Background
Pancreatic ductal adenocarcinoma (PDAC), which has the worst prognosis of human malignancies [
1], is poorly responsive to therapies and is histologically characterized by clusters of cancer cells surrounded by a dense desmoplastic stroma [
2]. The tissue stroma undergoes transformation during cancer progression, and the interplay between cancer cells and the stroma creates a favorable tumor microenvironment (TME) for successful tumor development [
3]. Previous efforts to deconstruct the desmoplastic stroma have been unsuccessful, which reflects the multifaceted nature of the TME, but a breakthrough could be achieved by exploring more biologically integrated targets to reshape the TME in PDAC.
Cancer-associated fibroblasts (CAFs) have long been considered to contribute most to malignant changes in the stroma by laying down extensive extracellular matrix around the tumor cells and secreting trophic factors that enable the formation of a protumorigenic niche [
4]. CAFs are generally spindle shaped and positive for several activated fibroblast markers, such as fibroblast activation protein (FAP), α-smooth muscle actin and platelet-derived growth factor (PDGF) receptor-α (PDGFRα) [
5,
6]. FAP is the first target used to assess the effect of cancer treatment on CAF inhibition. The unfavorable results of a phase II clinical trial using sibrotuzumab (a humanized monoclonal antibody) to target CAFs in patients with colorectal cancer prevented further investigation of this inhibitor [
7]. Equally disappointing results were reported by another phase II trial showing limited efficacy against advanced PDAC of a combination of an oral small molecule inhibitor of FAP (talabostat) and gemcitabine [
8]. Faced with unsuccessful attempts to inhibit CAFs, a more radical approach with genetic deletion of activated fibroblasts was proposed, which actually led to a more aggressive phenotype and worse overall survival (OS), indicative of the perplexity of the stroma-tumor interaction [
9‐
11] and emphasizing the urgent need to exploit the composition and function of the stroma in PDAC.
CAFs are derived from pancreatic stellate cells (PSCs), quiescent resident fibroblasts and bone marrow-derived mesenchymal stem cells (MSCs) through activation of multiple signaling pathways. The activation of PSCs is considered a very early event during PDAC tumorigenesis [
12,
13]. Recently, similar to the diversity in CAF origins, the heterogeneity in CAF fate and function has also attracted great attention and has indicated the possibility of targeting a subpopulation of CAFs to control malignancies. Researchers have discovered two CAF phenotypes, classified as myofibroblastic CAFs (myCAFs) and inflammatory CAFs (iCAFs), based on the expression levels of α-SMA [
14]. myCAFs, characterized by strong α-SMA expression, are mainly located in the periglandular region of the TME and induce desmoplasia by direct juxtacrine interactions with cancer cells. In contrast, significantly lower αSMA levels were found in iCAFs, which play critical roles in promoting tumor progression through a group of cytokines and chemokines [
14]. Accordingly, Giulia et al. discovered that iCAFs were stimulated by IL1 and provided a protumorigenic niche through JAK/STAT activation and IL6 secretion and shifts in the myCAF/iCAF ratio might have differential effects on PDAC progression [
15]. Subsequently, a third subtype of CAFs, characterized by MHC class II molecule expression, was identified, with the ability to present antigens to CD4+ T cells [
16], suggesting that the phenotypic and functional diversity of CAFs is still far from clarified.
Circular RNAs (circRNAs) are an interesting class of noncoding RNAs due to their unique generation by a noncanonical splicing event called backsplicing and their characteristics of covalently closed structures and exonuclease resistance [
17]. circRNAs have diverse biological functions by acting as microRNA (miRNA) or protein sponges and regulating protein function or translation into peptides [
18]. circRNAs have a potential role in tumorigenesis and act as oncogenes or tumor suppressors in different types of tumors [
19], including PDAC [
20,
21]. However, circRNAs have not been elucidated in the context of the TME in PDAC. Here, we demonstrated that the upregulation of circCUL2 expression in fibroblasts mediated the conversion of normal tissue-associated fibroblasts (NFs) into iCAFs and contributed to CAF heterogeneity in PDAC by inducing the MYD88-dependent NF-kb signaling pathway. These findings shed light on the development of strategies to selectively target CAFs that support tumor growth.
