Transparent, Nanoporous, and Transferable Membrane‐Based Cell–Cell Paracrine Signaling Assay

A transparent, nanoporous, and transferable membrane-based cell coculture platform is devised and developed to be used for the accurate and versatile control and analysis of paracrine communications between cocultured cells, as well as easy isolation of each cell line for further manipulation and analysis. On the basis of the high degree of freedom in stacking and destacking of the membranes and the controllable nanopores that efficiently allow exchanges of paracrine factors without physical contact between the different cell lines, the paracrine signaling between human breast metastatic cancer cells and three different types of stromal cell types are controlled and studied. Cells communicate with one another typically through direct contacts (i.e., gap junctions or tight junctions) or the exchange of soluble factors (i.e., paracrine and endocrine signaling through cytokines, chemokines, and growth factors) to maintain the functions of tissues and homeostasis. In particular, cancer development has been considered the result of cross talks between tumor cells and various types of stromal cells surrounding them, including fibroblasts, myoblasts, immune cells, and mesenchymal stem cells.1 Tumor cells constantly modulate their stromal environment by producing a range of growth factors that activate the stromal cells in a paracrine manner and disrupt normal tissue homeostasis, leading to the secretion of additional growth factors and proteases. In this regard, the reciprocal communication between tumor and stroma promotes cancer progression, near as well as far from the primary tumor site. For better understanding and use of the interactions between tumor cells and tissue environment for therapeutic applications, significant efforts have been made to explore the tumor development pathway and to determine key signaling molecules.2 Therefore, the development of highly efficient cell coculture platforms to systematically investigate the paracrine communications between cancerous and stromal cells has frequently been emphasized to fully understand and control the tumor cell behavior and to provide some pivotal and unprecedented information in cancer biology.3 At the same time, several nanoscale platforms that interface cells to reveal and control cell signaling mechanisms and functions have been developed and proven useful.4 Although previous approaches provided a platform for culturing multiple cell lines, there are several issues that need to be resolved. For example, they require extracellular matrix (ECM) proteins and a few days to form multiple cell layers on a scaffold.5 Temperature-responsive polymer layer was also introduced with an ECM protein layer to harvest the cell sheet. Further, it is difficult to analyze how neighboring cells affect other cells through monitoring the signaling molecules from each cell line separately. Recently, membrane platforms were used for paracrine signaling study, but these studies suffer from major drawbacks, such as the lack of in vivo like functionality, separation of specific cells of interest and low pore density, micrometer-scale film thickness, and optical non-transparency of the membrane.[1] On the contrary, direct cell coculture approach offers more in vivo like cell–cell communication environment,6 but this method also has many issues including cell cross contamination by xenogenic reaction and difficulties in manipulating and isolating each cell line for detailed individual analysis.7 Herein, we designed and prepared cellulose acetate (CA)-based nanoporous thin film as a transparent, nanoporous, and transferable (TNT) cell coculture membrane to address the aforementioned issues in cell coculture assays. The TNT membrane allows for cell adhesion within a short time and separation of single cell layer efficiently after coculturing of multiple cell lines through good affinity to cells and transferability of membrane, which can provide a new platform for controlling and analyzing cell–cell paracrine signaling. CA has been widely used as filter membranes due to its low static charges and high mechanical strength.8 Also, CA has received great attention in biomedical fields as cell scaffold materials that exhibit a wide range of properties, such as high modulus, adequate flexural, and tensile strength with biocompatibility.9 The porous structure of the CA membrane was realized by nonsolvent (water) vapor-induced phase separation (NVIPS) during spin coating in a closed chamber with controlled relative humidity (RH) (Figure 1a, and Figure S1, Supporting Information). In cell culture medium, the transparency, flexibility, and facile transferability of the CA-based TNT membrane originate from its nm-scale dimension in film thickness (482 ± 7 nm), small pore size (≤150 nm), and low water solubility (Figure 1b). To the best of our knowledge, this work is the first example in adopting spin-casted ultrathin CA membranes with controlled pore size and density for cell coculture and paracrine signaling assays. The TNT membrane allows accurate analysis of paracrine communication between cocultured cells as well as easy isolation of each cell line for further manipulation of individual cell lines. Also, the high degree of freedom in stacking and destacking of the membranes enable us not only to coculture three or more different cell lines