0.05), nucleocytoplasmic transport (P.adj < 0.05), spliceosome (P.adj < 0.05), and amyotrophic lateral sclerosis (P.adj < 0.05, Fig. 7D). GSEA analysis helped identify significant differences in pathway enrichment in samples with high levels of DDX21. The most highly enriched signaling pathways included pathways in cancer, PI3K/Akt signaling pathway (NES = 2.064, P.adj < 0.001, FDR < 0.001), focal adhesion PI3K/Akt/mTOR signaling pathway (NES = 2.149, P.adj < 0.001, FDR < 0.001), TGF-β signaling pathway (NES = 2.744, P.adj < 0.001, FDR < 0.001), HER2 pathway (NES = 2.234, P.adj = 0.007, FDR = 0.005), and EGF/EGFR signaling pathway (NES = 2.019, P.adj = 0.001, FDR < 0.001) were displayed (Fig. 7E–J).

Fig. 7.

Functional enrichment analysis of DDX21 and its co-expressed genes in PDAC. A The heat map showing the top 10 genes positively co-expressed with DDX21. B The heat map showing the top 10 genes negatively co-expressed with DDX21. C The top four BP, CC, and MF terms for DDX21 and its co-expressed genes as obtained from functional enrichment analysis. D The top five KEGG pathways associated with DDX21 and its co-expressed genes. E–J GSEA results for the group exhibiting elevated DDX21 expression. Spearman correlation analysis was performed. *** p < 0.001

Expression profile of DDX21 in PDAC cell lines and RNA interference (RNAi) efficacy of siDDX21

The dependency scores of DDX21 were assessed in PDAC cell lines using DepMap online tool. In the “RNAi (Achilles + DRIVE + Marcotte, DEMETER2)” dataset, the dependency scores for DDX21 across 14 PDAC cell lines were found to be negative (Fig. 8A). In the “Expression Public 23Q2” dataset, the expression levels of DDX21 in 28 PDAC cell lines were as shown in Fig. 8B. Since the dependency scores and expression levels for MIA PaCa-2 cells fell within the median range among all PDAC cells, MIA PaCa-2 cells could be a representative PDAC cell line. Therefore, the MIA PaCa-2 cell line was selected for further in vitro experiments. To investigate the effect of on these cells, DDX21 was silenced. The results showed that DDX21 expression at the mRNA level was significantly reduced in the siDDX21-1 (0.175 ± 0.067, p < 0.05), siDDX21-2 (0.052 ± 0.011, p < 0.05), and siDDX21-3 (0.054 ± 0.005, p < 0.05) groups compared to the NC (0.955 ± 0.057) and blank (1 ± 0.059) groups as indicated by the RT-PCR (Fig. 8C). Similar trends were observed in western blot assay (Fig. 8D and E). In particular, siDDX21-2 displayed a slightly better silencing efficiency and was used in subsequent experiments.

Fig. 8.

Screening representative PDAC cell lines and siDDX21. A Perturbation effect of DDX21 knockdown in 14 PDAC cell lines based on the DepMap database. B Expression Profile of DDX21 in 28 PDAC Cell Lines based on the DepMap database. C RT-PCR detection of the expression of DDX21 in cells transfected with siDDX21-1, siDDX21-2, and siDDX21-3. D Western blotting detection of the expression of DDX21 in cells transfected with siDDX21-1, siDDX21-2, and siDDX21-3. E The quantification of the Western blotting based on grayscale value. One-way ANOVA followed by Tukey multiple comparison test was performed. *** p < 0.001 among the compared groups. ns, not significant

Knockdown of DDX21 in MIA PaCa-2 cells inhibits cell proliferation, migration, and invasion