Methods
Patients and clinical samples
Tumor samples were collected from 161 patients with PDAC who underwent surgical resection at the Sun Yat-sen Memorial Hospital and Guangdong Provincial People’s Hospital between 2012 and 2020. None of the patients had received radiotherapy, chemotherapy, immunotherapy or targeted therapy before surgery. Tumor staging was determined using the 8th edition of tumor-node-metastasis (TNM) system of the American Joint Committee on Cancer. The clinical features of the enrolled patients are showed in Table
1. Overall survival (OS) was measured as the time interval from the date of randomization to the date of death or last follow-up evaluation (December 2020), and disease-free survival (DFS) was defined as the time interval between the date of randomization to the date of first disease-free failure event. All patients provided informed consent, and all related procedures were performed with the approval of the Ethical Committee of the indicated hospitals.
Table 1
Correlation between circCUL2 expression level and clinicopathologic characteristics of PDAC patients
Total cases
| 161 | 81 | 80 | |
Gender
| | | | 0.529 |
Male | 87 | 46 | 41 | |
Female | 74 | 35 | 39 | |
Age, mean ± SD, year
b
| | 60.10 ± 9.29 | 58.46 ± 7.96 | 0.232 |
Differentiation
| | | | 0.061 |
Well | 11 | 7 | 4 | |
Moderate | 88 | 50 | 38 | |
Poor | 62 | 24 | 38 | |
TNM stage (AJCC)
c
| | | | <0.001*** |
I | 31 | 25 | 6 | |
II | 103 | 50 | 53 | |
III | 27 | 6 | 21 | |
Lymph-node metastasis
| | | | <0.001*** |
Negative | 61 | 48 | 13 | |
Positive | 100 | 33 | 67 | |
Perineural invasion
| | | | 0.641 |
Negative | 21 | 12 | 9 | |
Positive | 140 | 69 | 71 | |
Long-term smoking
d
| | | | 0.093 |
No | 141 | 67 | 74 | |
Yes | 20 | 14 | 6 | |
High-fat diet
| | | | 0.269 |
No | 138 | 72 | 66 | |
Yes | 23 | 9 | 14 | |
Chronic pancreatitis
| | | | 0.831 |
No | 135 | 67 | 68 | |
Yes | 26 | 14 | 12 | |
Abbreviations: SD = standard deviation; TNM = tumor node metastasis; AJCC = American Joint Committee on Cancer. a Chi-square test (except for age), * p < 0.05, ** p < 0.01, *** p < 0.001. b Unpair t test, * p < 0.05, ** p < 0.01, *** p < 0.001. c Patients were staged in accordance with the 8th Edition of the AJCC Cancer’s’ TNM Classification. d Long-term smoking (duration ≥ 29 years). |
Primary cell isolation and culture
Cancer-associated fibroblasts (CAFs) and primary normal fibroblasts (NFs) were isolated from pancreatic ductal carcinoma and corresponding noncancerous tissues. Freshly isolated surgical resections were collected with informed consent from Sun Yat-sen Memorial Hospital and Guangdong Provincial People’s Hospital. For CAF isolation, carcinoma samples were cut (0.5 cm
3) from the core part of PDAC tissues using a sterile scalpel. Some of the isolated specimens were used for histological examination to confirm the diagnosis, and the remaining tissues were minced and dissociated by collagenase digestion medium (DMEM/F12, collagenase digestion 125 units/mg, insulin 10 mg/mL, hydrocortisone 0.5 mg/mL, penicillin 100 U/mL and streptomycin 100 mg/mL) [
22]. The sample was digested for 30 min then quenched in 10% FBS/DMEM. Then the dissociated tissues were incubated without shaking plated on a 6-cm dish for 10 min at 37 °C. The stromal cell-enriched supernatant was collected in a new tube and centrifuged at 250 g for 5 min. For NF isolation, similar anatomopathological techniques were used, freshly isolated normal adjacent pancreatic tissue were taken from areas at least 3 cm distal to primary PDAC tumor masses [
23]. All samples were taken from surgical resections, not from intrasurgical biopsies. Primary fibroblasts were cultured in Dulbecco’s modified Eagle’s medium (DMEM, GBICO) with 15% fetal bovine serum (FBS, GIBCO) and 1% penicillin/streptomycin at 37 °C in humidified air with 5% CO
2.
