A novel accessible in vitro model of proliferative vitreoretinal diseases shows facilitated epithelial mesenchymal transition through aquaporin-1 Authors: Azine Datlibagi, Clément Duveau, Anna Zein-El-Din, Nargis Bolaky, Vincent Lambert, Jason Perret, François Willermain, Christine Delporte, Elie Motulsky Published: 2025 ABSTRACT AbstractThis study aimed to establish a new in vitro model of proliferative vitreoretinal disease (PVD) using ARPE-19 cells and to evaluate the role of Aquaporin-1 (AQP1) in this model. ARPE-19 cells, stably transfected without or with AQP1, were differentiated in a nicotinamide-enriched medium (N + medium) prior to a 5 day treatment with either N + medium, medium devoid of nicotinamide (N−), or N− medium supplemented with TNF-α and TGF-β (TNT) to induce epithelial-mesenchymal transition (EMT). Human proliferative vitreoretinal disease (PVD) membranes and retinal pigmented cells isolated from ocular globes were collected. RNA sequencing was performed on both ARPE-19 treated cells and human samples. Cell migration assays and immunofluorescence analyses were performed to assess the effects of cell treatments. After differentiation in N + medium, control and AQP1-transfected ARPE-19 cells treated with TNT in N− medium (CT TNT N− and AQP1 TNT N−) exhibited features of EMT, including elongated spindle-like morphology and enhanced migratory capacity. AQP1 TNT N− cells shared the highest transcriptional similarities with human PVD membranes. Moreover, in the absence of TNT, AQP1-transfected ARPE-19 cells in N− medium exhibited increased migration and early EMT markers, suggesting a role for AQP1 in these processes. Importantly, AQP1 expression appeared to slightly upregulate matrix metallopeptidase 2 (MMP2) expression, which may promote cell invasion. Zinc finger E-box-binding homeobox-1 (ZEB1), a key EMT regulator, was upregulated in conditions containing TNT and AQP1. Our findings suggest that the AQP1 TNT N− model best mimics human PVD membranes, both in transcriptomic profile and cellular behavior.Supplementary InformationThe online version contains supplementary material available at 10.1038/s41598-025-28041-2.Keywords: Cellular culture, Proliferative vitreoretinal diseases, Retina, In vitro model, Aquaporin−1Subject terms: Biochemistry, Biological techniques, Cell biology, Molecular biology, Diseases, Eye diseases, Retinal diseases Introduction Under pathological conditions, human retinal architecture and function can be altered by the development of membranes such as proliferative vitreoretinopathy (PVR), proliferative diabetic retinopathy (PDR) and epiretinal membranes (ERM). These vision-threatening pathologies are usually clustered under the general proliferative vitreoretinal diseases (PVD) expression and are often linked to ocular trauma or retinal detachment in the case of PVR, and to inflammation or oxidative stress for membranes developing in PDR. Although the origin and proportion of cells found in membranes of PVD may vary, epithelial-mesenchymal transition (EMT) seems to be a key process occuring in all PVDs1–7. Many in vivo and in vitro models of PVD have been developed and used throughout the years to study the pathogenesis of PVD and evaluate potential therapeutic targets. However, most in vitro models have been progressively abandoned or are rarely used due to inherent flaws8. Accumulative evidence favors using human primary adult and stem cell-derived RPE cells as a basis for in vitro models of retinal diseases. However, increased costs, local ethical difficulties obtaining such cells and long differentiation protocols greatly limit accessibility and routine use of these models. ARPE-19, a spontaneously arising retinal pigment epithelial (RPE) cell line that is easily available and widely used, shows chromosomal abnormalities, loss of several RPE genes and proteins’ expression, and a lack