Tumor-Stroma Co-Cultures Modify the Invasive Properties of Human Pancreatic Adenocarcinoma Cells Through Different Patterns of Cytokine Secretion and Depending on the Type of Cell Lines.

Background: Pancreatic ductal adenocarcinoma (PDAC) is among the most lethal cancer worldwide. One key feature of PDAC is a dense desmoplastic reaction that has become recognized to play important roles in tumor growth, metastasis, and therapeutic resistance. We aim to study tumor-stromal interactions in an in vitro co-culture model between human PDAC cells and broblasts. Methods: Capan-1 or PL-45 human pancreatic tumor cell lines and LC5 broblasts were grown in direct or indirect co-cultures, and their relationship regarding migratory and invasive properties were studied by wound-healing and transwell migration assays. Confocal immunouorescence, ELISA, and western blotting were used to evaluate the expression of activation markers. Cytokines arrays were performed to identify secretome proles associated to migratory and invasive properties of tumor cells. Extracellular vesicles production was examined by ELISA and transmission electron microscopy. Results: Co-cultures conditions increased FGF-7 secretion and α-SMA expression, characterizing broblasts activation; and decreased epithelial marker E-cadherin in tumor cells, suggesting that they suffer epithelial-to-mesenchymal transition. The expression pattern of cytokine secretion differed under co-culture conditions versus monocultures, with GM-CSF, GRO a/b/g, GRO-α, IL-6, IL-8, MCP-1, and RANTES being the most affected. A greater number of exosomes with higher amount of proteins was quantied in co-cultures assays. Interestingly, tumor cells and broblasts migrate as a package, with tumor cells in the middle surrounded by broblasts, and maximizing the contact between cell to cell. Conclusions: We show different migratory properties in tumor and stromal cells, and we draw a different mechanism for tumor spread through a cooperative migration between tumor cells and activated broblasts. Furthermore, we found that the level of IL-6 mainly, in addition to other cytokines, change signicantly in co-cultures conditions with respect to monocultures, and this could affect the invasive and migratory capacities of cells. Targeting the interaction between tumor cells and tumor microenvironment might represent a growth factor/Fibroblast growth factor 7; LC5-GFP + : LC5 broblasts expressing Green uorescent protein; PDAC: Pancreatic ductal adenocarcinoma; TEM: Transmission electron microscopy; TGF-β: Transforming growth factor-β.

mechanism for tumor spread through a cooperative migration between tumor cells and activated broblasts. Furthermore, we found that the level of IL-6 mainly, in addition to other cytokines, change signi cantly in co-cultures conditions with respect to monocultures, and this could affect the invasive and migratory capacities of cells. Targeting the interaction between tumor cells and tumor microenvironment might represent a novel therapeutic approach to advance pancreatic carcinoma.