For human PDAC cell lines culture, PANC-1 and MiaPaCa-2 were purchased from American Type Culture Collection (ATCC, Manassas, VA). The cells were then cultured in DMEM medium supplemented with 10% fetal bovine serum at 37 °C and 5% CO2.
Flow cytometry assay
For identification of fibroblast population, CD31-FITC (4,220,516, BD Biosciences, USA), CD45-PE/Cy7 (561,868, BD Biosciences, USA), CD326 (EPCAM)-PE (2,088,498, Invitrogen, USA) were used to distinguish fibroblast from T cells, endothelial cells and epithelial cells. PDGFRα (3174, CST) and α-SMA (ab32575, Abcam) were used to distinguish iCAF from myCAF.
RNA isolation and quantitative real-time PCR (qRT–PCR)
Total RNA was isolated with TRIzol (Life, USA) and reverse-transcribed using the PrimeScript RT Reagent Kit (DRR037A, Takara, Japan). Then, cDNA was amplified by qRT–PCR on a Light Cycler 480 Detection System (Roche, Switzerland) using the TB Green Premix Ex TaqTM kit (RR820A, Takara, Japan) with GAPDH or U6 as an internal control. The 2−∆∆CT method was used to calculate relative gene expression levels in cells.
For expression in tissues, the levels were first normalized to GAPDH expression via the ∆CT method. To analyze the clinical significance of circCUL2, miR-203a-3p, IL6 and Myd88, tissues from 161 PDAC patients were divided into two groups—the low and high expression groups. Samples with normalized expression levels (∆CT) of these genes less than or equal to the median value were classified as the low expression group, and those with levels greater than the median value were classified as the high expression group.
For absolute quantification, 1 × 106 NFs and CAFs cells were used for copy number analysis. Absolute quantification was performed via 10 times gradient dilution of the reference standard. Based on the measured cycle threshold (CT) values of the standards and the known concentrations of standards, we built a standard curve for the log (copy number) and CT values. The standard curve was used to extrapolate the number of molecules of circCUL2 and miR-203-3p in NFs and CAFs.
All primers are listed in Supplementary Table S
1.
RNase R digestion and actinomycin D assay
For RNase R digestion assay, total RNA of NFs and CAFs were treated with or without 5 U/µg RNase R (RNR07250, Epicenter Technologies) and incubated at 37℃ for 30 min. For actinomycin D assay, or total RNA was treated with 2 µg/mL actinomycin D (Sigma, USA) for 0 h, 4 h, 8 h, 12 and 24 h. And qRT–PCR was used to detected circCUL2 and CUL2 expression levels. The experiments were performed three times.
Plasmid construction and transfection
circCUL2 was cloned into the pCD-ciR vector by IGE (Guangzhou, China). The luciferase reporter plasmids of circCUL2, the MyD88 3’ untranslated region (UTR) and mutant luciferase reporters were synthesized by IGE (Guangzhou, China). siRNA and miRNA mimic or inhibitor were purchased from IGE (Guangzhou, China). Plasmids and oligonucleotides were transfected using Lipofectamine 3000 (Invitrogen, USA) according to the manufacturer’s protocol. The targeted sequences of oligonucleotides are provided in Supplementary Table S
2.
Conditioned medium derived from human NFs and CAFs
After transfected, NFs and matched CAFs were cultured in 25 mL culture flasks at the same density, respectively. The supernatants were harvested after 48 h and centrifuged to removed cell pellets, then the conditioned medium (CM) was stored at -80 °C or incubation with PANC-1 or MiaPaCa-2 for 48 h.
5-Ethynyl-20-deoxyuridine (EdU) assay
Cell proliferation was measured by the EdU assay using BeyoClick EdU-555 detection kits (Beyotime, Shanghai, China), according to the manufacturer’s instructions. PDAC cells with indicated treatments were seeded into 24-well plates, followed by incubating with 50 µM EdU for 2 h. Cells were then fixed using 4% paraformaldehyde and sealed with Apollo reaction cocktail and Hoechst 33,342 in order. All images were photographed with a fluorescent microscope.
