Tissue samples of human PC were collected with the approval (approval no. 18138-4) of the Human Ethics Review Committee of the Graduate School of Medicine of Osaka University (Osaka, Japan) and the National Institutes of Biomedical Innovation, Health and Nutrition (approval no. 201-01; Osaka, Japan). Written informed consent for sample use was obtained from all patients before surgery. All tissue samples were collected at Osaka University Hospital (Suita, Japan) from June 2019 to March 2020. Human pancreatic stellate cells, the reported origin cells of CAFs, were cultured from the tissue samples as previously described (24,25). Briefly, the cancerous portion was cut from the tissue samples, and chopped into small pieces. These tissue pieces were washed with phosphate-buffered saline (PBS) containing 5% penicillin-streptomycin (Sigma-Aldrich; Merck KGaA), and then three pieces were sown in each well of a six-well plate (Corning, Inc.). The samples were cultured in Dulbecco's modified Eagle's medium (DMEM) containing 20% fetal bovine serum (FBS) for 2 days at 37°C, and thereafter in DMEM containing 10% FBS. Gradual proliferation of only spindle-shaped cells was observed as previously reported (Fig. S1), which was considered to be CAFs based on their cell morphology. This outgrowth method, which was previously reported by Bachem et al (24), was already well-established method for the CAF primary culture. It was indicated in the aforementioned study that α-smooth muscle actin (α-SMA), vimentin (VIM), and desmin were expressed in all primary cultured cells and immunocytochemistry of primary cultured cells was also performed in previous studies and it has been already confirmed that primary culture cells were true CAFs and there was no contamination of cancer cells (24,25). On the 14th day of culture, these CAFs were cryopreserved for use in further experiments. In the present study, primary CAFs were used without immortalization. This is because, to immortalize primary cultured cells, viral vector or gene induction such as telomerase reverse transcriptase (TERT) had to be used, but it is considered that such immortalized cells were very different from originally primary cultured cells. Therefore, primary cultured cells were established with great care and cell damage was minimized by using a programmed freezer when storing cells and accutase when passaging CAFs. In addition, all assays were performed using primary culture CAFs within three passages. This is due to the fact that culture time and passages could cause changes in cell morphology and gene expression in CAFs.

Primary culture of CAFs was performed using tissue samples collected from 8 PC patients. The clinical information for these 8 patients before the initiation of treatment are presented in Table SI. The median age was 67 years (range 53–77 years), 3 were male, and the median tumor diameter was 25.1 mm (range 9.5-34.3 mm). Of the 8 patients, 6 underwent preoperative chemotherapy (details presented in Table SII). The mean age of the male and female groups was 65.6 and 65.7 years, respectively, with no significant difference (P=0.993). The tumor diameters measured by contrast-enhanced CT before and after chemotherapy, and the calculated tumor size ratio (TSR) for each case: TSR=(Tumor size after neoadjuvant chemotherapy)/(Tumor size before neoadjuvant chemotherapy) are also included in Table SII. The macroscopic and microscopic pathology findings for each case are demonstrated in Fig. S2. The following procedure was used for H&E staining: First, tissues were fixed with neutral formalin 10% at room temperature for 2 days, embedded in paraffin, and manually sectioned with a microtome to obtain 3-µm thick paraffin sections. The sections were dewaxed with xylene and stained with hematoxylin for 6 min and eosin for 2 min and 30 sec at room temperature. Tissue-Tek^®Eosin and Tissue-Tek^® Mayer hematoxylin for Prisma (both from Sakura Finetek USA, Inc.) were used as staining solutions, and staining was performed on a Tissue-Tek^®DRS TM 2000 (Sakura Finetek USA, Inc.).

