Dynamic Microenvironment Induces Phenotypic Plasticity of Esophageal Cancer Cells Under Flow

Cancer microenvironment is a remarkably heterogeneous composition of cellular and non-cellular components, regulated by both external and intrinsic physical and chemical stimuli. Physical alterations driven by increased proliferation of neoplastic cells and angiogenesis in the cancer microenvironment result in the exposure of the cancer cells to elevated levels of flow-based shear stress. We developed a dynamic microfluidic cell culture platform utilizing eshopagael cancer cells as model cells to investigate the phenotypic changes of cancer cells upon exposure to fluid shear stress. We report the epithelial to hybrid epithelial/mesenchymal transition as a result of decreasing E-Cadherin and increasing N-Cadherin and vimentin expressions, higher clonogenicity and ALDH positive expression of cancer cells cultured in a dynamic microfluidic chip under laminar flow compared to the static culture condition. We also sought regulation of chemotherapeutics in cancer microenvironment towards phenotypic control of cancer cells. Such in vitro microfluidic system could potentially be used to monitor how the interstitial fluid dynamics affect cancer microenvironment and plasticity on a simple, highly controllable and inexpensive bioengineered platform.

nodules, which showed significantly different metastatic profiles. Similarly, microfluidic systems have been designed to recapitulate complex transport and drug responses at the tumor microenvironment that cannot be emulated on conventional static culture models that lack the dynamics of interstitial fluid flow [15][16][17] . Many studies show the effect of the flow-induced shear stress on the vascular endothelial cells and the changes on their cellular physiology 18 . However, a limited number of studies focus on the effect of flow-mediated dynamic culture conditions on cancer cells and more investigations are needed to better understand the cancer microenvironment 19 .
To further delineate how flow-based shear stress may affect the phenotypic plasticity in terms of switching from epithelial to mesenchymal character of cancer cells, we integrated in vitro cell culture techniques within a dynamic laminar flow-based microfluidic platform. We chose esophageal cancer due to its highly dynamic physiologic tumor microenvironment. The esophagus is exposed to peristalsis contractions during the movement of dietary contents to the stomach, and backward flow of stomach acids in the case of gastroesophageal reflux 20,21 . Moreover, it is continuously subjected to shear forces through its extensive lymphatics and vascular network 22 . We herein engineered a microfluidic system to evaluate the effect of shear stress on a model system to partially represent the microenvironment of esophageal pathologies and report the effects of fluid flow on the phenotypic plasticity of these cancer cells, in effort to demonstrate the efficacy of bioengineered systems as novel in vitro cancer models.

Results and Discussions
Microfluidic platform design for dynamic cancer cell culture. We have designed a microfluidic platform that accommodates cancer cells and optimize their sustained viability and growth. To accomplish this, we first theoretically evaluated and characterized the physical environmental parameters such as channel designs, flow rate and patterns in order to assess and predict their influences on the cells. It is critical that the cells seeded within the microfluidic channel are exposed to uniform and laminar fluid flow and thereby all feel the same physical stress through their membranes 23,24 . The flow in the microfluidic channel changes as a function of location. To evaluate the uniformity of fluid shear stress along the penetration path, we first derived a computational model ( Fig. 1B-D, for details of the model see methods section). Reynolds number (Re) describes whether the Scientific RepoRts | 6:38221 | DOI: 10.1038/srep38221 flow within a system is laminar (Stocks flow) or turbulent according to the ratio of the inertial forces and viscosity forces. Reynolds number in the presented bioengineered platform (Re = 0.0173) belongs to Stokes flow regime for microfluidic systems (Re < 0.1). In the Stokes flow, average wall shear stress at the middle of the channel is calculated as 1.44 × 10 −5 Pa. Mechanical microenvironmental factors such as flow-induced shear stress can modulate the physiology of cancer cells (e.g., cell cycle, metastatic character) 25 . To evaluate local wall shear stress distributions in a microfluidic channel, we performed computational fluid dynamics using finite element software, COMSOL Multiphysics TM . The steady state Navier-Stokes equation for incompressible Newtonian fluids was solved. The flow velocity and shear stress are illustrated in Fig. 1B-D. These results demonstrated that adherent cancer cells in the microchannels experience a simulated uniform wall shear stress of up to 25 μ Pa.
