Identification of autophagy-related immune targets for enhancing immunotherapy in pancreatic cancer aggressiveness

Abstract

Background

Pancreatic cancer (PC) presents significant challenges in oncology, with metastasis critically affecting patient outcomes. Autophagy-related genes (ARGs)’s involvement in influencing immune activity and metastasis in PC remains inadequately understood.

Aim

This study seeks to identify and validate five ARGs that could serve as immune targets, enhancing enhancing Pancreatic cancer metastasis (PCM)’s prognostic models and informing immunotherapy strategies.

Methods

ARGs that were diffentially expressed were screened, followed by Cox regression and LASSO analyses to pinpoint five genes linked to overall survival (OS). A prognostic model was developed and validated using ROC curves. Functional analyses, including GO and KEGG, were performed to elucidate ARG mechanisms. Immune infiltration and TFs/microRNA/mRNA networks were assessed to understand ARG-immune cell interactions. Experimental validation employed real-time PCR, IHC, and Western blotting, supported by TCGA data. Functional assays explored RHEB’s role in PC, particularly its interaction with LC3.

Results

Five ARGs (CASP1, RHEB, CHMP2B, MYC, and HDAC6) were identified, contributing to a robust prognostic model where low-risk individuals showed significantly longer OS. The model demonstrated high AUC scores, indicating strong prognostic capability. CD8 T cells and Treg cells’ elevated levels were observed in metastatic subjects. RHEB knockdown suppressed cancer cell proliferation and invasion, with a negative correlation between RHEB and LC3, suggesting a role in autophagy-mediated modulation of PC metastasis.

Conclusion

This study introduces a novel prognostic model incorporating five ARGs, highlighting their potential as immune targets for cancer immunotherapy. The negative correlation between RHEB and LC3 suggests a therapeutic pathway for PCM intervention, laying the groundwork for more effective anti-cancer strategies. These findings advance the identification of novel immune targets and signaling pathways, aligning with precision medicine goals in cancer treatment.

Supplementary Information

The online version contains supplementary material available at 10.1007/s12672-025-02190-2.

Introduction

Pancreatic cancer (PC) represents a considerable global health dilemma, ranking as the 12th most popular and 7th most deadly malignant tumor according to GLOBOCAN 2020 statistics [1]. The disease is characterized by a notably short median survival time and a low 5-year survival rate, primarily due to factors such as late diagnosis, high invasiveness, and resistance to chemotherapy [2]. These factors underscore the imperative need for creative strategies to advance PC prognosis and patient endings.

The possibility of improving the immune system’s ability to detect and battle cancer cells has been highlighted by recent developments in cancer immunotherapy. To design effective immune targets, it is essential to identify the distinct molecular properties of cancer cells, such as their genetic, epigenetic, and protein expression patterns. Within this framework, autophagy—a defense mechanism in eukaryotes that includes phases such as autophagic lysis and maturation—is essential for preserving the homeostasis of cells. A number of malignancies, including PC, have been related to dysregulated autophagy. This is because dysregulated autophagy is connected to enhanced invasion and metastasis, which worsens prognosis and treatment results [3–6].

Autophagy-related genes (ARGs) have emerged as key regulators of PC progression [5, 7], yet their precise mechanisms remain unclear. While ARGs have been explored in the prognosis of other cancers such as colorectal cancer, grade IV astrocytoma, non-small cell lung cancer, and prostate carcinoma, investigation specific to PC is limited [8–11]. Notably, autophagy may influence the tumor immune microenvironment, suggesting that ARGs could serve as novel immune targets for cancer immunotherapy [12]. This potential makes them attractive candidates for developing aimed therapies that coud strengthen the immune system’s ability to combat cancer.

In this study, we investigate the impact of ARGs on pancreatic cancer metastasis (PCM) and elucidate their mechanisms of action through functional enrichment assays, immune cell function analysis, and transcription factor (TF)/mRNA/microRNA (miRNA) network evaluations. We propose small-molecule compounds targeting ARGs as potential therapeutic agents for PC. Moreover, we build a prognostic model grounded in five ARGs to predict PC outcomes. Our findings reveal that the expression of these ARGs in PC cells or tissues aligns with both experimental and bioinformatic analyses, with RHEB identified as a key player in promoting PC proliferation and invasion. This research offers new insights into early clinical characteristics of PCM and suggests innovative approaches for more effective treatment strategies. By focusing on the identification of immune targets and signaling pathways, this introduction aligns with the special issue’s emphasis on cancer immunotherapy and immunoprecision medicine.

