Impact of the secretome of activated pancreatic stellate cells on growth and differentiation of pancreatic tumour cells

Pancreatic ductal adenocarcinoma (PDAC) exists in a complex desmoplastic microenvironment. As part of it, pancreatic stellate cells (PSCs) provide a fibrotic niche, stimulated by a dynamic communication between activated PSCs and tumour cells. Investigating how PSCs contribute to tumour development and for identifying proteins that the cells secrete during cancer progression, we studied by means of complex antibody microarrays the secretome of activated PSCs. A large number of secretome proteins were associated with cancer-related functions, such as cell apoptosis, cellular growth, proliferation and metastasis. Their effect on tumour cells could be confirmed by growing tumour cells in medium conditioned with activated PSC secretome. Analyses of the tumour cells’ proteome and mRNA revealed a strong inhibition of tumour cell apoptosis, but promotion of proliferation and migration. Many cellular proteins that exhibited variations were found to be under the regulatory control of eukaryotic translation initiation factor 4E (eIF4E), whose expression was triggered in tumour cells grown in the secretome of activated PSCs. Inhibition by an eIF4E siRNA blocked the effect, inhibiting tumour cell growth in vitro. Our findings show that activated PSCs acquire a pro-inflammatory phenotype and secret proteins that stimulate pancreatic cancer growth in an eIF4E-dependent manner, providing further insight into the role of stromal cells in pancreatic carcinogenesis and cancer progression.

experimental design and statistical rationale. The protein content of the secretome of activated and non-activated PSCs was studied by incubations on complex antibody microarrays. From the observed binding patterns, significant differences were defined and applied to predict functional changes that the secretome from activated PSCs could trigger in recipient tumour cells. The prediction was followed up by experiments, analysing in detail the changes occurring at the protein level within tumour cells grown in presence of secretome, again by means of antibody microarrays. These intracellular protein level variations were utilised to identify affected Cancer cell growth in conditioned medium. PT45P1 cells were incubated with serum-free IMDM medium (SFM) for 24 h. SFM was removed and PT45P1 cells were replenished with new serum-and phenol-red-free IMDM alone (control) or medium supplemented with either 100 µl concentrated supernatant from PSCs treated with TNF-α (activated PSCs) or untreated PSCs (non-activated PSC). After growth for 48 h, the PT45P1 cells were washed three times with ice-cold PBS and subjected to protein extraction as described below. All experiments were done in quintuplicates. Additionally, to insure that incubation for 48 h in serum-free medium was not affecting cell viability, cell growth was monitored at 24, 48 and 72 h for PSCs and PT45P1 cells grown in serum-free media (negative control). Viability of cells was 95% or higher at all tested time points. protein extraction. Protein extraction was performed as described previously 20 . In brief, after washing with ice-cooled Dulbecco's modified phosphate buffered saline (DPBS, Thermo Scientific), the PT45P1 cells were layered with a minimal volume of extraction/labelling solution composed of 50 mM bicine buffer, pH 8.5, 20% glycerol, 1.0 mM MgCl 2 , 5.0 mM EDTA, 1.0 mM phenylmethanesulfonyl fluoride, 1.0 IU/ml benzonase (Merck Bioscinces, Schwalbach, Germany), Halt Protease and Phosphatase Inhibitor Cocktail (Thermo Scientific), 0.5% Nonidet P-40 substitute, 1.0% cholic acid, 0.25% n-dodecyl-β-maltoside (GenaXXon Bioscience, Ulm, Germany), and 0.5% amidosulfobetaine-14. Flasks were kept on ice for 30 min with occasional shaking. Cell lysates were collected with a cell scrapper followed by centrifugation at 20,000 g and 4 °C for 20 min. The supernatant was collected and the protein concentration was determined with the Bicinchoninic Acid Protein Assay Reagent kit (Thermo Scientific).
