Standardization of esophageal adenocarcinoma in vitro model and its applicability for model drug testing

FLO-1 cell line represents an important tool in esophageal adenocarcinoma (EAC) research as a verified and authentic cell line to study the disease pathophysiology and antitumor drug screenings. Since in vitro characteristics of cells depend on the microenvironment and culturing conditions, we performed a thorough characterization of the FLO-1 cell line under different culturing conditions with the aim of (1) examining the effect of serum-free growth medium and air–liquid interface (A–L) culturing, which better reflect physiological conditions in vivo and (2) investigating the differentiation potential of FLO-1 cells to mimic the properties of the in vivo esophageal epithelium. Our study shows that the composition of the media influenced the morphological, ultrastructural and molecular characteristics of FLO-1 cells, such as the expression of junctional proteins. Importantly, FLO-1 cells formed spheres at the A–L interface, recapitulating key elements of tumors in the esophageal tube, i.e., direct contact with the gas phase and three-dimensional architecture. On the other hand, FLO-1 models exhibited high permeability to model drugs and zero permeability markers, and low transepithelial resistance, and therefore poorly mimicked normal esophageal epithelium. In conclusion, the identified effect of culture conditions on the characteristics of FLO-1 cells should be considered for standardization, data reproducibility and validity of the in vitro EAC model. Moreover, the sphere-forming ability of FLO-1 cells at the A–L interface should be considered in EAC tumor biology and anticancer drug studies as a reliable and straightforward model with the potential to increase the predictive efficiency of the current in vitro approaches.


Abbreviations
A-L Air-liquid EAC Esophageal adenocarcinoma UroM Epithelial medium L-L Liquid-liquid SE Standard error SEM Scanning electron microscopy TEM Transmission electron microscopy TER Transepithelial resistance 2D Two-dimensional 3D Three-dimensional Esophageal Adenocarcinoma (EAC) shows the steepest rise in incidence in Western countries over recent years than any other cancer, due to increased risk factors such as gastroesophageal reflux, smoking, and obesity [1][2][3][4][5][6] . The clinical outcome, however, remains poor. EAC is a highly aggressive and highly mutated cancer with a high degree of heterogeneity [7][8][9] . Main problems for the treatment of EAC presents late detection (the majority of Scientific Reports | (2021) 11:6664 | https://doi.org/10.1038/s41598-021-85530-w www.nature.com/scientificreports/ (Leica Microsystems GmbH, Wetzlar, Germany), objective L40 × /0.50 Ph2 and Nikon Eclipse TE300 (Melville, NY, USA), objective DIC Plan Fluoro L10 × /0.30 Ph1). Once 70-80% confluent, the cells were harvested. The cells were detached with TrypLE Select (Gibco), collected and centrifuged at 200 g for 5 min at room temperature. The cell pellet was then resuspended in culture medium, cells were counted with a haemocytometer and their viability was assessed with the Trypan blue dye (Gibco) exclusion method.
For the evaluation of FLO-1 cell models on different substrates and interfaces while maintained in three different media types, the cells were seeded at a density of 3 × 10 4 cells/cm 2 (according to the ECACC cell collection, recommended seeding density is 1-3 × 10 4 cells/cm 2 ) onto polystyrene Tissue Culture Flasks, Cell culture 9.2 cm 2 dishes (TPP, Trasadingen, Switzerland), 96-microplates (Microplate 96-W, PS, TC, Black/clear B, BD Falcon, Bedford, MA, USA, Cat. No. 34-115-5002-000) and glass coverslips. For successful culturing of FLO-1 cells on Polyethylene Terephthalate (PET) porous membranes (BD Falcon, Cat. No. 353180), the seeding density should be higher (6 × 10 4 cells/cm 2 ) (Supplemental Fig. S1). To establish the liquid-liquid (L-L) growth interface, an appropriate volume of culture medium was added to both the apical and basal compartments of porous membrane inserts, whereas for the air-liquid (A-L) interface, the culture medium was removed from the apical compartment 24 h after the seeding. The cells were maintained in culture for 1, 2, or 4 wks.
Cell viability. FLO-1 cells were seeded onto polystyrene Tissue Culture Flasks and maintained in DMEM, A-DMEM and UroM media. Subculturing was performed every 3-5 days and the viability of FLO-1 models was assessed with a haemocytometer and Trypan blue dye exclusion staining. Viability of FLO-1 models was also assessed with ATP measurements (CellTiter-Glo, Luminiscent Cell Viability Assay, Promega, Madison, WI, USA). The cells were seeded onto 96-microplates (Microplate 96-W, PS, TC, Black/clear B, BD Falcon, Cat. No. 34-115-5002-000) and maintained in D-MEM, A-DMEM or UroM for 1 and 2 wks. The experiment was performed for luminometric measurement of cell growth (viability) according to the standard protocol of the manufacturer. Briefly, 100 μL of the reagent was added to the 100 μL of medium-containing wells of 96-well plate. The reaction mixture was incubated at room temperature in the dark for 30 min to allow for whole microtissue lysis. The samples were then analyzed using a microplate reader Tecan Safire II (Tecan Safire2, Tecan Group Ltd, Männerdorf, Switzerland), which measured the amount of light generated by chemical reaction (luminescence), and produced a result expressed in arbitrary relative light units (RLUs). The intensity of the light is proportional to the amount of ATP in the sample.

