Individualized treatment and key prognostic biomarkers based on folate metabolism in patients with pancreatic cancer

Introduction

Pancreatic cancer (PC) is among the most malignant solid tumors. Its overall 5-year survival rate is 13% (1-3). Pancreatic ductal adenocarcinoma (PDAC) is the most common subtype, accounting for 90% of all PC cases, and is characterized by insidious early clinical symptoms, rapid tumor progression, and an aggressive and extremely poor prognosis (4-6). Traditional clinical diagnostic methods, including serum markers and radiological imaging (such as computed tomography and magnetic resonance imaging), often fail to detect precancerous lesions or predict patient prognosis. Therefore, there is an urgent need to identify effective biomarkers and potential therapeutic targets to reduce mortality (7).

Folate (vitamin B9) is involved in cancer cell proliferation and DNA metabolism. Folate participates in DNA synthesis and repair by carbon metabolism, thereby affecting cancer cell proliferation and epigenetic regulation (8). In addition, excessive folate accelerates tumor progression by promoting the metabolic reprogramming of cancer cells. High-dose folate (20 mg/kg) promotes the development of hepatocellular carcinoma by stabilizing methionine adenosine transferase 2A (MATIIα) (9). In colorectal cancer, the lack of folate increases the risk of DNA damage, and folate supplementation accelerates tumor growth (10). Folate activation requires folate metabolism. The core pathways of folate metabolism are divided into the folate and methionine cycles. Their core functions include nucleotide synthesis, methyl donation, and glutathione synthesis to maintain intracellular redox equilibrium (11,12). For example, in hepatocellular carcinoma, the intake of large amounts of folic acid stabilizes the MATIIα protein by deubiquitination of VCIP135, thereby accelerating the methionine cycle and promoting the proliferation of cancer cells. Therefore, high expression of MATIIα and VCIP135 in liver cancer tissue is significantly correlated with poor prognosis in patients (12). Folate metabolism is compartmentalized in the cytoplasm and mitochondria of PC cells. In the cytoplasm, the main functions of folate metabolism are the synthesis of nucleotides and tetrahydrofolate (THF) and the support of DNA replication (12). Additionally, the role of folate metabolism in shaping the tumor immune microenvironment (TIME) of PC is still ambiguous.

The TIME comprises tumor cells, tumor-infiltrating lymphocytes (TILs), fibroblasts, blood vessels, nerves, and various extracellular matrices (13). In PC, the TIME is mainly characterized by high immune suppression, which limits the ability of the immune system to eliminate tumor cells and plays a crucial role in the malignant development of PC (14). The major cellular components of the TIME include tumor-associated macrophages (TAMs), cancer-associated fibroblasts (CAFs), and TILs (13). TAMs are important immunosuppressive factors in the TIME. Cytokines secreted by PC cells (mainly IL-4 and IL-13) and colony-stimulating factor 1 can recruit macrophages and polarize them to the M2 type (11). M2 macrophages secrete many factors, such as Interleukin-10 (IL-10), arginase-1 (ARG-1), tumor necrosis factor-alpha (TNF-α), and vascular endothelial growth factor (VEGF). Interestingly, IL-10 can enhance M2 polarization through the JNK/STAT3 pathway, thus forming a feedback loop (15). Therefore, M2 macrophages are dominant in the TIME of PC, and the number of M2 macrophages also correlates with shorter overall survival (OS) and disease-free survival (DFS) in patients. Moreover, ARG-1, TNF-α, and VEGF promote immune escape, angiogenesis, and tumor invasiveness (16,17). Furthermore, M2 macrophages specifically express CD206 on the cell membrane surface; therefore, M2 macrophages can be identified based on the expression of CD206 in the TIME of PC (18). Carbohydrate antigen 19-9 (CA19-9) in PC can promote macrophage polarization to the M2 phenotype and affect the TIME; however, the specific mechanism remains unclear (19). In addition, TAMs in PC originate not only from monocytes in the blood circulation but also from yolk sac embryos derived from embryonic development (20,21). TAMs play a role in promoting fibrosis, and their role in promoting PC progression is stronger than that of monocyte-derived macrophages (22). CAFs are heterogeneous cells, comprising inflammatory fibroblasts (iCAFs), myofibroblasts (myCAFs), and antigen-presenting fibroblasts (23). Fibroblast activation protein (FAP) and alpha-smooth muscle actin (α-SMA) are expressed on myCAFs and iCAFs, respectively. FAP⁺ myCAFs promote tumor occurrence and development and reduce OS, whereas α-SMA⁺ iCAFs can inhibit tumor occurrence and development and prolong OS (24). TILs are also important components of the TIME. CD8⁺ and FOXP3⁺ cells are the two subsets of TILs. CD8⁺ TILs are key immune cells in tumor immunity that kill cancer cells by triggering apoptosis, whereas FOXP3⁺ TILs are immunosuppressive regulatory T cells (Tregs) that inhibit the proliferation and activation of CD8⁺ TILs and T helper cells and promote immune escape from tumors (25,26).

Therefore, this study aimed to investigate the relation between folate metabolism-related genes (FMGs) and the prognosis of PC and to explore the relation between key genes of important prognostic value in FMG and various components of the TIME. We present this article in accordance with the REMARK reporting checklist (available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-2070/rc).

Methods

Data collection and screening

This single-center retrospective study included 150 surgically resected and pathologically confirmed PC patients from The Affiliated Hospital of North Sichuan Medical College (January 2016–October 2024), with a median follow-up of 38 months. The inclusion criteria were as follows: (I) pathological diagnosis of primary PC. (II) Patients did not receive radiation therapy, chemotherapy, or immunotherapy before surgery. (III) Iconography revealed that the tumor was a resectable PC. (IV) Patients with PC remained free of distant metastases. (V) The tumor did not invade large blood vessels or more than four (including four) lymph nodes. Exclusion criteria: (I) patients received neoadjuvant therapy before surgery. (II) The pathological diagnosis was metastatic tumor. (III) Iconographic evaluation revealed borderline unresectable PC or locally advanced PC. (IV) The tumor had invaded nearby large vessels. (V) PC had spread to four or more lymph nodes. (VI) Distant metastases were also observed. Tumor classification was conducted according to the 2019 World Health Organization guidelines, and staging followed the American Joint Committee on Cancer Eight Edition criteria. DFS was measured as the time to first tumor recurrence or death from any cause, starting from the date of surgery. OS was determined as the time to death. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the Medical Ethics Committee of The Affiliated Hospital of North Sichuan Medical College (No. 2025ER257-1). Informed consent was taken from all the patients.