Protein labelling. Protein labelling was performed as described 21,22 . In brief, the protein concentration was adjusted to 1.0 mg/ml for the cell secretomes and to 2.0 mg/ml for cell lysates. The fluorescent NHS-ester dyes DY-649 or DY-549 (Dyomics, Jena, Germany) were then added to each sample at a molar ratio dye/protein of 7.5. Antibody microarray. The microarrays were produced as described in much detail earlier 21,22 . Briefly, the antibodies were spotted from 384-well plates onto epoxysilane-coated slides (Nexterion-E; Schott, Jena, Germany) using the contact printer MicroGrid-2 (BioRobotics, Cambridge, UK) and SMP3B pins (Telechem, Sunnyvale, USA) at a humidity of 50-60%. The spotting buffer was prepared by generating 15 µl in each well of 50 mM bicine buffer, pH 8.5, containing 1.0 mM MgCl 2 , 5% trehalose, 0.005% Tween-20 and 5 µg of the respective antibody. On each slide, the antibodies were printed in quadruplicates. After printing, the slides were equilibrated at a humidity of 50-60% overnight and stored in dry and dark conditions at 4 °C until use.
Protein profiling. The protein samples were investigated on antibody microarrays as described in detail before 23 . All steps were performed in the dark. Before incubation with labelled samples, slides were washed twice (for 10 and 5 min, respectively) with PBS containing 0.05% Tween-20 (PBST). Following washing, the slides were blocked with 10% non-fat dry milk (Biorad, Munich, Germany) in PBST at room temperature for 3 h. Each slide was incubated with 35 μg of DY-649 labelled sample and 35 μg of a common reference sample made of the pooled protein extracts of 23 PDAC cell lines labelled with DY-549 (Suppl. Table. S2) 19 . Incubation was done using Quadriperm chambers (Greiner Bio-one, Frickenhausen, Germany) in 5 ml of PBST supplemented with 10% milk at 4 °C overnight. The slides were washed 4 times for 5 min each with PBST, rinsed with deionized water and dried in a ventilated oven at room temperature. Scanning of the slides was performed using a Tecan power scanner (Tecan, Grödig, Austria) at constant laser power and PMT. Image acquisition and analysis were performed with GenePix Pro 6.0 software (Molecular Devices, Sunnyvale, USA), generating numerical values of signal intensities. Data analysis. The signal intensities from microarray experiments were analysed with the Chipster software 24 . The median of signal intensity with local background for each spot at both red (DY-649) and green (DY-549) channels was used to generate ratios. Ratios were normalized using the Loess method with background correction offset [0, 50] of the normexp method 25 . Test of significance between control and treatment groups was performed using the Empirical Bayes test with Bonferroni-Hochberg adjustment of p-values 26 . The empirical Bayes make use of a moderated t-statistic in which posterior residual standard deviations are applied rather www.nature.com/scientificreports www.nature.com/scientificreports/ than ordinary standard deviations, which give a far more stable inference when the number of arrays is small 26 . A p-value of 0.05 or less was considered significant. Multiple-set Venn diagrams were generated using the open-source software VENNTURE 27 . The bio-functional annotation of the differentially expressed proteins was performed with the Ingenuity Pathways Analysis (IPA) software (version 6.3; Ingenuity Systems, Redwood City, USA). Prediction of variations in biological functions was performed using a z-score of +2 or −2, respectively, as threshold for significance. Protein functional interaction networks were evaluated using the open-source software STRING 9.0 28 . For the proliferation assay, unpaired student t-test (two-tailed) was used to determine the significance of differences between the control (serum-free incubations) and each of the other treatments. The inter-and intra-assay coefficient of variance (CV) was always less than 20%.