Transepithelial electrical resistance measurements. When the FLO-1 cells, maintained in DMEM
and A-DMEM at the L-L and A-L interfaces, reached confluence on the porous membranes, TER was measured using an epithelial voltohmmeter (EVOM & EVOMX, Sarasota, Center Boulevard, FL, USA) with EndOhm-12 chamber. To conduct the measurements, 0.4 mL of fresh culture medium was added to the apical compartment of both, the A-L and L-L interface cultures for 3 min. Then the cell culture inserts were transferred into the EndOhm-12 chamber. TER was measured 5 days per week for 1 month. The measured TER values were corrected by subtracting the mean resistance of blank porous membranes [150 Ωcm 2 ]. Results are expressed as the resistance of a unit area Ωcm 2 (resistance (Ω) x effective membrane area (cm 2 )).
Immunocytochemistry. Indirect immunofluorescence was performed as previously described 44 . Cells grown on glass coverslips for 1 and 2 wks and porous membranes at the A-L and L-L interface for 1, 2 and 4 wks in DMEM, A-DMEM and UroM media were fixed at room temperature with cold (− 20 °C) 100% ethanol for 25 min. After fixation, cells were washed with PBS and incubated in blocking solution containing 1% Bovine Serum Albumin (BSA, Sigma) in PBS for 45 min at room temperature. Primary and secondary antibodies were diluted in 1% BSA in PBS and incubated overnight at 4 °C and 1.5 h at room temperature, respectively. Extensive washing with PBS was performed between each step. Finally, cells were mounted in DAPI-Vectashield medium (Vector Laboratories, Burlingame, CA, USA). For negative controls, incubation with primary antibodies was omitted.
Antibodies. Primary  Transmission electron microscopy. Transmission electron microscopy was carried out as previously described 40,45 . FLO-1 cells cultured at the A-L and L-L interfaces on porous membranes (BD Falcon) for 1, 2 or 4 wks and cell culture dishes for 1 or 2 wks in DMEM, A-DMEM and UroM were fixed with 4% (w/v) formal-

Permeability of model drugs and zero permeability markers.
For permeability assays with model drugs and zero permeability markers, the FLO-1 cells were seeded at a density of 3 × 10 4 cells/cm 2 on polyethylene terephthalate (PET) membranes (1 μm pores, 0.7 cm 2 ) in Millicell 24-well cell culture plates (Merck Millipore, Tullagreen, Ireland). The cells were cultured in two media: DMEM and A-DMEM at the L-L interface.
In order to establish L-L interface, adequate volumes of cell culture medium were added to the apical (400 μL medium/insert) and basolateral compartments (22-24 mL medium/feeder tray). FLO-1 cells were cultured for 1 wk. Prior to conducting the permeability assay, the FLO-1 cells were rinsed twice with assay buffer consisting of Hank's balanced salt solution (HBSS) (Thermo Fisher Scientific, Waltham, MA, USA) and 0.01 M HEPES (Thermo Fisher Scientific, Waltham, MA, USA), pre-warmed at 37 °C. Permeability assays with model drugs and zero permeability markers were carried out as previously described 53 . Model drugs, zero permeability markers and chemicals were obtained from Sigma-Aldrich (Munich, Germany). All solutions of the model drugs and zero permeability markers were prepared in assay buffer (HBSS with 0.01 M HEPES). When the desired concentrations of the model drugs could not be achieved due to low solubility of the compounds, assay buffer with 1% DMSO (Sigma-Aldrich (Munich, Germany)) was used. The investigated model drugs and zero permeability markers and the tested concentrations are shown in Table 1. For the transport studies in apical to basolateral (A-B) direction, 400 μL of pre-warmed test solutions and 800 μL of pre-warmed assay buffer were added to the apical and basolateral compartments, respectively. When assessing permeability in basolateral to apical (B-A) direction, 400 μL of pre-warmed assay buffer was added to the apical and 800 μL of pre-warmed test solutions were added to the basolateral compartment. The 24-well cell culture plates with FLO-1 cells were incubated at 37 °C in a humidified 5%CO 2 /95% air atmosphere throughout the bidirectional permeability assays. The 100 μL samples (except for azilsartan-150 μL) were withdrawn from the acceptor wells at predetermined time points (30,60,90,120, and 180 min), while 20 or 10 μL samples were withdrawn from the donor wells for the model drugs or zero permeability markers, respectively, at 0 and 180 min. The withdrawn volume was replenished with a fresh assay buffer or donor solution.  53 . Quantification of the amount of the zero permeability markers, i.e. FITC-dextrans (FDs) with molecular weights (MW) of 3-5 kDa, 10 kDa, 20 kDa, 40 kDa and 70 kDa, was done by measuring the fluorescence intensity at the appropriate excitation and emission wavelengths (λex/em = 495/515 nm), using microtiter plate reader (Infinite M1000, Tecan, Männedorf, Switzerland). The amount of diffused fluorescent compounds was calculated from a calibration curve.