In addition, datasets (TCGA-PC, GSE57495, GSE62452, GSE28735, ICGC-AU and E-MTAB-6134) from The Cancer Genome Atlas (TCGA), Gene Expression Omnibus (GEO), International Cancer Genome Consortium (ICGC), and ArrayExpress were collected and pooled using the SVA package to remove batch effects, resulting in 732 PC samples. Subsequently, 78 FMGs were extracted from the molecular signature database to form a gene set.

Comprehensive analysis of folate metabolism genes in pan carcinoma

FMGs were comprehensively explored and summarized by downloading multi-omics data from the TCGA pan-cancer cohort. Copy number variation (CNV), single nucleotide variation (SNV), and methylation of folate metabolism genes were analyzed in detail. At the transcriptomic level, we extensively investigated the expression of the FMGs in the TCGA pan-cancer cohort. We also evaluated the risk factors for the folate metabolism gene set in the TCGA pan-cancer cohort, providing a meaningful understanding of the regulatory mechanisms and clinical significance of folate metabolism from the biological perspective of pancreatic carcinoma.

Non-negative matrix factorization (NMF) clustering determination of folate metabolism genes modification subtypes and pathway enrichment analysis

We further investigated the relation between FMG expression and the clinical characteristics of PC. The NMF algorithm was used to cluster 732 PC samples according to folate metabolism gene expression. The NMF algorithm identifies potential gene expression models by decomposing the original matrix into two or more nonnegative matrices. The specific parameters of the NMF clustering algorithm are as follows: rank =2:10, method = “brunet”, nrun =100. The first point on the cophenetic curve, with the largest descending range, was selected as the classification number. We also performed a Kaplan-Meier (K-M) survival analysis using the “survival” package in R to better understand the significance of NMF clustering in PC clinical outcomes. In addition, we used the R package “GSVA” to calculate valid scores for pathway activity of each gene set and compared differences in folate metabolism scores between two groups of patients using the Wilcoxon test in R. We also used “GSVA” package to calculate the enrichment scores for 50 signature pathways to differentially identify potential signaling pathways between subtypes.

Analysis of TIME

To explore the immunological characteristics between the two cluster subtypes, we quantified ImmuneScore, StromalScore, EstimateScore, and tumor purity for C1 and C2 using the R package “estimate”. For visualization, the R package “ggplot2” and “ggpubr” were used. To comprehensively assess immune cell infiltration, we applied several immune-related algorithms, including TIMER, CIBERSORT, CIBERSORT-ABS, QUANTISEQ, MCPCOUNTER, XCELL, and EPIC from the “IOBR” package. Furthermore, tumor cells can employ immune checkpoints to evade immune system attacks. Therefore, we compared immune checkpoint expression levels between the subtypes of PC.

Analysis of drug sensitivity and differentially expressed genes

Using the R package “pRRophetic”, we forecasted drug sensitivity to enhance understanding of the relation between tumor drug treatment and the expression of folate metabolism genes. We also used R’s Wilcox test to identify genes that differ in folate metabolism between the two subgroups. Gene Expression Profiling Interactive Analysis (GEPIA) and TIME were used to analyze differences in gene expression between C1 and C2, as well as survival and associations with tumor immune cell infiltration.

Immunohistochemistry and evaluation

Tissue sections from PC patients were baked at 60 ℃ for 4 h, dewaxed with Van-clear (H-H101, Hongzi Biotechnology, Chongqing China) and graded ethanol, and subjected to antigen retrieval using Tris-EDTA (pH 9.0, MVS-0100, Maixin Biotechnology, Fuzhou, China) under boiling for 15 min. Immunostaining was performed using the UltraSensitive SP kit (KIT-9710, Maixin Biotechnology). Endogenous peroxidase and nonspecific binding were blocked using peroxidase blocker (Reagent 1) and protein block (Reagent 2) for 10 min at room temperature. Primary antibodies—including DHFR (AB124814, Abcam), CD206 (YT5640, Immunoway), CD8 (AF5126, Affinity), FOXP3 (AF6544, Affinity), FAP (AF5344, Affinity), and α-SMA (AF1032, Affinity)—were incubated overnight at 4 ℃. Sections were then treated with biotinylated secondary antibody (Reagent 3) and streptavidin-horseradish peroxidase (HRP) (Reagent 4) for 10 min each, followed by DAB (DAB-0031, Maixin Biotechnology) development (1–3 min), hematoxylin counterstaining, differentiation in acid alcohol (G1861, Solarbio), dehydration, and mounting. Three researchers, blinded to the clinicopathological data, evaluated the stained slides under a ×200 light microscope. For CD206, CD8, and FOXP3, positivity was determined by averaging counts from three tumor fields. DHFR, FAP, and α-SMA expression was scored as: 0 (negative), 1 (weak), 2 (moderate), or 3 (strong); scores ≥2 were considered positive.

Statistical analysis

Continuous variables that were normally distributed are expressed as mean ± standard deviation, whereas non-normally distributed data are expressed as median (interquartile range). Differences between continuous variables were assessed using the Student’s t-test or Mann-Whitney U test, whereas categorical variables were compared using the Chi-squared test or Fisher’s exact test. Logistic regression models were used to identify clinical factors associated with prognostic molecules. Prognostic factors were determined by Cox proportional hazards regression analysis, and variables with P<0.05 in the univariate analysis were incorporated into the multivariate analysis. The Kaplan-Meier method was employed to construct survival curves, and differences in survival were assessed by the log-rank test, with a P value <0.05 considered statistically significant. Correlation analysis was performed using the Wilcoxon rank sum test and chi-square test. The statistical significance level was set at P<0.05. All analyses were performed using IBM SPSS Statistics 23 software (version 26.0; Chicago, IL, USA) and GraphPad Prism 10 software (version 9.5; San Diego, CA, USA).