simultaneously but also to coculture cell lines in a sequentially different manner, which would offer a highly versatile in vitro platform to trace the history of stromal activations for metastatic cancer cells. In a typical proof-of-concept experiment, MDA-MB-231 (metastatic cancer cell; M) and human mesenchymal stem cell (hMSC; H) were first grown separately on a TNT membrane and cocultured by stacking the cell-culture TNT membranes (Figure 1c; see the Supporting Information for the experimental details). First, the cell–cell paracrine communications between MDA-MB-231 and hMSC via the TNT membrane were studied by luminescence-based cytokine assay for regulated on activation, normal T cell expressed and secreted (RANTES) expression.10 It has recently been reported that this RANTES induction required close physical contacts between hMSCs and MDA-MB-231 cells, and RANTES secretion disappeared when the two cell lines were separated by a commercially available membrane platform [polyethylene terephthalate (PET) membrane].[1] In this regard, RANTES expression results, when hMSCs are cocultured with MDA-MB-231 cells using the TNT membrane stacking assay, can demonstrate the feasibility of our TNT membrane platform for the analysis of cell–cell paracrine communications. For this purpose, we tested three different membranes with different pore size and PET membrane for RANTES expression, and the membranes were precoated with poly(dopamine) to enhance cell adhesion (Figure 1d,e and Figure S2,S3, Supporting Information). In the TNT co-stacking cell assay, the largest amount of RANTES was detected in the case of the CA membrane prepared at 65% RH, which has well-defined nanopores (diameter: 50–150 nm) with a maximum density of ≈40% and 482 nm film thickness, while the assay with the commercially available 10 μm thick PET membrane with low pore density resulted in much lower RANTES secretion (Figure 1f and Table S1, Supporting Information). Next, we checked the membrane thickness effect on the paracrine signaling between cells. First, the 10 μm thick 20 wt% CA membranes were prepared by NVIPS, and the result was compared with the data from the 480 nm thick 4 wt% CA membrane and the PET membrane cases (Figure S4, Supporting Information). It should be noted that the 20 wt% of CA film has almost the same porosity as the 4 wt% CA membrane but its thickness is largely different from 4 wt% CA membrane. The result showed that the fold induction of RANTES decreased five times in the case of 20 wt% CA film as compared with the 480 nm thick 4 wt% CA membrane, which means that distance between cells is critical for paracrine communication (Figure 2a). We further investigated the effect of the distance between cells on cytokine secretion by increasing the stacking number of the identical TNT membranes while maintaining pore size and distribution. Cells were cultured only on the top and bottom layers in this study, and this is different from the cell stacking assays that cells were cultured on each membrane layer from top to bottom. Overall, the results indicated that the amount of secreted RANTES as the result of cell–cell paracrine communications was strongly correlated with the cell–cell distance (Figure 2b). In the paracrine signaling, the signaling molecules released by a cell affect target cells in close proximity because they diffuse away from the signaling cell, forming a concentration gradient and also inducing different cellular responses depending on the distance of a target cell from the signal-releasing site. In this regard, the close proximity between cocultured cells is a major parameter in cell–cell signaling, and the TNT membrane offers such an environment by keeping the sub-micrometer intercellular distance between different cell lines. Furthermore, we checked the diffusion of epidermal growth factors (EGFs) through the TNT membrane with a two chamber system with one chamber filled only with buffer solution and the other chamber filled with EGFs in buffer solution (Figure 2c,d). The concentration gradient of EGFs across the membrane was clearly observed, and the amount of diffusing EGF decreased as the number of the stacking layers was increased. Importantly, the result from the sandwich-enzyme-linked immunosorbent assay (ELISA) for EGF that diffuses into the buffer chamber shows that the saturation of EGF in the buffer chamber occurs after 6 h incubation in phosphate buffered saline (PBS) with 5% fetal bovine serum (FBS), and the protein concentration remains the same even after 24 h incubation. The data support our hypothesis that the TNT membrane with porous structure and sub-micrometer thickness allows for efficient protein transfer between cells that reside in different compartments and also show that the first 6 h is critical for the protein diffusion through the membrane pores. It has also reported recently that the diffusion efficacy of proteins in the initial phase (first 4 h period) is critical for intercellular communications.11 All these validate our hypothesis and the results reported in this study. Finally, we analyzed the transport of calcein-AM from MDA-MB-231 to hMSC to check whether there were any physical contacts between different cell lines through the TNT membrane. Calcein-AM is transferable only through the gap junctions, which can be formed by direct contacts between cells.12 The data confirmed that there was no physical contact