Cell transfection. We used the siRNA gene silencer system (siRNA #6554) as well as a control siRNA (#6568) of Cell Signaling Technology (Danvers, USA) to perform the EIF4E gene silencing in the pancreatic cancer cell lines PT45P1, Panc-1 and Capan-1 according to the manufacturer's protocol. Briefly, RNA transfections were carried out in 6-well or 96-well plates using siPORT NeoFXTM (Ambion, Carlsbad, USA) reagent. siPORTTM NeoFXTM transfection agent and the RNA molecules were mixed and distributed on the culture plates and overlaid with the cells. The final transfection volume in a 6-well plate was 2.5 ml of medium containing 2 × 10 5 cells per well; in 96-well plates, it was 100 μl of medium containing 5 × 10 3 cells per well. The final concentration of the RNA molecules transfected was 100 nM. After this procedure, the plates were maintained at 37 °C and 5% CO 2 . After 48 h, cells were serum-starved overnight and either left untreated or treated with activated PSC secretome for 24 h.

eLIsA.
To determine the concentration of fibronectin and collagen, 100 μl PSC culture supernatant (20 µg/ ml) were coated onto 96-well microtiter plates (Nunc-Maxi Sorp, Langenselbold, Germany) in five replicate experiments and incubated overnight at 4 °C. Subsequently, the plates were blocked with 5% non-fat milk in PBST for 3 h prior to an incubation overnight at 4 °C with polyclonal rabbit-anti-human-collagen type I (Biomol, Hamburg, Germany) or polyclonal rabbit-anti-human fibronectin antibody. Wells were washed with PBST and incubated with HRP-conjugated secondary antibody (Santa Cruz Biotechnology, Germany). Antibody complexes were detected with the peroxidase substrate SureBlue TMB (KPL, Gaithersburg, Germany). Plates were read on a standard plate reader at 540 nm.
Proliferation assay. Changes in proliferation rates of cancer cells treated with PSC secretome were determined using the CyQUANT NF Cell Proliferation Assay kit (Invitrogen, Schwerte, Germany). The assay is based on the measurement of cellular DNA content via binding of a fluorescent dye. PT45P1 were seeded in 96-well flat bottom tissue culture plates at a density of approximately 5000 cells/well and allowed to attach for 5 h. Then cells were washed twice with PBS and incubated in SFM for 24 h. SFM was removed and cells were incubated for 48 h with SFM alone or medium conditioned with concentrated secretome of TNF-treated PSCs. After 48 h incubation, the medium was removed and cell proliferation was determined according to the manufacturer's recommendations. Briefly, 100 µl of the CyQUANT NF reagent was added to each well, and the plate was incubated at 37 °C for 1 h. Absorbance measurements were done at an excitation of 485 nm and emission at 530 nm. For transfected cells, cell proliferation was quantitated by using the cellTiter-Glo Luminescent cell viability assay kit (Promega, Madison, USA). Briefly, 5 × 10 3 PT45P1 and Panc1 cells were plated in 96-well plates and allowed to adhere. After transfected with siRNA and treated as above for another 24 h, cell proliferation was determined according to the instructions of the manufacturer.
Apoptosis assay. The effect of various conditions on PT45P1 cell apoptosis was conducted by measurement of a mitochondrial membrane potential assay using the Mito-ID Membrane Potential kit of Enzo Life Sciences (Lörrach, Germany). Briefly, PT45P1 cells were seeded at 2 × 10 4 cells per well and allowed to attach overnight. After that, cells were treated either with conditioned medium supplemented with supernatant from activated PSCs or inactivated PSCs, respectively. After 48 h, the Mito-ID Membrane Potential assay was used according to the manufacturer's instructions. A decrease in mitochondrial membrane potential is considered to be a function of accelerated apoptotic activity. For transfected cells, cell apoptosis was assessed using the Caspase-Glo3/7 assay kit (Promega) according to the manufacture's instruction. Briefly, after Panc-1 and PT45P1 cells had been transfected and treated as above, the dye solution was added to cells and incubation continued at 37 °C for 30 min. Cell apoptosis was determined by measuring the differences in luminescence.