Results

Pan-cancer overview of the folate-related genes

A pan-cancer analysis was conducted on FMGs comprising 78 genes, revealing that FMGs are generally characterized by genomic and epigenetic alterations, including CNVs, SNVs, risk factor and a predominant trend of promoter hypomethylation (Figure 1A-1E). When focusing on the pattern most relevant to PC, the results indicated that PC exhibited a distinct FMG copy number change spectrum. Additionally, it displayed one of the highest SNV mutation frequencies on the key tumor suppressor gene TP53 in folic acid metabolism. Moreover, it demonstrated a nearly universal trend of hypomethylation, with genes such as MTHFD1 and DHFR being notably affected. This collective multi-omics alteration pattern, integrating specific genomic instability with extensive epigenetic dysregulation, underscores a potent mechanism of folic acid metabolism dysregulation in PC.

Figure 1 Overview of FMGs in pan-carcinoma. (A) CNV in pan-cancer; red represents amplification of the copy. (B) CNV in pan-cancer; blue represents deletions. (C) SNV frequency in pan-carcinoma. (D) For folate metabolism-related gene survival rates in all cancer types, the white is considered to indicate P>0.05; the red represents oncogenes; and the green represents protective genes. (E) In various types of cancer, methylation levels of FMGs are shown in different colors. The change from red to blue represents the methylation level of the gene from high to low. CNV, copy number variation; FMGs, folate metabolism-related genes; SNV, single nucleotide variation.

NMF cluster typing based on FMG expression

We clustered 732 patients into the C1 and C2 groups based on FMG expression using the NMF algorithm (Figure 2A). Using GSVA and K-M analyses, we found that patients in C2 had poorer OS (P=0.002) and higher metabolic scores (P<0.001) than those in group C1 (Figure 2B,2C). Because of differences in OS and pathway activation between the two subgroups, we compared FMGs between the two subgroups and identified 33 differentially expressed genes (Figure 2D). In addition, pathway differences were observed between subtypes C1 and C2, with pathway activation being higher in subtype C2 than in subtype C1 (Figure 2E). For example, KRAS_SIGNALING_DN, PANCREAS_BETA_CELL, BILE_ACID_METABOLISM, SPERMATOGENESIS, XENOBIOTIC_METABOLISM, and FATTY_ACID_METABOLISM were more active in C2.

Figure 2 FMGs clusters analysis. (A) Based on NMF, 732 patients are divided into two subgroups (C1 and C2). (B) Compared with C1, C2 has a better survival rate in the survival curve (the abscissa is the number of years, while the ordinate is the survival rate). (C) A violin plot means differential folate-metabolism score between C1 and C2. (D) Comparison of classical pathways between C1 and C2. (E) Differential expression of FMGs was revealed between subgroups. *, P<0.05; **, P<0.01; ***, P<0.001; ****, P<0.0001. FMGs, folate metabolism-related genes; NMF, non-negative matrix factorization.

Correlation between cluster typing and TIME

We used transcriptome data to perform immunoscoring for subtypes C1 and C2 (Figure 3A-3D). The results showed that C2 had higher scores than those of C1 for ESTIMATEScor (P<0.0001), immune score (P<0.001), and StromalScor (P<0.0001), whereas C1 had higher tumor purity (P<0.0001) than that of C2. To further explore the relation between the two subgroups and components of the TIME (Figure 3E), we applied seven different algorithms. We found that the number of immune cells in subgroup C2 was higher than in subgroup C1. For example, the numbers of macrophages, CD8⁺ T cells, and CD4⁺ T cells were high in C1. In addition, comparison of immune checkpoint gene expression between the C1 and C2 subgroups revealed significant differences in CD80, JAK2, PDCD1, TNFRSF9, TNFSF4, and YTHDF1 (Figure 3F). Specifically, CD80 (P<0.001), JAK2 CD80 (P<0.0001), TNFRSF9 (P<0.0001), TNFSF4 (P<0.0001), YTHDF1 (P<0.0001) were more highly expressed in C2 than in C1, while PDCD1 (P<0.05) were all expressed at higher levels in C1. Therefore, it is the differential expression of immune checkpoint genes that provides crucial molecular insights for deciphering the complexity of the TIME, predicting responses to existing immunotherapies, and designing novel, precise combination treatment strategies.

Figure 3 Differences between two subgroups and TIME. (A-D) Violin and box plots showing comparisons of ESTMATEScore, ImmuneScore, StromalScore, and TumorPurity between C1 and C2. (E) Comparison of immune checkpoint genes expression in C1 and C2. (F) To explore the differences in tumor immune microenvironment between C1 and C2 through seven different algorithms. *, P<0.05; **, P<0.01; ***, P<0.001; ****, P<0.0001. ns, not significant; TIME, tumor immune microenvironment.

Comparison of drug susceptibility between C1 and C2

Molecular-targeted therapy has long been a focus in cancer treatment (Figure 4A-4G). We compared the responses of C1 and C2 to commonly used chemotherapy drugs using the “pRRophetic” package. The drugs included epothilone B, mitomycin, paclitaxel, bryostatin-1, cyclosamine, docetaxel, and doxorubicin. We found that sensitivity to these chemotherapy agents was higher in subgroup C1 than in C2. Therefore, the use of these drugs may be more beneficial in C1, which is associated with a poor prognosis.

Figure 4 We utilized the ‘pRRophetic’ package to predict drug sensitivity and enhance our understanding of the relationship between tumor drug treatment and the expression of FMGs. (A-G) The yellow box represents C1, while the green box represents C2. Each graph compares the IC50 of drugs (including epothilone B, mitomycin, paclitaxel, bryostatin.1, cyclosamine, docetaxel, and doxorubicin) between C1 and C2. FMGs, folate metabolism-related genes; IC50, half maximal inhibitory concentration.

