Although fibrotic stroma forms an integral component of pancreatic diseases, whether fibroblasts programmed by different types of pancreatic diseases are phenotypically distinct remains unknown. Here, we show that fibroblasts isolated from patients with pancreatic ductal adenocarcinoma (PDAC), chronic pancreatitis (CP), periampullary tumors, and adjacent normal (NA) tissue (N = 34) have distinct mRNA and miRNA profiles. Compared with NA fibroblasts, PDAC-associated fibroblasts were generally less sensitive to an antifibrotic stimulus (NPPB) and more responsive to positive regulators of activation such as TGFβ1 and WNT. Of the disease-associated fibroblasts examined, PDAC- and CP-derived fibroblasts shared greatest similarity, yet PDAC-associated fibroblasts expressed higher levels of tenascin C (TNC), a finding attributable to miR-137, a novel regulator of TNC. TNC protein and transcript levels were higher in PDAC tissue versus CP tissue and were associated with greater levels of stromal activation, and conditioned media from TNC-depleted PDAC-associated fibroblasts modestly increased both PDAC cell proliferation and PDAC cell migration, indicating that stromal TNC may have inhibitory effects on PDAC cells. Finally, circulating TNC levels were higher in patients with PDAC compared with CP. Our characterization of pancreatic fibroblast programming as disease-specific has consequences for therapeutic targeting and for the manner in which fibroblasts are used in research.

Significance:

Primary fibroblasts derived from various types of pancreatic diseases possess and retain distinct molecular and functional characteristics in culture, providing a series of cellular models for treatment development and disease-specific research.

The extensive fibrotic stroma associated with pancreatic diseases plays an integral role in their pathogenesis. Activated fibroblasts constitute the most significant component of disease-associated desmoplastic stroma (1). Pancreatic stellate cells (PSC), a major source of pancreatic fibroblasts, are resident lipid-storing cells, which in health are quiescent, and characterized by many lipid droplets rich in vitamin A. Pancreatic injury causes the activation of PSCs to myofibroblast-like cells. This is associated with loss of cytoplasmic lipid droplets, increased proliferation, and enhanced expression of α-smooth muscle actin (αSMA), as well as extracellular matrix (ECM) components (2). Where pancreatic injury is long-lasting, e.g., in chronic pancreatitis (CP) or pancreatic ductal adenocarcinoma (PDAC), persistent fibroblast activation leads to overproduction of collagen, fibronectin, laminin, and hyaluronic acid resulting in a dense fibrotic meshwork.

In mouse models of pancreatic cancer, PSCs increase tumor growth, promote metastasis, and accompany cancer cells to metastatic sites where they stimulate angiogenesis (3). There is evidence that the dense collagenous meshwork produced by activated fibroblasts physically prevents drug delivery to the tumor cells (4). Moreover, cancer-associated fibroblasts have been reported to scavenge the chemotherapeutic drug gemcitabine, reducing the levels of drug available to tumor cells (5). On the other hand, depleting pancreatic fibroblasts from PDAC tumors in mice yielded aggressive tumors and decreased survival (6, 7). Thus, a complex picture emerges of cells with the apparent contradictory capabilities of both protecting against tumor aggression and promoting tumor progression.

Understanding how best to target disease-associated pancreatic fibroblasts is a priority. An early study identified significant overlap in gene expression of desmoplastic stroma from PDAC and the fibrous stroma of CP (8). Consistent with this, isolated activated fibroblasts from either CP or PDAC share morphologic and functional characteristics (9). However, whether pancreatic fibroblasts, activated in the context of different pancreatic diseases or in culture, are similar or whether they are uniquely programmed by their environments remains an unanswered question in the field. Answering this question has important implications for how we therapeutically target these cells in different diseases and for how pancreatic fibroblasts are used in research.

Using a combination of miRNA and mRNA expression profiling of fibroblasts isolated from PDAC, CP, periampullary tumors (PAT), and areas of histologically normal pancreas, followed by comprehensive validation, we show that activated fibroblasts derived from different pancreatic disease types are distinct.

Pancreatic tissue and blood samples

Resected pancreatic tissue and blood samples were obtained following the approval of the UK Health Research Authority in accordance with the Declaration of Helsinki, including ethical approval (London – South East Research Ethics Committee, Ref: 16/LO/1630) and written-informed consent, from patients undergoing pancreatic surgery at the Royal Liverpool University Hospital. For discovery work, activated fibroblasts from areas of diseased pancreas were isolated from patients with PDAC (n = 6), PAT (n = 5) comprising ampullary and duodenal tumors, and CP (n = 5). Unmatched adjacent-morphologically normal pancreatic tissue from patients with mucinous cystic neoplasm (n = 1), intraductal papillary mucinous neoplasm (IPMN; n = 1), PDAC (n = 1), and PAT (n = 2) was used to isolate adjacent-normal PSCs. For independent validation, fibroblasts were isolated in the same manner as the discovery set from an additional 13 patients (PDAC, n = 9; CP, n = 3; and Normal, n = 1). Finally, commercially available human primary normal pancreatic fibroblasts (NF; Vitro Biopharma) were also analyzed. A portion of each specimen was formalin-fixed and stained with hematoxylin and eosin to confirm the histopathology.

Primary fibroblasts isolation

Diseased fibroblasts (PDAC, CP, and PAT) were isolated from pancreatic tissue using the outgrowth method (10) with modifications. Briefly, fibrotic pancreatic tissue, identified following gross analysis of the resected specimen by a pathologist, was collected in ice-cold PBS containing 1% penicillin–streptomycin. Tissue was cut into 1 to 2 mm3 pieces, seeded in a 6-well cell culture uncoated plate in the presence of Iscove's Modified Dulbecco's Medium (IMDM) supplemented with 20% FBS, 2% l-glutamine, and 1% penicillin–streptomycin, and maintained at 37°C in a humidified atmosphere of 5% CO2/95% air until fibroblasts grew out (2 to 4 weeks to reach 80% confluence). The tissue pieces were removed when fibroblasts reached about 30% confluence. The medium was changed twice weekly, and cells were grown to 80% confluence, harvested, and stored in liquid nitrogen.

Unmatched adjacent-morphologically normal pancreatic tissue was used to isolate adjacent-normal PSCs using the density gradient method (11). Isolated quiescent PSCs were cultured in a 6-well cell culture uncoated plate (and thus culture-activated) until cells reached 80% (2 to 4 weeks) to achieve adequate yield and purity as previously described by Sherman and colleagues (12) and are hereafter referred to as normal activated (NA). These cells were cultured in IMDM supplemented with 20% FBS, 2% l-glutamine, and 1% penicillin–streptomycin, and maintained at 37°C in a humidified atmosphere of 5% CO2/95% air. The medium was changed twice weekly, and cells were grown to 80% confluence, harvested, and stored in liquid nitrogen.

The commercially available normal fibroblasts (NF) were grown using commercial low serum medium (NF-LSM; MSC-GRO, Vitro Biopharma) and IMDM supplemented with 10% FBS, 2% l-glutamine, and 1% penicillin–streptomycin (NF-HSM) using the same conditions detailed above.

For mRNA/miRNA array experiments, diseased fibroblasts, NA fibroblasts, and commercially available NF from early passages (P1 or P2) were recovered from liquid nitrogen simultaneously and placed in a T75 flask with IMDM supplemented with 10% FBS, 2% l-glutamine, and 1% penicillin–streptomycin. The media were replenished twice weekly, and cells were grown to 80% confluence before RNA was isolated as described below.

Cell lines

Human PDAC cell lines, MIA Paca-2, PANC-1, and HeLa were obtained from the ATCC. Cells were validated by short tandem repeat profiling and tested for mycoplasma using e-Myco plus Mycoplasma PCR Detection Kit (iNTRON Biotechnology) following the manufacturer's instructions.

Microarray experiments

Total RNA isolated from the fibroblasts obtained was hybridized onto GeneChip Affymetrix miRNA 4.0 and Human Transcriptome Array 2.0 Affymetrix. The labeling, hybridization, scanning, and data extraction of mRNA and miRNA array experiments were performed at the Centre for Genomic Research, University of Liverpool.

Human transcriptome array

Note that 50 ng of total RNA (free from DNA) was labeled using the Ovation Pico WTA system V2 (NuGen Technologies, Inc.). This kit prepares amplified cDNA for gene expression analysis. Following second-strand synthesis, the cDNA was purified using Agencourt RNAClean XP beads prior to single primer isothermal amplification. The single primer isothermal–amplified cDNA was purified using the Qiagen QIAquick PCR Purification system. Utilizing the Encore Biotin Module, 5 μg of amplified cDNA was fragmented and Biotin-labeled before hybridization on Affymetrix GeneChip Human Transcriptome 2.0 arrays (HTA_2.0; Affymetrix, Inc.). Arrays were hybridized for 17 hours at 45°C and 60 rpm in an Affymetrix GeneChip hybridization oven 645. Following hybridization, the arrays were washed and stained on an Affymetrix GeneChip Fluidics Station 450 using a FS450_0001 script and scanned in an Affymetrix Genechip Scanner 3000G.

miRNA array

Note that 130 ng of total RNA (free from DNA) was labeled using the FlashTag Biotin HSR RNA Labeling Kit (Affymetrix). This process began with a brief tailing reaction followed by ligation of the biotinylated signal molecule to the target RNA sample. Biotin-labeled samples and controls were added to a hybridization mix and hybridized to Affymetrix GeneChip miRNA 4.0 arrays for 18 hours at 48°C and 60 rpm in an Affymetrix GeneChip hybridization oven 645. Following hybridization, the arrays were washed and stained on an Affymetrix GeneChip Fluidics Station 450 using a FS450_0002 script and scanned in an Affymetrix Genechip Scanner 3000G.