500 PDAC cells with indicated treatments were seeded into 6-well plates and cultured for 2 weeks. Then the colonies were fixed in 4% paraformaldehyde for 20 min, followed by staining with 0.1% crystal violet. Colonies were then manually counted. Three different independent experiments were performed.
Wound healing assay
PDAC cells with indicated treatments were seeded into 24-well plates. After 24 h, each well was wounded with a 20 µl pipette tip. The cell migration was photographed with an inverted microscope at the time points of 0 and 36 h. Three different independent experiments were performed.
Transwell assays
The pretreated cells were cultured with 200 µl of serum-free medium in the top chamber that had been loaded with or without Matrigel (Matrigel BD biosciences, NY, USA). And 600 µl complete medium was added to the bottom compartment. After incubation for 18 h, cells on the upper surface of the top chamber were removed, and invaded cells were fixed and stained by crystal violet. The number of invaded cells was counted and captured with a light microscope. Three different independent experiments were performed.
Cytokine array
The Proteome Profiler Human XL Cytokine Array Kit (R&D Systems, ARY022B, USA) was used to assess the cytokines secreted by fibroblast according to the manufacturer’s instructions. In brief, 400 µl of the indicated NF medium was incubated with an array membrane overnight. Then, detection antibody cocktails and streptavidin-HRP were added. The cytokine dots on X-ray films were scanned.
Enzyme-linked immunosorbent assay (ELISA)
Concentration of cytokines were determined by using the human IL6 ELISA Kit (SEKH-0013, Solarbio, China) according to the manufacturer’s instructions. In brief, 100 µl of indicated NF medium was incubated with plates at 37℃ for 90 min. Then detection antibody, streptavidin-HRP and TMB were added in order. The absorbance of each well was measured at 450 nm with the SPARK 10 M spectrophotometer (Tecan, Austria).
Western Blotting
Protein was extracted from the cells using RIPA lysis buffer (CWBIO, China), followed by subjected to SDS-polyacrylamide gels and transferred to polyvinylidene difluoride membranes. Corresponding primary antibodies including MyD88 (1:1000, 4283, CST), STAT3 (1:1,000, 9139, CST), p-STAS3(Tyr705) (1:1,000, 9145, CST), p65 (1:1,000, 8242, CST), pp65(Ser536) (1:1,000; 3033, CST), IkBα (1:1,000, 4812, CST), p-IkBα(Ser32) (1:1,000, 2859, CST), FAP (1:1,000, 28,244, Abcam), GAPDH (1:1,000; 132,004, Absin) were added to the membrane. HRP-conjugated secondary antibodies were used. The immunoreactive bands were detected by ECL detection system (Millipore, Germany) and photographed by Chemi XT4.
Isolation of cytoplasmic and nuclear RNA
Cytoplasmic and nuclear RNA of CAFs were isolated using NE-PER Nuclear and Cytoplasmic Extraction Reagents (Thermo Scientific, USA) according to the manufacturer’s instructions. Then, the ratio of cytoplasmic and nuclear was measured by qRT–PCR. U6 served as the nuclear control, and GAPDH served as the cytoplasmic control.
Fluorescence in situ hybridization (FISH)
FISH was performed using a In Situ Hybridization Kit (Gene Pharma, Guangzhou, China) according to the manufacturer’s instructions. FAM-labeled circCUL2 and Cy3-labeled hsa-miR-203a-3p probes (Gene Pharma, Guangzhou, China) were hybridized with cells overnight at 37℃. All images were captured by confocal microscopy. The targeted sequences of probes are provided in Supplementary Table S
2.
RNA pull-down assay
The biotinylated probes were synthesized by IGE (Guangzhou, China). Approximately 1 × 107 CAFs and NFs were collected and lysed, followed by incubation with the biotinylated probe or oligo probe overnight at 4℃. Streptavidin magnetic beads (Invitrogen, USA) were added and then incubated for 3 h. The RNA-beads were isolated by TRIzol and analyzed by qRT–PCR.