Permeability data analysis. The apparent permeability coefficients (P app ) were calculated using the Eq. (1): where k d (mol/s or mg/s) is the slope of the linear section in the amount or mass of investigated drug substance permeated to the acceptor side versus time plot, A is the exposed surface area of the cell monolayer (0.7 cm 2 ), C 0 (M or mg/mL) is the average initial concentration of drug substance in donor wells.
Data analysis and statistics. In the study, 2-3 independent experiments were performed to examine the effect of different substrates, interfaces and different media types on in vitro characteristics of FLO-1 cell line. Within each independent experiment, 3-4 technical replicates were examined with different experimental methods. The images shown in figures are representative of at least 2 independent experiments. For TEM and SEM analysis, the high-quality micrographs were assessed independently by two of the authors. The authors had no prior knowledge about the cell culture condition, and how long and in which growth medium the cells were maintained. The calculated ATP test values, TER values, and permeability analysis values are expressed as means ± standard error (SE) of 3-24 replicates. Data were analyzed using the Graph Pad Prism program, version 8.1.2 (GraphPad Software, LaJolla, San Diego, CA, USA). Differences between experimental groups were tested for significance using un-paired two-tailed Student t-test or two-way ANOVA with Tukey's multiple comparisons test. Statistical significance was accepted at P values: 0.05, 0.01, 0.001, and 0.0001, as indicated. All data are expressed as mean ± standard error (SE).

Results
Effect of cell culture medium type on cell proliferation. The esophageal adenocarcinoma cell line FLO-1 is routinely cultured in DMEM medium, supplemented with 10% FBS. In the present study, we tested the cell viability and proliferation of FLO-1 cells in two additional culture media containing a reduced FBS www.nature.com/scientificreports/ concentration. The A-DMEM medium was supplemented with 2.5% FBS for the entire time of culturing, while UroM medium was supplemented with 2.5% FBS only during the first 7 days. Cells were seeded onto polystyrene culture flasks at a seeding density of 3 × 10 4 c/cm 2 . Cells exhibited polygonal shape and successfully adhered and proliferated in all culture media types. The confluence was reached after a maximum of 4 days in vitro ( Fig. 1).
Effect of culture medium type on cell viability. Cells were sub-cultured after they reached 70-80% confluence and their viability was evaluated. In all media, the determined cell viability for passaging cells was very high. The average sub-culturing viability assessed for FLO-1 cells of passages 11-17 was 97.6 ± 0.3% when maintained in DMEM medium (n = 16), 96.6 ± 0.5% in A-DMEM medium (n = 10), and 96.3 ± 1.2% in UroM medium (n = 9). The viability of FLO-1 cells maintained in vitro for 1 or 2 wks was then examined using the ATP test. After 1 wk, the cultures retained a higher number of viable cells when maintained in A-DMEM compared to DMEM and UroM media (two-way ANOVA, Tukey's multiple comparisons test, Fig. 2A). The average number of viable FLO-1 cells was also the highest in A-DMEM medium when maintained in vitro for 2 wks and was significantly different from DMEM and UroM medium (two-way ANOVA, Tukey's multiple comparisons test, Fig. 2A).
Daily observation of the cultures under phase-contrast microscopy revealed that FLO-1 cells detached from the growth support surface with the culture time. At 3 wks, cell shedding was observed in cultures maintained in all media types ( Fig. 2B-D).
The effect of culture medium type on the ultrastructure of the air-liquid and liquid-liquid interface FLO-1 models. Cells successfully adhered and proliferated in all types of culture media when seeded onto the porous PET membrane. FLO-1 cells were seeded onto porous membrane supports at a higher seeding density (6 × 10 4 cells/cm 2 ) as onto culture flasks since, despite higher seeding density, cells reached the confluence after 7 days (in all culture media types; Supplemental Fig. S1), whereas cells seeded onto culture flasks reached confluence after 4 days in vitro (Fig. 1). However, in cultures maintained in UroM at the A-L interface cell shedding was observed soon after reaching confluence or even before confluence was reached (Supplemental Fig. S1), whereas, in DMEM and A-DMEM medium, cell shedding started between 10 and 14 days in vitro (Supplemental Fig. S2). On the contrary, when maintained at the L-L interface in UroM medium, cell shedding was observed only in the last days of culturing (in the last, fourth week of culturing, Supplemental Fig. S2).
The cell culture architecture was interface dependent in all media types. Light microscopy and SEM showed that cells formed spheres at the A-L interface which became more numerous and larger with the culturing time TEM analysis confirmed cell shape heterogeneity in all FLO-1 cell models, i.e., maintained in different media types and interfaces (Fig. 5, Supplemental Fig. S8-10). Cells displayed relatively large, irregularly shaped nuclei and a small amount of cytoplasm. The surface was either smooth or shaped in irregular protrusions or microvilli, or both (Fig. 5, Supplemental Fig. S8-10). Cell spheres were composed of round or oval cells. The cell surface of cells located on the periphery of the spheres was very often shaped in round protrusions, microvilli, or both, although some superficial cells also exhibited smooth surface (Fig. 5A,B). Conversely, cells located inside the spheres had a relatively smooth surface (Fig. 5A,C). Furthermore, cells inside the spheres were frequently more tightly attached via anchoring and adherent junctions and desmosomes (Figs. 5C, 6, Supplemental Fig. S4), while cells on the periphery of spheres with microvilli and protrusion were more loosely attached to other cells via cell protrusions and anchoring junctions (Figs. 5B, 6, Supplemental Fig. S4). Small cell clusters exhibited similar ultrastructural properties as the spheres (Fig. 5D, Supplemental Fig. S8-10).