DHFR as a prognostic marker for PC and key driver gene for the TIME

Based on the DEGs between C1 and C2 (Figures S1,S2), the prognostic analysis showed that only DHFR and ALDH1L1 exhibited significant differences in OS and DFS when divided into high- and low-expression groups by the median (Figure 5A,5B). High expression of DHFR or ALDH1L1 was associated with a poor prognosis. We explored the relation between the expression of these two genes and the TIME (Figure 5C,5D). DHFR expression was correlated with B cells, CD8⁺ T cells, macrophages, neutrophils, and dendritic cells. ALDH1L1 was associated only with CD8⁺ T cells. To further explore DHFR as a key driver of the TIME (Figure 5E), we used CIBERSORT, MCP-counter, EPIC, quanTIseq, and custom signatures to assess the relation between DHFR expression and various components of the TIME. DHFR expression was correlated with Tregs, M2 macrophages, dendritic cells, and neutrophils. When DHFR was highly expressed, Tregs, M2 macrophages, dendritic cells, and neutrophils occupied a larger proportion of the TIME. Additionally, DHFR and major components of the TIME were identified through immunohistochemical staining of 150 clinicopathological sections.

Figure 5 Based on significantly different genes between C1 and C2, we screened for a key gene related to OS and TIME in PC. (A,B) DHFR and ALDH1L1 were identified through OS and DFS analysis from a pool of 33 genes, distinguishing between two expression groups (blue for low DHFR expression and red for high DHFR expression) in cohorts C1 and C2. (C) The correlation between ALDH1L1 and the TIME was analyzed, including purity, B cell, CD8⁺ T cell, CD4⁺ T cell, macrophage, neutrophil, and dendritic cell. (D) The correlation between DHFR and the TIME was analyzed. (E) The alterations in diverse immune cell populations, including MDSCs, macrophages, T cells, and B cells, are depicted using a color spectrum ranging from blue to yellow. These changes are further categorized and visualized through hierarchical clustering. CAF, cancer-associated fibroblast; DFS, disease-free survival; DHFR, dihydrofolate reductase; HR, hazard ratio; ICB, immune checkpoint blockade; MDSCs, myeloid-derived suppressor cells; NK, natural killer; OS, overall survival; PC, pancreatic cancer; TAM, tumor-associated macrophage; TIME, tumor immune microenvironment; TPM, transcripts per million.

Patient of PC characteristics and clinical pathological data

In total, 150 patients with surgically resected PC were include. Among them, 96 (64%) developed recurrence, and 105 (70%) died from the disease. Positive DHFR immunoreactivity was observed in 102 tumor samples (68%), whereas 48 cases (32%) showed negative expression (Figure 6A). We also assessed the number of macrophages expressing CD206, TILs expressing CD8, and TILs expressing FOXP3, defining patients as having high or low expression based on a 50% threshold (Figure 6B-6F). Additionally, we evaluated the expression of FAP and α-SMA in CAFs. Among these, 92 cases were FAP-positive, 58 cases were FAP-negative; 88 cases were α-SMA-positive, and 62 cases were α-SMA-negative.

Figure 6 Immunohistochemical staining in PC specimens. (A) Expression of DHFR in PC cells. (B) Expression of CD206 in tumor-associated macrophage. (C) Expression of FAP in cancer-associated fibroblasts. (D) Expression of α-SMA in cancer-associated fibroblasts. (E) Expression of CD8 in tumor-infiltrating lymphocytes. (F) Expression of Foxp3 in tumor-infiltrating lymphocytes. α-SMA, alpha-smooth muscle actin; DHFR, dihydrofolate reductase; FAP, fibroblast activation protein; PC, pancreatic cancer.

Among the clinicopathological characteristics evaluated (age, sex, total bilirubin level, tumor size, T stage, N stage, degree of differentiation, tumor location, and vascular invasion), no significant differences were observed between the DHFR-negative and DHFR-positive groups. However, CA19-9 levels differed significantly between groups (Z=−2.007, P=0.04). These findings indicate that DHFR expression does not correlate with most clinical parameters in patients with PC but demonstrates a statistically significant association with CA19-9 levels (Z=−2.007, P=0.03) (Table 1). Additionally, univariate and multivariate Cox proportional hazards regression analyses identified histological differentiation [hazard ratio (HR) =0.554, 95% confidence interval (CI): 0.387–0.793, P=0.006] and DHFR expression (HR =1.832, 95% CI: 1.284–2.619, P=0.003) as independent prognostic factors for OS (Table 2).

Validating DHFR as a prognostic marker and its link to the TIME

Based on the immunohistochemical staining results, K-M analysis showed that positive DHFR expression was associated with poor OS (P=0.001, 95% CI: 1.32–3.32) and DFS (P=0.02, 95% CI: 1.065–2.981) (Figure 7A,7B). In addition, studies have shown that the expression of DHFR is significantly different from CD 206⁺ macrophages (P<0.001) (Figure 7C), and is also significantly different from Foxp3⁺ TIL (P=0.03) (Figure 7D). However, the differences with CD8⁺ TIL, α-SMA⁺ and FAP⁺ CAFs cells were not statistically significant (P>0.05) (Figure 7E-7G). Our data suggest that DHFR expression plays an important role in the TIME of PC. This may be a new marker for potential targeted therapies for PC.

Figure 7 We investigated the association between DHFR expression and OS, DFS, and TIME. (A,B) Patients with PC were categorized into two groups based on their DHFR protein expression levels: a low-expression group and a high-expression group (the x-axis represents the number of months, while the y-axis represents the overall survival probability and disease-free survival probability. Patients with positive DHFR expression are denoted by red, while those with negative DHFR expression are denoted by blue). (C,D) The relationship between DHFR expression and the presence of tumor-associated fibroblast markers FAP and α-SMA is depicted in the graph (DHFR expression levels are plotted on the x-axis, with red bars representing FAP-positive and α-SMA-positive cancer-associated fibroblasts, and green bars indicating FAP-negative and α-SMA-negative tumor fibroblasts). (E-G) The abscissa represents CD206, CD8, and Foxp3, while the ordinate indicates their expression levels. The red bars denote DHFR-positive cells, and the green bars represent DHFR-negative cells. α-SMA, alpha-smooth muscle actin; CI, confidence interval; DFS, disease-free survival; DHFR, dihydrofolate reductase; FAP, fibroblast activation protein; HR, hazard ratio; OS, overall survival; PC, pancreatic cancer; TIME, tumor immune microenvironment.