To explore the biological function of DDX21 in PDAC, the effect of DDX21 on the cell proliferation, migration, and invasion of MIA PaCa-2 cell lines was analyzed. The CCK-8 assay revealed that silencing DDX21 significantly decreased cell viability (1.452 ± 0.040, p < 0.05) compared to the NC (1.851 ± 0.038) or Blank (1.871 ± 0.044) groups (Fig. 9A). Furthermore, the colony formation assay showed that the silencing of DDX21 significantly (79.93 ± 3.36, p < 0.05) reduced MIA PaCa-2 proliferation rate (Fig. 9B). The scratched areas were significantly smaller in the siDDX21 group compared to the NC and Blank groups (Fig. 9C, 84.19 ± 1.28, p < 0.05). The MIA PaCa-2-siDDX21 cell line exhibited lower migration (Fig. 9D, 43.43 ± 3.68, p < 0.05) and invasion (Fig. 9E, 30.368 ± 3.16, p < 0.05) rates compared to the NC and Blank groups.

Fig. 9.

Significant inhibition of cell viability, proliferation, migration, and invasion by silencing DDX21 in MIA PaCa-2 cells. A DDX21 knockdown significantly reduced cell viability. B_1–2 DDX21 knockdown significantly reduced the number of cell colonies. C DDX21 knockdown significantly reduced the cell migration area. D DDX21 knockdown significantly suppressed cell migration. E DDX21 knockdown significantly suppressed cell invasion. One-way ANOVA followed by Tukey multiple comparison test was performed. * p < 0.05; ** p < 0.01; *** p < 0.001 among the compared groups, ns, not significant

Discussion

This study offers overall insights into the mRNA expression of the DDX genes. Seventeen out of 43 members of the DDX family, such as DDX10, DDX21, DDX60, and DDX60L, were identified as DEGs in PDAC. Concurrently, the expression of the 17 DDX genes was upregulated. DDX60 and DDX60L were also proved to be highly expressed in PDAC tissues or cells [29, 30]. However, DDX10 expression in PDAC tissues was downregulated compared to normal controls, as shown in the microarray dataset (MTAB-6690) [39]. This is contrary to our results based on the TCGA dataset. Subsequently, LASSO-Cox analysis revealed the significant clinical relevance of a four-gene prognostic signature (DDX10, DDX21, DDX60, and DDX60L) for OS in PDAC. ROC curves indicated acceptable predictive accuracy for the prognostic signature. Feng et al. constructed a seven-gene prognostic signature for OS, including DDX10, in PDAC [40]. Similarly, DDX10 and DDX60 upregulation were proved to be relatively credible risk indicators for OS in PDAC [30, 40]. However, the two studies did not show prognostic features of DDX10 and DDX60 for DFS in PDAC [30, 40]. The prognostic features of DDX21 and DDX60L in PDAC are rarely documented. In this study, our analysis reveals the significant predictive values of DDX10, DDX21 and DDX60L for both OS and DFS in patients with PDAC. In the current research, mRNA upregulation of DDX21 in PDAC tissues showed a significant association with shorter OS and DFS. Nevertheless, elevated DDX21 protein expression in CRC tissues showed a significant correlation with longer OS and DFS in patients with early-stage CRC [41]. Similarly, Zhang et al. reported that down-regulated mRNA expression of DDX21 is highly correlated with a dismal prognosis of breast cancer (BC) [42]. Moreover, the upregulation of DDX21 mRNA was not associated with clinical outcomes of patients with CRC or BC. DDX21 appears to exhibit inconsistent prognostic value and expression characteristics in different tumors. In general, the proliferative capacity of tumor cells is increased, cellular processes (such as the cell cycle and RNA metabolism) are heightened, and the expression or activity of most RNA helicases may correspondingly rise [16].