Real-time qpCR. Analysis of mRNA expressions was accomplished by RT-qPCR using primers from Qiagen (Hilden, Germany). Total RNA was extracted from PT45P1 cells using the Trizol Reagent (Invitrogen) following  Table 3. Expression of genes grouped according to biological role. www.nature.com/scientificreports www.nature.com/scientificreports/ the manufacturer's instructions. The cDNA was synthesised using the protoscript first strand cDNA synthesis kit (Cell Signaling Technologies). A total of 2 mg of total RNA was reverse transcribed using a SYBR Green PCR Kit (Invitrogen). For all steps, the samples were kept on ice. Following the guidelines of the manufacturer, PCR was performed at 95 °C for 10 min, followed by 50 cycles of 15 s at 95 °C and 1 min at 60 °C. The reactions were performed in triplicate.

Results
The effect of TNF-α on the psC secretome. We had previously examined the effect of pro-inflammatory factors on the biological functions of PSCs, taking advantage of the very antibody microarray utilised here. TNF-α was identified as the prime factor responsible for PSC activation 6 . Consequently, we used TNF-α in this study accordingly. The secretome released by TNF-α activated PSCs was collected and analysed on antibody microarrays. To this end, the growth media of activated and non-activated PSCs were collected individually in three independent technical replicates each. Only serum-free media were used for the experimental set-up in order to avoid any interference from growth factors and cytokines contained in added serum. Even low concentrations would be detected by the high sensitivity of the affinity-based antibody microarray assay. In addition, signal variations could have arisen from the effect of serum-soluble factors on the PSCs. In the secretome of TNF-α treated PSCs, 309 secreted proteins were found to be differentially abundant in the cell culture medium compared to samples from non-activated cells (Suppl. Table. S3). Only 18% of the identified proteins were known to be secreted via classical secretory mechanism, and 17% were categorized as plasma membrane proteins (Suppl. Fig. 1). Particular protein abundance variations detected by the antibody microarray analyses were validated by Western blot or ELISA, or both. In accordance with the microarrays data, the levels of proteins SERPINE, IL-4, IL-1ß, and FGF-1, for example, were found to be increased in the secretome of activated PSCs. The same was observed for ECM proteins, such as FN-1 and Col-1 (Fig. 2), which are known to be preferentially expressed by activated PSCs and reported to be involved functionally in cell proliferation and angiogenesis 29   www.nature.com/scientificreports www.nature.com/scientificreports/ proteins in cancer and tumour-associated fibroblast processes as well as a close association with cancer-related functions, such as cell apoptosis, cellular growth and proliferation (Table 1). In addition, cell mobility aspects were strongly represented, such as metastasis, cellular movement and adhesion of eukaryotic cells. Many ECM proteins were regulated in the secretome of TNF-α activated PSCs, including matrix-metalloproteases (MMPs) and tissue inhibitor of metalloproteases (TIMPs). Examples are FN-1, MMP-10, MMP-11 and MMP-14, whose  and Panc-1 were transfected with 100 nM of an eIF4E-specific siRNA or a control siRNA of scrambled sequence for 48 h, followed by serum starvation overnight. Subsequently, they were grown in presence or absence of activated PSC secretome for 24 h. The effect of the various conditions on eIF4E is shown, determined by Western blotting. Also, the results of caspase-3/7 assays are shown for each column (c1 to c6). Note: the two lanes c1 and c2 are exchanged in the Western blots generated from Panc-1 cells.
www.nature.com/scientificreports www.nature.com/scientificreports/ concentrations were significantly elevated. To predict potential activities of the secreted proteins upon uptake by recipient tumour cells, a network map was constructed on the basis of the secretome protein profile using the IPA software (Suppl. Fig. 2). The predicted cellular functions triggered by the protein variations observed in the PSC secretome include activation and proliferation of fibroblasts, proliferation and migration of tumour cells as well as invasion. In addition, apoptosis of cancer cells was predicted to be strongly inhibited.