of mature RPE morphology over the years9–12. However, limitations to the use of ARPE-19 cells due to their lack of typical RPE phenotype and functionality may be alleviated using a differentiation protocol recently described, involving nicotinamide supplementation in the culture medium to obtain mature RPE showing a cobblestone morphology within a few weeks13. On the other hand, aquaporin-1 (AQP1) is expressed in PVD membranes14 and has been described to play a significant role in cell proliferation and EMT in various cell types, suggesting a potential role in PVD development through a yet unknown mechanism15–22. As there is conflicting evidence showing expression of AQP1 in native adult human RPE cells, but that its expression is significantly increased during differentiation and maturation of human pluripotent stem cells into RPE and PVDs, it has become necessary to investigate its role as either an active key player of EMT or a secondary regulated protein during the EMT process leading to PVD development23. Our work aimed to establish a new in vitro cell model of PVD by using for the first time differentiated ARPE-19 cells prior to inducing EMT through the addition of both TNF-α and TGF-β in the culture medium24,25. Furthermore, we aimed to evaluate the role and effect of Aquaporin-1 (AQP1) expression in this in vitro PVD model. Finally, we compared the differential transcriptome profile of our in vitro PVD models to that of human PVD membranes. Materials and methods Cell culture The human retinal pigment epithelial cell line (ARPE-19) was obtained from the American Type Culture Collection (ATCC, CRL-2302™) and authenticated by short tandem repeat analysis with Powerplex 16HS kit (Promega, Madison, WI, USA) according to manufacturer’s instructions (ECACC, Salisbury, UK). ARPE-19 cells were stably transfected by electroporation (270 V, 950µF, using a GenePulser II System) without (ARPE-19) or with 8 µg plasmid encoding human AQP1 (ARPE-19 + AQP1). Cells were selected for 7 days using 800 µg/ml of G418 antibiotic and were maintained at 37 °C and 5% CO2 in Dulbecco’s Modified Eagle Medium (DMEM)/F12 supplemented with 10% fetal bovine serum (FBS), 100 U/mL penicillin/streptomycin, and 2 mM L-glutamine (Thermo Fisher Scientific, Waltham, MA, USA). Cells were then maintained in a medium made of MEM alpha, 2 mM GlutaMax, 1% FBS, 100 U/mL penicillin/streptomycin, and 0.1 mM non-essential amino acids (N− medium), or in N− medium supplemented with 0.25 ng/mL taurine, 0.013 ng/mL triiodo-L-thyronine, 20 ng/mL hydrocortisone, 10 mM nicotinamide, and 1% N1 supplement (N + medium). ARPE-19 and ARPE-19 + AQP1 cells were first differentiated for 28 days in N + differentiation medium. After 28 days of differentiation, ARPE-19 cells were treated for 5 days with either N + medium (CT CT N+) or N− medium containing both 10 ng/mL TNF-α and 10 ng/mL TGF-β (referred hereafter as TNT) (CT TNT N−) or without TNT (CT CT N−). After 28 days of differentiation, ARPE-19 + AQP1 cells were treated for 5 days with either N− medium (AQP1 CT N−) or N− medium containing TNT (AQP1 TNT N−) (Fig. 1). Transcriptome analysis was conducted in four independent experiments for each treatment condition (n = 4). ARPE-19 cell treatments. ARPE-19 and ARPE-19 + AQP1 cells were first differentiated in N + medium for 28 days, followed by a further 5 days of differentiation in N + medium for the control condition ARPE-19 CT N+ (referred to as CT CT N+). Alternatively, cells were subjected to dedifferentiation for 5 days in N− medium, without or with TNT, under the following conditions: ARPE-19 + AQP1 CT N− (referred to as AQP1 CT N−), ARPE-19 TNT N− (referred to as CT TNT N−), and ARPE-19 + AQP1 TNT N− (referred to as AQP1 TNT N−). After 33 days of culture, cells were harvested for RNA extraction and RNA sequencing to determine their transcriptome and identify differentially expressed