Discussion

PC is one of the most malignant solid tumors worldwide. Its occurrence and development are unknown, and effective treatment methods and prognostic markers are lacking. Folic acid is the key to tumor proliferation, and folate metabolism, as the key to folate activation, is particularly important for the occurrence and development of cancer. Our pan-cancer analysis established a foundational landscape of FMGs across tumors. We observed that FMGs are frequently altered at the genomic (CNV, SNV) and epigenetic (methylation) levels in most cancer types, with widespread promoter hypomethylation emerging as a particularly common event. Notably, PC exhibited significant CNV alterations and contained FMGs with among the highest mutation frequencies (e.g., TP53). These findings underscore the general importance of folate metabolism dysregulation in oncology. Significant differences in survival analysis, metabolic scores, metabolic pathways, and TIME were observed between the C1 and C2 subgroups. In an analysis comparing the susceptibility to commonly used chemotherapeutic agents, we found that C1 was more susceptible than C2, suggesting a potential for enhanced therapeutic efficacy. DHFR is not only an independent prognostic risk factor for PC but also a key immunosuppressive gene in the TIME. Our study also showed that high DHFR expression was strongly associated with a significant increase in the number of CD206⁺ macrophages and FOXP3⁺ TILs. CD206⁺ macrophages are often expressed in M2-type TAMs, and FOXP3⁺ TILs are often expressed in Treg cells. Both TAMs and Tregs exhibit tumor immunosuppression. Therefore, DHFR may be a key regulator of TIME immunosuppression in patients with PC. To the best of our knowledge, this study is the first to clarify the prognostic role of DHFR in PC and its impact on the TIME.

DHFR is a key enzyme in the folate metabolism cycle that catalyzes the reduction of dihydrofolate to THF (27,28). Because cell proliferation depends on one-carbon units, THF is an important cofactor in the synthesis of one-carbon units (29,30). Therefore, the role of DHFR in driving the folate metabolic cycle is crucial. Cancer cell proliferation requires numerous nucleotides. As a key enzyme in one-carbon unit metabolism, DHFR plays an indispensable role in the proliferation of cancer cells because it catalyzes the production of THF, which provides sufficient raw materials for the synthesis of nucleotides necessary for cancer cell proliferation (31,32). Therefore, the inhibition of DHFR has become a therapeutic target for the treatment of various malignant tumors. For example, increased expression levels of DHFR mRNA and protein in glioblastoma are strongly associated with a poor prognosis, and the DHFR inhibitor methotrexate plays an important role in reducing the ability of glioma cells to self-renew (33). In addition, the high expression of DHFR in colorectal cancer, osteosarcoma, cervical cancer, and hepatocellular carcinoma can promote the occurrence, development, and invasion of cancer and affect the TIME (30,34-36). To date, no studies have examined the prognostic value of DHFR in PC and relation between DHFR expression and the TIME in PC.

Given that the immunosuppressive TIME in PC is characterized by tumor-promoting immunosuppression and that TAMs and Tregs represent the principal cellular components mediating this immunosuppressive phenotype, these cell populations have garnered significant attention in preclinical and translational research (37). Tumor cells orchestrate TAM activation and lipid metabolic reprogramming through the secretion of lipid and non-lipid mediators (38). For instance, in lung cancer, tumor-derived exosomes carrying long-chain fatty acids or TRIM59 promote TAM lipid metabolism and M2 phenotypic polarization (39). In ovarian cancer, arachidonic acid and its metabolites secreted by tumor cells modulate TAM functional polarization via multiple signaling cascades (40). In breast cancer, tumor cells utilize Hedgehog signaling and secrete S1P/microRNAs to regulate TAM lipid metabolism and M2 polarization (41-43). In PC, DHFR diminishes cellular susceptibility to ferroptosis by catalyzing the reduction of BH2 to BH4, thereby reducing lipid peroxides. This elevated resistance to oxidative stress and lipid peroxidation not only facilitates tumor growth but also equips tumor cells to withstand reactive oxygen species (ROS)-mediated cytotoxicity from activated immune effectors, such as cytotoxic T cells and NK cells (44). Thus, DHFR confers a cell-autonomous foundation for immune evasion. Beyond cell-intrinsic effects, DHFR likely orchestrates immunosuppressive niche formation. We observed a significant enrichment of DHFR in the p53 signaling pathway [normalized enrichment score (NES): 1.66, adj. P<0.001], suggesting its modulation of this axis (Figure S3). Notably, recent studies indicate that p53 dysregulation in PDAC alters lipid metabolism, leading to fatty acid accumulation that drives macrophage polarization toward the M2 phenotype (45-47). We therefore propose that DHFR may disrupt p53-mediated lipid homeostasis, thereby enhancing the recruitment and polarization of tumor-promoting M2-type TAMs. In addition, we found a significant correlation between DHFR expression and serum levels of CA19-9, a known driver of M2 macrophage polarization (19). This points to a potential synergy between DHFR and CA19-9-associated pathways in establishing an M2-TAM-dominated immunosuppressive milieu. The immunosuppressive landscape is further reinforced through cellular crosstalk. We confirmed that FOXP3⁺ Treg infiltration correlates with DHFR expression. Literature delineates a vicious cycle between M2 TAMs and Tregs: M2 TAMs secrete IL-10 and TGF-β, which suppress effector T cells and expand Tregs, while Tregs in turn enhance TAMs’ lipid metabolism and M2 function (48,49). DHFR appears to be a critical driver of this feedback loop. By facilitating M2 TAM generation (as above) and shaping a supportive microenvironment, DHFR promotes Treg survival and proliferation. Tregs then directly inhibit effector immune cells via receptors like TIGIT (50). This explains the paradoxical phenotype in DHFR-high tumors: an enriched overall immune infiltration coupled with a dominant proportion of inhibitory cells and compromised anti-tumor immunity. Notably, our analysis also revealed DHFR enrichment in the insulin signaling pathway, leading to a novel metabolic hypothesis. We speculate that DHFR-overexpressing PC cells may exacerbate the Warburg effect, consuming glucose voraciously and creating a localized “hypoglycemic” niche (51). Intriguingly, Tregs uniquely maintain robust proliferation under low glucose, whereas effector T cells depend on glycolysis. Thus, DHFR-driven metabolic stress may selectively favor Treg expansion while impairing effector T cell function, offering a systemic metabolic perspective on how DHFR tilts the immune cell balance. Collectively, these mechanisms—ranging from intrinsic ferroptosis resistance, extrinsic regulation of immune cell polarization and crosstalk, to potential metabolic reprogramming—paint a comprehensive picture of how DHFR fosters an immunosuppressive TME that fuels PDAC progression.