The effect of PSC secretomes on tumour cells. For elucidating the actual effect of PSC secretome on
tumour cells, pancreatic adenocarcinoma PT45P1 cells were grown in presence of secretome from either activated PSCs (PT45P1-act) or non-activated PSCs (PT45P1-non-act). In addition, cells were grown in absence of any substitution (PT45P1-ctrl). Cell viability was tested throughout and was at 95% to 100% for all cultures at any time. Prior to using supernatants of PSC cultures for conditioning media of tumour cell cultures, the supernatants were subjected to a buffer exchange step during the ultrafiltration process that was applied for increasing protein concentration. Only by this measure we could prevent the accumulation of small molecules, such as metabolites or other cellular by-products, which might have influenced cell function in the tumour cell cultures. After growth for 48 h, the PT45P1 cells were collected and washed to remove any secretome contamination. Subsequently, their cellular proteomes were isolated. Compared to PT45P1-ctrl, a substantial number of proteins differed significantly in abundance in all secretome-treated PT45P1 samples (Suppl . Tables S5, S6). This indicates that PSCs -not surprisingly -affect neighbouring tumour cells by their secretome. For the identification of differences that are due to the activation state of the PSCs, from which the secretomes were collected, the PT45P1-act and PT45P1-non-act protein samples were compared, yielding 247 cellular proteins that differed in abundance between them (Suppl. Table. S7). They mainly cluster in 12 functional categories, which are associated with processes such as metabolism of reactive oxygen species, cell death, cancer cell apoptosis and differentiation (Table 2).
To confirm the effect of the secretome released by activated PSCs on the apoptosis of pancreatic cancer cells, PT45P1 cells were grown for 48 h in medium conditioned with such secretome in comparison to growth in presence of secretome from non-activated PSCs. The cells were then harvested and both protein and total RNA were extracted, which were studied by either Western blot or real-time PCR analyses (Table 3). Conditioning the growth medium with the secretome of activated PSCs resulted in a down-regulation of pro-apoptotic factors, such as BAX, CASP-9, NFKB-1, NFKB-2, FASTK, and a concurrent up-regulation of the anti-apoptotic factor BCL-2, for example (Table 3; Fig. 3A). In addition, we confirmed that the tumour cells significantly decreased their apoptosis activity by measuring the mitochondrial membrane potential (representing extrinsic pathways) and caspase-3/7 activity (representing intrinsic pathways) in comparison to cells treated with non-activated PSC secretome (Fig. 3B). Mitochondrial membrane potential inversely correlates with apoptotic activity. Both its increase and the decrease of caspase-3/7 activity indicate a reduction of apoptotic activity. These findings confirm the functional prediction based on the protein variations observed on the antibody microarrays. Furthermore, also the functional prediction based on the protein variations detected in the PSC secretomes could be corroborated.
To determine whether secretome from activated PSCs was also inhibiting tumour cell differentiation, as predicted, we examined the expression of particular differentiation proteins. The mRNA levels of LMNA and RPS-19, for example, were reduced in PT45P1-act (Table 3). Furthermore, most proteins of the purine metabolism and translation initiation were up-regulated (e.g., IMPDH, eIF2B-1 and eIF4E) ( Table 3; Fig. 3C). Expression profiling further indicated up-regulation of proto-oncogenes or oncogenes, such as c-Jun, and down-regulation of tumour suppressor gene RB-1. Cell-cycle regulator genes that activate cell-cycle progression, for example CCNA-2, CCNB-1, CD-72, and GAS1, were also up-regulated, whereas expression of cell-cycle inhibitors, including CDKN2A, decreased (Table 3; Fig. 3C).