Luciferase reporter assay
The circCUL2/MyD88 wild-type or mutant plasmids and miR-203a-3p mimic were co-transfected into CAFs cells. Then the transfected Cells were seeded into 96-well plates and luciferase activities were determined by dual-luciferase reporter assay system (Promega, USA) according to the manufacturer’s instructions.
Animal experiments
Animal experiments were conducted according to guidelines approved by the Animal Experimental Research Ethics Committee of South China University of Technology. BALB/c nude mice aged 4 to 5 weeks were purchased from South China University of Technology.
For the lung metastasis model, 100 µL suspension of 5 × 106 PANC-1 cells or MiaPaCa-2 stably expressing luciferase (luc-PANC-1/luc-MiaPaCa-2) in PBS were injected into the tail vein of BALB/c nude mice. The mice were randomly divided into three groups, which were injected with luc-PANC-1 or luc-MiaPaCa-2 incubated with conditioned medium from (1) NFs transduced with empty vector for 48 h, (2) NFs transduced with circCUL2 vector for 48 h, (3) NFs transduced with circCUL2 vector in the presence of IL6 neutralizing antibodies (50 ng/mL, MAB206, R&D) for 48 h. The mice were imaged with an In Vivo Imaging System (IVIS Lumina XR Series III) 30 days later. Lung tissues were resected and were examined for metastatic foci via hematoxylin and eosin (HE) staining.
For the orthotopic model, the mice were randomly divided into three groups: (1) luc-tumor cells/empty vector-transduced NFs treated with PBS every three days, (2) luc-tumor cells/circCUL2-transduced NFs treated with PBS every three days, and (3) luc-tumor cells/circCUL2 transduced NFs treated with neutralizing antibodies against IL6 (2 mg/kg) every three days. BALB/c nude mice were anesthetized with ketamine (100 mg/kg) and xylazine (10 mg/kg). A 1-cm left subcostal incision was made to expose the pancreas, and a 50-µL suspension of luc-tumor cells/indicated NFs (1:1, 1 × 106) in PBS was injected with a 30-G needle. After orthotopic implantation, the incision was closed with monofilament sutures. The mice were imaged with an In Vivo Imaging System (IVIS Lumina XR Series III) 30 days later to assess primary tumor size and metastasis.
For patient-derived xenograft (PDX) mouse models, fresh PDAC samples obtained from 3 patients were cut into small pieces and then implanted into subcutaneous in 4-week-old NSG mice (F1). Xenografts were removed from from F1 mice, cut into small pieces and replanted to other mice (F2). When tumors reached about 1500 mm
3, they were excised and cut again into small pieces and replanted to other mice (F3). Seven days after xenografts reached about 100 mm
3, F3 mice were divided into three groups randomly (n = 5 per group) and treated respectively with: (1) in vivo-optimized si-Control (50 mg/kg, RiboBio, Supplementary Table S
2), i.v. injection every three days for 3 weeks; (2) in vivo-optimized si-circCUL2 (50 mg/kg, RiboBio, Supplementary Table S
2), i.v. injection every 3 days for 3 weeks; (3) neutralizing antibodies against IL6. Tumor volume (length × width
2/2) was monitored every 6 days.
RNA-seq
Total RNA was isolated and purified using TRIzol (Life, USA) following the manufacturer’s procedure. After the quality inspection of Agilent 2100 Bioanalyzer (Agilent, USA) and NanoPhotometer (Implen, Germany), ribosomal RNA was removed from 1 µg total RNA. VAHTS Universal V6 RNA-seq Library Prep Kit for Illumina (Vazyme, China) was used for lncRNA library construction following the manufacturer’s protocol. Each library was sequenced on an Illumina Novaseq 6000 (Illumina Corporation, USA) in 150PE mode following the vendor’s recommended protocol by Guangzhou Huayin Health Medical Group CO.,Ltd. (Guangzhou, China).
Statistical Analysis
All experimental data were expressed as mean ± standard deviation (SD) using GraphPad Prism 8.0. The differences between parametric variables were determined by Student’s t-test or one-way analysis of variance (ANOVA), and nonparametric variables were determined by Mann-Whitney U test. Statistical significance of survival was estimated by Kaplan-Meier analysis and the log-rank test, and correlation analysis was performed by two-sided Pearson’s correlation. Multivariate analysis of the relative risk was performed by cox regression. Correlation analysis was examined with two-sided Pearson’s correlation. p<0.05 was used as an indicator of statistical significance.