between cocultivated cell lines (Figure 2e). Therefore, we can safely conclude that the TNT membrane-based cell-stacking assay can exclusively allow the study of cell paracrine signaling between MDA-MB-231 and hMSC. Next, we investigated the paracrine interactions between human breast metastatic cancer cells [MDA-MB-231 (M)] and each of three different stromal cell lines [hMSC (H), fibroblast NIH-3T3 (N), and myoblast C2C12 (C)] using the TNT membrane platform (Figure 3a). The growing tumor cells actively recruit hMSCs through the secretion of various endocrine and paracrine signals.13 Fibroblasts and myoblasts also have significant effects on the progression of tumors via cytokine release.14 Accordingly, we chose four different cell lines [metastatic cell line (M) and three stromal cell lines (H, N, and C)] to investigate the concerted interactions between cancer cells and stromal cells. The results of the cell co-stacking assay using the Luminex bead-based suspension array system showed that there was a significant increase in RANTES levels for the MbottomHtop co-stacking case, while noticeably increased vascular endothelial growth factor (VEGF) amount was observed for both the MbottomNtop and MbottomCtop co-stacking cases (Figure 3b). The results shown here are in good agreement with direct coculture-based literature results, since ≈60-fold increase in RANTES after 3 day direct coculture of MDA-MB-231 and hMSC and >5-fold increase in VEGF in the case of fibroblast direct coculture with breast cancer cells for 2 days have been reported.2, 15 In addition, we studied the effect of stacked-TNT membrane co-stacking-based cell coculture on the migration and invasiveness of MDA-MB-231 cells by using the transwell migration assay (Merck Millipore, MA, USA) and gelatin zymography for detection of matrix metalloproteinase (MMP), which degrades and remodels the ECM (Figure 3c). The optical density (OD) at 560 nm represents the degree of migration. In the migration assay, the ODs of cocultured samples (MbottomHtop, MbottomNtop, and MbottomCtop) were higher than the ODs in single cell line-cultured cases (M, N, and C), indicating that the migration of MDA-MB-231 cells is positively influenced by stromal cell coculture with the TNT membranes. In the gelatin zymography results, coactivation interactions between MDA-MB-231 cells and stromal cells generated stronger MMP bands than did the single cell line-culture cases. It is known that RANTES acts on cancer cells in a paracrine fashion to increase their mobility, invasion and metastasis, and an increase in RANTES secretion could also promote the expression of MMP 9.10 VEGF is an important factor for tumor proliferation and migration by stimulating angiogenesis, which is typically secreted from mammary or stromal (myo)fibroblasts.16 Our results also suggested that the increased VEGF amount from the MbottomNtop and MbottomCtop cocultures played important roles in activating the migration and invasiveness of the MDA-MB-231 cells. Furthermore, as the TNT membrane-based cell-stacking assay is applicable to any cell types, it could selectively identify key cytokine or growth factor for inducing a specific cell phenotypic result via cell–cell paracrine signaling between many different types of cells. We further tested the TNT membranes as cell coculture platforms with three different cell lines to overcome the limitations of conventional coculture platforms, which were possible due to the optical transparency, transferability, and structural flexibility of the TNT membranes. The transparent characteristic of the TNT membranes allows for the in situ monitoring of cell morphology on each stacked layer by bright-field optical micro­scopy and confocal laser scanning microscopy (CLSM) (Figure 3d and Figure S5, Supporting Information). We also checked for cell apoptosis that could be induced by the pressure on the bottom layer of cells and a gradient of nutrients and oxygen in the multicellular co-stacking condition.17 We performed the TUNEL assay that is based on the incorporation of modified dUTPs by terminal deoxynucleotidyl transferase at the 3′-OH ends of fragmented DNA, the hallmark of apoptosis (Figure S6, Supporting Information). We also compared the actin levels of cells on different layers by labeling and measuring the fluorescence signals from actins by staining with phalloidin-TRITC (Figure S7, Supporting Information). Our results clearly verify that all the cells were alive and maintained their proliferation rate without cross contamination during the cell coculture with up to 4 TNT membrane layers. The cytokine assay results showed that different expression levels of various cytokines and growth factors were observed when M cell line was sandwiched with different stromal cell lines (Figure 3e). Interestingly, RANTES was predominantly observed in the HbottomMmiddleNtop case, while VEGF was the most highly secreted signaling mole­cule in the case of NbottomMmiddleCtop. It is likely that VEGF was mostly secreted by (myo) fibroblasts because it is known that hMSC is a major source of RANTES. Importantly, this result supports the finding that the plasma concentration of RANTES is significantly correlated with the VEGF level in breast cancer patients.18 During multiple cell coculture in which cells are exposed to multiple paracrine signaling cues, each cellular function of interest would be enhanced or reduced