Materials and methods

Data Collection and Differentially Expressed ARGs (DEARGs) Screening ARGs associated with PCM were selected from GEO’s gene expression profile GSE19279 dataset [13, 14]. mRNA profiles from 15 samples (3 normal pancreatic samples, 4 primary PDAC samples (primary group), 3 normal liver samples, and 5 liver metastasis samples (metastatic group)) were collected. To determine how ARGs affect metastatic PC patients’ outcomes, RNA-seq-batch effects data and clinicopathological information were collected [15] (https://xenabrowser.net/datapages/). ARGs were obtained from HADb [16]. GEO2R was used for data analysis. Differentially expressed mRNAs (DEmRNAs) were characterized with the standard: (adj. p-values) < 0.05 and |log₂FC (fold-change)|> 1. DEARGs were obtained by intersecting 243 human ARGs in HADbs with the DEmRNAs.

The 5-ARGs prognostic model’s analysis

For the purpose of determining how OS affects DEARGs, Kaplan–Meier (KM) plots were used [17]. Univariate Cox was applied to retain OS-related DEARGs and LASSO Cox was applied to delve OS-related DEARGs. Expression of hub genes was confirmed in TCGA-PAAD. “ggplot2” package in R (version 4.2.1) was used. A risk score was provided to every subject, and its median was selected to separate subjects into low- or high-risk groups. Survival difference was used to assess the model’s effectiveness.

Subsequently, a prognostic model grounded in five autophagy-related genes (ARGs) was developed. The risk scores for this model were computed by the formula: $\text{Risk Score}=\sum _{t=1}^n\text{Coefi}\times \text{Xi}$, where Coefi represents the coefficient of each of the five ARGs, and Xi denotes the expression level of every gene. The prognostic value of the TCGA cohort was evaluated using R. Survival of PC patients was evaluated by conducting KM analysis. Time-ROC plots were created. AUC was determined to evaluate this model’s prognostic ability. Additionally, PC patients were divided into subgroups based on: age, gender, stage, and T staging. Survival in subgroups was performed using the KM curve. To comparatively evaluate the prognostic ability between clinical characteristics and risk score, independent risk factors were identified using Xiantao study forest plot tool. ROC curves were created by “pROC”, and AUC was used to evaluate the prognostic ability. DCA was performed and calibration curves were used to experiment the prognostic ability of the model using the “stdca.R” and “rms” packages in R.

Predicting DEARGs’ mechanism of action

GO and KEGG enrichment analyses were used to predict hub genes’ function using Metascape [18]. To predict PC drugs, HDAC6 was placed in the downregulation group, the other four were placed in the upregulation group, and they were uploaded to L1000FWD [19]. A list that includes the top 10 potential compounds and the top 3 drugs was uploaded to PubChem [20] for visualization. Next, to elucidate the roles of DEARGs on PCM, a TF/miRNA/mRNA network was created. miRNA-mRNA data were extracted from the online platform Starbase [21], and TF-mRNA data were from chEA3 [22]. Cytoscape was used to create the whole TF/miRNA/mRNA network. Additionally, Pearson analyses were performed to determine the relationship between autophagy’s key component and ARGs. The correlation between ARGs and MAP1LC3B was evaluated by “ggplot2”.

Immuno-microenvironment assay

Estimation of STromal and Immune cells in MAlignant Tumor tissues using Expression data (ESTIMATE) was performed by “estimate” of R. Ratio of the stromal-immune element in TME was assessed to get stromal score, immune score, ESTIMATE score (the total of stromal and immune score), and tumor purity (# of tumor cells). A lower score indicated a lower ratio of TME. To evaluate immunoinfiltration, CIBERSORT was used to calculate 22 subsets via CIBERSORTx [23]. The relationship between ARGs and levels of infiltrating subsets was evaluated via Pearson analyses using “ggplot2”.