Polygonal cells forming a monolayer or bilayer very often had a relatively smooth surface with few microvilli (Fig. 5E,F). They were also more tightly attached to neighboring cells via adherent junctions and desmosomes or, less frequently, via tight junctions (Fig. 6, Supplemental Fig. S4). The medium type also influenced the development of cell junctions in FLO-1 models. Namely, FLO-1 cells maintained in A-DMEM and UroM were more www.nature.com/scientificreports/ often tightly connected to neighboring cells by adherent junctions and desmosomes, and even with individual tight junctions, than cells maintained in DMEM (Fig. 6, Supplemental Fig. S4). Except for the increased number of vacuolar structures mainly in the cells in spheres and clusters, no other peculiarities in the intracellular ultrastructure were observed (Fig. 5.).

The effect of culture medium type on cell junction protein expression and TER of air-liquid and liquid-liquid interfaces FLO-1 models. Since FLO-1 models maintained in A-DMEM and UroM
(both containing a lower serum concentration than DMEM medium) have similar ultrastructural properties, in the following experiments, the cells were maintained only in A-DMEM and compared with cells maintained in DMEM.
The expression of junctional proteins in FLO-1 models was further investigated with immunofluorescent labeling. In all media types, regions of cultures with high and low occludin expression were found (Supplemental Fig. S5). Areas with high occludin expression, where occludin was localized at apicolateral borders of adjacent cells, consisted of polygonal FLO-1 cells forming a monolayer or bilayer (Fig. 7A-H). In cell spheres and clusters, however, occludin expression was weak, with the immunofluorescence signal diffusely distributed throughout the cells and not located at the apicolateral borders (Supplemental Fig. S5). Consistent with the TEM observations, FLO-1 models maintained in A-DMEM exhibited higher occludin expression levels than FLO-1 models maintained in DMEM. In both media types, the expression of occludin was higher at 4 wks than at 2 wks of culturing (Fig. 7A-H).
E-cadherin, the protein of adherent junctions, was expressed in all FLO-1 models regardless of interface or medium type (Fig. 7I-P). E-cadherin was localized as a continuous line at the apicolateral border of adjacent cells www.nature.com/scientificreports/ in spheres and clusters as well as in regions where the cells formed a monolayer or bilayer ( Fig. 7I-P). Individual cells not connected to adjacent cells did not display E-cadherin immunoreactivity. The TER of established FLO-1 models was very low, regardless of the interface and the type of culture medium. The TER measurements started when the cells reached confluence (i.e., 1 week after the seeding). At that time, the measured TER values ranged between 10 in 20 Ωcm 2 , with the highest values in FLO-1 cultures maintained in the A-DMEM medium at the L-L interface. The TER values then slowly decreased at the L-L interface over the next two wks and then increased again in the last wk (fourth) of measurements, while at the A-L interface the TER slowly increased over the next 3 wks (Fig. 7R). Throughout the course of the TER measurements, the models cultured at the L-L interface showed higher TER values compared to the A-L interface, except for the last day of measurements, when the A-L interface model showed the highest TER values (Fig. 7R). Nevertheless, all FLO-1 models exhibited very low TER values compared to normal esophageal epithelia, which exhibit TER between 280 and 1000 Ωcm 22,55-57 .

Differentiation potential of FLO-1 cells under optimized conditions. Specific cancer cell lines
have the potency to differentiate under optimized culture conditions and can partially mimic the properties of particular epithelia in vivo [33][34][35][36] . In this study, we have determined that cell viability is increased, and the formation of tight junction enhanced when FLO-1 cells were maintained in A-DMEM (supplemented with 2.5% FBS) ). Therefore, we performed further molecular, ultrastructural, and functional (barrier function) characterizations of FLO-1 models maintained in A-DMEM medium to evaluate their differentiation potential. Cells were maintained in vitro at the A-L and L-L interfaces for short-term period, i.e., 1 wk, since the longer time of culturing resulted in cell shedding (Fig. 2) and because the TER values are relatively high at 1 wk in culture (compared to other examined time points; Fig. 7R. Short-term FLO-1 models were examined for cell junction proteins and cytokeratin 7 (CK-7) by double immunofluorescence labelling (Fig. 8). FLO-1 cells expressed the tight junction protein occludin, but not claudin-2, regardless of the growth interface (Fig. 8A-F). Occludin expression was higher when cells were grown at the A-L interface ( Fig. 8B-C,E-F). It was expressed at apicolateral borders of adjacent cells in some culture areas. Cells growing in clusters or not connected to adjacent cells did not express occludin at the apicolateral borders; the immunosignal was dispersed throughout the cell. FLO-1 cells did not express claudin-4 ( Fig. 8H-I,K-L), nor claudin-8 (Supplemental Fig. S6), regardless of the interface. Next, the expression of CK-7 was examined. CK-7 is a ductal marker expressed in the glandular mucosa in the esophagus 58 . CK-7 was expressed in the basal layers of FLO-1 models at the A-L and L-L interface, mainly localized near the cell periphery (Fig. 8G,I,J,L). Analysis of adherent junctional protein expression showed that both the A-L and L-L interfaces short-term FLO-1 models exhibited E-cadherin expression at the lateral borders (Fig. 8M,O,P,S). In contrast, cells did not express N-cadherin, regardless of their growth at the A-L or L-L interface (Fig. 8N-O,R-S).