depending on the presence of synergistic or antagonistic factors as well as the potency of a specific cytokine in competition, often different from its individual action.19 In this regard, the triggers that induce a specific cytokine to be produced differ between initial cell coculture combinations and cell number ratios, and our approach allows the study (Figure S8, Supporting Information). It should be noted that the understanding of cell–cell interactions between three or more different types of cells is important in identifying a relevant in vitro model to mimic an in vivo-like microenvironment.20 On the basis of the transferability of TNT membranes, we designed and performed the TNT membrane-based cell transfer and shuffling (TNT-CTS) assay, which enabled us to trace the history of sequential stromal activations and to control the sequence of stromal cell coculture with metastatic cancer cells. The sequential coculture of metastatic cancer cells with different types of stromal cells offers systematic ways of understanding of cancer progress by mimicking and simplifying complex in vivo tumor microenvironment via controllable sequential signaling between metastatic cells and different types of stromal cells during metastasis (Figure 4a). First, we confirmed the stability of cell adhesion to TNT membrane and the absence of cross contamination between different cell lines during the sequential coculture (Figure S9, Supporting Information). It has been documented that endothelial cells activated by tumor cells in coculture conditions induce the self-activation of non-activated endothelial cells through autocrine loops.21 On top of this, the TNT-CTS assay can provide a way of sequential activation of different types of stromal cells during cancer metastasis. Overall, the different expressions of various cytokines and growth factors were detected when the same stromal cell lines were used with different coculture sequences. In particular, the major protein species and amount produced from MH, largely varied depending on the culture sequences. RANTES was the major protein when there was no preculture (Figure 4b), but fibroblast growth factor 2 (FGF-2) and interleukin 5 (IL-5) were the major proteins when M was precultured with N and C, respectively (Figure 4c,d). Release of FGF-2 was enhanced in the MH configuration under the sequence effect of MN interaction, whereas IL-5 was stimulated in the same MH configuration under the influence of first MC communications. Notably, the levels of FGF2 and IL-5 amounts in each condition reflected the synergistic interactions between stromal cells in the sequential coculture since the amounts were higher than those produced by M culture alone. It was also reported that FGF-2 is involved in both angiogenesis and hMSC growth,22 and FGF-2 stimulates the growth of hMSC and helps maintain the potential for multidifferentiation of hMSC.23 This further shows how all these signaling are involved in a complicated manner and the TNT-CTS assay can be a versatile and controllable handle in exploring these signalings. Although further work is required to systematically characterize the network of cytokine and growth factor-based paracrine signaling, we suggest that this TNT-CTS method could serve as a powerful tool in studying paracrine signaling-based cancer development and finding new target molecules for cancer diagnosis and treatment. In summary, TNT membrane-based coculture platform facilitates versatile and controllable cell coculture assays and allows for the quantitative analysis of paracrine communications between cancer cells and different types of stromal cells. Moreover, TNT-membrane-based system provides the handles to observe and analyze the specific cell types of interest when different cell types are cocultured simultaneously. We also expect that this TNT membrane cell co-stacking system can be applied to the studies for various types of cell assays and tissue engineering applications. For example, the differentiation of stem cells and cell–cell communications in nerve system can be tuned by changing the density and size of the membrane nanopores and the types and coculture sequences of surrounding cells in a highly controlled and analyzable manner. Furthermore, TNT membrane can be improved and designed to mimic in vivo-like environment, which could offer more in vivo-like results in identifying key signaling molecules between cells and therapeutic targets of a disease. K.C. acknowledges the financial support from the National Research Foundation of Korea (NRF) grant funded by the Korean Government (MSIP) (The National Creative Research Initiative Program for “Intelligent Hybrids Research Center” (Grant No. 2010–0018290)). J.-M.N. was supported by a National Research Foundation (NRF) of Korea (Grant No. 2011–0018198) and BioNano Health-Guard Research Center funded by the Ministry of Science, ICT & Future Planning (MSIP) of Korea as Global Frontier Project (H-GUARD_2013M3A6B2078947). Y.J. and H.L. contributed equally to this work. As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials are peer reviewed and may be re-organized for online delivery, but are not copy-edited or typeset. Technical support issues arising from supporting information (other than missing files) should be to the authors. The is not for the or of any supporting information supplied by the authors. (other than missing should be to the for the