Cell culture and tissue specimens

Capan-1, COLO357, PANC-1, CFPAC-1, MIA PaCa-2, and HPDE6-C7 from the Chinese Academy of Sciences Cell Bank were maintained in DMEM with 10% FBS at 37 °C. The tissue samples used for RT-qPCR, including primary tumor samples and paired normal samples (n = 5), were collected from the 1st Affiliated Hospital of Chongqing Medical University. This study has the approval from the hospital’s ethics committee (2023–0104).

RNA extraction, qPCR, and RNA interference

RNAs were isolated by Trizol (Biyotime). RT-PCR was carried out with 1 µg of total RNA. qPCR assays were carried out by a SYBR kit and a 7500 cycler (ABI). Each sample had three replicates. The result was calculated by the 2^−ΔΔCt formula. Primers were designed for the 5 hub genes (MYC, CHMP2B, RHEB, and CASP1, Supplementary Table 1). The siRNA that targeted human RHEB was purchased from Tsingke Biotech. PANC-1 and MIA PaCa-2 cells were transfected with siRHEB and siNC Lipo 2000 (Invitrogen) for 48 h following the user manual. Cells were gathered 48 h later. WB assays were carried out to estimate the knockdown efficiency, using GAPDH as a loading control. The RHEB-siRNA sequences used were as follows:

RHEB-SiRNA1: GTGTATTCTGTTACATCAA.

RHEB-SiRNA2: CACAGTAAATGGACAAGAA.

RHEB-SiRNA3: GAAAGACCTGCATATGGAA.

Immunohistochemistry (IHC)

The paraffin-embedded tissues were used to perform immunohistochemical analysis. Sections were incubated with anti-RHEB antibody (ab25873) overnight at 4 °C. After secondary antibody detection, the antigen was visualized by standard DAB staining (Dako, Sweden) followed by hematoxylin counterstaining.

Immunofluorescence (IF)

The tissue slices were used to perform immunofluorescence staining by incubating them with anti-RHEB antibodies (Abcam, ab25873) at a dilution of 1:1000, LC3 antibodies (Wanleibio, WL01506) at a dilution of 1:400, and the second antibodies at a dilution of 1:5000. The stained samples were counterstained with DAPI (2 µg/mL, Beyotime). Finally, slices were observed with a fluorescence microscope (Leica Microsystems GmbH).

Western blotting (WB) analysis

Whole-cell lysates were put on ice with RIPA buffer and quantified using a BCA Kit (Beyotime, China). The extracts were mixed with a 5 × loading buffer (Mengbio, China) and denatured by boiling. Lysates (40 µg) underwent 10% SDS-PAGE (Epizyme, China) and were transferred to PVDF membranes (Invitrogen, USA). After blocking with a buffer (Beyotime Technology, China), primary antibodies against RHEB (Abcam, ab25873) and GAPDH (Huabio, 12D6) were incubated for 12 h at 4 °C. Samples were then treated with HRP-conjugated secondary antibodies (Servicebio, G1213, China) for 2 h. Blots were visualized using the HRP Substrate kit (Merck Millipore Corporation, Germany).

Cell proliferation and colony formation assays

CCK-8 (Biosharp, China) was added to 96-well plates containing 2000 cells per well, and the cells were incubated for 2 h at 0/24/48/72 h. At 450 nm, absorbance was measured. For 10–14 days, the cells that were implanted in six-well plates were grown at densities of 1000 cells/well. The surviving colonies were scanned and counted using a CanoScan 8800F MOEL-85 scanner (CanoScan, Japan) following fixation with 4% paraformaldehyde (Servicebio, China). The colonies were then visualized using Gentian violet (Servicebio, China). There were three independent duplicates of each sample.

Transwell cell migration and aggression experiments

Transwell chambers with an 8-µm pore size (Corning Inc.) were utilized to assess cell migration and aggression. For the aggression assay, a Matrigel layer (Becton, Dickinson, and Company, USA) was applied to the Transwell membrane. After 48 h of cell incubation, the migrated or invaded cells were photographed and documented using a Leica camera (Germany). Each experiment was conducted three times.