genes (DEGs) in each treatment condition relative to the control condition (CT CT N+). AQP1: Aquaporin-1; ARPE-19: retinal pigment epithelium (RPE) cell line from post-mortem ocular globes; DEGs: differentially expressed genes; RPE: retinal pigment epithelium; PVR: proliferative vitreoretinal diseases; N+: nicotinamide-enriched medium; N−: medium without nicotinamide; EMT: epithelial-mesenchymal transition; TNT: 10 ng/mL of TNF-α and 10 ng/mL of TGF-β in the culture medium. Pictures of the different culture conditions were captured at the end of the differentiation process (after 28 days in N + medium) and daily during the subsequent 5-day treatment period using a Motic AE31-E (Leica, Wetzlar, Germany) inverted trinocular microscope equipped with the Motic Images Plus 3.0 ML system. Imaging was performed in grayscale using a Plan-Achromat 10x/0.25 objective lens. Human samples Forty-four human PVD samples from patients undergoing surgery and thirty-nine post-mortem ocular globes were obtained with the approval of the ULB-Erasme Ethical committee (P2019/066 A2022/136) and of the Eye Bank (CHU Liège). RPE cells were isolated from post-mortem ocular globes following a previously described protocol26. Briefly, the protocol involved mechanical removal of the eye’s anterior segment, iris and lens, detachment of the vitreous and neuroretina, multiple rinsing with D-PBS and incubation of the shell with a solution of Trypsin-EDTA containing DNase I followed by a gentle brushing of the RPE cells to detach them from the Bruch membrane and collect them. Ethics declaration The study was conducted with the approval of the ULB-Erasme Ethical committee (P2019/066 A2022/136) and of the Eye Bank (CHU Liège), with no need for informed consent in compliance with Belgian law. The study was performed in accordance with the Declaration of Helsinki. The human tissue experiments complied with the guidelines of the ARVO Best Practices for Using Human Eye Tissue in Research (Nov2021).” Information about patients and post-mortem ocular globe are presented in supplemental Table S1 and supplemental Table S2. RNA sequencing and bioinformatics analysis RNA from treated cells and human samples was isolated using an AllPrep DNA/RNA/Protein Mini or AllPrep DNA/RNA micro kit (Qiagen, Hilden, Germany). RNA quality was verified using a 5200 Fragment Analyzer (Agilent Technologies, Santa Clara, CA, USA). Indexed cDNA libraries were prepared using the TrueSeq Stranded mRNA Sample Preparation Kit (Illumina, San Diego, CA, USA) for treated cells or the Ovation Solo RNA-seq Kit (Tecan, Mannedorf, Switzerland) for human samples. Multiplexed libraries were loaded onto a NovaSeq 6000 system (Illumina, San Diego, CA, USA), and sequences were generated using a 200-cycle kit. Paired-end reads were aligned to the human reference genome GRCh38.90.gtf from sftp.Ensembl.org using the STAR 2.5.3a software27. After transcript assembly, gene-level counts were obtained with HTSeq-0.9.128 and normalized to 20 million aligned sequences. Differentially expressed genes (DEGs) with a log2 fold change (log2FC) > 1 or < -1 and a false discovery rate (FDR) < 0.05 were identified using the EdgeR quasi-likelihood method in Degust software (version 4.1.1), applying a minimum of 10 counts per million (cpm) in at least 20 samples for MB/RPE and a minimum of 0.5 cpm in at least 1 sample for cells. Functional annotation-based enrichment analysis of DEGs was performed using ToppFun, part of the ToppGene suite29. Gene Ontology (GO) terms with an FDR < 0.05 were considered significantly enriched among the DEGs. Cell migration assay ARPE-19 and ARPE-19 + AQP1 cells were seeded in 96-well Essen ImageLock plate (10⁴ cells/well; Essen BioScience, Ann Arbor, MI, USA) and treated as described above. Two independent experiments were performed with 24 replicates per treatment condition. Cell-free scratches, 