Esophageal adenocarcinoma cells were seeded in microfluidic channels and supplied with continuous laminar fluid flow through gas permeable tubing ( Fig. 1A and Figure S1). The microfluidic system is incubated at 37 °C and 5% CO 2 atmosphere as we reported earlier under constant flow 14 . The interstitial flow values for carcinoma has been adapted from literature 13 . Cells were exposed to 5 μ m/s continuous laminar fluid flow for up to 7 days and the alterations on the plasticity related with epithelial and/or mesenchymal phenotypes of esophageal cancer were monitored by immunocytochemistry for E-cadherin, N-cadherin and vimentin. An identical system at static/no flow conditions where the culture media has been replaced manually twice a day was used as a control.
Microfluidic platform sustains the cell viability and proliferation. Next, viability and proliferation analysis were performed to determine the effects of long-term microfluidic cell culture and to validate engineered dynamic microenvironment as a suitable niche for esophageal cancer cells. Cell viability was assessed for up to 9 days of dynamic culture and compared with static control condition using fluorescent Live/Dead assay. Cell viability in flow conditions was observed to be over 95% compared to the controls (p-value > 0.05, Mann-Whitney Test) (Fig. 1E). Proliferation capacity was evaluated with the immunocytochemical expression levels of nuclear proliferation specific marker Ki-67. Both flow and control groups showed similar levels of proliferation (p > 0.05, Mann-Whitney Test) (Fig. 1E). These results further demonstrated that microfluidic platform presented here can host and sustain cancer cells within a flow-based dynamic culture model.

Switch between epithelial and mesenchymal phenotypes under microfluidic flow. Stochastic
switching of phenotypic states within populations of cancer cells can contribute to malignant features, including chemoresistance and metastasis 14,26 . Metastatic cells, which lose cell-cell adhesion, tend to gain tendency to enter the vascular or lymphatic circulation through epithelial to mesenchymal transition (EMT) 26 . Therefore, we examined how shear flow affects the phenotype of cancer cells by evaluating markers of both epithelial and mesenchymal states. For example, acquisition of a motile mesenchymal phenotype needed for the migration of cancer cells occurs through decrease in E-cadherin, elevation of N-cadherin and vimentin expressions 27 . Esophageal adenocarcinoma cells were exposed to continuous laminar flow for a period of 7 days in microchannels prior to immunocytochemical analysis of mesenchymal markers (E-cadherin, N-cadherin and vimentin) on days 1, 3, 5 and 7.
Cancer cells under flow condition demonstrated a statistically significant decrease in cell-cell adhesion mediated by a decrease in E-cadherin expression and an increase of transendothelial migration capacity through elevation of N-cadherin expression. The phenotypic switch could be observed from the third day of culture under fluid flow ( Fig. 2A,B). Immunocytochemical analysis showed that under flow conditions there is 21.6% decrease in E-cadherin expression at day 1 and 54.8% decrease at day 7 compared to the static control conditions ( Fig. 2A,B).
In parallel, we observed flow mediated 1.24 -times increase in N-cadherin expression at day 1 and 2.41-times increase at day 7 compared to static control conditions ( Fig. 2A,B). Furthermore, we confirmed the gain of mesenchymal features under fluidic flow induced shear stress by detection of the intracellular expression of vimentin. At day 1, vimentin expression under flow conditions was 2.15-times higher than the static controls and this increase continued to a 3.54-times change at day 7 ( Fig. 2A,B). Intriguingly, under laminar fluid shear stress esophageal adenocarcinoma cells present a hybrid epithelial/mesenchymal (E/M) state without any EMT induction through cytokines such as TGF-β . Gaining a hybrid epithelial-mesenchymal (E/M) state, which includes both epithelial and mesenchymal phenotype (epithelial E-cadherin is expressed simultaneously with mesenchymal markers N-cadherin and vimentin), can occur during the switch between the epithelial to mesenchymal transition and the mesenchymal to epithelial transition 28,29 . Indeed, the hybrid, E/M phenotype is associated with malignant features such as an acquired drug resistant state 30 . Our results indicate that fluid flow triggers the EMT in the absence of any type of chemical or biological stimuli within the bioengineered microenvironment.