Data analysis

Data were analyzed by SPSS 26.0 R (4.2.1), and GraphPad Prism 9.0. Differences in categorical data were determined by Fisher’s exact test or the Chi-Square test. Distinctions in continuous data were determined by the Wilcoxon test or Student’s t-test. The distinctions in all groups of continuous variables were determined by ANOVA. Survival was analyzed by regression assay. Correlation was performed by Pearson analysis. p < 0.05 was designated as significance.

Results

Study flowchart

The study’s flowchart is illustrated in Fig. 1a. Initially, we selected a dataset focused on liver metastasis in PC to analyze differentially expressed genes, intersecting them with the autophagy database HADb to identify 22 ARGs. Through regression analysis, five genes closely linked to overall survival (OS) were identified. These genes were then used for survival analysis and to construct a risk model for diagnostic efficacy. Additionally, we conducted functional enrichment analysis, constructed a TF/mRNA/miRNA regulatory network, and assessed immune infiltration to understand its mechanism of action. We also identified promising small-molecule compounds, such as Apicidin, Simvastatin, and PD-0325901, targeting ARGs.

Fig. 1.

Flow chart and hierarchical clustering analysis of autophagy-related genes (ARGs). a Study design. b Venn diagrams highlighting ARGs common to both GSE19279 and HADb datasets. c Heatmaps of differentially expressed ARGs (DEARGs), with orange indicating up-regulation and gray indicating down-regulation

DE ARGs and ca 5-ARGs prognostic model

From the GSE19279 dataset, 1113 differentially expressed mRNAs (DEmRNAs) were identified. After intersecting with 243 human ARGs from HADb, 22 differentially expressed ARGs (DEARGs) were obtained (Fig. 1b, c). The Kaplan–Meier (KM) method and Cox regression analysis distinguished seven DEARGs (Fig. 2a–h). LASSO Cox regression further refined this to five prognostic ARGs: CASP1, RHEB, CHMP2B, MYC, and HDAC6, which were included as hub genes in this model (Fig. 2i–j).

Fig. 2.

Clinical correlation of ARGs. a–g Kaplan–Meier (KM) plots. h Forest plot. i Univariate Cox regression analysis of hub genes. j LASSO coefficients profiling

The 5-ARGs prognostic model’s application

Risk score distribution, survival, and heat maps for the five ARGs were created (Fig. 3a–c). The data indicated that higher risk scores correlated with increased mortality. Four ARGs, except HDAC6, had lower levels in low-risk subjects. KM curves demonstrated significantly higher OS in low-risk subjects (Fig. 3d). Risk scores for 1-, 3-, and 5-year OS had AUCs of 0.68, 0.69, and 0.70, respectively (Fig. 3e). Stratification based on risk scores effectively differentiated between favorable and unfavorable prognoses for PC patients, with low-risk groups showing better outcomes across different genders, stages, and T staging (Supplementary Fig. 1). Thus, the 5-ARGs prognostic model accurately predicted PC patient survival.

Fig. 3.

Evaluation of the prognostic model. a Distribution of risk scores. b Correlation between ARGs. c Correlation between survival rates. d KM survival assay. e Risk scores

The 5-ARGs model and clinicopathological features’ prognostic value

For the purpose of estimating predictive values, univariate Cox regression identified prognostic factors. Forest plots indicated that only risk score and stage were typical independent risk factors (Fig. 4a). ROC curves are presented in Fig. 4b, excluding T/M staging due to insufficient data. AUCs for pathological indices were lower than risk scores. Decision curve analysis (DCA) suggested that the risk score was a high quality prognostic marker at 1, 3, and 5 years (Supplementary Fig. 2). Calibration curves, plotted to assess predictive accuracy, revealed that at 1, 2, and 3 years, the model’s predictions closely matched the criteria curve, demonstrating excellent prognostic capability (Fig. 4c).

Fig. 4.