The co-culture assay was performed in a non-contact manner, using a Transwell with 0.4-µm pores (Corning, Inc.). First, CAFs (1.0×10⁴ per well) were plated on the top layer and incubated overnight in DMEM containing 10% FBS. Once the CAFs had settled in the Transwell, PSN-1 cells (a human PC cell line, p53 mutant) were plated at 1.0×10⁵ per well on the lower layer of a 24-well culture plate (Corning, Inc.). PSN-1 cells were obtained from the American Type Culture Collection. After an additional 48 h of co-culture in DMEM containing 0.5% FBS, the Transwell was removed and PSN-1 cell proliferation was evaluated.

PSN-1 cell proliferation was evaluated by fluorescence staining using Hoechst 33342 (Thermo Fisher Scientific, Inc.). Briefly, after removing the Transwell and the culture medium, diluted Hoechst 33342 was added and allowed to react for 10 min. Then the PSN-1 cells were washed three times with PBS, and the fluorescence intensity was measured using the EnSpire Multimode Plate Reader (PerkinElmer, Inc.). An assay was performed to compare the effects of CAFs from 8 different patients on the proliferation of PSN-1 cells. The fluorescence intensity ratio (FIR) was calculated based on the fluorescence intensity of the canonical CAFs (#1) showing the lowest fluorescence intensity: FIR (#X)=Fluorescence intensity (#X)/Fluorescence intensity (#1).

RT-qPCR arrays were performed using two kits: QIAGEN RT² Profiler™ PCR Array ‘Human Cellular Senescence’ (GeneGlobe ID-PAHS-050Z) and ‘Human Cancer Inflammation and Immunity Crosstalk’ (GeneGlobe ID-PAHS-181Z) following the supplier's instructions (http://www.sabiosciences.com). After CAFs were settled on the plate, they were incubated at 37°C for 1 day under serum-free conditions, and then incubated at 37°C for 12 h with or without serum stimulation (2% human AB serum), respectively, before mRNA extraction was performed. RNA was extracted from CAFs of 1.0×10⁴ cells under both serum unstimulated and serum stimulated conditions using the RNeasy Micro kit (Qiagen GmbH), following the supplier's protocol. At the same time, gene expression of actin alpha 2 (ACTA2: alias α-SMA) and fibroblast activation protein (FAP) were also examined for CAFs. The following primers were used for gene expression analysis of human ACTA2 and FAP: ACTA2 forward, 5′-GTGTTGCCCCTGAAGAGCAT-3′ and reverse, 5′-GCTGGGACATTGAAAGTCTCA-3′; FAP forward, 5′-TCTAAGGAAAGAAAGGTGCCAA-3′ and reverse, 5′-GATCAGTGCGTCCATCATGAAG-3′. Expression level was calculated using the 2^−ΔΔCq method based on the expression level of GAPDH, the housekeeping gene. Additionally, the results were analyzed by the 2^−ΔΔCq method, using the sample without serum stimulation as a reference (26). Gene expression levels were evaluated as the ratio of expression with serum stimulation to expression without serum stimulation: RNA level (fold change)=(Expression level with serum stimulation)/(Expression level without serum stimulation).

Pathway enrichment analysis was performed using two databases: Kyoto Encyclopedia of Genes and Genomes (KEGG: http://www.genome.ad.jp/kegg) and Reactome (https://reactome.org/). Enriched pathways were identified according to the cut-off value of a false discovery rate (FDR) <0.05.

First, CAFs were plated and incubated overnight in DMEM containing 10% FBS. Once the CAFs had settled in the plate, the medium was changed and CAFs were cultured in DMEM containing 10% FBS for 5 days with the p53 inhibitor pifithrin-alpha (PFT-alpha) (Sigma-Aldrich; Merck KGaA), or with the same amount of DMSO as a control. A PFT-alpha concentration of 10 µM was used based on previous a previous study (27). After 5 days of incubation, the cells were thoroughly washed and seeded in the top layer of Transwell, and then co-cultured with PSN-1 cells for 2 days, as in the aforementioned co-culture assay, to assess proliferation of the PSN-1 cells. Notably, viability of CAFs at the start of co-culture was measured using Guava^® Muse cell analyzer (Luminex) as described in the instructions to confirm that no problems with CAF viability had occurred after 5 days of incubation. This assay was conducted using CAFs from three patients, #2, #7 and #8. The experiments of #2, #7, and #8 were performed simultaneously on the same plate. Data for a control, a sample without CAF is shared.