Identifying effective therapeutic targets for tumors has always been a topic of great interest. DHFR is a key enzyme involved in one-carbon metabolism. It catalyzes the reduction of dihydrofolate to THF, which is crucial for the synthesis of purines, pyrimidines, and amino acids within cells (52). In the tumor microenvironment, THF transports one-carbon units to construct purine rings. In the synthesis of pyrimidines, particularly thymidine monophosphate, THF donates methyl groups to facilitate thymidine monophosphate production, which in turn supports tumor growth and proliferation (22,53). DHFR promotes tumor cell growth by influencing folate metabolism in several types of cancers, including breast cancer, chondrosarcoma, colorectal cancer, and acute lymphoblastic leukemia (54). Specifically, DHFR expression is significantly higher in chondrosarcomas than in normal tissues and correlates with the malignancy grade of chondrosarcomas (55). A study on acute lymphoblastic leukemia showed that related gene variants affect the metabolism of methotrexate, which exerts anticancer effects by inhibiting DHFR (30,56). Notably, in breast cancer, the activity of DHFR and breast cancer cells was significantly inhibited by cycloguanil. This indicated that the inhibition of DHFR was closely associated with the suppression of cell growth, thus confirming the role of DHFR in breast cancer cell metabolism (57,58). In addition, 1,3-diamino-7H-pyrrolo[3,2-f]quinazoline derivatives (DHFR inhibitors) exert antitumor effects and outperform paclitaxel in some models (59). This further emphasizes the pivotal role of DHFR in tumor metabolism and provides new ideas for the treatment of PC.

Our study used a large sample size for analysis and cross-validation using bioinformatics and immunohistochemical experiments. We revealed the association between FMGs and the immune microenvironment in PC and highlighted DHFR as a key prognostic marker and therapeutic target in PC. In addition, our research was a single-center study. And we lack further functional experiments to prove how DHFR affects the correlation and the number of CD206⁺ TAMs and Foxp3⁺ T cells, which constitutes a core objective of our future research.

Conclusions

We divided PC into two subgroups according to FMG expression levels, revealing that FMG expression levels partially affect TIME, drug responsiveness, and OS in individuals with PC. We then screened DHFR for genes critical for folate metabolism and identified its association with TIME components as a prognostic biomarker of PC. Finally, we discuss the importance of DHFR in TIME.

Acknowledgments

The author(s) would like to thank the Institute of Hepato-Biliary-Pancreatic-Intestinal Disease, Affiliated Hospital of North Sichuan Medical College, Nanchong, China; And North Sichuan Medical College Innovation Centre for Science and Technology; Editage (www.editage.cn) for English language editing.

Footnote

Reporting Checklist: The authors have completed the REMARK reporting checklist. Available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-2070/rc

Data Sharing Statement: Available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-2070/dss

Peer Review File: Available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-2070/prf

Funding: This work was supported by the National Natural Science Foundation of China (No. 82300737), the Doctoral Start-up Fund of The Affiliated Hospital of North Sichuan Medical College (No. 2023GC010), the Office of Science and Technology of Nanchong (No. 22SXQT0110) and The Affiliated Hospital of North Sichuan Medical College Foundation (No. 2024LC011).