Comparison of diagnostic efficiency. a Regression analyses of prognostic factors, specifically T stage and N stage. b AUC values. c Calibration curves

Functional enrichment analysis and screening small-molecule drugs

GO and KEGG analysis identified macroautophagy, autophagy, and autophagic processes as the most enriched biological processes (BPs). The most enriched cellular components (CCs), molecular functions (MFs), and KEGG pathways are shown in Fig. 5a, suggesting that hub genes regulate tumor progression by modulating autophagy. To identify promising drugs, the five hub genes were submitted to L1000FWD. The most promising small-molecule drugs, including Apicidin, Simvastatin, and PD-0325901, are listed in Table 1. These drugs, with their 3D/2D structures modeled using PubChem (Supplementary Fig. 3), can downregulate ARG expression by inducing autophagy, thereby advancing targeted PC treatment.

Fig. 5.

Mechanism of action of hub genes. a GO and KEGG analyses. b TF/mRNA/miRNA networks, with light purple representing TFs, green for mRNAs, and deep purple for miRNAs. c Relationships between hub genes and LC3

Table 1.

The screened drugs for PAAD treatment

Drug	Similarity score	p-value	q-value	Z-score	Combined score	MOA	Predicted MOA
Apicidin	− 0.6	6.90E-05	2.74E-01	1.59	− 6.6	Unknown	HDAC inhibitor
Simvastatin	− 0.6	5.36E-05	2.74E-01	1.63	− 6.98	HMGCR inhibitor	Glucocorticoid receptor
PD-0325901	− 0.6	4.55E-05	2.74E-01	1.76	− 7.63	MEK inhibitor	MEK inhibitor
Auranofin	− 0.6	5.11E-05	2.74E-01	1.81	− 7.78	NFkB panthway inhibitor	Calcium channel
Panobinostat	− 0.4	3.69E-03	2.76E-01	1.55	− 3.77	HDAC inhibitor	HDAC inhibitor
PP-2	− 0.4	2.83E-03	2.74E-01	1.62	− 4.13	Scr inhibitor	Cyclooxyenase inhibitor
BMS-387032	− 0.4	3.12E-03	2.74E-01	1.6	− 4.01	CDK inhibitor, MCL1 inhibitor	CDK inhibitor
LY-294002	− 0.4	2.94E-03	2.74E-01	1.65	− 4.19	mTOR inhibitor, PI3K inhibitor	PI3K inhibitor
Olvanil	− 0.4	3.11E-03	2.74E-01	1.68	− 4.2	TRPV agonist	Adrenergic receptor
THM-I-94	− 0.4	3.49E-03	2.74E-01	1.64	− 4.04	HDAC inhibitor	HDAC inhibitor

TF/miRNA/mRNA regulatory network and correlation between ARGs and MAP1LC3B

To explore how ARGs influence PCM, interactions between miRNA-DEARG and TF-DEARG were analyzed using Starbase and chEA3, resulting in a TF/miRNA/mRNA network (Fig. 5b). This network indicated that most hub genes are regulated by FOS and MYC. MAP1LC3B is indentified as a crucial component of autophagy [24]. Pearson correlation analysis between MAP1LC3B and five ARGs (Fig. 5c) showed significant associations with RHEB, CHMP2B, MYC, and CASP1, implying their potential role in regulating PCM through autophagy.

Assay of immunoinfiltration and co-expression of ARGs and metastasis-related immune cells

We examined the principle that five autophagy-related genes (ARGs) influence pancreatic cancer (PC) prognosis. Autophagy is linked to immune infiltration [5, 25, 26] and tumor progression [27, 28]. Using Sangerbox, we analyzed the relationship between ARG expression and immune components. The data revealed that metastatic subjects had significantly lower stromal, immune, and ESTIMATE scores (Fig. 6a–c) but higher tumor purity (Fig. 6d), suggesting a connection between hub gene levels and the tumor microenvironment (TME). CIBERSORT analysis identified differences in 22 immune cell subsets between elementary and metastatic cases (Fig. 6e). T cell follicular helpers and CD8 T cells were prevalent in primary subjects, while CD8 T cells and Treg cells dominated in metastatic subjects (Fig. 6i). Principal component analysis (PCA) indicated big differences in immune cell percentages, with plasma cells or Tregs being higher in primary or metastatic subjects, respectively (Fig. 6f). We applied Pearson analysis to exploring the relevance between immune cells and ARGs. We examined 22 immune cell subsets (Fig. 6g), revealing a link with infiltration. Specifically, CASP1, HDAC6, and RHEB showed a strong correlation with infiltrating plasma cells, while RHEB, MYC, and CHMP2B were closely associated with infiltrating Treg cells (Fig. 6h). This represents that these hub genes function in modulating immune cell infiltration, particularly Treg cells, in pancreatic cancer.