700–800 μm wide, were created using a WoundMaker tool (Essen BioScience, Ann Arbor, MI, USA). After two washes with 0.01 M PBS to remove detached cells and debris, cells were maintained in their respective treatment media for the duration of the experiment (24 h). Phase-contrast images of each scratch were automatically captured by the optical module of the Incucyte S3 system (Essen BioScience, Ann Arbor, MI, USA) using a 10x objective at 45-minute intervals over 24 h. Digital image analysis was carried out using IncuCyte 2021 A software to measure the percentage of scratch closure over time relative to the initial scratch width. Samples with non-uniform initial scratches were excluded from the analysis. Based on this, analyses were conducted on 13/48, 31/48, 30/48, and 42/48 wells for the conditions CT CT N+, AQP1 CT N−, CT TNT N−, and AQP1 TNT N−, respectively. Immunofluorescence ARPE-19 and ARPE-19 + AQP1 cells were grown in 8-well Millicell plate (10⁴ cells/well; Merck Millipore, Carringtohill, Ireland) for 28 days. Cells were rinsed with 0.01 M PBS, fixed in 4% paraformaldehyde, incubated for 60 min in 0.01 M PBS containing 5% goat or donkey serum (Sigma-Aldrich, St. Louis, MI, USA) and 0.1% Triton X-100 (Sigma-Aldrich, St. Louis, MI, USA), and incubated overnight at 4 °C with primary antibodies diluted in 0.01 M PBS with 5% goat or donkey serum (Sigma-Aldrich, St. Louis, MI, USA) (see Table 1). The following day, cells were washed with 0.01 M PBS and incubated for 60 min at room temperature with secondary antibodies diluted in 0.01 M PBS containing 5% goat or donkey serum (Table 1). Table 1List of primary and secondary antibodies and microscope settings.AntibodyReferenceCompanyDilutionHost speciesMaster gainLaser powerVimentin (VIM)5741Cell Signaling, Danvers, MA, USA1/100Rabbit710 V0.7mWN-Cadherin (CDH2)13,116Cell Signaling, Danvers, MA, USA1/400Rabbit750 V1mWMMP-240,994Cell Signaling, Danvers, MA, USA1/200Rabbit810 V2.5mWCOL1A172,026Cell Signaling, Danvers, MA, USA1/200Rabbit780 V2.5mWRPE65ab235950Abcam, Cambridge, UK1/100Rabbit720 V0.95mWMiTFab3201Abcam, Cambridge, UK1/100Mouse820 V2.2mWZEB170,512 SCell Signaling, Danvers, MA, USA1/800Rabbit800 V2.8mWAQP1SBG003Custom-made Eurogentec, Seraing, Belgium1/200Goat700 V1.3mWGoat anti-rabbit8889Cell Signaling, Danvers, MA, USA1/1000Goat//Goat anti-mouse4410Cell Signaling, Danvers, MA, USA1/1000Goat//Donkey anti-goatA32758Thermo Fisher Scientific, Waltham, MA, USA1/500Donkey// List of primary and secondary antibodies and microscope settings. After a final PBS wash, nuclei were stained with 1 µg/mL DAPI (4’,6-diamidino-2-phenylindole) diluted in 0.01 M PBS for 30 min. A negative staining control was performed for each experiment by omitting the primary antibody. Goat anti-AQP1 antibodies were produced by goat immunization using a 26 amino acid peptide corresponding to the 25 C-terminal amino acids of hAQP1 with an additional cysteine residue on the N-terminal for coupling to the keyhole limpet hemocyanin carrier protein, and affinity purified (Eurogentec, Seraing, Belgium). Images were captured using a Zeiss LSM 900 confocal microscope equipped with an Axiocam 705 system and lasers (Carl Zeiss, Oberkochen, Germany). Imaging utilized a Plan-Apochromat 20x/0.8 objective and spectral separation based on a prism and photomultiplier tubes, configured for the following detection ranges: DAPI: 400–535 nm, AF594: 580–630 nm, and AF647: 630–700 nm. For DAPI, AF594, and AF647, pinhole aperture sizes of 2.21 AU, 1.74 AU, and 1.55 AU were used, capturing optical sections of 48 μm, 53.2 μm, and 55.9 μm, respectively. Imaging parameters included a pixel dwell time of 1.03 µs, a line averaging of 1, and an 8-bit depth. The main gain and laser power settings were individually optimized for each antibody but remained constant across all culture conditions for each specific antibody (Table 1). Statistical