Cellular morphology and stiffness changes upon fluid shear stress. Physical microenvironmental
stimuli can dramatically influence the cellular metabolism and cell membrane properties such as cell stiffness 31,32 . We evaluated the physical effects of shear stress on cell membrane and cell morphology. The cell membrane of cancer cells cultured under flow for 7 days was analyzed with atomic force microscopy (AFM). Importantly, to avoid bias due to the fixative driven stiffness, AFM analysis was performed on viable cells. We observed that cells exposed to 7 days of flow showed an increase in cellular stiffness compared to a static condition (Fig. 2C). The membrane stiffness for flow-conditioned cells was measured as 2.24 ± 1.02 kPa and for static controls as 1.87 ± 1.02 kPa. The cell stiffness is often associated with the remodeling of the actin cytoskeleton 31,33 . Actin filament network in the cellular organization, determines the shape, stiffness and cell motility as well as plays a crucial role in transduction of mechanical signals, muscle contraction, cell division, invasion and EMT [34][35][36] . The presence of stressed F-actin fibers and microtubules may induce fibroblast-like changes in cell shape during EMT 37,38 . We used fluorescently labeled actin binding marker Phalloidin to visualize stressed actin filaments in select subcellular locations and fibroblast-like changes in cell morphology after 7 days of cell culture under flow conditions (Fig. 2D). The switch between epithelial and mesenchymal states of cancer cells, is not only a biological process at a molecular level but also involves physical and mechanical regulations at each step of the cascade. Our data from cancer cells exposed only to flow-based shear stress is in agreement with the previously reported correlation between biochemical and physical changes at phenotype transformation phase 17 . Changes in actin cytoskeletal network of esophageal adenocarcinoma cells represent the alteration of expressions profile of both epithelial and mesenchymal markers. These results indicate that flow-based shear stresses on cancer cells may allow the existence of a hybrid epithelial/mesenchymal phenotype (E/M) 39,40 . Dynamic microenvironment promotes stem cell like character of esophagus cancer cells. Drug detoxifying enzyme aldehyde dehydrogenase (ALDH) positivity, as a poor prognostic factor, is identified in lung, prostate, breast, colon and esophageal cancers and has been linked to mesenchymal fate of cells and EMT 41,42 . We observed a statistically significant shift in ALDH-positive population once cultured under laminar shear stress and their ALDH level has been monitored up to 7 days (Fig. 3A,B and Figure S2). Under flow conditions the ALDH-positive population increases from 6.0% to 20.9%, whereas the increase under static controls remains from 4.4% to 10.8%. ALDH-positive cells have greater colony forming ability, which can be associated with higher metastatic progression and higher number of tumor-initiating cells 43,44 . Indeed, the colony formation assay under flow conditions demonstrated 2.2-fold higher capacity than the static controls (Fig. 3C,D). Cancer cells cultured under static conditions express ALDH 45,46 . In our experimental design, both cancer cells in static and flow conditions proliferate (thereby cell number increases) over 7 days of culture period. The culture conditions such as seeding density and duration is attributed to change the physiology of cells in vitro 47 . The increase in ALDH-positive cells under static condition is attributed to the increased cell-cell contact over the time course of the experiment. The proportion of ALDH-positive cancer cells changes with cell culture density due to the expression of different ALDH isoforms 48 . When static and flow conditions were compared to each other at time points (days 1, 3, 5 and 7) only the ALDH-positive increase in flow mediated cells has been observed as significant. The physical fluid shear stress induces the elevation of ALDH-positive cell fraction together with hybrid E/M state, suggesting a shift towards aggressive and cancer stem cell-like phenotype 49 .