Fig. 6.

Immune phenotype of PAAD, hub genes’ co-expression patterns, and immune cell subsets. a Stromal score. b Immune score. c ESTIMATE score. d Tumor purity. e PCA of GEO samples. f Immune cell abundance in PAAD. g Correlations among different infiltrating immune cells. h Relationships between ARGs and infiltrating immune cells. i Relative proportions of 22 immune cells

Expression characteristics of the 5 hub genes in PC cells or tissues

We assessed the levels of five differentially expressed ARGs in PC and normal tissues (n = 179) using TCGA + GTEx-PAAD data. CASP1, CHMP2B, MYC, and RHEB were significantly elevated in PC tissues, while HDAC6 was decreased (Fig. 7a). Five ARGs were measured for their mRNA levels using qPCR in five pancreatic cancer (PC) cell lines, five normal pancreatic cell lines, and PC and normal tissues. The outcomes of the research indicated that PC cells and tissues had considerably higher levels of the mRNAs of four ARGs, with the exception of HDAC6 (Fig. 7b–k). These results are consistent with the five differentially expressed ARGs (DEARGs) that we identified during our first screening.

Fig. 7.

Validation of hub genes expression. a Gene expressions in TCGA + GTEx-PAAD. b–f mRNA levels of five ARGs in five PC cell lines. g–k mRNA levels in five paired PC tissues

RHEB silencing suppresses proliferation, migration, and aggression of pancreatic cells

Given the significant increase in RHEB mRNA, we selected it to validate its role in pancreatic cancer (PC) cells. Immunohistochemistry (IHC) and Western blot (WB) assays confirmed that RHEB protein expression was markedly more in PC tissues and cells in comparison to normal counterparts (Fig. 8a, b). Upon silencing RHEB (siRHEB) (Fig. 8c), there was a notable reduction in PC cell growth (Fig. 8d–g). Furthermore, RHEB silencing led to significantly fewer cells traversing the membrane and reduced invasion of the Matrigel barrier compared to controls (Fig. 8h–i). These findings represent that RHEB may facilitate the proliferation, migration, and aggression of PC cells.

Fig. 8.

RHEB enhances proliferation, invasion, and migration in PC cells. a RHEB immunohistochemical staining. b Protein levels of RHEB. c siRNA knockdown of RHEB. d–e Cell proliferation. f–g Representative images and quantitative analyses of colony formation and invasion (*p < 0.05, **p < 0.01, ***p < 0.001)

RHEB silencing upregulates LC3

In our final analysis, we explored the relationship between RHEB and the autophagy marker LC3. Immunofluorescence experiments revealed that RHEB and LC3 co-localize on the cell membrane, with LC3 expression significantly increasing when RHEB is silenced (Fig. 9a). Western blot assays further confirmed elevated LC3 levels in RHEB-silenced cells compared to controls (Fig. 9b). This negative correlation between RHEB and LC3 implies that RHEB may influence pancreatic cancer cell proliferation and metastasis through autophagy, positioning RHEB as a potential target for pancreatic cancer therapy.

Fig. 9.

RHEB silencing upregulates the expression of LC3. a Immunofluorescence experiments of RHEB and LC3 in the PANC-1–1 and MIA PaCa-2 cells transfected with the siRNA or the siRHEB. b Western blot assays of RHEB and LC3 in the PANC-1–1 and MIA PaCa-2 cells transfected with the siRNA or the siRHEB

Discussion

PC is still a daunting challenge due to its aggressive character, with patients facing a grim 5–11-month survival period and a mere 9% 5-year survival rate [1]. Metastasis is a major contributor to the high mortality rate associated with this disease. In our study, we identified differentially expressed autophagy-related genes linked to pancreatic cancer metastasis and prognosis. Through regression analysis, we pinpointed five ARGs that either promote or inhibit PCM. Our survival curve analysis demonstrated that this model provides a more precise prognosis than traditional clinicopathological data by accurately classifying patients based on risk. We employed enrichment assays, network assays, and immune infiltration assays, which indicated that these ARGs influence PCM by modulating autophagy and immune infiltration. Additionally, we proposed several small-molecule drugs targeting these ARGs.