Given the clinical significance of DDX10, DDX21, DDX60, and DDX60L for OS or DFS, as well as their up-regulated mRNA expression trend in this study, the protein profiles of these four DDX genes were subsequently explored in CPTAC and HPA databases. This protein expression trend was generally aligned with the mRNA expression level. Nevertheless, this positive mRNA-protein correlation was not observed in DDX10, and DDX60L. It has been widely reported that there is not an absolute positive relationship between mRNA abundance and protein abundance [43, 44]. This incoherence is most common among various enzymes and transcription factors due to the post-transcriptional modifications of RNA and post-transcriptional regulations of protein; the copy number of mRNA is not a reliable indicator for the final amount of protein [44, 45]. Besides, the abundance of various types of RNA, including mRNA, is strongly influenced by both internal and external environmental factors, like sampling time and temperature [45]. If the mRNA expression of a certain gene is consistent with the corresponding protein expression, it would be more reasonable as a disease-related biomarker. In this regard, at the protein level, DDX21 and DDX60 are more likely to be better biomarkers of PDAC. Combining their prognostic values for both OS and DFS with their mRNA-protein correlations, DDX21 was ultimately identified as a particularly promising biomarker worthy of further investigation. Thus, we further confirmed the expression feature of DDX21 in 40 paired PDAC and normal pancreatic tissues collected from our center. The protein expression levels of DDX21 were higher than those in normal tissues. This is consistent with the data in the CPTAC and HPA databases.

Furthermore, we explored the correlation between DDX21 upregulation and unfavorable prognosis in human tumors, with a particular focus on PDAC. First, enrichment analysis of DDX21 co-expressed genes showed that most of the top 10 positive and negative genes have the potential for oncogenes. In particular, eukaryotic translation initiation factor 3a (EIF3A) was found to be overexpressed in PDAC, and knockdown of EIF3A could significantly decrease PDAC cell proliferation and motility [46]. Vacuolar protein sorting associated protein 26 A (Elevated VPS26A) upregulation was confirmed to be associated with poor prognosis and advanced stage of PDAC [47]. Upregulation of importin 7 (IPO7) was confirmed to be correlated with the poor prognosis of patients with PDAC. Besides, silencing IPO7 could significantly decrease PDAC cell metastasis and proliferation [46]. NPEPL1 (aminopeptidase-like 1) is a member of the aminopeptidase family, which was reported to play an essential role in the occurrence and development of CRC [48] and BC [49]. Literature indicated that YPEL3 (yippee-like 3) plays a key role in tumorigenesis. Especially, YPEL3 could suppress nasopharyngeal carcinoma (NPC) metastasis by Wnt/β-cantenin signaling pathway [50]. Studies involving mitochondrial (MT) genes have demonstrated their potential role in CRC. MT-ND1, MT-ND6, MT-CYB, and MT-CO1 exhibited differential expression and were involved in the development of CRC [53]. Hence, we hypothesized that DDX21 may have a synergistic effect with its co-expressed genes in the occurrence and development of PDAC.

In addition, GO and KEGG analysis of DDX21 and its co-expressed genes indicated that they were mainly involved in the cell cycle, ribosome biogenesis, and the spliceosome. The in vitro experiments in this study also confirmed the carcinogenic role of DDX21 in cell viability, proliferation, migration, and invasion. Similarly, a study demonstrated that DDX21 could promote gastric cancer proliferation by regulating the cell cycle [51]. Ribosome biogenesis is a highly complex and extraordinarily energy-consuming process. It involves several distinct proteins and maturation factors that are responsible for protein synthesis [52]. As we know, the characteristic of tumor cells is high ribosome production, which is necessary for maintaining enhanced growth and cell division. The increased ribosome production is related to abnormal ribosome biogenesis homeostasis. Furthermore, increasing evidence demonstrates that there is a close association between the deregulation of ribosome proteins (RB), rRNA synthesis, and the occurrence and development of tumors [52]. Recently, ribosomal protein L10 (RPL10) was reported to play a key role in PDAC development [53]. More importantly, RB is responsible for regulating cellular functions such as the cell cycle, cellular proliferation, and DNA-damage repair, which are key processes in tumor development [54–56]. The splicing of precursor messenger mRNA (pre-mRNA) is performed by the spliceosome, which is a ribonucleoprotein complex. The pre-mRNA splicing plays a key role in eukaryotic cells. Mutations in some components of the spliceosome and altered pre-mRNA expression were found to be correlated with various tumors [57]. Another study confirmed that DDX21 could promote the splicing of some essential key pro-differentiation genes, which indicated that DDX21 helicase is necessary for tissue differentiation [58].