Conditioned media experiments using PDAC cell lines

PANC-1 and MIA PaCa-2 cells were seeded at a density of 7.5 × 105 cells per 75-cm2 flasks using DMEM + GlutaMAX supplemented with 10% FBS and 1% penicillin–streptomycin. After 24 hours, media were replaced and cells were grown for an additional 72 hours, at which point, cells had reached 80% to 90% confluence. Conditioned media (CM) were collected and filtered using a 0.22 μm filter and stored at −80°C until use. Flasks without cells were processed similarly, as controls. Fibroblasts were seeded in 6-well plates at 8.0 × 104 cells per well using DMEM + GlutaMAX, supplemented with 10% FBS and 1% penicillin—streptomycin, and after reaching approximately 70%, confluence cells were incubated with CM diluted 1:2 with fresh DMEM for 48 hours before RNA was extracted for qRT-PCR analysis.

NPPB treatment

R2796 and R2797 (NA fibroblasts) in addition to R2910 and R3104 (PDAC-associated fibroblasts) were seeded on a 24-well plate at 3.0 × 104 cells per well using DMEM + GlutaMAX supplemented with 10% FBS and 1% penicillin–streptomycin. The next day, cells were washed with PBS twice and further incubated in DMEM + GlutaMAX serum-free media supplemented with 1% penicillin–streptomycin while being treated with 100 nmol/L of human NPPB (Bachem) and water (control) 3 times a day for 48 hours. This dosing protocol was followed to maintain active levels of NPPB in culture (13).

TGFβ1 treatment

R2796 and R2797 (NA fibroblasts) in addition to R2910, R3104, and R2875 (PDAC-associated fibroblasts) were seeded on a 24-well plate at 3.0 × 104 cells per well using DMEM + GlutaMAX supplemented with 10% FBS and 1% penicillin–streptomycin. The next day cells were washed with PBS twice and further incubated in DMEM + GlutaMAX serum-free media supplemented with 1% penicillin–streptomycin while being treated with 5 ng/mL or 10 ng/mL of human TGFβ1 (Millipore) and water (control) for 24 and/or 48 hours.

Wnt pathway activation

Wnt pathway stimulation was achieved using an L-Wnt-3A cell line (14). In order to generate L-Wnt-3A CM, 1 × 106 cells per T75 flask were seeded using DMEM + GlutaMAX, 10% FBS, and 1% geneticin. After 24 hours, media were replaced and cells were grown for an additional 72 hours, at which point, cells had reached 80% to 90% confluence. Then, CM were collected and filtered using a 0.22 μm filter. In addition, a flask without cells was also filled with culture medium for 72 hours to generate control media and processed as before. R2796, R2797, and R2951 (NA fibroblasts) in addition to R3008 and R3072 (PDAC-associated fibroblasts) and R3030 and R334 (PAT-associated fibroblasts) were seeded at 50,000 cells on a 6-well plate for 24 hours before media were replaced with L-Wnt-3A CM for an additional 24 hours before RNA was extracted to measure AXIN2 levels using qRT-PCR.

Tenascin C knockdown and collection of CM

Immortalized PDAC-associated and NA fibroblasts were seeded on a 6-well plate at 6 × 104 cells per well in DMEM + GlutaMAX supplemented with 10% FBS without antibiotics. After 24 hours, cells were transfected using Opti-MEM I–reduced serum media followed by addition of Lipofectamine (Thermo Fisher Scientific) according to the manufacturer's instructions. Two different siRNAs targeting tenascin C (TNC; ON-TARGET plus siRNA Human TNC: #J-009298-07-0005 and #J-009298-05-0005) or a nontargeting control (ON-TARGET plus control pool: #D-001810-10-20; Dharmacon) with a final siRNA concentration of 30 nmol/L were used. The day after the transfection medium was replaced by serum-free media (DMEM + GlutaMAX without antibiotics). Cells were then grown for a further 48 hours to reach 80% to 90% confluence. CM were collected and filtered using a 0.22 μm filter and stored at −80°C until use for cell migration and proliferation experiments. In addition, cells were lysed to obtain protein.

Migration assays

Transwell cell migration assays were performed using 8 μm pore size ThinCertTM cell culture inserts (Greiner Bio-One) in a 24-well format. MIA PaCa-2 cells were preincubated in serum-free DMEM + GlutaMAX media overnight. Note that 100 μL of cell suspension containing 6 × 104 cells was placed in the insert. The outer chamber was filled with 500 μL of CM from the fibroblasts TNC knockdown experiment including controls (diluted 1:2 with DMEM + GlutaMAX and supplemented with 2% FBS) and cells were allowed to migrate for 24 hours. After incubation, nonmigrating cells were removed from the top of the membrane with cotton swab. Migrated cells were fixed with 70% ethanol and stained with 0.3% crystal violet. Quantification of migration was performed using QuPath (version 0.2.0-m2; ref. 15). Images were taken using a Nikon Eclipse E600 at 4x. The area of migrated cells (crystal violet) and cell-free area were quantified using the pixel classifier tool (applying the Random Trees classifier and using a downsample resolution between 4 and 8). Cell migration of treated cells was expressed as a percentage of control.

Proliferation assays

MIA PaCa-2 cells were plated on a 96-well plate at 5 × 103 cells per well using DMEM + GlutaMAX supplemented with 10% FBS. The following day cells were washed twice with PBS and treated with CM from the normal or PDAC-associated fibroblasts TNC knockdown experiment including controls (diluted 1:2 with DMEM + GlutaMAX and supplemented with 2% FBS). After 48 hours, cell proliferation was measure using the EZ4U assay system (Biomedica) and a Multiskan FC microplate reader (Thermo Fisher Scientific) according to the manufacturer's instructions.

miRNA mimics

Immortalized PDAC-associated fibroblasts were seeded on a 24-well plate at 2.0 × 104 cells per well using DMEM + GlutaMAX supplemented with 10% FBS without antibiotics. After 24 hours, miRNA mimics including hsa-miR-137 and hsa-miR-212-3p (30 nmol/L; Qiagen) or AllStars off-target control (30 nmol/L; Qiagen) were diluted in Opti-MEM I–reduced serum media (Thermo Fisher Scientific) followed by addition of HiPerFect (Qiagen) transfection reagent according to the manufacturer's instructions. Complexes were added drop-wise onto the cells and incubated for 48 hours before RNA or CM were extracted for qRT-PCR and immunoblotting respectively.

Tissue microarray construction

Formalin-fixed paraffin-embedded PDAC and CP specimens were obtained from the Royal Liverpool University Hospital. Matched hematoxylin and eosin–stained slides were marked by a specialist histopathologist (F. Campbell) for areas of tumor and fibrosis in the PDAC and CP cases, respectively. Two tissue microarrays (TMA) were constructed from PDAC specimens, PDAC TMA #1 (n = 41) and PDAC TMA #2 (n = 48), with three 0.6 mm cores extracted per patient. One TMA was constructed from CP specimens (n = 47), with four 0.6 mm cores extracted for each patient.

Serum biomarker analysis by multiplex and ELISA analysis

A discovery cohort of serum samples was obtained from patients undergoing pancreatic surgery at the Royal Liverpool University Hospital (termed UoL, n = 65) and healthy donors (n = 15). The discovery cohort was subjected to multiplex quantification of 101 cancer-associated proteins (Human OncologyMAP v1.0, Myriad Rules-Based Medicine).

ELISAs for TNC (Large FNIII-B, IBL international) were performed according to the manufacturer's instructions and assessed for reproducibility in the UoL discovery cohort (n = 75). Upon confirmation, TNC concentrations were determined in an independent cohort of UoL samples (n = 155).

Statistical analysis

All statistical tests were performed using R software V.3.5.2. Continues variables were expressed as mean ± SEM and compared using a one-way ANOVA or Kruskal–Wallis Test, paired or unpaired t test, and Mann–Whitney U test as appropriate; categorical variables were compared using the χ2 test. Results were considered significant at P < 0.05.

Supplementary data

Supplementary Methods are described in Supplementary Materials and Methods.

Disease-associated fibroblasts from pancreatic pathologies show distinct gene expression profiles

To determine the extent to which fibroblasts are programmed by their distinct disease environments, we isolated pancreatic fibroblasts and analyzed uniquely low passage (P1 or P2) from a total of 34 individuals, enabling us to compare PDAC, PAT, CP, and adjacent normal-derived cells (Fig. 1A; Supplementary Table S1). Consistent with others, we found that disease-associated fibroblasts exhibited myofibroblast-like morphology with characteristic expression of αSMA, desmin, vimentin (Fig. 1B), and other common markers of fibroblasts (Supplementary Fig. S1A). One day after isolation, primary normal PSCs contained characteristic lipid droplets as measured by Oil Red O (Fig. 1B); however, following expansion in culture, they became activated, gaining expression of αSMA, and are hereafter called NA. None of the isolated cells expressed macrophage (CD68) or epithelial (pan-cytokeratin and EPCAM; Fig. 1B; Supplementary Fig. 1A) markers. KRAS mutational status was successfully determined for 33 isolates. One isolate, from a PDAC tumor, was found to be heterozygous for mutant KRAS (G12R) and was excluded from the study.