Performing ultrastructural analysis of short-term FLO-1 models we have demonstrated that cells mostly grew in a monolayer, composed of polygonal cells with a smooth surface, at both interfaces ( Fig. 9A-D). Cells were often connected to neighboring cells by adherent junctions and desmosome and immature tight junctions (Fig. 9E-H), which were more frequently found at the A-L interface. They also formed weak cell junction, i.e., anchoring junctions (Fig. 9E-G). Small clusters of round/oval FLO-1 cells with rough surface were also observed, but were less frequent compared to long-term (3-4 wks) FLO-1 models (Fig. 9A-D).
Further, cells grown in A-DMEM at the A-L or L-L interface were incubated with fluorescent iNANO-RITC nanoparticles. The measured permeability and fluorescence of endocytosed iNANO-RITC nanoparticles of www.nature.com/scientificreports/ FLO-1 models were then compared to highly differentiated in vitro model of the normal urothelium, which expresses tight junctions, has a high TER and forms the permeability barrier 40,41,59-61 . The interface did not play a significant role in iNANO-RITC nanoparticles permeability in the short-term 1wk FLO-1 models (Fig. 9I). However, the measured fluorescence intensity of endocytosed nanoparticles in cells revealed a higher amount of iNANO-RITC nanoparticles in FLO-1 cells maintained at the A-L interface (Fig. 9I). In comparison to urothelial model, FLO-1 cells exhibited approximately 60% higher permeability (58 ± 3.6%, Fig. 9J-K) and 30% higher fluorescence intensity (30.3 ± 4.5%, Fig. 9J-K) of endocytosed iNANO-RITC nanoparticles, indicating low barrier integrity of FLO-1 models. Taken together, FLO-1 cells poorly mimic the properties of normal esophageal epithelium.  (Fig. 2) and higher TER values obtained in comparison with the A-L model (Fig. 7R), respectively. The apparent permeability coefficients (P app s) for the dextrans with different MW are shown in Fig. 10A. The lowest and the highest P app values in FLO-1 models cultured in DMEM were obtained for the FD 70 kDa and FD 3-5 kDa, respectively, which is in line with their MW. The P app s for all the tested FDs, except for FD 70 kDa, are > 1.0 × 10 −6 cm/s, indicating leakiness of the FLO-1 cell model.
Taking into account the recommendations for demonstrating the suitability of in vitro permeability methods for drug permeability classification from the recent ICH and FDA guidelines 62,63 , the permeability of several model drugs from the low, moderate and high permeability categories was assessed across the FLO-1 cells cultured in DMEM or A-DMEM (Fig. 10B-C). The highly permeable drugs permeate across the cell barrier by passive diffusion, while some of the low and moderately permeable drugs utilize the paracellular pathway or are known to be subject of active transport mechanisms. The obtained P app values for the short-term FLO-1 models in the DMEM and in A-DMEM are shown in Supplemental Table 1 and in Fig. 10B and C, respectively. www.nature.com/scientificreports/ The determined efflux ratios for all tested compounds were close to 1 (0.72-1.33), suggesting that passive diffusion, but not active (efflux) transport, plays the primary role in the permeability of compounds through FLO-1 epithelium.
In the FLO-1 models cultured in DMEM, only a 1.5-fold difference was observed between the obtained P app values for the highly permeable compounds propranolol and antipyrine and the P app values of metoprolol, and the low and moderately permeable drugs. Metoprolol is suggested to be suitable permeability reference standard in intestinal permeability models, having permeability in the proximity of the low/high Biopharmaceutics Classification System (BCS) permeability class boundary 64,65 .
The P app values for the low and moderately permeable model drugs determined in the FLO-1 model cultured in A-DMEM are significantly lower (p < 0.05) than the respective P app values in the model cultured in DMEM. This is in agreement with the higher expression of tight junction proteins when A-DMEM is used as a cell culture medium (Figs. 6, 7A-H). The P app value for highly permeable drug antipyrine is 2-threefold higher than values for the low and moderately permeable model drugs.