GSEA also confirmed that DDX21 was involved in various pathways, including cancer pathways, the PI3K/Akt signaling pathway, focal adhesion, the PI3K/Akt/mTOR signaling pathway, the TGF-β signaling pathway, the HER2 pathway, and the EGF/EGFR signaling pathway. Previous literature has clarified that the TGF-β signaling pathway plays a crucial role in tumor occurrence and development [59, 60]. TGF-β is overexpressed in most cases of PDAC. Besides, many patients with PDAC have alterations in TGF-β signaling pathway-related genes [57, 61, 62]. Studies have reported that the PI3K/Akt signaling pathway plays a crucial role in the viability, migration, invasion, and angiogenesis of tumor cells. Also, it plays a crucial role in tumor progression [63, 64]. A recent study demonstrated that curcumin could suppress EGF-induced migration and invasion of PDAC cells by inhibiting the EGR/EGFR signaling pathway [65]. The functional enrichment analysis of DDX21 was performed using two different tools, both of which contributed to the identification of similar properties of DDX21 in tumor occurrence and progression. These findings related to DDX21 suggested that its upregulation plays a carcinogenic role in the initiation and development of PDAC.

Our study may also have some therapeutic significance. Indeed, previous studies indicated that alternative splicing is involved in the prognosis of patients with bladder cancer and revealed pivotal players of alternative splicing events, including DDX21, in the context of tumor immune microenvironment and the response to chemotherapy and immunotherapy [66]. Another research indicated that using certain chemotherapy medications could result in increased lethality in cancer cells that have been made fragile due to genome fragility induced by DDX21, providing a possible approach for treating tumors with elevated DDX21 expression [67]. Further studies are needed to explore the effect of DDX genes, especially DDX21 in the development of therapeutic drugs for PDAC.

Notably, this study has some limitations. First, the small sample size of the clinical PDAC tissue samples could introduce bias in the expression profile. Second, the clinical significance of DDX21 at the protein level has not been demonstrated. Third, only one cell line (MiaPaCa2) was utilized for functional experiments; to confirm findings, it would be important to analyze more pancreatic cancer cell lines. Further, the functional experiments (proliferation, invasion, and migration) are rather preliminary, and the expression characteristic and biological function of DDX21 have not yet been validated in vivo. Therefore, in our future experiments, we will deeply focus on the effect of DDX21 on different aspects of PDAC in vitro, in vivo, and clinical samples and the underlying molecular mechanisms.

Conclusion

Significant relationships between the expression of DDX genes and the prognosis of patients with PDAC have been identified. Four prognostic signature-related genes (DDX10, DDX21, DDX60, and DDX60L) can predict the prognosis of patients with PDAC with relative precision. Upregulation of DDX21 may suggest an unfavorable prognosis due to the worsening of the biological properties of PDAC cells. DDX21 may serve as a candidate target for developing immunotherapeutics and targeted therapies for PDAC.

Author contributions

Lei Qin contributed to the conceptualization, funding acquisition, project administration, and manuscript review. Shaohan Wu, Xiaofang Sun, Ruheng Hua, and Chundong Hu contributed to data acquisition, formal analysis, investigation, methodology, resources, software, supervision, validation, and visualization of the study. The original draft was written by Shaohan Wu. All authors commented on previous versions of the manuscript. All authors approved the final manuscript.

Funding

This research was supported in by the Science and Technology Program of Jiaxing (grant number: 2023AD31036) and the Postgraduate Research & Practice Innovation Program of Jiangsu Province (grant number: SJCX23_1669).

Data availability

Data is provided within the manuscript or supplementary information files.

Declarations

Ethics approval and consent to participate

The study protocol was approved by the Ethical Committee of the Second Affiliated Hospital of Jiaxing University (Approval No. 2023-ZFYJ-049-01). This study conformed to the guidelines set by the Declaration of Helsinki. All participants provided written informed consent. All participants also provided informed consent for publication.

Competing interests

The authors declare no competing interests.