Figure 1.

Isolation, characterization, and clustering by gene expression profiling of pancreatic fibroblasts. A, Schematic showing fibroblast isolates from different pancreatic diseases using the outgrowth (top) or from unmatched normal appearing tissue using the density gradient methods (bottom) including the total number of isolates per fibroblast type (discovery set plus independent set). B, Representative images of primary PDAC-associated fibroblasts confirming expression of αSMA, desmin, and vimentin. Fibroblasts did not express CD68 or cytokeratin. Scale bar, 50 μm. In addition, the image shows a primary normal PSC containing lipid droplets visible by Oil Red O staining 1 day after isolation. C and D, PCA with 25,195 gene transcripts (C) and 6,631 miRNA genes (D) separated samples into 5 subsets. Each sphere (PDAC = 6; CP = 5; PAT = 5; NA = 5) represents one primary sample. Commercial primary NFs cultured with high serum medium (HSM) or low serum medium (LSM) were analyzed in triplicate. E, Venn diagram depicting the number of differentially expressed genes (mRNA array) between disease-associated fibroblasts from PDAC, CP, and PAT versus NA fibroblasts, including the number of genes up- or downregulated per contrast (the percentage is indicated in the parenthesis). F, Top canonical pathways enrichment analysis of differentially expressed genes in all disease-associated fibroblasts versus NA fibroblasts, performed using IPA. Fisher exact test was used to determine P values, with significance set at P = 0.05, which translates to –log (P value) of 1.3.

Figure 1.

Isolation, characterization, and clustering by gene expression profiling of pancreatic fibroblasts. A, Schematic showing fibroblast isolates from different pancreatic diseases using the outgrowth (top) or from unmatched normal appearing tissue using the density gradient methods (bottom) including the total number of isolates per fibroblast type (discovery set plus independent set). B, Representative images of primary PDAC-associated fibroblasts confirming expression of αSMA, desmin, and vimentin. Fibroblasts did not express CD68 or cytokeratin. Scale bar, 50 μm. In addition, the image shows a primary normal PSC containing lipid droplets visible by Oil Red O staining 1 day after isolation. C and D, PCA with 25,195 gene transcripts (C) and 6,631 miRNA genes (D) separated samples into 5 subsets. Each sphere (PDAC = 6; CP = 5; PAT = 5; NA = 5) represents one primary sample. Commercial primary NFs cultured with high serum medium (HSM) or low serum medium (LSM) were analyzed in triplicate. E, Venn diagram depicting the number of differentially expressed genes (mRNA array) between disease-associated fibroblasts from PDAC, CP, and PAT versus NA fibroblasts, including the number of genes up- or downregulated per contrast (the percentage is indicated in the parenthesis). F, Top canonical pathways enrichment analysis of differentially expressed genes in all disease-associated fibroblasts versus NA fibroblasts, performed using IPA. Fisher exact test was used to determine P values, with significance set at P = 0.05, which translates to –log (P value) of 1.3.

Close modal

mRNA and miRNA gene expression profiles of isolated fibroblasts, as well as commercially available NF grown in low or high serum, were generated. qRT-PCR analysis of eight randomly selected genes was undertaken to validate our microarray profiling experiments. The expression patterns observed in array data for four mRNAs (Supplementary Fig. S1B) and two miRNAs (Supplementary Fig. S1C) were confirmed by qRT-PCR in the discovery samples and for a further two mRNA candidates (Supplementary Fig. S1D) in analyses, which included independent primary fibroblasts (Independent set, Supplementary Table S1).

Principal component analysis (PCA) of both mRNA (Fig. 1C) and miRNA (Fig. 1D) expression showed evidence of clustering based on the disease group from which fibroblasts were derived, indicating differences in the nature of activated fibroblasts between disease types.

To explore this further, we compared the extent to which fibroblasts programmed by different pancreatic diseases differ from NA fibroblasts. We found 1,469 genes differentially expressed between PDAC-associated fibroblasts and NA, compared with only 776 genes differentially expressed between PAT-associated fibroblasts and NA. This suggested that distinct malignancies program fibroblasts differently. Chronic pancreatitis fell between PDAC and PAT with 925 differentially expressed genes compared with NA (Fig. 1E). Surprisingly, only 24% of genes were commonly dysregulated across the three different diseases compared with NA (Fig. 1E), although this rose to 48.5% when the disease-associated fibroblasts groups were compared with NF (Supplementary Fig. S2A).

Almost one third (32.2%) of genes significantly altered between PDAC and NA were unique to this comparison group, whereas only 6.6% and 7.3% respectively were exclusive for the CP versus NA and PAT versus NA comparisons (Fig. 1E). Interestingly, PDAC and CP shared over 41.5% of genes dysregulated when compared with NA, indicating similarities in the programming of fibroblasts by these disease types.

Consistent with recent reports (16–18), the most enriched canonical pathways associated with disease-associated fibroblasts compared with NA were phagosome maturation, autophagy, and endocytosis signaling (Fig. 1F). The most common biological functions differentially regulated between disease-associated fibroblasts and NA or NF corresponded to cell movement and migration with key upstream regulators such as TNF, TP53, and TGFβ1 (19–21) altered in these comparison groups (Supplementary Fig. S2B and S2C).

Only five miRNAs were significantly dysregulated between PDAC-associated fibroblasts and NA (Supplementary Fig. S2D). Similarly, five miRNAs were altered between CP and NA, and four of these were also dysregulated between PDAC-associated fibroblasts and NA. Fibroblasts programmed by PAT had two dysregulated miRNAs compared with NA, one of which was unique to PAT (hsa-miR-92a-1-5p), and the other, hsa-miR-138-1-3p, was also dysregulated in CP and PDAC. Nine miRNAs were dysregulated in all three-disease types compared with NF (Supplementary Fig. S2E).

Taken together, our data indicate that although there is commonality between fibroblasts cultured from different pancreatic disease types or normal tissue, the groups are distinct. Given that our PCA analysis showed NA to be more closely related to the other primary fibroblast isolates than the commercially available NF were, we focused subsequent analyses on NA.

PDAC-associated fibroblasts are less responsive to an anti-fibrotic stimulus and more responsive to positive regulators of fibroblast activation than NA fibroblasts

Next we sought to understand how PDAC-associated fibroblasts differed from all other disease-associated or NA fibroblasts. Our analysis (Fig. 2A) revealed 224 differentially regulated genes between PDAC- and PAT-derived fibroblasts and 90 differentially regulated genes between PDAC and CP. Twelve genes were uniquely dysregulated in PDAC-associated fibroblasts compared with all other groups. One of these, NPPB, also known as brain natriuretic peptide, was among the most highly upregulated transcripts between groups in both discovery (Fig. 2B), and validation samples (Supplementary Fig. S2F). Analysis of data from the Gene Expression Omnibus database (Fig. 2C, https://www.ncbi.nlm.nih.gov/geo/geo2r/, GEO accession: GSE21440; ref. 22) corroborated our observation. NPPB mRNA levels in fibroblasts derived from PDAC [cancer-associated fibroblast (CAF); n = 9] were higher than from control fibroblast lines (HPNE and NF1-3) and from an IPMN (SC-2; ref. 22).

Figure 2.

PDAC-associated fibroblasts display greater resistance to the antifibrotic effects of NPPB and greater responsiveness to TGFβ1 and WNT signaling compared with NA. A, Venn diagram depicting the number of differentially expressed genes (mRNA array) between PDAC-associated fibroblasts and the other groups. The 12 genes dysregulated between PDAC and all other groups are named, with NPPB highlighted in red as the most upregulated gene in the list. B,NPPB mRNA array expression measured as Robust Multi-array Average (RMA) in discovery fibroblast isolates (PDAC = 6; CP = 5; PAT = 5) and normal PSC culture–activated as described in Material and Methods (NA = 5). C,NPPB mRNA expression values from Gene Chips data in PDAC- (CAF), CP- (NF1-3), pancreatic IPMN tumor– (SC-2), and normal pancreas immortalized–associated fibroblasts (HPNE). Data collected from GEO database accession numbers GSM535920 to GSM535931 D, qRT-PCR analysis of NPPB expression levels in PDAC-associated fibroblasts (PDAC-F: R2928 and R2875) and NA-associated fibroblasts (NA-F: R2796 and R2797) following incubation for 48 hours with CM from PANC-1 and MIA PaCa-2 cells. Expression was normalized to GAPDH using unconditioned DMEM as control. Data are shown as mean ±SEM; n = 3 independent experiments; P value determined by one-way ANOVA with post hoc Dunnett's test. n.d., nondetectable expression. E, qRT-PCR analysis of ACTA2 expression levels in discovery and independent fibroblast samples. Two PDAC- (R2910 and R3104) and two NA-associated fibroblasts (R2796 and R2797) isolates are highlighted and analyzed in F by qRT-PCR for ACTA2 after incubation with recombinant human NPPB for 48 hours. Expression levels were compared with control (DMEM) and normalized to GAPDH. Error bars depict mean ±SEM of technical replicates from two NA- and two PDAC-associated fibroblast isolates. P value determined by unpaired t test using DMEM as control. G, The top 5 upstream regulators (with their predicted activation state) significantly enriched in all disease-associated fibroblasts versus NA fibroblasts, as determined using IPA (left plot). qRT-PCR for ACTA2 after incubation with recombinant human TGFβ1 for 24 hours. Expression levels were compared with control (DMEM) and normalized to GAPDH. Error bars depict mean ± SEM of technical replicates from two NA- (R2796 and R2797) and three PDAC-associated fibroblast isolates (R2910, R3104, and R2875). P value determined by unpaired t test (right plot). H, Network analysis employing the IPA molecular activity prediction tool to assess directionality of changes in the Wnt signaling pathway for differentially expressed genes in all disease-associated fibroblasts versus NA fibroblasts. I, qRT-PCR analysis of AXIN2 expression levels following incubation with CM from L-Wnt-3A mouse fibroblast cell line for 24 hours. Expression levels were compared with unconditioned media as control (DMEM) and normalized to GAPDH. Error bars depict mean ± SEM of technical replicates from three NA- (NA-F: R2796, R2797 and R2951), two PDAC- (PDAC-F: R3008 and R3072), and two PAT-associated fibroblasts (PAT-F: R3030 and R3334). P values determined by unpaired t test.