Finally, we also studied the effect of secretome from activated PSCs on the proliferation and migration of PT45P1 cells. Measuring proliferation by an assay that is based on quantifying the cellular DNA content demonstrated that exposure to secretome of activated PSCs significantly increased the rate of proliferation of the tumour cells (Fig. 4A). For an initial analysis of migration capacity, a wound healing scratch assay was employed (Fig. 4B). Cell migration was significantly increased in PT45P1 cells after a 24 h treatment with secretome from activated PSCs as compared to secretome from non-activated PSCs (p < 0.05). Activated PSCs promote pancreatic tumour cell growth through the eIF4E pathway. A pathway analysis of the proteins that exhibited variations in PT45P1-act indicated that the eukaryotic translation initiation factor 4E (eIF4E) could act as an upstream regulator of many of the proteins that showed changes in their expression. At the same time, eIF4E was found to be affected by various proteins of the secretome of activated PSCs (Fig. 5). eIF4E has been reported to affect oncogenes and growth factors promoting PDAC 30,31 and was found strongly up-regulated in tumour cells. Therefore, we postulated that regulation of the eIF4E signalling pathway might be involved in functional changes observed for PT45P1-act. If so, the effect of conditioning the growth medium of tumour cells with the secretome of activated PSCs should be suppressed by inhibiting eIF4E expression. To test this, two tumour cell lines -PT45P1 and Panc-1 -were transfected with EIF4E-siRNA or a control-siRNA of scrambled sequence and cultured for 24 h. Subsequently, the cells were grown in medium for another 24 h that was supplemented with the secretome of activated PSCs or relevant control conditions (Fig. 6).
Western blot analyses showed that there was little eIF4E expression in absence of secretome, while the presence of activated PSC secretome led to a strong induction of eIF4E expression (Fig. 6, columns c1 & c2). Suppression of eiF4E expression was very specifically achieved with the EIF4E-siRNA, counteracting the secretome induction (Fig. 6, c4). An siRNA with unspecific sequence had no such effect (Fig. 6, c3). Functionally, the effect of inhibiting eIF4E expression was tested by looking at apoptosis, employing the caspase-3/7 assay as a means for (2019) 9:5303 | https://doi.org/10.1038/s41598-019-41740-x www.nature.com/scientificreports www.nature.com/scientificreports/ measuring the combined effect of siRNA and conditioned medium. In accordance with the eIF4E levels, apoptosis changed as expected (Fig. 6). The same is true also for proliferation (not shown). For both apoptosis and proliferation, the cellular response to the addition of secretome from activated PSCs could be reversed by silencing specifically EIF4E. Next, we analysed whether inhibition of eIF4E expression also impacted tumour cell migration. For this assay, another PDAC cell line was used -Capan-1 -since it acts as a model for pancreatic cancer metastasis. In a scratch assay, cell growth in serum-free medium resulted in little migration. Addition of secretome from activated PSCs changed this strongly; within 48 h, the gap was completely filled with cells (Fig. 7). Transfecting the cells with a scrambled siRNA did not make a difference. However, upon transfection with an EIF4E-specific siRNA, the migration stopped entirely, although the cells were grown in the presence of secretome. Inhibition of eIF4E expression strongly suppressed the invasive capacity of the Capan-1 cells, suggesting that elevated eIF4E expression is associated with positive regulation of cell migration.

Discussion
Proteins secreted by stroma or tumour cells represent a major part of the molecules involved in intercellular communication, cell adhesion, motility and invasion 32 . We had previously examined the effect of proinflammatory factors on the biological functions of PSCs and identified TNF-α is the prime molecule responsible for their activation 6 . In the study reported here, we followed up this work by looking at regulatory mechanisms in tumour cells that are governed by molecules released from activated PSCs. Comparing the secretome of activated and non-activated PSCs identified proteins that were secreted by the former cells but not their quiescent equivalent. These proteins are predominantly involved in biological functions that are associated with cancer pathologies. We confirmed the functional predictions by growing tumour cells in media supplemented with PSC secretome. Simultaneously, we looked at the variations occurring in the proteome of such tumour cells. The eukaryotic translation initiation factor 4E (eIF4E) was found to act as a connecting regulative factor. We used EIF4E-silencing experiments to corroborate its involvement in the actual functional regulation.