We created a prognostic model grounded in these five ARGs and utilized it for functional enrichment analysis. The oncogene MYC, known for its amplification in various tumors, lacks a fully understood mechanism [29]. CASP1, part of the cysteinyl aspartate protease family, is implicated in tumor metastasis [30, 31]. Previous studies have shown CASP1’s role in promoting liver cancer metastasis, potentially through autophagy [32]. CHMP2B, associated with receptor degradation or recycling, has limited information regarding its role in cancer [33], although it has been suggested as a prognostic marker for PC [34]. Our findings indicate that CHMP2B exacerbates PCM, likely through autophagy, leading to poor prognosis. HDAC6 is involved in cell adhesion, invasion, and migration through deacetylation processes [35]. While its relationship with PC is less explored compared to other tumors, some studies have established a connection between HDACs and PC [36, 37]. Our results suggest that HDAC6 is downregulated in PCM tissues, indicating its potential inhibitory role in PCM. Notably, three of the 10 proposed small-molecule drugs are HDAC inhibitors. Further research is required to elucidate its precise mechanism of action.

RHEB, a GTPase known for activating mTORC1, has been identified as an oncogenic protein [38]. Previous studies have demonstrated that silencing RHEB can restrain the growth of colon tumor cells and accelerate apoptosis by suppressing the mTOR pathway [39]. Moreover, elevated levels of RHEB have been linked to poor prognosis in PC [40]. Our study corroborates these findings, showing that RHEB silencing reduces the proliferation and metastatic potential of PC cells. Although our study confirmed the role of RHEB in driving PC advance and metastasis, the precise mechanisms remain to be elucidated. Future research will focus on unraveling these pathways to enhance our understanding of immune signaling in cancer. It is still important to mention that our study’s sample size was limited, and the clinical data was predominantly derived from TCGA, which mainly includes Caucasian populations. Additionally, immune cell infiltration was assessed using a single algorithm, necessitating further verification. By addressing these limitations, we aim to help to indentify new immune targets and signaling pathways, aligning with the goals of improving cancer immunotherapy.

In our research, enrichment assays of five ARGs indicated significant enrichment in necroptosis, senescence, and the thyroid hormone (TH) axis, which affects mitophagy and other biological processes [41]. The THR axis, known to play a role in hepatic autophagy, is also implicated in liver cancer [42]. This suggests that ARGs might influence PCM through these pathways. Furthermore, transcription factors and microRNAs have been found to be tightly associated with PC progression [43–52]. Our network assays revealed that most of the five ARGs are linked to FOS/MYC and regulated by five miRNAs, supporting the idea that FOS or miR-369-3p could significantly impact PCM by modulating RHEB, a hypothesis we plan to explore in future research.

Immune infiltration is crucial for tumor progression, as it aids in immune surveillance evasion and recruitment of immunosuppressive cells, which suppress tumor immunogenicity [53]. Although immune infiltration and autophagy are strongly associated with PC evolution [5], little research has focused on immune infiltration in metastatic PC patients. Our comprehensive study of immune infiltration mechanisms and ARGs in PCM revealed increased Tregs and decreased plasma cells in metastatic subjects, with the five ARGs linked to various cell subsets. Dendritic cells, B cells, and cytotoxic lymphocytes, components of tertiary lymphoid structures (TLSs), are generally associated with better cancer prognoses [54]^. Consistent with this, decreased B cell numbers have been linked to poor PC prognosis, a trend observed in other cancers as well [55–57]. While the role of B cells in autophagy remains unclear, Tregs are known to be influenced by autophagic depletion and rely heavily on autophagy to function [58]. Despite limited studies on Tregs in PCM, data suggest that in pancreatic ductal adenocarcinoma, Tregs are upregulated early but restricted to paracancerous lymph nodes in advanced stages [59]. Tregs modulate CD4 + T cell infiltration in the tumor microenvironment through a CTLA-4/CD8-dependent axis [59], though further verification is needed. Our findings suggest a regulatory relationship between immune infiltration and 5 ARG levels, indicating that PC cells might modulate immune infiltration via autophagy during invasion or metastasis, potentially attracting immunosuppressive cells to facilitate immune escape and aggression. This notion could explain the bad prognosis observed in metastatic PC patients, though further experiments are necessary to validate this hypothesis.