The P app values for the low and moderately permeable drugs obtained in both cell models are of the same order of magnitude (× 10 −5 cm/s) as the P app values for the highly permeable drugs. This indicates the FLO-1 model is very leaky, which is substantiated by the measured low TER values (< 20 Ωcm 2 , Fig. 7R). The permeability of the tested model drugs was shown to be from 1.2-fold to 3.4-fold lower than the respective permeability across the PET membranes without cells. Thus, these results do not support the possibility of utilizing the FLO-1 cell line as a model of the esophageal epithelial barrier for drug permeability studies. FLO-1 model in submerged conditions. In the last part of the study FLO-1 models were maintained in submerged conditions (seeded onto cell culture dishes) for 1 or 2 wks in DMEM or A-DMEM and then examined for junctional protein expression and ultrastructure. Like FLO-1 cells grown on porous membranes at the L-L interface, cells formed regions of polygonal-shaped cells and areas with cells growing in multi-layered (3-5 layers) clusters (Fig. 11A-D and Supplemental Fig. S7). Double immunofluorescence labelling revealed that FLO-1 cells expressed the tight junctional protein occludin in culture areas with polygonal cells, while claudin-2 was not expressed, regardless of the medium type. However, occludin expression was higher when cells were grown in A-DMEM (Fig. 11A-D), which is in agreement with the occludin expression in FLO-1 models grown on porous membranes (Figs. 6, 7). Occludin was expressed at apicolateral borders of adjacent cells in regions with polygonal FLO-1 cells forming a monolayer/bilayer (Fig. 11A-D, Supplemental Fig. S7). Cells growing in clusters or not connected to adjacent cells did not express occludin (Supplemental Fig. S7). The immunosignal was dispersed throughout the cells. In both media types, expression of occludin was higher at 1 wk compared to 2 wks of culturing ( Fig. 11A-D, Supplemental Fig. S7). FLO-1 cells exhibited adherent protein E-cadherin at the apicolateral borders of adjacent cells in regions with polygonal cells as well as in clusters in both media (Fig. 11E-H). In contrast, low expression of N-cadherin was detected, regardless of their growth in DMEM or A-DMEM media (Fig. 11E-H).
Ultrastructural analysis of submerged FLO-1 models with TEM substantiated morphology heterogeneity shown with immunolabeling of junctional proteins. Cells exhibited either polygonal morphology, usually forming mono-or bilayer, or round to oval morphology, usually forming cell clusters (Supplemental Fig. S8-10). Cells were connected by protrusions, anchoring junctions, adherent junctions, and desmosomes and occasionally with developing tight junctions (Fig. 12). Immature tight junctions were more frequently seen in the cells maintained in A-DMEM than in DMEM (Fig. 12), supporting higher occludin expression in the FLO-1 model maintained in A-DMEM. Tight and adherent junctions and desmosomes were usually found between polygonal cells and less frequently between adjacent cells in spheres. Polygonal cells often exhibited a relatively smooth surface, with few microvilli. Cells inside spheres were usually tightly attached with anchoring junctions, while cells on the periphery of spheres loosely connected to other cells via cell protrusions (Fig. 12, Supplemental Fig. S8-10). No other peculiarities in the intracellular ultrastructure were observed (Fig. 12).

Discussion
The effect of culturing conditions on the EAC cell line FLO-1 has not yet been evaluated. Therefore, the aim of this study was to characterize this cell line further and evaluate the culture conditions for its optimization and standardization as an in vitro model for EAC.
Cancer cell lines have been widely used to study the signaling pathways of the disease, to define potential molecular markers, and for the development, screening, and characterization of cancer therapeutics. They have also been recognized by biomedical and pharmaceutical companies as a model for drug testing and translational studies 66 . However, every cell line must be characterized with respect to culture conditions and environment in order to optimally mimic the in vivo situation for the investigated process. The optimization of culture conditions thus increases the validity of the in vitro model 67 .
The representability of human cancer cell lines of the in vivo cancer cell population can be augmented by simulation of physiological conditions as closely as possible in contrast to adapting the culturing conditions for maximum cell proliferation 67 . For example, animal-derived serum has been used as a medium supplement for the in vitro propagation of many normal as well as cancer cell lines. However, most cells are not exposed to serum under physiological conditions 68 . Furthermore, serum supplementation of culture media has additional disadvantages. It is a known source of variability, which makes it difficult to standardize experimental protocols. Proteomic and metabolomic studies showed that serum consists of about 1800 proteins and more than 4000 metabolites. Serum composition display qualitative, quantitative, geographical and seasonal batch-to-batch variations, originating in genetic diversity of source herds, animal nutrition and the serum production process 69,70 . The use of serum-free media is, therefore, key to obtaining more accurate results with cancer cell line models,  71 . Moreover, the serum is a potential source of contamination [72][73][74] . FBS may contain adverse factors such as endotoxins, mycoplasmas, viral contaminants, or prion proteins 70 . As cancer cell lines can be used directly for the production of new biological medicinal products In the majority of the studies, FLO-1 cells are maintained in growth medium supplemented with 10% serum (DMEM with 10% FBS [14][15][16]18,19 ). We have successfully cultured FLO-1 cells with high viability using substantially lower serum concentration. Firstly, we have utilized A-DMEM medium, which is enhanced basal medium formulation of DMEM enriched with normal serum components, allowing the cell cultivation at reduced serum concentrations. FLO-1 cells were successfully cultured in A-DMEM medium supplemented with only 2.5% FBS (Figs. 1, 2, 3, Supplemental Fig. S1-2). Secondly, cell growth was tested in serum-free defined medium formulated for epithelial cells, UroM. It consisted of equal amounts of MCDB153 and A-DMEM medium, both allowing growth in low serum concentrations, supplemented by several cell growth-supporting factors such as insulin and hydrocortisone. After an initial cell expansion in the 2.5% FBS supplemented UroM in the first week of culturing, the cells were successfully maintained for additional 1-3 wks in the serum-free UroM (Figs. 1, 2, 3). Taken together, our study demonstrates that cancer cell line FLO-1 can be cultured in substantially lower serum concentrations, which is an important step towards standardization and quality assurance.