Figure 2.

PDAC-associated fibroblasts display greater resistance to the antifibrotic effects of NPPB and greater responsiveness to TGFβ1 and WNT signaling compared with NA. A, Venn diagram depicting the number of differentially expressed genes (mRNA array) between PDAC-associated fibroblasts and the other groups. The 12 genes dysregulated between PDAC and all other groups are named, with NPPB highlighted in red as the most upregulated gene in the list. B,NPPB mRNA array expression measured as Robust Multi-array Average (RMA) in discovery fibroblast isolates (PDAC = 6; CP = 5; PAT = 5) and normal PSC culture–activated as described in Material and Methods (NA = 5). C,NPPB mRNA expression values from Gene Chips data in PDAC- (CAF), CP- (NF1-3), pancreatic IPMN tumor– (SC-2), and normal pancreas immortalized–associated fibroblasts (HPNE). Data collected from GEO database accession numbers GSM535920 to GSM535931 D, qRT-PCR analysis of NPPB expression levels in PDAC-associated fibroblasts (PDAC-F: R2928 and R2875) and NA-associated fibroblasts (NA-F: R2796 and R2797) following incubation for 48 hours with CM from PANC-1 and MIA PaCa-2 cells. Expression was normalized to GAPDH using unconditioned DMEM as control. Data are shown as mean ±SEM; n = 3 independent experiments; P value determined by one-way ANOVA with post hoc Dunnett's test. n.d., nondetectable expression. E, qRT-PCR analysis of ACTA2 expression levels in discovery and independent fibroblast samples. Two PDAC- (R2910 and R3104) and two NA-associated fibroblasts (R2796 and R2797) isolates are highlighted and analyzed in F by qRT-PCR for ACTA2 after incubation with recombinant human NPPB for 48 hours. Expression levels were compared with control (DMEM) and normalized to GAPDH. Error bars depict mean ±SEM of technical replicates from two NA- and two PDAC-associated fibroblast isolates. P value determined by unpaired t test using DMEM as control. G, The top 5 upstream regulators (with their predicted activation state) significantly enriched in all disease-associated fibroblasts versus NA fibroblasts, as determined using IPA (left plot). qRT-PCR for ACTA2 after incubation with recombinant human TGFβ1 for 24 hours. Expression levels were compared with control (DMEM) and normalized to GAPDH. Error bars depict mean ± SEM of technical replicates from two NA- (R2796 and R2797) and three PDAC-associated fibroblast isolates (R2910, R3104, and R2875). P value determined by unpaired t test (right plot). H, Network analysis employing the IPA molecular activity prediction tool to assess directionality of changes in the Wnt signaling pathway for differentially expressed genes in all disease-associated fibroblasts versus NA fibroblasts. I, qRT-PCR analysis of AXIN2 expression levels following incubation with CM from L-Wnt-3A mouse fibroblast cell line for 24 hours. Expression levels were compared with unconditioned media as control (DMEM) and normalized to GAPDH. Error bars depict mean ± SEM of technical replicates from three NA- (NA-F: R2796, R2797 and R2951), two PDAC- (PDAC-F: R3008 and R3072), and two PAT-associated fibroblasts (PAT-F: R3030 and R3334). P values determined by unpaired t test.

Close modal

We postulated that the high levels of NPPB in PDAC-associated fibroblasts were the result of in vivo tumor–stroma paracrine signaling prior to fibroblast isolation. Indeed, stimulating primary PDAC-derived fibroblasts with CM from PANC-1 or MIA PaCa-2 cells led to increased NPPB transcription in the fibroblasts (Fig. 2D). By contrast, NPPB expression was undetectable in two independent isolates of NA fibroblasts, and consistently remained undetectable after treatment with CM from pancreatic cancer cell lines. This suggested differences in the ways in which PDAC and NA fibroblasts respond to external stimuli. Curiously, for one of the most highly regulated molecules in PDAC-associated fibroblasts, NPPB's function may relate to reducing fibrosis. NPPB is reported to have antifibrotic properties in the heart (13) and kidney (23) and inhibits liver fibrosis by preventing activation of hepatic fibroblasts (24). We examined whether incubation of activated pancreatic fibroblasts with NPPB could reverse their activation, leading to diminished αSMA gene (ACTA2) expression. To avoid bias, we selected NA- and PDAC-associated fibroblasts respectively with high (R2796 and R2910) and low (R2797 and R3104) pretreatment αSMA levels (Fig. 2E). Treatment of fibroblasts with recombinant NPPB protein led to a decrease in αSMA expression in both NA and PDAC high and low αSMA expressers (Fig. 2F), although the greatest antifibrotic effects were observed in NA.

We next evaluated positive regulators of fibroblast activation. Consistent with its prominent role in pancreatic fibrosis (25), Ingenuity Pathway Core Analysis identified TGFβ1 among the top three activated upstream regulators in pancreatic disease–associated fibroblasts compared with NA (Fig. 2G, left plot). Treatment of high (R2796 and R2910) and low (R2797, R3104, R2875) αSMA–expressing fibroblasts with TGFβ1 increased the activation state, as measured by αSMA expression, in both PDAC- and NA-associated fibroblasts (Fig. 2G, right plot), with a significant induction of αSMA observed in PDAC-derived fibroblasts. The canonical Wnt pathway mediates fibroblast activation through interaction with TGFβ (26). The molecular activity predictor tool in Ingenuity Pathway Analysis (IPA) predicted that the Wnt pathway was activated in PDAC-associated fibroblasts compared with NA (Fig. 2H). To evaluate Wnt signaling, we incubated NA, PDAC-associated fibroblasts and PAT-associated fibroblasts with CM from a cell line that constitutively secretes Wnt-3A (14). The resulting induction of AXIN2 expression was observed in both NA and the cancer-associated fibroblasts, with the highest induction seen in cancer-associated fibroblasts (Fig. 2I). Collectively, our data on NPPB, TGFβ, and WNT signaling demonstrate that fibroblasts activated in vitro by the culture process (NA) are functionally distinct from fibroblasts programmed in vivo by cancer. PDAC-associated fibroblasts were less readily inactivated than NA, and considerably more responsive to a positive trigger of activation.

Disease-associated fibroblasts from distinct pancreatic disorders exhibit subtype-specific genetic profiles

We next narrowed our focus to fibroblasts programmed by pancreatic diseases, examining differences between fibroblasts from PDAC, CP, and PAT (Fig. 3A). The greatest overlap in gene expression was between PDAC and CP with only 90 differentially expressed genes. By contrast, PDAC and PAT differed by 224 genes, and there were 272 genes differentially regulated between PAT and CP. Interestingly, the most enriched pathway identified by IPA, the Hepatic Stellate Cell Activation Pathway was not significantly differentially enriched between PDAC and CP, indicating commonality between the activation of fibroblasts in these two disease types (Fig. 3B and C). By contrast, this pathway was significantly upregulated in PDAC compared with PAT, and downregulated in PAT compared with CP (Fig. 3B and C), highlighting differences in the way in which PAT fibroblasts are activated compared with both PDAC and CP. Gene expression alterations in this pathway can be attributed to upstream regulators such as TGFβ1, p38 MAPK, MAPK, and HIF1A, (27, 28), which were significantly modulated between these comparison groups (Supplementary Table S2). Four of the top eight pathways identified by IPA were immune-related (Fig. 3B). Of note, the genes most commonly dysregulated in these pathways belonged to the MHC class I, namely HLA-A, HLA-B, and HLA-C, which were downregulated in PDAC-versus-CP and in PDAC-versus-PAT (Fig. 3D). In addition, the Cell Cycle Control of Chromosomal Replication and the ATM signaling pathways, which showed no differences between PDAC and CP (Fig. 3B), were significantly dysregulated in PAT compared with CP (Fig. 3E and F).

Figure 3.

Diseased pancreatic fibroblasts retain identity ex vivo. A, Venn diagram depicting the number of differentially expressed genes (mRNA array) between disease-associated fibroblast groups including the number of genes up- or downregulated per contrast (the percentage is indicated in the parenthesis). B, Top canonical pathway enrichment analysis of differentially expressed genes between disease-associated fibroblast groups using IPA comparison analysis. Fisher exact test was used to determine P values, with the dashed line indicating P < 0.05. C–F, Fold change of gene expression in selected pathways for contrasts analyzed in B.