analysis Cell migration data were analyzed and graphically represented using GraphPad Prism version 9 (RRID: SCR_002798). Results are presented as mean ± SD. Data were analyzed using a two-way ANOVA followed by Bonferroni’s post hoc test. A p-value less than 0.05 was considered statistically significant. Results Cell morphology After 28 days of differentiation in the N + medium (Day 0), ARPE-19 cells formed a cuboidal epithelium typical of RPE cells despite the absence of pigmentation (CT CT N+), while ARPE19 + AQP1 cells exhibited a more elongated morphology, lacking the classical RPE cobblestone appearance (AQP1 CT N−) (Fig. 2). Morphology of treated cells. Images of the different culture conditions (taken using grayscale imaging at 10x magnification with a Motic AE31-E trinocular inverted microscope) at the end of cell differentiation (i.e., after 28 days in N + medium, labeled as Day 0), and subsequently after 1 day and 5 days of additional CT or TNT treatment. AQP1: Aquaporin-1; ARPE-19: RPE cell line derived from post-morterm ocular globes; CT: control; N+: nicotinamide-enriched medium; N−: nicotinamide-free medium; TNT: 10 ng/ml TNF-α and 10 ng/ml TGF-β in the culture medium.Scale bar: 100 μm. ARPE-19 cells differentiated into mature RPE and cultured for an additional 5 days (Day 5) in N + medium retained their cobblestone morphology (CT CT N+). However, cells treated in N− medium supplemented with TNT underwent EMT, displaying elongated, spindle-like morphology and reduced intercellular adhesion (CT TNT N−) (Fig. 2.). ARPE-19 + AQP1 cells exposed to N− medium (AQP1 CT N−) for 5 days (Day 5) exhibited mild morphological differences compared to control cells (ARPE-19 treated for 5 days in N + medium; CT CT N+). Their epithelial conformation was altered during dedifferentiation, though to a lesser extent than in the presence of TNT (AQP1 TNT N−). Importantly, no morphological differences were observed between ARPE-19 (CT TNT N−) and ARPE-19 + AQP1 (AQP1 TNT N−) cells subjected to dedifferentiation for 5 days with TNT (Day 5). Furthermore, the EMT process occurs very rapidly, with visible changes observed as early as the first day of treatment (Fig. 2). To distinguish the effect of N + removal on differentiation from that of AQP1, ARPE differentiated into mature RPE and cultured for an additional 5 days in N− medium (CT CT N) were compared to the CT CT N + and AQP1 CT N− conditions. Removal of N + medium did not induce a marked loss of mature RPE phenotype (Supplementary Fig. S1). Transcriptome analysis A bioinformatics analysis was conducted on RNAseq data to identify DEGs between the conditions CT TNT N−, AQP1 TNT N− and AQP1 CT N− compared to the control condition (CT CT N+), as well as between human PVDs membranes (MB) and human RPE from eye donors (Fig. 3A). A Venn diagram was then generated to determine the common DEGs among different conditions (Fig. 3A). Transcriptome analysis. (A) Venn diagram identifying the common DEGs between different conditions. (B) 10 most significantly enriched biological processes (Top10) and epithelial-mesenchymal transition (“EMT”) -related enriched pathways based on pathway enrichment analysis performed on the 2232 DEGs common between Mb and AQP1 TNT N−. AQP1: Aquaporin-1; ARPE19: retinal pigment epithelium cell line derived from post-mortem ocular globes; CT: control; DEGs: differentially expressed genes; MB: human membranes from proliferative vitreoretinal diseases; RPE: human retinal pigment epithelium; N+: nicotinamide-enriched; N−: nicotinamide-free; TNT: 10ng/ml of TNF-α and 10ng/ml of TGF-β in the culture medium. This transcriptomic analysis revealed that the AQP1 TNT N− cellular model shares approximately 49% DEGs with the MB (2232/4572 DEGs), compared to 28% (1291/4572) or 44% (2000/4572) of DEGs shared between AQP1 CT N− or CT TNT N− and MB, respectively. Among the 2232 DEGs