Background
Pancreatic ductal adenocarcinoma (PDAC) is among the most lethal tumors worldwide, with a 5-year overall survival around 9% (1). Unfortunately, many chemotherapeutic agents, radiation and targeted therapies have failed to improve survival. This lack of success has been mainly associated with late diagnosis and chemoresistance of the disease (2,3). The increasing knowledge of cellular and molecular mechanisms involved in pancreatic carcinogenesis has also failed to translate into new therapeutic approaches. This may be due in part, to the fact that most studies focus mainly on targeting tumor cells and obviate the complex surrounding microenvironment, which plays an important role in the evolution of the disease (4,5).
One of the hallmarks of PDAC is the formation of an abundant and dense stroma wrapping tumor cells, termed desmoplastic reaction, which leads to an increase in the interstitial pressure, and hence, to poor tissue perfusion and hypoxia. This tumor-associated stroma may represent up to 80% of the tumor mass in the majority of patients (6). Although initially it was thought that this component was a barrier limiting tumor progression, today its pro-tumor properties have been widely recognized (5)(6)(7)(8)(9)(10). The surrounding stroma or tumor microenvironment is composed of an altered extracellular matrix, and a wide variety of cell types, including resident epithelial cells, broblasts, smooth muscle cells, immune and in ammatory cells, and lymphatic and vascular endothelial cells (11,12). A population of surrounding broblasts known as cancer-associated broblasts (CAFs) play a critical role in tumor progression by stimulating survival and proliferation signaling pathways, through heterotypic signaling with cancer cells (13,14).
Therefore, PDAC is formed by neoplastic cancer cells and stromal cells, and the development of the tumor requires reciprocal communication and interactions among all cell types, which cause alterations in each other by mechanisms that remain poorly understood (15). Moreover, during tumor progression, microenvironment progressively changes to a cancer-associated state that promotes tumor growth and metastasis. On the other hand, pancreatic cancer is a disease characterized by a highly oncogenic genetic pro le (16)(17)(18). Consequently, tumors are heterogeneous tissues composed of different cells with distinct characteristics, and individual tumor cells of the same tumor have different tumorigenic potential (11,19).
Biological information between cells can also be transferred through small extracellular vesicles (EV; exosomes, microparticles, oncosomes, etc.) released into interstitial spaces, which can be considered a snapshot of the cell's inner contents. They are being intensively studied as promising biomarkers for treatment selection and monitoring for several cancers (20)(21)(22).
In this study, to elucidate the role of tumor-stromal interactions in modulating pancreatic cancer progression, the effects on migration and invasion properties of different co-cultures conditions between human PDAC cell lines (Capan-1 and PL45) and broblasts (LC5), have been studied.
LC5 cell line was infected for 24h with 4 mL of virus-containing medium in presence of 10 µg/mL polybrene. Selection and isolation of broblasts-GFP + was performed with Sony Synergy Cell Sorter separation device (Sony Biotechnology Inc.), and based of GFP expression, and incorporation of Propidium Iodide (PI) for the exclusion of dead cells. All plasmids required were from non-pro t plasmid repository Addgene (http://www.addgene.org/).

Co-cultures of broblasts and pancreas cancer cell lines
For direct co-cultures between pancreatic tumor cell lines and broblasts, cells were cultured in a ratio 1:1, in RPMI medium supplemented with 10% FBS. Tumor pancreatic cells were seeded 72 hours before broblasts were added to the culture. The optimum time for these direct co-cultures was experimentally established in 72h, to obtain a nal ratio of tumor cells: broblasts of 20:80, re ecting what is described in vivo. For functional analysis, cells were separated by cell sorting based of GFP expression and incorporation of PI.
Conditioned media (CM) were prepared from broblasts and pancreatic cancer cell lines. Cells were grown in 75cm asks until semicon uent conditions, washed three times with phosphate-buffered saline (PBS), and then incubated in 1% of FBS medium for another 48 h. Next, supernatants were collected, centrifuged at 2000rpm for 10 min, sterile ltered (0.45 µm pore), and immediately used for assays of indirect co-cultures.
Wound healing, migration, and invasion assays For wound-healing assay, cells were seeded on 6-well plates, and left to reach 80% con uence. A single wound per well was scratched with a sterile 200 µL pipette tip within the monolayers and then washed with PBS to remove detached cells. CM from tumor cells was added to broblasts, and similarly CM from broblasts was added to tumor cells, and the rate of wound closure was observed in the following 24, 48, and 72 hours. Wound closure was monitored using a Nikon TE2000-S microscope.
Transwell migration assays were carried out using 8.0 µm polycarbonate cell culture inserts (Transwell® Permeable Supports, Corning Inc., NY). Cells were serum-starved 24h before the start of the experiments. Tumor cells were seeded at 1x10e5 per well in 2 mL growth medium in the upper chamber (cell culture inserts) of 6-well plates. The same number of broblasts was seeded in the lower wells. Cells were allowed to migrate for 72 hours toward the underside of the membrane to the lower chamber. After that, cells attached to the upper side of the lter were removed by wiping with a cotton swab, and the migrated cells on the lower surface were xed with formaldehyde 4% in PBS and permeabilized with absolute methanol in darkness. After several washes with PBS, cell nuclei were stained with 4'-6-diamino-2phenylindole (DAPI).
We used Ibidi culture-insert 2 well in µ-dish 35 mm plates, with a de ned 500 µm cell free gap (Inycom Biotech, Zaragoza, ES) for migration assays. Tumor cells were seeded on one side of the insert and broblasts-GFP + on the opposite side. After 24h for cells adhesion, the insert was removed, and migration was allowed. This system allows us to observe the migratory behavior of each cell type under the indirect in uence of the other.