Figure 3.

Diseased pancreatic fibroblasts retain identity ex vivo. A, Venn diagram depicting the number of differentially expressed genes (mRNA array) between disease-associated fibroblast groups including the number of genes up- or downregulated per contrast (the percentage is indicated in the parenthesis). B, Top canonical pathway enrichment analysis of differentially expressed genes between disease-associated fibroblast groups using IPA comparison analysis. Fisher exact test was used to determine P values, with the dashed line indicating P < 0.05. C–F, Fold change of gene expression in selected pathways for contrasts analyzed in B.

Close modal

Taken together, these data indicate significant differences in the pathways activated in disease-associated pancreatic fibroblasts from different pathologies.

Tenascin C levels are elevated in PDAC-associated fibroblasts compared with CP-associated fibroblasts

Although this study was ongoing, in parallel research we sought differences in serum proteins from patients with CP and PDAC using a multiplex quantification of 101 cancer-associated proteins (Myriad Rules-Based Medicine's Human OncologyMAP v1.0 platform). Apart from the well-established carbohydrate antigen tumor markers, CA19-9 and CA125, only circulating TNC was significantly upregulated in PDAC compared with CP patients in discovery analysis (Fig. 4A) and in an independent validation cohort (Fig. 4B; Supplementary Table S3). We considered this finding highly significant, as circulating proteins capable of distinguishing CP from PDAC are extremely rare. Moreover, TNC is not widely expressed in adults, except in the ECM of diseased or injured tissues. Our serum data drove us to question whether TNC expression was more pronounced in fibroblasts isolated from PDAC than CP. Although not identified as significantly differentially expressed using the stringent criteria of our array analysis, our array data nonetheless indicated lower TNC in CP than in PDAC (Fig. 4C). Moreover, qRT-PCR confirmed that TNC transcripts were significantly upregulated in PDAC-associated fibroblasts compared with CP-associated fibroblasts (Fig. 4D). IHC for TNC in resected PDAC and CP specimens was performed using the same antibody we used in our serum analysis (clone 4C8MS, TNC-FN III-B) and revealed heterogeneous TNC staining, with TNC protein variously present in areas of stroma, benign ducts, and tumor cells. Notably, regions of strong staining were observed in the desmoplastic stroma of tumors even when neighboring epithelia lacked expression (Fig. 4E). TMAs from PDAC and CP cases (Supplementary Table S4) revealed more frequent TNC protein in the stromal compartments of PDAC cases compared with CP cases (60.4% compared with 13.9%, respectively, Fig. 4F). TNC-null mice exhibit attenuation of skin and lung fibrosis (29), and TNC has recently emerged as a key regulator of CAFs as part of a Twist1-Prrx1-TNC feedback loop that operates as an ON/OFF switch regulating fibroblast activation (30). We measured the ratio of αSMA to collagen, defined as the activated stroma index (Fig. 4G; ref. 31). PDAC cases with TNC+ stroma had significantly greater activated stroma index values than those lacking stromal TNC (Fig. 4H). RNA in situ hybridization for TNC mRNA showed that of 79 PDAC cases examined, 82.3% contained TNC mRNA–expressing cells in the stroma, compared with 47.2% of CP cases (n = 36; Fig. 4I and J). Together, our data indicate that fibroblasts programmed in the context of CP are distinct from those programmed in the context of PDAC, and this is exemplified by the lower expression of TNC in CP-associated fibroblasts than PDAC fibroblasts.

Figure 4.

High levels of TNC identified in primary PDAC- compared with CP-associated fibroblasts reflect levels present in tissue and blood. A, Serum levels of TNC in the discovery cohort of samples collected at the Royal Liverpool University Hospital, as measured by Myriad RBM's Human Oncology MAP. Samples were obtained from individuals with PDAC with or without obstructive jaundice (PDAC high bilirubin and PDAC low bilirubin, n = 15 and n = 20, respectively), as well as healthy individuals (n = 15) and individuals with benign biliary obstruction (n = 10) and CP (n = 15). P values determined by Mann–Whitney U test. B, Serum levels of TNC in independent samples from individuals with PDAC and high bilirubin (n = 35), PDAC and low bilirubin (n = 30), benign biliary obstruction (n = 27), and CP (n = 35), as well as healthy individuals (n = 28), as measured by ELISA. P values determined by Mann–Whitney U test. C,TNC mRNA array expression measured as Robust Multi-array Average (RMA) in discovery fibroblast isolates (PDAC, n = 6; CP, n = 5). D, qRT-PCR analysis of TNC expression in discovery and independent PDAC- (n = 15) and CP-associated (n = 8) fibroblasts. Expression was normalized to GAPDH using CP-associated fibroblasts as control. P value determined by Mann–Whitney U test. E, Representative image of TNC protein immunostaining in PDAC and CP. Scale bar, 100 μm (low magnification, left; high magnification of dashed line, right). F, Proportion of PDAC (n = 53) and CP (n = 36) patients with TNC protein expression in the fibrotic stroma, as measured on a TMA. P value determined by χ2 test. G, Example of color deconvolution of αSMA– and Masson's trichrome-stained PDAC tissue for calculation of activated stroma index. Scale bar, 100 μm. H, Activated stroma index in PDAC cases (n = 24) in G classified by TNC stromal protein expression. P value determined by Mann–Whitney U test. I, Representative image of TNC mRNA expression in PDAC stroma as determined by RNA in situ hybridization. Scale bar, 25 μm. J, Proportion of PDAC (n = 79) and CP (n = 36) patients with TNC mRNA expression in the fibrotic stroma, as measured on a TMA. P value determined by χ2 test.

Figure 4.

High levels of TNC identified in primary PDAC- compared with CP-associated fibroblasts reflect levels present in tissue and blood. A, Serum levels of TNC in the discovery cohort of samples collected at the Royal Liverpool University Hospital, as measured by Myriad RBM's Human Oncology MAP. Samples were obtained from individuals with PDAC with or without obstructive jaundice (PDAC high bilirubin and PDAC low bilirubin, n = 15 and n = 20, respectively), as well as healthy individuals (n = 15) and individuals with benign biliary obstruction (n = 10) and CP (n = 15). P values determined by Mann–Whitney U test. B, Serum levels of TNC in independent samples from individuals with PDAC and high bilirubin (n = 35), PDAC and low bilirubin (n = 30), benign biliary obstruction (n = 27), and CP (n = 35), as well as healthy individuals (n = 28), as measured by ELISA. P values determined by Mann–Whitney U test. C,TNC mRNA array expression measured as Robust Multi-array Average (RMA) in discovery fibroblast isolates (PDAC, n = 6; CP, n = 5). D, qRT-PCR analysis of TNC expression in discovery and independent PDAC- (n = 15) and CP-associated (n = 8) fibroblasts. Expression was normalized to GAPDH using CP-associated fibroblasts as control. P value determined by Mann–Whitney U test. E, Representative image of TNC protein immunostaining in PDAC and CP. Scale bar, 100 μm (low magnification, left; high magnification of dashed line, right). F, Proportion of PDAC (n = 53) and CP (n = 36) patients with TNC protein expression in the fibrotic stroma, as measured on a TMA. P value determined by χ2 test. G, Example of color deconvolution of αSMA– and Masson's trichrome-stained PDAC tissue for calculation of activated stroma index. Scale bar, 100 μm. H, Activated stroma index in PDAC cases (n = 24) in G classified by TNC stromal protein expression. P value determined by Mann–Whitney U test. I, Representative image of TNC mRNA expression in PDAC stroma as determined by RNA in situ hybridization. Scale bar, 25 μm. J, Proportion of PDAC (n = 79) and CP (n = 36) patients with TNC mRNA expression in the fibrotic stroma, as measured on a TMA. P value determined by χ2 test.

Close modal

Depletion of TNC in PDAC-associated fibroblasts promotes migration of cancer cells

We next sought to understand the functional significance, for PDAC cells, of high TNC expression in PDAC-associated fibroblasts (Fig. 5AC). We found that CM from TNC-depleted CAFs but not TNC-depleted NA fibroblasts (Fig. 5D) caused significantly greater migration of MIA PaCa-2 cells when compared with CM from control siRNA–treated fibroblasts. In addition, we found a modest 8% increase in MIA PaCa-2 cell proliferation following exposure to CM from TNC-depleted CAF but not TNC-depleted NA fibroblasts (Fig. 5E). Our data indicate that CAF-derived TNC decreases the migration of PDAC cells and has a minor negative effect on PDAC cell growth.

Figure 5.

TNC gene silencing in PDAC-associated fibroblasts leads to a diminished TGFβ1 expression and promotes migration of pancreatic cancer cells. A and B, Experimental design detailing the collection of CM from PDAC-associated fibroblasts and NA fibroblasts (A) to study the functional significance of stromal TNC on MIA PaCa-2 cells using a transwell cell migration assay (B). The images depict representative photographs of transwell migration assays conducted on MIA PaCa-2 cells with CM from PDAC-associated fibroblasts treated with nontargeting control siRNA and two different siRNAs targeting TNC (n = 7). Cell migration was quantified using QuPath. C, Representative immunoblot analysis after TNC knockdown in PDAC-associated fibroblasts and NA fibroblasts to confirm effective knockdown before cell migration or proliferation experiments were conducted. β-Actin was used as a loading control. D and E, Cell migration (D) and proliferation (E) of MIA PaCa-2 cells following treatment with CM from NA and PDAC-associated fibroblasts with siRNA-depleted TNC. Data are expressed as mean percentage of migration or proliferation relative to control ±SEM; for the migration assays, PDAC = 7 and NA = 6 independent experiments were performed, and for proliferation, PDAC = 4 and NA = 2 independent experiments were completed. P value determined by t test assuming unequal variance. F and G, Representative immunoblot analysis of TGFβ1 after TNC knockdown in PDAC-associated fibroblasts (F) and NA fibroblasts (G). Densitometry quantification of TGFβ1 normalized to β-actin is shown for each fibroblast type analyzed (n = 2). P value determined by the Student t test.