The IL-6 concentration was analyzed in the co-culture supernatant using an enzyme-linked immunosorbent assay (ELISA) kit (cat. no 3460-1A-6; Mabtech, Inc.), following the manufacturer's instructions. Co-culture supernatants were collected at the end of the 2-day co-culture in the aforementioned experiment. When performing the analysis, the co-culture supernatant was diluted 10-fold. This assay was also conducted using CAFs from three patients, #2, #7 and #8.

Numerical data are presented as mean ± standard deviation. Statistical analysis between two groups was performed using a two-tailed Student's t-test (unpaired t-test). Statistical analysis among three groups or more was performed using Dunnett's t-test, with the test level α=0.05. Correlation analysis was performed using Pearson's correlation coefficient. P<0.05 was considered to indicate a statistically significant difference. JMP software (JMP^® Pro 16.2.0, SAS Institute, Inc.) was used as the software for these analyses.

To examine whether differences in CAFs affected PC cell proliferation, 8 CAF samples were primarily cultured and co-cultured with PSN-1 cells. The PSN-1 showed different proliferation ability when cultured with each CAFs from different patients (Fig. 1). In particular, the mean fluorescence intensity of #8 was significantly higher than that of #1. The fluorescence intensity rate (FIR) of #8 was ~1.4 (P<0.001), indicating that differences in CAFs could affect the proliferation ability of PC cells.

It was examined how the results of co-culture assays were associated with the original clinical data of patients. The correlation between FIR in co-culture assays and five clinical data is demonstrated in Fig. 2. Unfortunately, tumor size, serum CA19-9 level, and SUV max before the initiation of treatment were not significantly associated with the FIR (Fig. 2A) and this may be partially because sample size (only 8 patients) was too small and these factors depended on the time of diagnosis. There was also no significant association between sex and co-culture assays results. Fluorescence intensity (mean ± SD) in the co-culture experiment was 5,139.4±682.6 for females and 5,063.4±536.8 for males (P=0.779).

Thus, next we examined the correlation between TSR and FIR in six patients who received preoperative treatment to clarify the clinical impact of CAFs on the effect of preoperative treatment. As revealed in Fig. 2B, the FIR was significantly correlated with the TSR (R²=0.821, P=0.013), indicating that the efficacy of chemotherapy may be determined by the malignant potential of CAFs. Notably, further examination of the relationship between FIR and the patients' original clinical data revealed that patient's age was marginally related to the FIR (P=0.051), as shown in Fig. 2C, raising the suspicion that it was possible that the senescence of CAFs may be related to its malignant potential to grow cancer cells.