Western blot analysis
Whole cell lysates were prepared using MCL1 lysis buffer in the presence of protease and phosphatase cocktail inhibitor (Sigma-Aldrich, St Louis, MO) following the manufacter's protocol. Protein lysates were subjected to SDS-PAGE on 10% polyacrylamide gel and transferred onto PVDF membranes. Membranes were probed for the protein levels of E-cadherin, and α-tubulin as loading control, using speci c primary antibodies (#4065, Cell Signaling Tech. Inc., Danvers, MA, and #sc-5286, Santa Cruz Biotechnology, Inc., Dallas, TX, respectively). Speci c bands for each protein were visualized by WesternBright™ kit (Advansta, San Jose, CA).
Enzyme-linked Immunosorbent Assay (ELISA) and human cytokine array.
Culture supernatant was collected from each growth condition, and keratinocyte growth factor (KGF-7) concentration was determined using Quantikine human KGF/FGF-7 Immunoassay (

Statistical analysis
Each experiment was performed in triplicate in at least three independent experiments. The analysis of differences was performed using Mann-Whitney U test and variance analysis for repeated measurement data. Results were analyzed using software GraphPad Prism version 6.0 (GraphPad software, San Diego, CA) and SPSS v21.0 (SPSS Inc., Chicago, IL). The value P<0.05 was considered as signi cant.

Co-cultures of pancreatic tumor cells and broblasts-GFP +
First, we established a broblast cell line constitutively expressing GFP protein (LC5-GFP + ) in which the percentage of broblasts-GFP + was higher than 97% (Supplementary Figure 1).
Experimental design and images from direct co-cultures between PDAC and LC5-GFP + cells at 72h, are shown in Fig. 1. It is noteworthy that broblasts in the vicinity of the tumor cells surface established close contacts to these cells, being so arranged around the tumor cells to maximize the contact between cell to cell.
With the aim of evaluating the importance of the cell-cell adhesion and juxtacrine signaling versus paracrine signaling between cancer and stromal cells in the pathophysiology of pancreatic cancer, directand indirect co-cultures with CMs, were used in the following assays.

Co-culture modi es the expression of activation markers in tumor cells and broblasts
To check activation of broblasts, LC5-GFP + cells were seeded in direct co-cultures with tumor cells, or indirect co-cultures with CM from Capan-1 or PL-45 cells. Representative pictures of α-SMA immuno uorescence staining are shown in Fig. 2a. LC5-GFP + , Capan-1, and PL-45 cells do not express the α-SMA activation marker when they grow alone at monocultures (rows 1, 2, and 5). However, we can see that broblasts express high level of α-SMA both in direct-and indirect-culture with tumor cells (rows 3, 4, 6, and 7). Quanti cation of LC5-GFP + cells expressing α-SMA is shown in Fig. 2b. However, this broblast activation did not result in increased tumor cell proliferation, since no signi cant differences between indirect-and monoculture conditions were found in cell proliferation assays (Fig. 2c).
In addition, different FGF-7 expression levels between cells grown alone or in co-cultures were found. Thus, FGF-7 levels in co-culture supernatants were higher than in monoculture supernatants were (Fig. 3a,  P<0.0001). These results show that broblasts under the co-culture conditions here described, acquire the so-called myo broblast or CAF phenotype.
Similarly, tumor cells were also affected by the presence of broblasts in co-culture. To check the induction of epithelial mesenchymal transition (EMT) process, we isolated tumor cells by sorting, and quanti ed the protein levels of E-cadherin by immunoblotting (Fig. 3b-c). E-cadherin is a cell adhesion molecule necessary for maintenance of intercellular contacts and cellular polarity in epithelial tissues, and a key feature of EMT. We can see that in both direct and indirect co-cultures, pancreatic tumor cells decrease their expression of E-cadherin, which would be attributed to an EMT process.