Further, we investigated the effect of physical stimulation through flow-based shear stress on well-studied cancer cell populations. CD44, a multifunctional class I transmembrane glycoprotein, and CD24, a small cell surface protein, are widely expressed in various cancer cell populations, and are described as cancer stem-like cells capable of initiating invasion, metastasis, therapeutic resistance, heterogeneity and plasticity in tumors 30,50-52 . Our results demonstrate that CD44 low /CD24 low cells show phenotype switching to CD44 high /CD24 low and CD44 high / CD24 high under both static and flow conditions towards day 7 (Fig. 3E-G and Figure S3). Cell surface markers are dynamically expressed, even under static conditions, as a result of active signaling mechanisms and microenvironmental conditions. A decrease in the expression of CD44 low /CD24 low cells was statistically significant on the first and third day of culture under flow when compared with the results under static condition. The percentages of CD44 high /CD24 low (mesenchymal-like; M) and CD44 high /CD24 high (hybrid phenotype; EM) cell populations under flow conditions were significantly higher than under static conditions. Fluid flow could be a driver of phenotypic switching into the EM and M phenotypic states. These states are shown as one of the main reasons that cancer therapy is often unsuccessful 2,30,53 . In in vitro esophageal cancer studies, a CD44 high /CD24 low sub-population predicts a reduced response to chemotherapy, radiation and a high capacity for metastasis, e.g. with breast cancer, because of the increased aggressive behavior of the tumor 54,55 . Recent evidence now suggests that CD44 High /CD24 High (M) phenotype, between epithelial and mesenchymal states, can be significantly more metastatic and indicative of an adaptive resistance to chemotherapy 30 . Based on these reports, it is possible to gain insights into CD44 and CD24 expression and their possible roles in metastasis and resistance to cancer therapies. However, no studies have been reported on the status of these markers under flow conditions [54][55][56] .
In Fig. 3E-G we monitor how flow changes the CD24 and CD44 expressions on esophagus cancer cells. JH-EsoAd1 esophagus cancer cell line is a heterogeneous cell line, which consists of multiple sub-populations, and has been extensively characterized elsewhere [57][58][59] . The in vitro culture course of the cells leads alterations on their physiology and cell surface markers such as CD24 and CD44. In this study, we have analyzed the CD44 high CD24 low , CD44 high CD24 high and CD44 low CD24 low populations under flow and static conditions. The CD44 high CD24 low sub-population has been attributed to more tumorogenic and cancer stem cell character as well as exhibiting EMT characteristics 60,61 . Our results in Fig. 3E shows that under flow conditions CD44 high CD24 low population is significantly higher than the static condition for 1 week of culture period. This observation correlates with findings in Fig. 2A,B as flow induces the EMT. Similarly to slight increase in ALDH expression under (n ≥ 5), star (*) indicates p < 0.05. Immunocytochemical analysis showed that fluid flow caused 21.6% decrease in E-cadherin expression at day 1 and 54.8% decrease at day 7 compared to the static control conditions ( Fig. 2A,B). In parallel, we observed flow mediated 1.24-times increase in N-cadherin expression at day 1 and 2.41-times increase at day 7 compared to control conditions ( Fig. 2A,B). Furthermore, we confirmed the gain of mesenchymal features under fluidic flow induced shear stress by detection of the intracellular expression of vimentin. At day 1, vimentin expression under flow conditions was 2.15-times higher than the static controls and this increase continued to a 3.54-times at day 7 ( Fig. 2A,B). To define the changes in experimental groups, the ratio of flow and static mean fluorescence intensity values of static samples were used. (C) Atomic force microscopy (AFM) was used to measure the cell stiffness. Bar graphs demonstrating the Young's Modulus of cells under static control and flow conditions. (D) Esophageal cancer cells were stained with Phalloidin to visualize the actin microfilaments, and with 4′ , 6-diamidino-2-phenylindole (DAPI) to visualize nuclei under static and laminar microfluidic flow conditions. (Scale bars: A and C, 50 μ m). All data were expressed as mean ± S.E.M., n = 3, star (*) indicates, p < 0.05.