Discussion
Due to the development of high-throughput sequencing technology and bioinformatics, circRNAs have attracted increasing attention in the noncoding RNA field and have been discovered and reported as critical regulators of various biological processes, including multiple cancers [
27,
28]. Currently, the roles of circRNAs in the progression of pancreatic cancer have been revealed in many studies [
29]. Recently, our team reported that circBFAR promotes the progression of PDAC via the miR-34b-5p/MET/Akt axis and that circNFIB1 inhibits lymphangiogenesis and lymphatic metastasis via the miR-486-5p/PIK3R1/VEGF-C axis in pancreatic cancer [
20,
21]. However, little is known about the role and underlying mechanisms of circRNAs in the context of the TME in PDAC. Herein, we identified differentially expressed circRNAs in primary isolated fibroblasts and demonstrated that circCUL2 was highly expressed in iCAFs and that its enrichment in tumor tissues was correlated with the prognosis of patients with PDAC. Overexpression of circCUL2 in NFs drove the cells toward the iCAF phenotype, and then iCAFs promoted the progression of PDAC by secreting IL6. Mechanistically, circCUL2 functioned as a ceRNA and modulated the miR-203a-3p/MyD88/NF-κB/IL6 axis, thereby further activating the STAT3 signaling pathway in pancreatic cancer cells to induce the progression of PDAC. Regarding the function of circCUL2 in gastrointestinal tumors, circCUL2 was found to promote tumor malignancy and metastasis in hepatocellular carcinoma [
30], while recently, it was revealed that it may function as a tumor suppressor and mediator of cisplatin sensitivity in gastric cancer [
31]. Our study found that circCUL2 can promote the platinum resistance of pancreatic cancer through the activation of iCAFs, which is inconsistent with the conclusions of the recently finding in gastric cancer [
31]. These findings indicate the heterogeneity of the expression and biological function of circCUL2 in different tissues in the context of the complexity of the TME. To our knowledge, this is the first report to provide insight into the biological significance of the circRNA-mediated phenotypic transition of iCAFs in PDAC and highlight that circCUL2 may serve as a novel therapeutic target for iCAFs that support tumor growth.
For years, the mechanisms of genetic and epigenetic modifications in pancreatic cancer cells have been deeply explored, while the complex TME is far from clear. circRNAs have recently attracted enormous attention, and their underlying cancer-associated mechanism in altering the TME is gradually being revealed. By performing a comprehensive bioinformatics analysis of one of the most studied circRNAs, CDR1as, Zou et al. showed that CDR1as may play a specific role in stromal and immune cell infiltration in tumor tissue and could promote cancer progression by altering the TME [
32]. Recently, Kristensen et al. reported that CDR1as was undetectable in colon cancer cells in vivo but was abundantly expressed within tumor stromal cells, which highlights the importance of spatially resolving the expression patterns of circRNAs in the TME [
33]. Determining the specific mechanisms that control the fate and function of CAFs in PDAC is needed for the development of therapeutic regimens that selectively target tumor-promoting CAFs. For the first time, we explored the potential involvement of circRNAs in primary fibroblasts from human PDAC tissues. Based on circRNA profiling in primary CAFs and NFs from human PDAC adjacent normal tissue by next-generation sequencing, we focused on circCUL2, and the high circCUL2 level in stoma was positively correlated with lymph node metastasis and advanced clinical stage and independently correlated with OS and DFS. At present, many studies have shown that the correlation between tumor gene expression and prognosis rarely provides in situ data to assess the potential intratumor heterogeneity of target gene expression. Especially in PDAC, which has a fibroblastic population comprising 90% of the whole tumor mass [
34], the expression patterns of circRNAs in stromal cells are far from elucidated. Our study is the first to link circRNA expression in CAFs with the prognosis of PDAC patients, and the results indicated that the heterogeneity of gene expression in stromal cells also plays a key role in tumor progression, suggesting the potential of targeting circRNAs in the TME to improve therapeutic outcomes.