In this study, we have characterized the influence of different culturing conditions on the morphological, ultrastructural and functional properties of FLO-1 cells in vitro; i.e., different media types, culturing interfaces and culturing times. The A-L interface culturing of FLO-1 cells was examined, as air is present in the esophagus 26,27 and compared to the L-L interface culturing. We have demonstrated that the culture interface significantly affects the architecture of FLO-1 cell culture (Fig. 3). The latter plays an important role in the prediction efficacy of in vitro tests of new cancer therapies. In particular, the cell culture architecture considerably influences the response to anti-cancer drugs in in vitro cell line models [78][79][80] . Although the in vitro drug testing depends primarily on the use of monolayered subconfluent cell lines, they poorly imitate the conditions in vivo. For a better simulation of in vivo microenvironment, 3D cell cultures are used in cancer studies, such as spherical cultures and 3D tumor spheroids 81,82 . Multi-layered cell cultures, formed by cancer cell lines, which can grow in several layers after they reach confluence, also exhibit many properties of solid tumors, such as drug distribution into the tumor tissue [29][30][31][32] . For example, Movia et al. (2018) have demonstrated that post-confluent multilayered human adenocarcinoma alveolar basal epithelial cells A549 maintained at the A-L interface exhibit increased resistance to investigated benchmarking chemotherapeutics compared to sub-confluent monolayer 2D culture 80 . The increased resistance was comparable to 3D hypoxic tumor spheroids.
We have shown in this study that post-confluent culturing of FLO-1 cells at the A-L interface results in the formation of multi-layered spheres of grape-like appearance, with rough-surface (microvilli and round protrusions) cells predominantly located on the periphery of the spheres and smooth-surface cells inside of the spheres (Figs. 3, 4, 5). Large spheres of FLO-1 cells (100-200 µm) were formed in all media types, but at the A-L interface only, when maintained in culture for 4 wks. Shorter time of culturing i.e., 1-2 wks at the A-L interface resulted in cell forming small clusters of round/oval cells, surrounded by polygonal flattened cells with a relatively smooth   (Figs. 3, 4). Similarly, at the L-L interface, small clusters of cells were formatted (up to 10 cells, diameter 50 µm) among areas of polygonal flattened cells forming a monolayer or bilayer. Culturing time did not affect cell architecture at the L-L interface. In submerged conditions, FLO-1 cultures exhibited similar cell architecture as when maintained at the L-L interface (Fig. 12, Supplemental Fig. S8-10). Morphology and ultrastructure of FLO-1 spheres at the A-L interface were very similar to sphere formations of other cancer cell lines, such as pancreatic ductal adenocarcinoma cells (PDAC) 83 . It has been demonstrated that the formation of cancer cell spheres can be induced by ultra-low-attachment surfaces, hanging-drop method and scaffold culturing [84][85][86] . Here, we have demonstrated that cancer sphere formation can also be induced by the A-L interface culturing. Furthermore, studies have shown that such multicellular three-dimensional spheres enable the growth and proliferation of cancer stem cells, which are defined as a subpopulation (small proportion) of cancer cells that possess a high tumorigenic potential and are responsible for tumor initiation, growth, and metastasis 83,[85][86][87][88] . They are resistant to cytotoxic chemotherapy and ionizing radiation and are believed to be responsible for tumor recurrence and are therefore an important therapeutic target 83,[88][89][90][91] .
Field of EAC is underrepresented in 3D in vitro models and the most common EAC in vitro research models are two-dimensional (2D) cell line cultures. The latter have been used in many studies to evaluate the therapeutic efficacy of different agents and mechanisms of intrinsic drug resistance (reviewed in 92 ). However, they are faced with the problem of standardized protocols and conditions for establishing EAC cell lines 92 . Furthermore, cross-contamination of cell lines and their mistaken identities have been major dilemmas for esophageal cancer research 11 . On the other hand, only a handful of 3D EAC in vitro models can be found in the literature. These include organotypic cultures 93 , matrix-embedded 3D cultures 94 , hollow scaffold-based approaches 95 , tumor explants 96 , multicellular tumor spheroids 97 and recently established EAC organoids 98,99 . Nevertheless, most of the 3D tumor models are capable of mimicking only certain aspects of a tumor environment and thus are not an accurate representation of the complexity and heterogeneity of tumors. 3D models also face inconsistencies in size which impact the reproducibility of the data. Next, the limitations of 3D models are also non-uniform cell attachment and lack of high-throughput methods for tumor model formation 100,101 . There is therefore an urgent need for a high-throughput, relatively inexpensive 3D cell model which is fast and easy to develop, with controllable physical and spatial parameters, that can consistently and accurately represent the tumor microenvironment and provide reliable and reproducible data when used for in vitro drug testing 100 . This is especially true for the field of EAC research, due to the lack of 3D EAC in vitro models.