Scientific RepoRts | 6:38221 | DOI: 10.1038/srep38221 static conditions the shift from CD24 low to CD24 high in static conditions can be attributed to elevation of cell density within the confined environment over time 62,63 . Microfluidic cancer cell culture platform as test bed for chemotherapeutic administration. The presented microfluidic-based dynamic cancer microenvironment can be potentially utilized in chemotherapeutic administration studies. The phenotypic plasticity and metastasis capacity can be altered by the interactions between cancer cells and dynamic microenvironment. We evaluated the response of esophageal cancer cells to a chemotherapeutic agent (e.g., docetaxel) in well-defined and controlled administrative regimes. Docetaxel (DTX) is used as a second line therapy in aggressive cases of esophageal adenocarcinoma and has been successfully evaluated in the clinic for over a decade. Miyanaga et al. have reported that docetaxel is not cytotoxic for low concentrations (< 10 nM) and enhances cell death in concentration dependent manner 64 . We applied 3 nM DTX in our microfluidic platform to observe the chemotherapeutic response of cancer cells, in particular to evaluate the CD44 and CD24 expression. Continuous delivery of DTX decreases the expression of CD44 high /CD24 low population from 46.54 ± 3.1% to 22.58 ± 0.4% and increases CD44 high /CD24 high populations from 45.6 ± 6.4% to 56.8 ± 5.8% under flow (Fig. 4A-C). Under DTX administration when the fluid flow is applied CD44 low population is increasing from 7.8 ± 3.2% to 18.1 ± 8.7%. We observe that CD24 expression at flow conditions increases both in combination with DTX administration and its absence. The correlation of the increase in CD24 expression with the increase of malignancy of the cancer cells has been previously reported 52 . During disease progression, cancer cells are able to demonstrate adaptation and phenotypic changes such as escape from apoptosis, resistance to chemotherapeutics and increase in motility, invasiveness and migration, which are triggered by the transition between epithelial and mesenchymal phenotype 49,53 . Flow-based phenotypic plasticity we observe here can be a potential mechanism for enabling cells to survive and adapt to environmental changes. Understanding the nature of cancer cells in their native dynamic microenvironment, in terms of physical and physiological components, is essential 65,66 . Intrinsic mechanical forces such as interstitial fluid flow-based shear stress might be one of the reasons for inconsistencies between in vitro experimental and clinical observations. Bioengineered in vitro cancer microenvironments and model platforms could potentially provide more reliable information than traditional in vitro static culture systems. This study demonstrates how flow can direct the phenotypic plasticity in esophageal cancer cells in terms of modulating the EMT based molecular expression signatures, clonogenic potential, ALDH, and CD44/CD24 expressions. The presented platform can be further utilized in cancer research for understanding transcriptomic and genomic variations (e.g., signaling mechanisms, cellular heterogeneity, mutations) and controlling physiochemical parameters (e.g., pH and hypoxia level, matrix stiffness). Investigating flow-based plasticity and its effects on cell fate, through microfluidic model platforms will potentially further accelerate the progress of cancer management and pharmaceutical applications. In particular, with advances in microfluidic cancer cell culture techniques, developing new personalized therapy strategies investigating the unknowns in tumor heterogeneity and chemotherapeutic resistance are interesting for future research in cancer.

Methods
Cell culture. The esophageal adenocarcinoma cell line, JH-EsoAd1 57 , a kind gift from Dr. James R. Eshleman (The Johns Hopkins University, Baltimore, Maryland, USA), which was previously characterized by Dr.Eshleman, was maintained in complete culture medium containing high-glucose Dulbecco's Modified Eagle Medium (DMEM with 4 g/L glucose, 4.0 mM L-Glutamine) (Gibco by Life Technologies, Grand Island, NY, USA) supplemented with 10% heat-inactivated FBS (Gibco Life Technologies, Grand Island, NY, USA), and 1% Penicillin-Streptomycin (Gibco by Life Technologies, Grand Island, NY, USA) at 37 °C with 5% CO 2 in microfluidic chips. In the static culture group, the media was changed manually two times per day to minimize the potential effects of hypoxia. In the flow group, adherent cells were continuously supplied with fresh medium using a programmable syringe pump (NewEra Pump Systems Inc., NY) at a flow rate of 2 μ l/min. The aldehyde dehydrogenase (ALDH), CD44, CD24 measurements and epithelial to mesenchymal marker stainings were performed on days 1, 3, 5 and 7 to determine the phenotypic plasticity under flow and static conditions.  Figure S1). Polystyrene layers from petri dishes were used for the bottom of the chips to achieve effective cell attachment. Outer dimensions of microfluidic chips components were cut to 25 × 41 mm dimensions (VersaLASER; Universal Laser Systems Inc., AZ, USA). Three channels (width: 4 mm, length: 27 mm) were cut from 1.5 mm thick PMMA plates to support the cells with cell culture media in the channels. The channel length is referred to as the distance between the inlet and outlet ports. The polystyrene bottom layer and a 1.5 mm thick PMMA layer (with channels) were bound by a DSA film. The upper part of the chip, with inlet and outlet ports, was cut from 3.175 mm thick PMMA sheets. The diameter of the inlet and outlet ports was 0.7 mm. All components of the microfluidic chip were sterilized with 70% ethanol followed with ultraviolet light exposure under laminar flow hood for an hour. Entire assembly process was performed under aseptic conditions. Before connecting the silicon-based tubing to the inlet/outlet ports, a suspension of JH-EsoAd1 cells was manually inoculated into the channels of the microfluidic chip.