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-2070/coif). The authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the Medical Ethics Committee of The Affiliated Hospital of North Sichuan Medical College, Nanchong (No. 2025ER257-1). Informed consent was taken from all the patients.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Yu B, Shao S, Ma W. Frontiers in pancreatic cancer on biomarkers, microenvironment, and immunotherapy. Cancer Lett 2025;610:217350. [Crossref] [PubMed]
2. Siegel RL, Giaquinto AN, Jemal A. Cancer statistics, 2024. CA Cancer J Clin 2024;74:12-49. [Crossref] [PubMed]
3. Conroy T, Bachet JB, Ayav A, et al. Current standards and new innovative approaches for treatment of pancreatic cancer. Eur J Cancer 2016;57:10-22. [Crossref] [PubMed]
4. Cocca S, Pontillo G, Lupo M, et al. Pancreatic cancer: Future challenges and new perspectives for an early diagnosis. World J Clin Oncol 2025;16:97248. [Crossref] [PubMed]
5. Pergolizzi RG, Brower ST. Molecular Targets for the Diagnosis and Treatment of Pancreatic Cancer. Int J Mol Sci 2024;25:10843. [Crossref] [PubMed]
6. Rawla P, Sunkara T, Gaduputi V. Epidemiology of Pancreatic Cancer: Global Trends, Etiology and Risk Factors. World J Oncol 2019;10:10-27. [Crossref] [PubMed]
7. Brozos-Vázquez E, Toledano-Fonseca M, Costa-Fraga N, et al. Pancreatic cancer biomarkers: A pathway to advance in personalized treatment selection. Cancer Treat Rev 2024;125:102719. [Crossref] [PubMed]
8. Zwart NRK, Kok DE, Ariyaratne SNH, et al. Folate-Associated DNA Methylation and Chemotherapy-Induced Toxicities in Patients With Colorectal Cancer. Mol Nutr Food Res 2025;69:e70127. [Crossref] [PubMed]
9. Li JT, Yang H, Lei MZ, et al. Dietary folate drives methionine metabolism to promote cancer development by stabilizing MAT IIA. Signal Transduct Target Ther 2022;7:192. [Crossref] [PubMed]
10. Sanderson SM, Gao X, Dai Z, et al. Methionine metabolism in health and cancer: a nexus of diet and precision medicine. Nat Rev Cancer 2019;19:625-37. [Crossref] [PubMed]
11. Kim YI, Pogribny IP, Basnakian AG, et al. Folate deficiency in rats induces DNA strand breaks and hypomethylation within the p53 tumor suppressor gene. Am J Clin Nutr 1997;65:46-52. [Crossref] [PubMed]
12. Zarou MM, Vazquez A, Vignir Helgason G. Folate metabolism: a re-emerging therapeutic target in haematological cancers. Leukemia 2021;35:1539-51. [Crossref] [PubMed]
13. Sherman MH, Beatty GL. Tumor Microenvironment in Pancreatic Cancer Pathogenesis and Therapeutic Resistance. Annu Rev Pathol 2023;18:123-48. [Crossref] [PubMed]
14. Luo W, Wen T, Qu X. Tumor immune microenvironment-based therapies in pancreatic ductal adenocarcinoma: time to update the concept. J Exp Clin Cancer Res 2024;43:8. [Crossref] [PubMed]
15. Yang S, Liu Q, Liao Q. Tumor-Associated Macrophages in Pancreatic Ductal Adenocarcinoma: Origin, Polarization, Function, and Reprogramming. Front Cell Dev Biol 2020;8:607209. [Crossref] [PubMed]
16. Pan Y, Yu Y, Wang X, et al. Tumor-Associated Macrophages in Tumor Immunity. Front Immunol 2020;11:583084. [Crossref] [PubMed]
17. Thielman NRJ, Funes V, Davuluri S, et al. Semaphorin 3D promotes pancreatic ductal adenocarcinoma progression and metastasis through macrophage reprogramming. Sci Adv 2024;10:eadp0684. [Crossref] [PubMed]
18. Korc M. Heparin Prowess: Favorable Vascular-Immune Reprogramming in Pancreatic Cancer. Clin Cancer Res 2023;29:2348-50. [Crossref] [PubMed]
19. Zhang D, Cui F, Zheng K, et al. Single-cell RNA sequencing reveals the process of CA19-9 production and dynamics of the immune microenvironment between CA19-9 (+) and CA19-9 (-) PDAC. Chin Med J (Engl) 2024;137:2415-28. [Crossref] [PubMed]
20. Wynn TA, Vannella KM. Macrophages in Tissue Repair, Regeneration, and Fibrosis. Immunity 2016;44:450-62. [Crossref] [PubMed]
21. Saeed AF. Tumor-Associated Macrophages: Polarization, Immunoregulation, and Immunotherapy. Cells 2025;14:741. [Crossref] [PubMed]
22. Zhu Y, Herndon JM, Sojka DK, et al. Tissue-Resident Macrophages in Pancreatic Ductal Adenocarcinoma Originate from Embryonic Hematopoiesis and Promote Tumor Progression. Immunity 2017;47:323-338.e6. [Crossref] [PubMed]
23. Lo A, Li CP, Buza EL, et al. Fibroblast activation protein augments progression and metastasis of pancreatic ductal adenocarcinoma. JCI Insight 2017;2:e92232. [Crossref] [PubMed]
24. McAndrews KM, Chen Y, Darpolor JK, et al. Identification of Functional Heterogeneity of Carcinoma-Associated Fibroblasts with Distinct IL6-Mediated Therapy Resistance in Pancreatic Cancer. Cancer Discov 2022;12:1580-97. [Crossref] [PubMed]
25. Miyahara Y, Takano S, Sogawa K, et al. Prosaposin, tumor-secreted protein, promotes pancreatic cancer progression by decreasing tumor-infiltrating lymphocytes. Cancer Sci 2022;113:2548-59. [Crossref] [PubMed]
26. Kim JE, Kim H, Kim B, et al. Comprehensive immunoprofile analysis of prognostic markers in pancreaticobiliary tract cancers. Cancer Med 2023;12:7748-61. [Crossref] [PubMed]
27. Liu H, Qin Y, Zhai D, et al. Antimalarial Drug Pyrimethamine Plays a Dual Role in Antitumor Proliferation and Metastasis through Targeting DHFR and TP. Mol Cancer Ther 2019;18:541-55. [Crossref] [PubMed]
28. Sehrawat R, Rathee P, Khatkar S, et al. DihydrofolateReductase (DHFR) Inhibitors: A Comprehensive Review. Curr Med Chem 2023; Epub ahead of print. [Crossref]
29. Bhagat K, Kumar N, Kaur Gulati H, et al. Dihydrofolate reductase inhibitors: patent landscape and phases of clinical development (2001-2021). Expert Opin Ther Pat 2022;32:1079-95. [Crossref] [PubMed]