For the key role of circCUL2 in fibroblasts, our gain- and loss-of-function experiments demonstrated that circCUL2 was critical to maintaining CAF protumor properties and that the transduction of circCUL2 into NFs was sufficient to induce a distinct CAF subtype characterized by an inflammatory phenotype. Our GSEA of mRNA-seq in NFs and circCUL2-transduced NFs showed that the hallmarks of the inflammatory response and IL6/JAK/STAT3 signaling were both significantly enriched. Furthermore, both qRT–PCR and flow cytometry assays confirmed the activation of iCAF markers (IL6, TNFα, IL1α and PDGFRα) but not myCAF markers (Acta2, Axin2 and α-SMA). In addition, a panel of inflammatory cytokines abundantly secreted by circCUL2-transduced NFs was also confirmed by cytokine array and ELISA. Intriguingly, our data revealed a significantly negative enrichment of circCUL2 overexpression and hallmarks of myogenesis. Consistent with recent work from Biffi et al., IL1R1 expression leads to NF-κB/LIF/JAK/STAT activation and iCAF formation [
15,
35]. Moreover, JAK inhibition or tumor-secreted TGF-β significantly increased the myCAF/iCAF ratio in treated tumors, further confirming that iCAFs and myCAFs are interconvertible cell states rather than endpoints in differentiation [
15]. Considering the key role of the circCUL2-induced cascade in the activation of the NF-κB/IL6 signaling pathway, targeting circCUL2 could potentially convert iCAFs into a more myofibroblastic state that has been previously suggested to inhibit tumor progression.
Accumulating reports have revealed that circRNAs always function as efficient miRNA sponges to then modulate miRNA target gene expression. In our study, miR-203a-5p was selected as the candidate target miRNA of circCUL2, and the circCUL2/miR-203a-5p interaction was confirmed by FISH, RNA pulldown and luciferase reporter assays. Interestingly, several studies identified miR-203 as a matrix stiffness-repressed transcript [
36‐
38], and the highly stiff ECM leads to a low miR-203 level in the TME, increasing breast cancer risk [
38], which indicates the key role of miR-203 in mediating the crosstalk between cancer cells and stroma. Our study provided the first evidence that the downregulation of miR-203 expression in NFs leads to the activation and transition of the iCAF state and revealed specific signal transduction mediated by miR-203 in stromal cells.
Intersection of the prediction results of three databases identified MyD88 as the downstream target of miR-203a-3p, which was further verified by qRT–PCR, luciferase assays and 3’UTR mutation experiments. MyD88 plays a key role in NF-κB signal transduction and is involved in both oncogene-induced intrinsic and extrinsic inflammation in cells [
39]. A large number of studies have shown that the adaptor protein MyD88 contributes to carcinogenesis, including cancer of the skin, liver, pancreas, and colon, by acting downstream of Toll-like receptors (TLRs) or the IL1 family [
39]. Recently, Biffi et al. demonstrated a signaling cascade induced by the IL1 family that leads to NF-κB/JAK/STAT activation to generate iCAFs, and an IL1 receptor antagonist might be an efficient strategy to target iCAFs in vivo [
15]. Most often associated with the IL1 family, MyD88 signaling mediates a proinflammatory feedback mechanism that is involved in the intrinsic inflammation associated with oncogene activation, cell transformation, and senescence [
40]. Consistent with these findings, our study underlined the key role of MyD88 in activating the NF-κB/IL6 signaling cascade to mediate the transition of iCAFs. Although the abundance of cytokines, such as CCL18 and IL1b, in cancer cells may be responsible for initiating NF-kB signaling through activation of IKK in the CAF subset [
41,
42], numerous feedback loops in the TME lead to the failure of therapeutic strategies to target inflammatory factors. Our study indicates a more meaningful strategy of antagonizing MYD88 endogenously by circCUL2.