A number of signalling pathways lead to PSC activation in response to cytokines. During inflammation, cytokines and growth factors like TGF-ß1, PDGF, interleukin-1ß (IL-1ß), IL-6 and TNF-α are released in the pancreatic tumour microenvironment causing stimulation and activation of PSCs 33 . Evidence from this laboratory and others suggests that TNF-α is a key element in PSC activation 6,[34][35][36] . In fact, TNF-α is known to be highly www.nature.com/scientificreports www.nature.com/scientificreports/ upregulated in pancreatitis with severe pancreatic fibrosis 37 . TNF-α is also a well-known mediator in the transformation and differentiation of myofibroblasts 38 . In pancreatic tumours the transformation is usually associated with induced proliferation of PSCs, expression of α-SMA, and synthesis of fibronectin. TNF-α is an inflammatory cytokine produced mainly by macrophages/monocytes as well as tumour cells during acute inflammation and responsible for a diverse range of signalling events within cells 39 . For this reason, we focussed our analysis on activation mediated by TNF-α.
Among the proteins detected in the secretome of activated PSCs were multiple cytokines and growth factors. Molecules such as IL-1ß 40 , S100 calcium binding protein A4 (S100-A4) 41 , TNF-α 42 , connective tissue growth factor (CTGF) 43 and activin 44 have been implicated to be involved in proliferation, migration or invasion of pancreatic cancer cells. IL-1ß, for instance, is known to up-regulate the expression of vascular cell adhesion protein 1 (VCAM-1) which mediates pro-metastatic tumour-stromal interactions 45 , promoting cancer cell adhesion and metastases 46 . PSCs also released platelet derived growth factor (PDGF) 11 , vascular endothelial growth factor (VEGF) 12 , TGF-β 47 and cyclooxygenase-2 (COX-2) 48 , which promote cancer cell growth. Other molecules secreted by activated stellate cells were fibroblast growth factor 1 (FGF1), IL-6, IL-4 and insulin-like growth factor I (IGF-1). Proliferation of pancreatic cancer cells is known to be mediated by PDGF and FGF1 signalling pathways 11 . Studies have shown that these pathways are influenced by the interplay of tumour cells and their microenvironment 49 . Furthermore, IL-1ß, IL-6 and TNF-α are present in the microenvironment of cancer, contributing to metastasis and drug resistance 36 . The increased IL-6 secretion by activated stellate cells strongly suggests a role in regulating cell growth in the cancer microenvironment. Another protein identified was plasminogen activator inhibitor 1 RNA-binding protein (SERPINE), whose deregulation has been associated with the progression of various cancers, including PDAC 50 . In addition, SERPINE co-localizes with hypoxic regions in tumours 51 . Also thrombospondin 3 (THBS-3) was specific for activated PSC secretome. THBS-3 is a member of adhesive glycoproteins that mediate cell-to-cell and cell-to-matrix interactions. THBS-3 binds to ECM proteins, including fibronectin, laminin and type V collagen 52 . Other ECM proteins that were found, such as FN1 and Col1, have previously been reported to be involved in pancreatic cancer progression. Their high abundance in the secretome of activated PSCs suggests a de novo induction of expression and subsequent secretion upon PSC activation. The release of ECM proteins represents a pivotal role of activated PSCs in tumorigenicity. They have important roles in resistance to anticancer drugs and cell proliferation of pancreatic cancer cells 53 .