30. Nie J, Huang L, Shen Y, et al. Methotrexate resistance and its regulatory mechanisms in pediatric tumors and beyond. Drug Resist Updat 2025;81:101225. [Crossref] [PubMed]
31. Raimondi MV, Randazzo O, La Franca M, et al. DHFR Inhibitors: Reading the Past for Discovering Novel Anticancer Agents. Molecules 2019;24:1140. [Crossref] [PubMed]
32. Newman AC, Maddocks ODK. One-carbon metabolism in cancer. Br J Cancer 2017;116:1499-504. [Crossref] [PubMed]
33. Fawal MA, Jungas T, Davy A. Inhibition of DHFR targets the self-renewing potential of brain tumor initiating cells. Cancer Lett 2021;503:129-37. [Crossref] [PubMed]
34. Meng XN, Ma JF, Liu YH, et al. Dynamic genomic changes in methotrexate-resistant human cancer cell lines beyond DHFR amplification suggest potential new targets for preventing drug resistance. Br J Cancer 2024;130:1819-27. [Crossref] [PubMed]
35. Murugesan B, Rathinavel T, Kumar RS, et al. Atranorin from lichens act as potential inhibitor for cervical tumor proteins TGFbeta and kinase enzyme CDK4/CyclinD3: An anti-cancer intervention therapy. Int J Biol Macromol 2025;320:145896. [Crossref] [PubMed]
36. Peng Y, Li Z, Chen S, et al. DHFR silence alleviated the development of liver fibrosis by affecting the crosstalk between hepatic stellate cells and macrophages. J Cell Mol Med 2021;25:10049-60. [Crossref] [PubMed]
37. Ghaddar B, Biswas A, Harris C, et al. Tumor microbiome links cellular programs and immunity in pancreatic cancer. Cancer Cell 2022;40:1240-1253.e5. [Crossref] [PubMed]
38. Shao N, Qiu H, Liu J, et al. Targeting lipid metabolism of macrophages: A new strategy for tumor therapy. J Adv Res 2025;68:99-114. [Crossref] [PubMed]
39. Schmall A, Al-Tamari HM, Herold S, et al. Macrophage and cancer cell cross-talk via CCR2 and CX3CR1 is a fundamental mechanism driving lung cancer. Am J Respir Crit Care Med 2015;191:437-47. [Crossref] [PubMed]
40. Zhang S, Xie B, Wang L, et al. Macrophage-mediated vascular permeability via VLA4/VCAM1 pathway dictates ascites development in ovarian cancer. J Clin Invest 2021;131:e140315. [Crossref] [PubMed]
41. Hinshaw DC, Hanna A, Lama-Sherpa T, et al. Hedgehog Signaling Regulates Metabolism and Polarization of Mammary Tumor-Associated Macrophages. Cancer Res 2021;81:5425-37. [Crossref] [PubMed]
42. Jung M, Ören B, Mora J, et al. Lipocalin 2 from macrophages stimulated by tumor cell-derived sphingosine 1-phosphate promotes lymphangiogenesis and tumor metastasis. Sci Signal 2016;9:ra64. [Crossref] [PubMed]
43. Frank AC, Raue R, Fuhrmann DC, et al. Lactate dehydrogenase B regulates macrophage metabolism in the tumor microenvironment. Theranostics 2021;11:7570-88. [Crossref] [PubMed]
44. Ma Y, Yi M, Wang W, et al. Oxidative degradation of dihydrofolate reductase increases CD38-mediated ferroptosis susceptibility. Cell Death Dis 2022;13:944. [Crossref] [PubMed]
45. Butera A, Roy M, Zampieri C, et al. p53-driven lipidome influences non-cell-autonomous lysophospholipids in pancreatic cancer. Biol Direct 2022;17:6. [Crossref] [PubMed]
46. Loughran RM, Emerling BM. Mechanistic roles of mutant p53 governing lipid metabolism. Adv Biol Regul 2022;83:100839. [Crossref] [PubMed]
47. Su P, Wang Q, Bi E, et al. Enhanced Lipid Accumulation and Metabolism Are Required for the Differentiation and Activation of Tumor-Associated Macrophages. Cancer Res 2020;80:1438-50. [Crossref] [PubMed]
48. Weagel E, Smith C, Liu PG, et al. Macrophage polarization and its role in cancer. J Clin Cell Immunol 2015;6:338.
49. Cui Zhou D, Jayasinghe RG, Chen S, et al. Spatially restricted drivers and transitional cell populations cooperate with the microenvironment in untreated and chemo-resistant pancreatic cancer. Nat Genet 2022;54:1390-405. [Crossref] [PubMed]
50. Steele NG, Carpenter ES, Kemp SB, et al. Multimodal Mapping of the Tumor and Peripheral Blood Immune Landscape in Human Pancreatic Cancer. Nat Cancer 2020;1:1097-112. [Crossref] [PubMed]
51. Watson MJ, Vignali PDA, Mullett SJ, et al. Metabolic support of tumour-infiltrating regulatory T cells by lactic acid. Nature 2021;591:645-51. [Crossref] [PubMed]
52. Ducker GS, Rabinowitz JD. One-Carbon Metabolism in Health and Disease. Cell Metab 2017;25:27-42. [Crossref] [PubMed]
53. Petrova B, Maynard AG, Wang P, et al. Regulatory mechanisms of one-carbon metabolism enzymes. J Biol Chem 2023;299:105457. [Crossref] [PubMed]
54. Chen L, Zhang Z, Hoshino A, et al. NADPH production by the oxidative pentose-phosphate pathway supports folate metabolism. Nat Metab 2019;1:404-15.
55. He L, Shi X, Liu Z, et al. Roles Of EAAT1, DHFR, And Fetuin-A In The Pathogenesis, Progression And Prognosis Of Chondrosarcoma. Onco Targets Ther 2019;12:8411-20. [Crossref] [PubMed]
56. Tulstrup M, Moriyama T, Jiang C, et al. Effects of germline DHFR and FPGS variants on methotrexate metabolism and relapse of leukemia. Blood 2020;136:1161-8. [Crossref] [PubMed]
57. Galassi R, Luciani L, Gambini V, et al. Multi-Targeted Anticancer Activity of Imidazolate Phosphane Gold(I) Compounds by Inhibition of DHFR and TrxR in Breast Cancer Cells. Front Chem 2020;8:602845. [Crossref] [PubMed]
58. Brown JI, Wang P, Wong AYL, et al. Cycloguanil and Analogues Potently Target DHFR in Cancer Cells to Elicit Anti-Cancer Activity. Metabolites 2023;13:151. [Crossref] [PubMed]
59. Zhu Z, Chen C, Zhang J, et al. Exploration and Biological Evaluation of 1,3-Diamino-7H-pyrrol[3,2-f]quinazoline Derivatives as Dihydrofolate Reductase Inhibitors. J Med Chem 2023;66:13946-67.