Figure 5.

TNC gene silencing in PDAC-associated fibroblasts leads to a diminished TGFβ1 expression and promotes migration of pancreatic cancer cells. A and B, Experimental design detailing the collection of CM from PDAC-associated fibroblasts and NA fibroblasts (A) to study the functional significance of stromal TNC on MIA PaCa-2 cells using a transwell cell migration assay (B). The images depict representative photographs of transwell migration assays conducted on MIA PaCa-2 cells with CM from PDAC-associated fibroblasts treated with nontargeting control siRNA and two different siRNAs targeting TNC (n = 7). Cell migration was quantified using QuPath. C, Representative immunoblot analysis after TNC knockdown in PDAC-associated fibroblasts and NA fibroblasts to confirm effective knockdown before cell migration or proliferation experiments were conducted. β-Actin was used as a loading control. D and E, Cell migration (D) and proliferation (E) of MIA PaCa-2 cells following treatment with CM from NA and PDAC-associated fibroblasts with siRNA-depleted TNC. Data are expressed as mean percentage of migration or proliferation relative to control ±SEM; for the migration assays, PDAC = 7 and NA = 6 independent experiments were performed, and for proliferation, PDAC = 4 and NA = 2 independent experiments were completed. P value determined by t test assuming unequal variance. F and G, Representative immunoblot analysis of TGFβ1 after TNC knockdown in PDAC-associated fibroblasts (F) and NA fibroblasts (G). Densitometry quantification of TGFβ1 normalized to β-actin is shown for each fibroblast type analyzed (n = 2). P value determined by the Student t test.

Close modal

Finally, given that TGFβ1 was one of the top activated upstream regulators in pancreatic disease–associated fibroblasts compared with NA (Fig. 2G), along with the well-established function of TGFβ in several key steps of tumorigenesis (32), we questioned whether TNC levels in pancreatic fibroblasts influenced TGFβ1 levels. We found that TNC depletion in CAF (Fig. 5F) but not NA (Fig. 5G) was accompanied by significant reduction in TGFβ1 expression. Taken together, our data support the notion that TNC deficiency in CAFs promotes migration of cancer cells, and that this is associated with a parallel reduction in TGFβ1 expression.

miR-137 regulates TNC expression

To explore mechanisms underlying the difference in expression of TNC in PDAC-associated stroma compared with CP stroma, an in silico analysis was performed to identify miRNAs with the potential to target the 3′ untranslated region (UTR) of TNC mRNA using http://www.microrna.org/microrna/getGeneForm.do. Of 24 candidate miRNAs, two, miR-212-3p and miR-137, tended toward downregulation in PDAC compared with CP (Fig. 6A and B), the inverse of the observed TNC mRNA levels in PDAC and CP (Fig. 4C and D). One potential binding site in the UTR of TNC mRNA was identified for hsa-miR-212-3p, and two sites were identified for hsa-miR-137 and ranked by mirSVR scoring (Fig. 6C). To test whether these miRNAs regulated the expression of TNC, PDAC-associated fibroblasts were treated with mimics of the selected candidate miRNAs. Transfection of a miR-137 mimic but not a miR-212-3p mimic significantly reduced TNC expression levels, at both the mRNA (Fig. 6D) and protein levels (Fig. 6E). A luciferase reporter assay utilizing the TNC 3′ UTR binding site for miR-137 with the best mirSVR score value (position 444; mirSVR -1.2304) confirmed that miR-137 targets this region (Fig. 6F). Notably, the decrease of luciferase activity was abolished when the TNC 3′ UTR target site was mutated. Our data point toward miRNA-mediated fine-tuning of pancreatic fibroblast gene expression contributing to distinct fibroblasts programing between CP and PDAC.

Figure 6.

TNC is modulated by miR-137 and TGFβ1 in pancreatic fibroblasts. A and B, Gene array expression data of hsa-miR-212-3p and hsa-miR-137 generated using Robust Multi-array Average (RMA) from discovery fibroblast isolates (PDAC = 6; CP = 5). Data are shown as mean ± SEM. C, Potential binding sites (bases in bold) of hsa-miR-212-3p and hsa-miR-137 predicted in the 3′UTR of TNC mRNA, alongside their mirSVR score. The more negative the score, the greater the predicted effect on mRNA. Lines represent complementary base pairing, whereas the gray shading depicts the seed sequence of the miRNAs shown. D, qRT-PCR analysis of TNC expression in immortalized PDAC-associated fibroblasts transfected with miR-212-3p mimics (30 nmol/L), miR-137 mimics (30 nmol/L), or off-target mimics (30 nmol/L) for 48 hours, plus untreated cells. Expression was normalized to GAPDH using off-target siRNA as control. Data are shown as mean ±SEM; n = 2 independent experiments; P value determined by paired t test. E, Western blot of TNC in resin-concentrated CM and cell lysates from experiment shown in D. F, Normalized luciferase activity of TNC 3′-UTR cloned into pGL3-Control vector containing either wild-type (WT) or mutant (MUT) binding sites of hsa-miR-137 (Supplementary Fig. 2G) after cotransfection with miR-137 mimics and off-target negative control in HeLa cells. Data are fold changes of Firefly pGL3-Control constructs/Renilla pRL-SV40 vector activity ratio normalized to off-target control ±SEM; n = 2 independent experiments; P value determined by paired t test. G and H, qRT-PCR for TNC (G) and mir-137 (H) after incubation of PDAC-associated and NA fibroblasts with recombinant human TGFβ1 for 48 hours. Expression levels were compared with control (DMEM) and normalized to GAPDH and RNU6-2, respectively. Error bars depict mean ± SEM; n = 3 independent experiments. P value determined by t test assuming unequal variance.

Figure 6.

TNC is modulated by miR-137 and TGFβ1 in pancreatic fibroblasts. A and B, Gene array expression data of hsa-miR-212-3p and hsa-miR-137 generated using Robust Multi-array Average (RMA) from discovery fibroblast isolates (PDAC = 6; CP = 5). Data are shown as mean ± SEM. C, Potential binding sites (bases in bold) of hsa-miR-212-3p and hsa-miR-137 predicted in the 3′UTR of TNC mRNA, alongside their mirSVR score. The more negative the score, the greater the predicted effect on mRNA. Lines represent complementary base pairing, whereas the gray shading depicts the seed sequence of the miRNAs shown. D, qRT-PCR analysis of TNC expression in immortalized PDAC-associated fibroblasts transfected with miR-212-3p mimics (30 nmol/L), miR-137 mimics (30 nmol/L), or off-target mimics (30 nmol/L) for 48 hours, plus untreated cells. Expression was normalized to GAPDH using off-target siRNA as control. Data are shown as mean ±SEM; n = 2 independent experiments; P value determined by paired t test. E, Western blot of TNC in resin-concentrated CM and cell lysates from experiment shown in D. F, Normalized luciferase activity of TNC 3′-UTR cloned into pGL3-Control vector containing either wild-type (WT) or mutant (MUT) binding sites of hsa-miR-137 (Supplementary Fig. 2G) after cotransfection with miR-137 mimics and off-target negative control in HeLa cells. Data are fold changes of Firefly pGL3-Control constructs/Renilla pRL-SV40 vector activity ratio normalized to off-target control ±SEM; n = 2 independent experiments; P value determined by paired t test. G and H, qRT-PCR for TNC (G) and mir-137 (H) after incubation of PDAC-associated and NA fibroblasts with recombinant human TGFβ1 for 48 hours. Expression levels were compared with control (DMEM) and normalized to GAPDH and RNU6-2, respectively. Error bars depict mean ± SEM; n = 3 independent experiments. P value determined by t test assuming unequal variance.

Close modal

To determine whether TGFβ1 controls the miR-137-TNC axis, we treated CAF and NA with TGFβ1 and measured the expression of both TNC and miR-137. We observed a significant increase in TNC expression in both CAF and NA following TGFβ1 treatment (Fig. 6G). Moreover in NA, TGFβ1 treatment was accompanied by significant downregulation in the expression of miR-137 (Fig. 6H), demonstrating that TGFβ can control this axis. Downregulation of miR-137 was not observed in CAFs following treatment by TGFβ1. Thus, it would appear that other factors may also underpin the regulation of this axis.