In our pursuit of advancing cancer immunotherapy, we explored the potential of small-molecule drugs as modulators of ARGs to identify novel therapeutic biomarkers. Gene alterations present a promising avenue for altering tumor therapy outcomes and targeting ARGs could unveil new immune targets for precision medicine. For instance, Apicidin, a histone deacetylase inhibitor, interacts with class I HDACs, disrupting deacetylation processes and modifying protein acetylation levels, thereby exerting therapeutic effects on PC [60]. Similarly, simvastatin, known for its lipid-lowering properties, has shown potential in counteracting gemcitabine resistance in PC by disrupting the TGF-β1/Gfi-1 signaling pathway [61], although its direct effects on PC remain to be amply explored. These findings underscore the need for further validation of small-molecule drugs in the context of cancer immunotherapy.

Conclusion

In this study, we have identified and validated five ARGs that are significantly associated with pancreatic cancer metastasis and prognosis. Through rigorous RT-qPCR and Western blot experiments, we confirmed that these genes’ expression levels align with our initial findings. Notably, our research highlights the negative correlation between RHEB and LC3, suggesting that RHEB may enhance pancreatic cancer cells’ proliferation, aggression and migration. This insight into the molecular interplay between RHEB and LC3 could unveil novel immune targets for cancer immunotherapy. Moreover, we created an innovative prognostic model grounded in these five ARGs, which holds promise for advancing precision medicine and enhancing the efficacy of anti-cancer treatments. This model could serve as a pivotal tool in identifying new immune targets and signaling pathways, aligning with the goals of improving cancer immunotherapy and personalized immune medicine.

Abbreviations

ARGs

Autophagy-related genes

DEARGs

Differentially expressed autophagy-related genes

GEO

Gene Expression Omnibus

TCGA

The cancer genome database

PDAC

Pancreatic ductal adenocarcinoma

UCSC

The University of California Santa Cruz

HADb

Human Autophagy Database

DEmRNAs

The mRNAs with different expression levels

OS

Overall survival

LASSO

Least absolute shrinkage and selection operator

PAAD

Pancreatic ductal adenocarcinoma

time-ROC

Time-dependent Receiver Operating characteristic curves

AUC

The area under the ROC curve

DCA

Decision curve analysis

GO

Gene Ontology

KEGG

Kyoto Encyclopedia of Genes and Genomes

BP

Biological process(s)

CC

Cellular

MF

Molecular function

L1000FWD

L1000 fireworks display

PubChem

Published Chemical information component

ESTIMATE

The Estimation of STromal and Immune cells in MAlignant Tumors tissues using Expression data

TME

The tumor microenvironment

RT-PCR

Reverse transcriptase-polymerase chain reaction

Author contributions

QD, ZJ: conception and design. QD, JH and FW: performed experiments. QD, LW: collect data and analyzed data. QD, LW: drafted the manuscript. ZJ, JH: reviewed data and finalized the manuscript. All authors reviewed and approved the final version.

Funding

This study was jointly supported by Sichuan Medical and Health Care Promotion Institute Youth Research Project (KY2023QN0129), Research Project of Early Gastrointestinal Cancer Physician Co-Growth Program (GTCZ-2023-SC-06) and The Third Hospital of Mianyang Research Project (202206). The funding agencies had no role in study design, data collection, and analysis, decision to publish, or preparation of the manuscript.

Data availability

The datasets used and/or analysed during the current study are available from NCBI Gene Expression Omnibus (GEO: GSE19279) https://www.ncbi.nlm.nih.gov/geo/ and the cancer genome database (TCGA -PAAD) https://portal.gdc.cancer.gov.

Declarations

Ethics approval and consent to participate

Informed consent was obtained from each participant included in the study. This study was ratified by the Ethics Committee of the Third Hospital of Mianyang on September 27, 2022 (Approved notice: 2022-14-2), and the experimental procedures followed the Declaration of Helsinki.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.