Cytokines secretion changes under co-culture conditions
It has been reported that cytokines and growth factors secreted by CAF or tumor cells are associated with tumor invasion and migration. Thus, the effects of direct-and indirect-cocultivation of broblasts and tumor cells on cytokine production were examined next. We collected the supernatants from monocultures, direct-and indirect co-cultures, and quanti ed the differences in protein expression of 23 speci c human cytokines by a human cytokine antibody array (Supplementary Figure 2, Fig. 3 d-g). Relative levels of the cytokines differentially expressed between broblasts, Capan-1 and PL-45 monocultures are shown in Fig. 3d. Then, we compared the expression of these proteins in direct-and indirect-cocultures versus respective monocultures (Fig. 3e, versus broblasts monoculture; 3f, versus Capan-1 monoculture; 3g, versus PL-45 monoculture). We found that the GM-CSF growth factor, and cytokines GRO a/b/g, GRO α, IL-6, IL-8 and RANTES were markedly increased in LC5-cancer cells cocultures compared to LC5 monoculture. Curiously, GM-CSF, GRO a/b/g and GRO α, show a considerable decrease only when LC5 cells are grown with CM of PL-45. It is noteworthy the high expression of IL-6 in co-cultures of broblasts and tumor cells compared to tumor cells monocultures (Fig. 3e-g). However, we observed an increase in chemokine RANTES and growth factor GM-CSF levels in co-culture of broblasts with Capan-1 cells, but a decrease in co-culture of broblast and PL-45 cells. The cocultivation of broblasts and PL-45 cells was the one that showed the most differences in the pattern of cytokine secretion depending on the co-culture conditions, mainly in relation to IL-8, GRO α and GRO a/b/g levels.

Co-culture modi es motility of tumor cells and broblasts
The migratory behavior of cells in direct-or indirect co-culture was examined by wound-healing and transwell migration assays. Fig. 4a-c shows representative images of Capan-1 cells grown in direct co-culture with broblasts-GFP + . Both cell types migrate into the wound, but curiously, broblasts migrate predominantly as single cells, whereas tumor cells migrated as a tightly packed sheet surrounded by broblasts.
To explore whether cell-cell contact is necessary for migration activity, we cultured Capan-1 or PL-45 tumor cells with broblasts CM for 3 days, and vice versa, LC5 with CM from each tumor cell line, and assayed their migration activity to wound-healing (Fig. 4e-f). Only PL-45 cells exposed to LC5 supernatant showed a weak migration increase at 48h (Fig. 4e).
On the other hand, Capan-1-and PL-45-CM differed in their ability to affect the movement of LC5 cells, with Capan-1-CM being a more potent inducer (Fig. 4f). It is noteworthy that broblasts grown with Capan-1-CM travel to the center of the wound, invading the entire eld. In contrast, broblasts cultured with PL-45-CM do not leave the edge of the wound, suggesting that wound healing occurs rather by cell proliferation than by broblasts migration.
Next, Capan-1 or PL-45 cells were cultured in the upper chamber of transwell migration system for 3 days.
The number of invasive tumor cells increased signi cantly when broblasts were seeded in the lower chamber, compared to complete growth medium as control. This is substantial in Capan-1 cell line, but even more so in the case of PL-45 tumor cell line (Fig. 4d).
Last, we carry out migration assays in the Ibidi culture-insert plates. Tumor cells were seeded on one side of the insert and LC5-GFP + on the other one. As a control, LC5-GFP + cells were seeded on both sides of the insert ( rst column, Fig. 4g). Fibroblasts in co-culture with PL-45 cells have a migration similar to that observed in the control (Fig. 4g, column 4 and 5). However, broblasts in co-culture with Capan-1 cells increase their migratory capacity, leaving the edge of the culture and moving towards the tumor cells (Fig.   4g, column 2 and 3). Moreover, in the vicinity of contact between both cell types, broblasts are arranged so as to increase the surface area of contact with tumor cells. These results are in agreement with those obtained previously in direct-and indirect co-cultures.