To elucidate the senescence of CAFs really affected the malignant potential of CAFs, a PCR array was performed among the 8 CAFs by using the ‘Human Cellular Senescence’ PCR array kit (GeneGlobe ID-PAHS-050Z QUIAGEN) and ‘Human Cancer Inflammation and Immunity Crosstalk’ PCR array kit (GeneGlobe ID-PAHS-181Z QUIAGEN). First, the expression of the CAF markers, ACTA-2 and FAP, was examined simultaneously with the analysis using these kits. Fortunately, the genes analyzed in these kits included genes known to be marker of CAF, VIM. Therefore, the expression of these genes in 8 primary CAFs was investigated and it was found that the expression of these 3 genes was positive in all CAFs (Fig. S3). By contrast, the expression of TERT, which is expressed in 85–90% of cancer tissues and is also known to be expressed in PC (28,29), was extremely low, suggesting that the cells in primary culture in the present study were CAFs without cancer cells. The gene expression of CAFs in relation to cancer cell growth was subsequently examined by investigating the relationship between the data of each gene expression analyzed by RT-qPCR and the results of the co-culture assay. Unfortunately, each gene expression with and without serum stimulation did not significantly correlate with the results of the co-culture assay. Then, the fold-change of RNA level of each gene was calculated using the 2^−ΔΔCq method and the relationship between fold-change of RNA level and co-culture assay data was analyzed (Table SIII and Fig. 3). The top 30 genes that exhibited a significant correlation between the fold-change of RNA level and FIR are presented in Fig. 3A. Table SIII shows the results for all 162 genes. The plot of absolute value of the correlation coefficient versus the P-value for each gene showed that 38 genes were significantly correlated with the FIR (Fig. 3B). These 38 genes included cellular senescence pathway-associated genes [such as tumor protein p53 (TP53), cyclin-dependent kinase inhibitor 1A (CDKN1A)/p21 and cyclin-dependent kinase inhibitor 2A (CDKN2A)/p16] and cytokines and chemokines known as SASP factors (such as IL1B, IL6, and CXCL12). Next, to clarify which pathway was related to the malignant potential of CAFs, pathway analysis for the top 30 genes was performed by using the KEGG and Reactome databases. The cellular senescence pathway was significantly enriched in both the KEGG and Reactome databases (Fig. 3C). Furthermore, Reactome pathway enrichment analysis revealed that the ‘SASP’ pathway showed the 6th strongest relationship with malignancy-related genes in CAFs.

As revealed in Fig. 4, for the 8 CAF samples, the relationship between FIR and the RNA levels of genes (selected from the top 30 genes) was analyzed. Ataxia telangiectasia mutated (ATM), TP53, CDKN1A/p21, and CDKN2A/p16 were related to the p53/p21 pathway or p16/RB pathway, both of which are known as cellular senescence pathways (30–32) (Fig. 4A). SASP-related genes, such as nuclear factor kappa B subunit 1 (NFKB1), IL6, and CXCL12 are revealed in Fig. 4B. NFKB1, which is reportedly essential for SASP induction (33,34), showed a strong correlation with FIR (R²=0.703, P=0.009). Additionally, the expression of these genes without and with serum stimulation, respectively, is shown in Fig. S4. Notably, most of these genes were significantly related to the TSR (Fig. 5A and B) and to the serum CA19-9 level at the time of surgery which is related to postoperative prognosis (Fig. S5). These findings indicated that these genes may be associated with the malignant potential of CAFs.

Based on the present results, focus was next addressed on the p53-mediated cellular senescence of CAFs. It was examined how CAFs treatment with the p53 inhibitor PFT-alpha altered the PC cell proliferation in co-culture assays. After 5 days of exposure to PFT-alpha, each CAF samples retained its spindle-like shape (Fig. S6) and exhibited no decrease of viability (all >90%). Upon co-culture of PSN-1 cells with DMSO-treated CAFs, a significant increase of fluorescence intensity compared with PSN-1 cells cultured alone was observed. This indicated that DMSO-treated CAFs also enhanced PSN-1 proliferation similar to untreated native CAFs. By contrast, PSN-1 co-cultured with PFT-alpha-treated CAFs showed significantly reduced fluorescence intensity compared with PSN-1 with DMSO-treated CAFs (Fig. 6A). This trend was observed in all 3 different CAFs. To further examine why the p53 inhibitor affected proliferative potential, the IL-6 concentration in the co-culture supernatant was measured. As revealed in Fig. 6B, IL-6 level significantly increased in PSN-1 cells with DMSO-treated CAFs compared with PSN-1 alone. By contrast, IL-6 level in PSN-1 co-cultured with PFT-alpha-treated CAFs was significantly reduced almost to the level of PSN-1 alone, indicating that PFT-alpha could suppress IL-6 secretion from CAFs by inhibiting p53-mediated cellular senescence, resulting in inhibition of PSN-1 proliferation.