Next to studying the content of the PSC secretome and predicting functional consequences it may have on recipient tumour cells, we performed incubations of PDAC cells in growth medium conditioned with PSC secretome from activated and non-activated cells. Intracellular expression was analysed by means of the antibody microarray and qRT-PCR. As a response to growth in presence of activated PSC secretome, we found a significantly inhibition of apoptosis and a simultaneous promotion of tumour cell migration. Furthermore, proliferation of the pancreatic cancer cells was induced. This fits very well to the predictions made on the basis of the variations in the PSC secretome. This result suggests a direct impact on the tumour cells by activated PSCs in addition to their involvement in the desmoplastic events within the tumour microenvironment. In response to activated PSC secretome, apoptotic factors of the extrinsic and intrinsic pathway were regulated. Such an effect was not observed with non-activated PSC secretome. Among proteins that exhibited significant differences in abundance was nuclear factor kappa B subunit 1 (NFKB-1). This protein is not expressed by normal cells, but strongly expressed in a number of human cancers including non-small cell lung carcinoma, pancreatic cancer, colon cancer, prostate cancer, breast cancer, bone cancer and brain cancer 54 . In pancreatic cancer cells, reduced expression of NFKB-1 activating molecules can obstruct tumour progression in vitro and in vivo 55 . In addition, NFKB-1 prompts the expression of inflammatory cytokines, adhesion molecules, key enzymes in the prostaglandin synthase pathway (COX-2), nitric oxide (NO) synthase and angiogenic factors. Furthermore, by inducing anti-apoptotic genes (e.g. BCL-2), it promotes survival in tumour cells and epithelial cells targeted by carcinogens 56,57 .
Many proteins involved in tumour cell differentiation, such as ribosomal protein 19 (RPS-19), lamin A/C (LMNA) and immunoglobulin heavy constant mu (IGHM), exhibited decreased expression upon cell treatment with activated PSC secretome. The results support previous observations that PSCs can reduce differentiation of PDAC cells [58][59][60] . The molecular changes triggered in the tumour cells could be responsible for dissuading cells from both differentiation and apoptotic processes 61,62 . In response to treatment with secretome from activated PSCs, expression profiling further indicated down-regulation of proto-oncogene or oncogenes, such as the retinoblastoma-associated tumour suppressor gene RB-1. Cell-cycle regulators that activate cell cycle progression, such as cyclin-A2 (CCNA-2), were also up-regulated, whereas expression of cell-cycle inhibitors, including the CDK-inhibitor 2 A (CDKN2A), decreased in PDAC as compared to controls. RB-1 is a crucial regulator of appropriate cell cycle progression, including G1 to S and G2 to M phase transitions 63 . The activity of RB-1 is mainly regulated by the upstream CDKN2A/CCND-1 pathway 64 . In accordance with the important role of RB-1 as a cell cycle regulator, RB-1 deregulation is observed in multiple types of cancers 65 . Functional loss of RB-1 causes accelerated, E2F-transcription factor 1 (E2F1) mediated transactivation, followed by uncontrolled cell cycle progression 66 .
Even though the role of PSCs in the microenvironment is well-established 10,11 , the cellular mechanisms mediating crosstalk between PSCs and PDAC cells and the signalling molecules involved remain a complex network awaiting to be untangled. Tumour cell growth in secretome of activated PSCs caused a significant increase in the expression of proteins of translation initiation, such as translation initiation factor eIF-2B subunit alpha (eIF2B-1) and particularly eukaryotic translation initiation factor 4E (eIF4E). Protein eIF4E triggers the synthesis of a variety of molecules involved in cell growth, proliferation and invasion, including the cell cycle regulatory protein cyclin D1 (CCND-1) 67 , the transcription factor c-Myc, growth factors such as the vascular endothelial growth factor (VEGF) and fibroblast growth factor 2 (FGF-2) 68 , as well as the anti-apoptotic protein MCL-1 69 . Recent studies have demonstrated that progression and resistance to standard therapies in tumour is closely related to the eIF4E pathway 30,31,70 . In our analysis, we confirmed that eIF4E acts as a key signalling molecule that mediates