Cell seeding in microfluidic chips. Cells at 5.58 × 10 4 cells/channel seeding density were loaded in each microfluidic channel and incubated for 24 hours to achieve cell adhesion and homeostasis. Gas-permeable silicon tubing 0.010" inner diameter × 0.030" outer diameter (Tygon, ColeParmer, USA) was attached to the inlet and outlet ports of the microfluidic chips using epoxy glue (Devcon, Danvers, MA, USA). Syringes containing cell culture medium were connected to the tubing. For the static condition, the culture medium was replaced twice per day by manual pipetting.
Computational modeling of flow in the microfluidic platform Platform. Theoretical investigation of the fluid system was performed to evaluate the shear stress effect on the adherent cells. The Reynolds number in the microchannel was calculated from Equation (1): where Q is the volumetric flow rate (2 μ L min −1 ), D H is the hydraulic diameter of the pipe, ρ is the fluid density, μ is the dynamic viscosity of the fluid and A is the cross-sectional area of the pipe. For rectangular microchannels with comparable heights and widths, the hydraulic diameter can be expressed as Equation (2): The boundary layer thickness of this fluid system at the middle of the microchannel (L/2) was calculated to be 41 mm, according to the following Equation (4): To evaluate the local wall shear stress distribution, we performed computational fluid dynamics using a finite element software, e.g. Comsol Multiphysics TM . The steady state Navier-Stokes equation for incompressible Newtonian fluids was solved. As the boundary layer thickness of this microfluidic system at middle of the microchannel (L/2) is larger than the height of the microchannel, the shear stress in the fluid can be expressed by Equation (5) τ µ = ∂ ∂ u y (5) where u is the fluid velocity along the microchannel, and y is the distance from the no-slip wall at the bottom of the microchannel.
Viability and proliferation analysis.. Cell viability was determined using a LIVE/DEAD Viability/ Cytotoxicity kit (Life Technologies, USA), which provides a two-color fluorescence assay for the determination of live (green) and dead (red) cells. Both static and flow conditions were analyzed with ImageJ (NIH, Bethesda, MD, USA) for a minimum of 5 images for each condition. Proliferation of the cells was measured by staining using a mouse monoclonal antibody against human Ki-67 (Abcam, ab15580, Cambridge, MA, USA). The cells were visualized using a fluorescence microscope. ALDEFLUOR assay. Aldehyde dehydrogenase (ALDH) activity was determined using an ALDEFLUOR Kit (Stem Cell Technologies, Vancouver, Canada) according to the manufacturer's instructions. Cells were suspended in ALDEFLUOR assay buffer containing ALDH substrate (boron dipyrromethene aminoacetaldehyde (BAAA)) and incubated for 45 minutes at 37 °C. As a negative control, an aliquot from each sample was treated with an ALDH-specific inhibitor (diethylaminobenzaldehyde (DEAB)). Stained cells were resuspended in ALDEFLUOR assay buffer and analyzed using a BD FACSAria flow cytometry (BD Biosciences, San Jose, CA, USA).
Colony forming unit assay. Colony forming assays were performed to evaluate the colonogenecity of the cell population under flow and static control conditions. Three hundred viable esophageal cancer cells were seeded into each channel of the microfluidic chips. On the day 4, silicon tubing was connected to the flow chips. After 3 weeks of incubation, the colonies were fixed with 4% PFA and stained with 0.01% crystal violet (Sigma Aldrich, St. Louis, MO). The colonies were counted using a microscope. Statistics. All experiments in this study were performed at least three times unless other specified. All graphs and numerical values in the figures are presented as the mean ± the standard error of the mean (SEM).
A non-parametric Mann Whitney U test was used to determine the significance of the differences between flow and static samples. Repeated measure ANOVA was used to analyze the trend of differences among different time points. Data were analyzed using SPSS V.21.0 for Windows (SPSS, Chicago, Illinois, USA). The level of statistical significance was set at p ≤ 0.05.