Identi cation and quanti cation of extracellular vesicles (EV) in co-cultures
Since there is ample evidence that EVs play a crucial role in intercellular communication, we decided to study the secretion of these vesicles in our co-culture assays. Visualization was realized by TEM (Fig. 5ae). Although TEM is not a quantitative technique, and we cannot differentiate between tumor cells and broblasts in co-cultures, we found a higher number of EVs per cell under co-culture versus monoculture conditions. EVs displayed typical spherical morphology and fell within the expected size range of exosomes (30-150 nm). We therefore isolated and quantify EVs from supernatants samples, and we could see that exosomes from co-cultures have a higher amount of protein than those from monoculture (Fig. 5f).

Discussion
Pancreatic tumors are heterogeneous tissues composed of tumor and non-tumor cells, embedded within a dense desmoplastic stroma (6,23). There are clear evidences that stroma is not just a static mechanical barrier but is thought to be an active participant involved in tumor initiation, progression, and metastasis (24). Thus, the study of pancreatic cancer cells behavior in monocultures does not faithfully re ect what is observed in vivo, given that all tumor components have a signi cant impact on the biological properties of malignancy. In this regard, in the work here exposed we use in vitro co-culture models, in an attempt to recapitulate some of the stromal-tumor connections, as a mean to assess the impact of this partnership on migration and metastasis initiation in PDAC.
Fibroblasts are the most common component of tumor stroma, and CAFs provide a favorable microenvironment with tumor-promoting properties, stimulating the progression of primary tumors via paracrine signaling (25), and serving as a niche to support the metastatic colonization (26)(27)(28). Thus, the association between cancer cells and CAFs leads to alteration of the biological properties of both types of cells through bi-directional tumor stroma crosstalk.
In our co-culture model, KGF/FGF-7 secretion is upregulated and it has been postulated than CAFs could increase the viability and migration of tumor cells by a paracrine manner through this growth factor (29). In addition, α-SMA is the most widely used marker to asses CAFs activation status and is associated with worse clinical outcomes (30,31).
Cell migration is an important factor in the spread of pancreatic cancer and is a key step in the series of events that lead to metastasis. In the wound-healing assays of direct co-culture, we can see that groups of cells move into the wound. In these clusters, broblasts are arranged surrounding tumor cells. Besides this curious migration formation, it is also noteworthy the amoeboid characteristics of tumor cells movement. Amoeboid migration facilitates cell position change without cell remodeling but move by low adhesion force and high cell deformability. Instead, collective migration takes place when cells retain cellcell adhesions and coordination among them. These migration modes are interconvertible, and tumor cells use both mechanisms in an adaptive way, exhibiting its great plasticity (32). In our work, this peculiar way of migrating set, tumor cells and surrounding broblasts in the middle of wound, could indicate that both types of cells actively cooperate to migrate to other organs, as a single pack of invasion. Previous reports have shown that single cancer cells may proliferate via interactions with stromal cells derived from other organs when they form metastatic tumors (33). However, these mechanisms are not mutually exclusive, and broblasts could come from either the primary tumor or metastatic sites.
Using traditional migration assay setup, we show that only Capan-1 cells were able to induce an important CAFs migration as individually moving cells. Similarly, Ishii et al., have reported that lung broblasts co-cultured with human cancer cell line Capan-1 showed signi cantly higher migration activity than broblasts alone (34). In addition, Karagiannis et al. showed in co-cultures of colon cancer cell lines and broblasts that the CAFs migratory behavior was context-dependent, varying from totally individual to collective con guration (35). Therefore, broblasts would have dissimilar migratory responses upon stimulation with different pancreatic cancer cells, such as Capan-1 or PL-45.
Activated broblasts can promote invasive growth by cell-cell contacts or by paracrine diffusible signals, which also can induce EMT of tumor cells (36,37). On the other hand, we have shown an important decrease in the expression of the epithelial marker E-cadherin in tumor cells from our co-culture model, suggesting that EMT occurs in this system and would facilitate the initiation of metastasis (4). However, variable expression of E-cadherin in human pancreatic cancer samples and commercial cell lines has been con rmed in several studies, supporting the hypothesis that PDAC cells exhibit a hybrid EMT-related phenotype (38).