Thus, the A-L interface effect on post-confluent FLO-1 cell culture architecture can be implemented when using FLO-1 cell line in tumor biology and anticancer drug studies as a reliable and straightforward approach with the potential to increase the prediction efficiency of the model.
While the culturing interface influenced the cell culture architecture, the type of culture medium affected the viability and expression of the junctional proteins. Namely, the number of viable cells and expression of the tight junction protein occludin were increased in FLO-1 models maintained in the A-DMEM and UroM (Figs. 2,6,7,11). Ultrastructural TEM analysis confirmed that tight junctions are occasionally developing between neighboring cells of FLO-1 cultures in areas of flattened polygonal cells in A-DMEM and UroM medium, whereas tight junctions were rarely found in DMEM medium (Figs. 6, 12). Nevertheless, all FLO-1 models displayed very low TER values (less than 20 Ωcm 2 , Fig. 7), suggesting that regardless of the growth medium type and growth interface FLO-1 cells do not form a tight barrier, as demonstrated for several cancer cell lines under optimized culture conditions, such as Calu3, Caco-2, and TR146 33,34,37,38 . In comparison, normal esophageal epithelium, which has a barrier function to protect the underlying tissue from the harmful intraluminal contents, exhibits TER values between 280 and 1000 Ωcm 22,[55][56][57] . In accordance with low TER values, FLO-1 models displayed high permeability of all model drugs with low MW (< 500 Da) (Fig. 10) and low expression of claudin-4, required for the regulation of the paracellular permeability of epithelial cells, which is inversely related to claudin-4 expression 102 . However, the obtained P app values in the FLO-1 model for the zero permeability markers (FDs) having high MW (3-70 kDa) indicate that this cell line, despite its high leakiness, can distinguish between different high MW compounds.
Since our study could not demonstrate ability of this cell line to differentiate between the high and low permeable model drugs with low MW, the FLO-1 cell line is not suitable for utilization in drug permeability assays, i.e., when testing the permeability of potential drug candidates or investigating the effect of different composition of drug formulations on drug permeability. Nevertheless, other possibilities for utilizing the FLO-1 model in anticancer drug studies remain an option, such as testing the cytotoxic potential of anticancer drugs and efficacy in esophageal cancer treatment 19 , bearing in mind the aforementioned A-L interface effect on post-confluent FLO-1 cell culture architecture. Recently, a new therapeutic concept for esophageal cancer has been presented, in which a chemopreventive drug is applied to the distal esophagus in the form of an aerosol under pressure, i.e., pressurized intraluminal aerosol chemotherapy (PILAC) 103 . Furthermore, the A-L interface in vitro model maintained on the semipermeable porous membranes allows for easy post-treatment ultrastructural analysis, exposure of semi-dry apical cell surfaces to investigational formulations, and collection of secretion products from basal compartments. A-L interface EAC in vitro models may therefore represent an important tool in drug screening for the evaluation of aerosol-based therapeutics in the future.
Considering the importance of cadherin expression in tumor growth 104 , we have shown in our study that independent of the medium type, FLO-1 models expressed the adherent junction protein E-cadherin, an epithelial marker. It was located at the lateral borders between adjacent cells, while the mesenchymal marker N-cadherin, a feature of cells with high motility, was expressed cytoplasmatically in a more diffuse way (Fig. 11). Our results are in agreement with Liu et al. (2016), who demonstrated E-cadherin and N-cadherin expression in FLO-1 cells. In cancer cells, external stimuli of different microenvironmental conditions within the tumor during growth and metastasis induce epithelial-to mesenchymal transition (EMT) or mesenchymal-to-epithelial transition (MET) during which expression of N-cadherin and E-cadherin are induced or diminished 14  www.nature.com/scientificreports/ are in a state of high plasticity, which is suggested to play an important role in the metastatic dissemination of carcinoma 105-108 . Liu et al. (2016) have shown that FLO-1 cells undergo EMT and metastasize following subcutaneous injection in mice 14 .
In conclusion, FLO-1 cells represent an important tool in EAC research as verified, authentic and noncontaminated cell line, which is easy to grow and manipulate in vitro. In this study, we have shown that its use as an in vitro model of EAC can be enhanced by optimized culturing conditions. Firstly, low serum and serum-free conditions can be implemented to achieve higher data reproducibility and to increase the validity of in vitro FLO-1 model, as the serum supplement does not represent physiological conditions. Secondly, post-confluent and A-L interface culture should be considered when the cell line is used in drug testing due to their effect on 3D architecture, which considerably influences the response to anti-cancer drugs in vitro. The characterization of the FLO-1 in vitro model, optimization of culturing conditions and their standardization therefore represent an important step towards the improved utility of the existing model, providing insights into the molecular mechanisms of tumor growth and metastasis of EAC and for the development of drug therapies.

Data availability
The datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request.