As a classic proinflammatory pathway, NF-κB signaling has been implicated as a hallmark of cancer progression and a potential therapeutic target. In recent years, the role of NF-κB signaling in mediating the reciprocal interplay between cancer cells and stroma has gradually been revealed [
41]. During the early preneoplastic stages of tumorigenesis, resting fibroblasts are activated to express a proinflammatory gene signature, thereby promoting cancer progression in an NF-κB-dependent manner [
41]. Nearly 70% of PDAC cases show NF-κB activation [
43], and several studies have revealed that NF-κB activation is required for the formation of iCAFs [
14‐
16]. Our study provides novel results that the upregulation of circCUL2 expression in fibroblasts increases the expression of MyD88, the key receptor upstream of the NF-kb pathway, thereby contributing to the function of various proinflammatory factors from the TME in generating iCAFs.
In our study, we revealed that IL6 is the key downstream mediator of circCUL2. circCUL2- induced iCAFs contributed to the tumorigenesis and metastasis of PDAC through increased secretion of IL6 and further activation of the STAT3 signaling pathway in pancreatic cancer cells. Previous studies have revealed that IL6 secreted by iCAFs is involved in increasing proliferation and invasion of PDAC mouse cells [
15,
44]. We reported for the first time that circCUL2 plays a key role in promoting the development of IL6-related iCAF phenotypes. Impressively, it has been gradually elucidated that IL6 secreted by CAFs mediates immunosuppression in pancreatic cancer. Recently, Thomas et al. demonstrated that IL6 blockade enhanced the efficacy of anti-PD-L1 therapy by promoting the T helper cell differentiation and increasing CD8+ T cell infiltration into PDAC tumors mice [
45]. IL6 can also contribute to immunosuppression by promoting the differentiation of macrophages, and myeloid-derived suppressor cells and driving the apoptosis of type 1 conventional dendritic cells [
46‐
48]. As circCUL2 contributes to iCAF phenotype development and the enrichment of the cytokine IL6 in pancreatic cancer, further exploration is needed to reveal its role in the immunosuppression of pancreatic cancer. Signal transduction downstream of IL6 includes STAT3, MAPK and PI3K carcinogenic pathways [
13,
49,
50]. Marina et al. revealed that the IL6/Stat3 pathway is strongly activated and contributes to the progression of pancreatic intraepithelial neoplasia progression and the development of PDAC [
13]. Zhang et al. reported that IL6 was able to synergize with Kras to facilitate the progression of pancreatic cancer precursor lesions by activating the MAPK signaling cascade [
51]. Our data highlighted that the IL6/STAT3 pathway is the key mediator of the protumorigenic properties of iCAFs in PDAC, but we do not exclude the involvement of other pathways, because these downstream signaling pathways form a complex regulatory network to induce synergistic effects [
52]. Although most recent studies emphasized the effect of IL6/STAT3 pathway [
53], other signaling pathways in this network, including MAPK and PI3K, still need to be further explored.
For the clinical implications of the circCUL2/miR-203a-3p/MyD88/IL6 axis in PDAC patients, we revealed that high levels of circCUL2, MyD88 and IL6 expression in tumor tissues are associated with poor prognosis in patients with pancreatic cancer, and high miR-203a-3p expression indicates better OS and DFS in PDAC patients. By conducting correlation analyses on the circCUL2/miR-203a-3p/IL6 axis, a positive correlation was found between circCUL2 and IL6 (R = 0.42, P < 0.001) or MyD88 (R = 0.44, P < 0.001), and negative correlations were found between miR-203a-3p and circCUL2 (R = -0.43, P < 0.001) or IL6 (R = -0.34, P < 0.001). The R values indicate that these correlations are not strong (|R| < 0.60) [
54], which implies that the clear significance based on the p value may be supported by a high number of samples. One limitation of this study is that the RNA levels of the genes were detected in tissue samples, not the corresponding primary fibroblasts, which may affect the accuracy of the correlation analysis results. Gene expression correlation analysis in primary isolated cells may provide stronger evidence for the prognostic value of the circCUL2/miR-203a-3p/MyD88/IL6 axis. The RNA expression level of circCUL2 in tissues can be used to differentiate the pancreatic cancer tissues and normal tissues to a certain extent. The clinical translational value of circCUL2 can be further explored by detecting the expression levels of circCUL2 in the serum of pancreatic cancer patients and healthy people.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.