In the crosstalk between broblasts and cancer cells, two closely pathway have been described: one pathway where stroma cells affect cancer cells response and other where cancer cells trigger a response in the stroma cells (37,39). Both responses could be mediated by different secretion of cytokines from the cells involved. Furthermore, we found that the level of IL-6 mainly, in addition to GM-CSF and RANTES, changed signi cantly in co-cultures conditions with respect to monocultures, and this could affect the invasive and migratory capacities of cells. Increased IL-6 level has previously been related to protumorigenic function of CAFs, promoting contractibility and extracellular matrix remodeling (40,41). Curiously, we have not found change in TGF-β level between the different co-culture conditions. Even more when IL-6 is also a common cytokine that enhances TGF-β signaling, resulting in EMT and stimulates tumor progression. However, TGF-β signaling has multifaceted and cell type speci c in PDAC (42). Thus, into the heterogeneous cancer-associated broblasts population, there may be a subset of in ammatory CAFs expressing pro-in ammatory cytokines, which could be majority in our co-culture models.
Exosomes have recently been described as communication tools between cells, and they could mediate interactions between tumor cells and the microenvironment (22,43). They are secreted from almost all cell types although tumor cells secrete more EVs compared to normal cells, playing an essential role in inducing metastasis and remodeling the primary microenvironment. Thus, it has been shown that EVs from broblasts promote invasive and chemoresistant behavior in cancer cells (44,45), whereas EVs from tumor cells can reprogram broblasts into CAFs. In addition, it has been described that an increased amount of protein per exosomes predict disease progression (46,47). Our results are in agreement with previously published data since they indicate that cells in co-cultures produce a greater number of exosomes, and these possess a higher amount of protein that cells in monoculture.
A growing body of literature dealing on the tumor-stroma interactions, point to the future approach for cancer treatments based against both malignant and microenvironmental cells (24,48,49). Since pancreatic cancer cells are surrounded by a thick and poorly perfused stroma, which halts penetration of otherwise effective treatments, it would be reasonable to think that incorporating a stromal depleting agent into therapeutic treatments, or disrupting the crosstalk between stroma and cancer cells, would sensitize chemoresistant tumors and improve the outcome of patients (50)(51)(52). Similarly, exosomes could be used in antitumor therapies by exploiting their characteristic properties (53).
Finally, an important limitation of our study is that it focuses on cancer-associated broblasts as the leading stromal element interacting with pancreatic tumor cells. However, many other cell types in the stroma are known to in uence tumor cells behavior, such as immune and in ammatory cells, adipocytes, and endothelial cells. Two different PDAC cell lines were used in our co-culture model, and we can not generalize the results to the entire spectrum of pancreatic cancer. In this study, we have achieved some relevant results, but also accepting the incomplete approximation of the in vitro set-ups, as they are not always able to simulate the high tissue complexity found in vivo.

Conclusions
Collectively, we have shown a co-culture model of tumor cells and broblasts, which recapitulates some features occurring in human PDAC, and might be used for a better understanding of tumor-stromal interactions and identify new targets for pharmacological intervention. We show different migratory properties in tumor and stromal cells, and a different mechanism for tumor spread through the joint migration of tumor cells and broblasts. Taken together, our results suggest that direct contact between broblasts and Capan-1 cells is involved in the progression of tumor, in addition to the paracrine promoting effect of growth factor or chemokines such as GM-CSF or IL-6. However, CAFs could lead PL-45 migration mainly though paracrine mechanisms, besides juxtacrine ones. Although cancer-stroma interaction may be a promising target for new therapeutic approaches, these should focus on each tumor histotype, because of the variability of tumor cells and the role played by CAFs in each tumor.

Declarations
Ethics approval and consent to participate Not applicable.

Consent for publication
Not applicable.