In the present study, we show the first direct comparison of pancreatic fibroblasts isolated from distinct pancreatic diseases and provide evidence that pancreatic fibroblasts are programmed by their unique disease-specific environments and that their distinct gene expression persists following isolation. The challenges associated with isolating and characterizing human primary pancreatic fibroblasts has led to this fundamental question remaining unanswered until now. Although all of the disease-associated fibroblasts compared in our study exhibited myofibroblast-like morphology and expressed characteristic markers of activation, less than one quarter of differentially expressed genes were commonly shared between disease groups, compared with NA. When comparisons were restricted to disease-associated fibroblasts only, less than 1% of differentially expressed genes were commonly altered across disease types. This argues against a single common phenotype of pancreatic fibroblast activation and emphasizes the importance of using appropriate fibroblasts to address disease-specific research questions. Similarly, all of the primary fibroblasts isolated for this study, including NA, were subjected to culture, prior to RNA extraction for gene profiling. Although one might expect this to have a “normalizing” effect, isolated fibroblasts exhibited distinct characteristics of the disease group to which they belonged. Notably, NA were more responsive than PDAC-associated fibroblasts to the antifibrotic effects of NPPB (13, 23, 24), showing large reductions of the fibroblast activation marker αSMA. Similarly, tumor fibroblasts responded with greater intensity than NA to the activation stimuli of TGFβ and WNT signaling.

The global impact of miRNA regulation in quiescent versus activated PSCs has thus far been performed in rats only (33). Our comparison of primary human NA fibroblasts with human disease–activated fibroblasts revealed only a small proportion of differentially expressed miRNAs, with no difference found between PDAC and CP or PDAC and PAT. This reflects a tight regulation of miRNAs across pancreatic disease fibroblasts in contrast to the high number of mRNA transcripts differentially regulated.

The activation of PAT-associated fibroblasts was distinctly different from that of PDAC-associated fibroblasts as indicated by the upregulation of genes associated with the Hepatic Stellate Cell Activation Pathway in PDAC-associated fibroblasts compared with PAT- but not CP-associated fibroblasts. Ampullary and duodenal cancer–associated fibroblasts are hardly characterized and warrant further study. Overlap observed between PDAC- and CP-associated fibroblasts may reflect the fact that both diseases can present concurrently, and that fibroblasts in the context of PDAC are educated by both the tumor and CP. Nonetheless, there were significant differences between PDAC- and CP-derived fibroblasts, as exemplified by TNC. The higher TNC transcript and protein levels in PDAC-derived fibroblasts and PDAC resected specimens compared with CP and the positive link between TNC and activated stroma suggests enhanced fibroblast activation in PDAC versus CP in our study.

The consequences, however, of high fibroblast levels of TNC on cancer cell behavior have hardly been examined to date. Our finding that CM derived from TNC-depleted PDAC-associated fibroblasts enhanced the motility of MIA PaCa-2 cells is intriguing, and implies that the high levels of TNC in PDAC-associated fibroblasts may serve to repel tumoral cells. Consistent with this, TNC knockout in breast carcinomas allowed the attraction of macrophages (34).

Uncovering miR-137 as a regulator of TNC in disease-associated fibroblasts provides a potential candidate for targeting TNC. Previous studies have reported downregulation of miR-137 in pancreatic cancer tissues compared with their normal counterpart with its overexpression leading to inhibition of cancer cell invasion and promotion of senescence (35, 36). In a study of acute lung injury, TNC-null lung fibroblasts exhibited impaired responsiveness to TGFβ (37). Our findings that TNC depletion diminishes TGFβ1 expression in pancreatic CAF while TGFβ1 stimulation of CAFs and NA increased TNC expression evidence the important interplay between TNC and TGFβ1 in pancreatic fibrosis. Finally, the identification of TNC as a biomarker with potential to distinguish PDAC from CP warrants further investigation as such circulating biomarkers are lacking.

In conclusion, we present evidence that in vivo programming of human fibroblasts is disease-specific and is, at least to an important extent, maintained in culture. Single-cell technologies have helped elucidate how stromal CAFs contribute to proliferative and invasive transcriptional programs of PDAC cells, with CAF-secreted TGFβ identified as a key mediator of PDAC cell heterogeneity (38). Moreover, the heterogeneity of human PDAC-associated fibroblasts themselves has been described, with CAFs exhibiting myofibroblast (myCAF), inflammatory (iCAF), or antigen presenting (apCAF) phenotypes (39, 40). Future work addressing specific subpopulations of pancreatic fibroblasts from different pancreatic diseases is now required. The prevailing perception that activated pancreatic fibroblasts are a uniform entity that can be used interchangeably in research requires revision. Furthermore, the development of treatments should take into consideration disease-specific features of activated pancreatic fibroblasts.

No potential conflicts of interest were disclosed.

Conception and design: L.N. Barrera, A. Evans, B. Lane, M. Jalali, Q. Nunes, C. Halloran, W. Greenhalf, J.P. Neoptolemos, E. Costello

Development of methodology: L.N. Barrera, A. Evans, B. Lane, S. Brumskill, F.E. Oldfield, P.A. Perez-Mancera, T. Liloglou, M. Jalali, Q. Nunes, P.A. Phillips, J.P. Neoptolemos, E. Costello

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): L.N. Barrera, A. Evans, S. Brumskill, F.E. Oldfield, F. Campbell, T. Andrews, Z. Lu, T. Liloglou, M. Ashworth, M. Jalali, Q. Nunes, J.F. Timms, C. Halloran, J.P. Neoptolemos, E. Costello

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): L.N. Barrera, A. Evans, B. Lane, S. Brumskill, F. Campbell, T. Andrews, Z. Lu, T. Liloglou, M. Ashworth, R. Dawson, P.A. Phillips, W. Greenhalf, E. Costello

Writing, review, and/or revision of the manuscript: L.N. Barrera, A. Evans, S. Brumskill, F. Campbell, T. Andrews, P.A. Perez-Mancera, Q. Nunes, P.A. Phillips, J.F. Timms, C. Halloran, W. Greenhalf, J.P. Neoptolemos, E. Costello

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): L.N. Barrera, A. Evans, B. Lane, E. Costello

Study supervision: J.P. Neoptolemos, E. Costello

Others (Bioinformatics): B. Lane

This work was funded by the Pancreatic Cancer Research Fund to E. Costello; Rosetrees Trust to E. Costello and L.N. Barrera (M384-F1); Pancreatic Cancer UK to E. Costello, L.N. Barrera, and A. Evans (2011_Grant-Costello/Timms and RIF2014_03_Costello); the National Institute for Health Research Liverpool Pancreas Biomedical Research Unit to E. Costello; North West Cancer Research, UK to E. Costello (CR1142); and Cancer Research UK to E. Costello. Researchers at UCL were supported by the NIHR University College London Hospitals (UCLH) Biomedical Research Centre to J.F. Timms. The authors would like to thank Dr. Lucille Rainbow at the Centre for Genomic Research for her assistance with microarray preparation, Dr. Nicholas Harper for providing essential reagents and help with the luciferase reporter construct, and Luke Taylor and Ben Crosby for their help in searching medical records. We would also like to thank Katie Bullock, Dr. Amelia Acha-Sagredo, Luke Wilkinson, Roberta Sanna, Dr. Neal Rimmer, Elizabeth Garner, and Dr. Katharine Hand for excellent technical assistance.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1.
Evans
A
,
Costello
E
. 
The role of inflammatory cells in fostering pancreatic cancer cell growth and invasion
.
Front Physiol
2012
;
3
:
270
.
2.
Pang
TCY
,
Wilson
JS
,
Apte
MV
. 
Pancreatic stellate cells: what's new?
Curr Opin Gastroenterol
2017
;
33
:
366
73
.
3.
Xu
Z
,
Vonlaufen
A
,
Phillips
PA
,
Fiala-Beer
E
,
Zhang
X
,
Yang
L
, et al
Role of pancreatic stellate cells in pancreatic cancer metastasis
.
Am J Pathol
2010
;
177
:
2585
96
.
4.
Kalluri
R
. 
The biology and function of fibroblasts in cancer
.
Nat Rev Cancer
2016
;
16
:
582
98
.
5.
Hessmann
E
,
Patzak
MS
,
Klein
L
,
Chen
N
,
Kari
V
,
Ramu
I
, et al
Fibroblast drug scavenging increases intratumoural gemcitabine accumulation in murine pancreas cancer
.
Gut
2018
;
67
:
497
507
.
6.
Rhim
AD
,
Oberstein
PE
,
Thomas
DH
,
Mirek
ET
,
Palermo
CF
,
Sastra
SA
, et al
Stromal elements act to restrain, rather than support, pancreatic ductal adenocarcinoma
.
Cancer Cell
2014
;
25
:
735
47
.
7.
Ozdemir
BC
,
Pentcheva-Hoang
T
,
Carstens
JL
,
Zheng
X
,
Wu
CC
,
Simpson
TR
, et al
Depletion of carcinoma-associated fibroblasts and fibrosis induces immunosuppression and accelerates pancreas cancer with reduced survival
.
Cancer Cell
2014
;
25
:
719
34
.
8.
Binkley
CE
,
Zhang
L
,
Greenson
JK
,
Giordano
TJ
,
Kuick
R
,
Misek
D
, et al
The molecular basis of pancreatic fibrosis: common stromal gene expression in chronic pancreatitis and pancreatic adenocarcinoma
.
Pancreas
2004
;
29
:
254
63
.
9.
Erkan
M
,
Weis
N
,
Pan
Z
,
Schwager
C
,
Samkharadze
T
,
Jiang
X
, et al
Organ-, inflammation- and cancer specific transcriptional fingerprints of pancreatic and hepatic stellate cells
.
Mol Cancer
2010
;
9
:
88
.
10.
Bachem
MG
,
Schneider
E
,
Gross
H
,
Weidenbach
H
,
Schmid
RM
,
Menke
A
, et al
Identification, culture, and characterization of pancreatic stellate cells in rats and humans
.
Gastroenterology
1998
;
115
:
421
32
.
11.
Vonlaufen
A
,
Phillips
PA
,
Yang
L
,
Xu
Z
,
Fiala-Beer
E
,
Zhang
X
, et al
Isolation of quiescent human pancreatic stellate cells: a promising in vitro tool for studies of human pancreatic stellate cell biology
.
Pancreatology
2010
;
10
:
434
43
.
12.
Sherman
MH
,
Yu
RT
,
Engle
DD
,
Ding
N
,
Atkins
AR
,
Tiriac
H
, et al
Vitamin D receptor-mediated stromal reprogramming suppresses pancreatitis and enhances pancreatic cancer therapy
.
Cell
2014
;
159
:
80
93
.
13.
Kapoun
AM
,
Liang
F
,
O'Young
G
,
Damm
DL
,
Quon
D
,
White
RT
, et al
B-type natriuretic peptide exerts broad functional opposition to transforming growth factor-beta in primary human cardiac fibroblasts: fibrosis, myofibroblast conversion, proliferation, and inflammation
.
Circ Res
2004
;
94
:
453
61
.
14.
Willert
K
,
Brown
JD
,
Danenberg
E
,
Duncan
AW
,
Weissman
IL
,
Reya
T
, et al
Wnt proteins are lipid-modified and can act as stem cell growth factors
.
Nature
2003
;
423
:
448
52
.
15.
Bankhead
P
,
Loughrey
MB
,
Fernandez
JA
,
Dombrowski
Y
,
McArt
DG
,
Dunne
PD
, et al
QuPath: open source software for digital pathology image analysis
.
Sci Rep
2017
;
7
:
16878
.
16.
Endo
S
,
Nakata
K
,
Ohuchida
K
,
Takesue
S
,
Nakayama
H
,
Abe
T
, et al
Autophagy is required for activation of pancreatic stellate cells, associated with pancreatic cancer progression and promotes growth of pancreatic tumors in mice
.
Gastroenterology
2017
;
152
:
1492
506
e24
.
17.
Bi
Y
,
Mukhopadhyay
D
,
Drinane
M
,
Ji
B
,
Li
X
,
Cao
S
, et al
Endocytosis of collagen by hepatic stellate cells regulates extracellular matrix dynamics
.
Am J Physiol Cell Physiol
2014
;
307
:
C622
33
.
18.
Bauer
AS
,
Nazarov
PV
,
Giese
NA
,
Beghelli
S
,
Heller
A
,
Greenhalf
W
, et al
Transcriptional variations in the wider peritumoral tissue environment of pancreatic cancer
.
Int J Cancer
2018
;
142
:
1010
21
.
19.
Egberts
JH
,
Cloosters
V
,
Noack
A
,
Schniewind
B
,
Thon
L
,
Klose
S
, et al
Anti-tumor necrosis factor therapy inhibits pancreatic tumor growth and metastasis
.
Cancer Res
2008
;
68
:
1443
50
.
20.
Arandkar
S
,
Furth
N
,
Elisha
Y
,
Nataraj
NB
,
van der Kuip
H
,
Yarden
Y
, et al
Altered p53 functionality in cancer-associated fibroblasts contributes to their cancer-supporting features
.
Proc Natl Acad Sci U S A
2018
;
115
:
6410
5
.
21.
Bhowmick
NA
,
Chytil
A
,
Plieth
D
,
Gorska
AE
,
Dumont
N
,
Shappell
S
, et al
TGF-beta signaling in fibroblasts modulates the oncogenic potential of adjacent epithelia
.
Science
2004
;
303
:
848
51
.
22.
Walter
K
,
Omura
N
,
Hong
SM
,
Griffith
M
,
Vincent
A
,
Borges
M
, et al
Overexpression of smoothened activates the sonic hedgehog signaling pathway in pancreatic cancer-associated fibroblasts
.
Clin Cancer Res
2010
;
16
:
1781
9
.
23.
Suganami
T
,
Mukoyama
M
,
Sugawara
A
,
Mori
K
,
Nagae
T
,
Kasahara
M
, et al
Overexpression of brain natriuretic peptide in mice ameliorates immune-mediated renal injury
.
J Am Soc Nephrol
2001
;
12
:
2652
63
.
24.
Sonoyama
T
,
Tamura
N
,
Miyashita
K
,
Park
K
,
Oyamada
N
,
Taura
D
, et al
Inhibition of hepatic damage and liver fibrosis by brain natriuretic peptide
.
FEBS Lett
2009
;
583
:
2067
70
.
25.
Vogelmann
R
,
Ruf
D
,
Wagner
M
,
Adler
G
,
Menke
A
. 
Effects of fibrogenic mediators on the development of pancreatic fibrosis in a TGF-beta1 transgenic mouse model
.
Am J Physiol Gastrointest Liver Physiol
2001
;
280
:
G164
72
.
26.
Akhmetshina
A
,
Palumbo
K
,
Dees
C
,
Bergmann
C
,
Venalis
P
,
Zerr
P
, et al
Activation of canonical wnt signalling is required for TGF-beta-mediated fibrosis
.
Nat Commun
2012
;
3
:
735
.
27.
Tsukada
S
,
Westwick
JK
,
Ikejima
K
,
Sato
N
,
Rippe
RA
. 
SMAD and p38 MAPK signaling pathways independently regulate alpha1(I) collagen gene expression in unstimulated and transforming growth factor-beta-stimulated hepatic stellate cells
.
J Biol Chem
2005
;
280
:
10055
64
.
28.
Wang
Y
,
Huang
Y
,
Guan
F
,
Xiao
Y
,
Deng
J
,
Chen
H
, et al
Hypoxia-inducible factor-1alpha and MAPK co-regulate activation of hepatic stellate cells upon hypoxia stimulation
.
PLoS One
2013
;
8
:
e74051
.
29.
Bhattacharyya
S
,
Wang
W
,
Morales-Nebreda
L
,
Feng
G
,
Wu
M
,
Zhou
X
, et al
Tenascin-C drives persistence of organ fibrosis
.
Nat Commun
2016
;
7
:
11703
.
30.
Yeo
SY
,
Lee
KW
,
Shin
D
,
An
S
,
Cho
KH
,
Kim
SH
. 
A positive feedback loop bi-stably activates fibroblasts
.
Nat Commun
2018
;
9
:
3016
.
31.
Erkan
M
,
Michalski
CW
,
Rieder
S
,
Reiser-Erkan
C
,
Abiatari
I
,
Kolb
A
, et al
The activated stroma index is a novel and independent prognostic marker in pancreatic ductal adenocarcinoma
.
Clin Gastroenterol Hepatol
2008
;
6
:
1155
61
.
32.
David
CJ
,
Massague
J
. 
Contextual determinants of TGFbeta action in development, immunity and cancer
.
Nat Rev Mol Cell Biol
2018
;
19
:
419
35
.
33.
Masamune
A
,
Nakano
E
,
Hamada
S
,
Takikawa
T
,
Yoshida
N
,
Shimosegawa
T
. 
Alteration of the microRNA expression profile during the activation of pancreatic stellate cells
.
Scand J Gastroenterol
2014
;
49
:
323
31
.
34.
Talts
JF
,
Wirl
G
,
Dictor
M
,
Muller
WJ
,
Fassler
R
. 
Tenascin-C modulates tumor stroma and monocyte/macrophage recruitment but not tumor growth or metastasis in a mouse strain with spontaneous mammary cancer
.
J Cell Sci
1999
;
112
:
1855
64
.
35.
Neault
M
,
Mallette
FA
,
Richard
S
. 
miR-137 modulates a tumor suppressor network-inducing senescence in pancreatic cancer cells
.
Cell Rep
2016
;
14
:
1966
78
.
36.
Xiao
J
,
Peng
F
,
Yu
C
,
Wang
M
,
Li
X
,
Li
Z
, et al
microRNA-137 modulates pancreatic cancer cells tumor growth, invasion and sensitivity to chemotherapy
.
Int J Clin Exp Pathol
2014
;
7
:
7442
50
.
37.
Carey
WA
,
Taylor
GD
,
Dean
WB
,
Bristow
JD
. 
Tenascin-C deficiency attenuates TGF-β-mediated fibrosis following murine lung injury
.
Am J Physiol Lung Cell Mol Physiol
2010
;
299
:
L785
93
.
38.
Ligorio
M
,
Sil
S
,
Malagon-Lopez
J
,
Nieman
LT
,
Misale
S
,
Di Pilato
M
, et al
Stromal microenvironment shapes the intratumoral architecture of pancreatic cancer
.
Cell
2019
;
178
:
160
75
e27
.
39.
Elyada
E
,
Bolisetty
M
,
Laise
P
,
Flynn
WF
,
Courtois
ET
,
Burkhart
RA
, et al
Cross-species single-cell analysis of pancreatic ductal adenocarcinoma reveals antigen-presenting cancer-associated fibroblasts
.
Cancer Discov
2019
;
9
:
1102
23
.
40.
Ohlund
D
,
Handly-Santana
A
,
Biffi
G
,
Elyada
E
,
Almeida
AS
,
Ponz-Sarvise
M
, et al
Distinct populations of inflammatory fibroblasts and myofibroblasts in pancreatic cancer
.
J Exp Med
2017
;
214
:
579
96
.