shared between the AQP1 TNT N− condition and MB, further analysis revealed that 846 DEGs had a Log2FC < 1 in both conditions, while 872 DEGs had a Log2FC > 1 in both. Additionally, 175 DEGs exhibited a Log2FC > 1 in AQP1 TNT N− but < 1 in MB, whereas 339 DEGs showed a Log2FC < 1 in AQP1 TNT N− but > 1 in MB. These findings indicate that 1718 out of the 2232 DEGs, representing 76.97% of the total, are regulated in the same direction in both AQP1 TNT N− and MB, suggesting a strong similarity in gene expression profiles between the cell line model and the human clinical samples. A total of 1146 DEGs were identified when comparing CT CT N− to CT CT N+, with 10 most significant enriched biological processes mainly related to cell adhesion, morphogenesis, extracellular matrix, and with four additional significant enriched biological pathways related to EMT (Supplementary Fig. S2A). When comparing AQP1 CT N− to CT CT N−, 1830 DEGs were identified, with 10 most significant enriched biological pathways mainly related to cell adhesion and morphogenesis, and with three additional significant enriched biological pathways related to EMT (Supplementary Fig. S2B). Furthermore, the comparison of AQP1 CT N + to CT CT N + showed 1730 DEGs, with 10 most significant enriched biological pathways mainly related to cell adhesion and morphogenesis, and with one additional significant enriched biological pathways related to EMT (Supplementary Fig. S2C). These data highligh that both nicotinamide removal and AQP1 expression induce transcriptomic modifications associated with EMT. A pathway enrichment analysis was then performed on the DEGs common to both MB samples and the AQP1 TNT N− model. This analysis revealed that the 10 most significantly enriched biological processes were related to cell adhesion, morphogenesis, development and locomotion. Additionally, several pathways associated with EMT were also significantly enriched, with a total of 98 non-redundant DEGs (Fig. 3B). Cell migration analysis Considering that the transcriptome analysis showed enriched biological processes related to cell adhesion and locomotion, cell migration was evaluated in all cell culture conditions. When comparing conditions in pairs, the effect of treatment on cell migration is statistically significant (p < 0.05) for each comparison, except for CT TNT N− versus AQP1 TNT N−, after a variable time delay depending on the conditions (Fig. 4). CT CT N− also showed an increased migration speed compared to CT CT N+, but AQP1 CT N− exhibited an even higher migration speed than the CT CT N− condition, suggesting a stronger effect of AQP1 on migration than the partial loss of differentiation induced by the removal of N + medium (Supplementary Fig. S3). CT CT N− showed an intermediate migration rate between CT CT N + and AQP1 CT N−. Similarly, AQP1 CT N + cells showed a higher migration rate than CT CT N + cells (Supplementary Fig. S3). Cell migration. Filling rate (%) over time (in hours) of the scratch under different conditions. The color of the curves is specific to a condition: blue for CT CT N+; purple for AQP1 CT N−; green for CT TNT N−; red for AQP1 TNT N−. Results are the mean ± SD (standard deviation) of filling rate (%) (n = 2; with 24 replicates/treatment condition). Data were analyzed using a two-way ANOVA followed by Bonferroni post-hoc test. The statistically significance is indicated as: *: p < 0.05; **: p < 0.01; ***: p < 0.001; ****: p < 0.0001. The conditions with the highest migration rates were CT TNT N− and AQP1 TNT N−. Compared to time zero, the effect of time on cell migration becomes significant at 2h15min, 3 h, 4h30min, and 14h15min for AQP1 TNT N−, CT TNT N−, AQP1 CT N−, and CT CT N+, respectively. Expression and localization of proteins As the transcriptome analysis showed enriched biological processes related to EMT, the expression and localization of proteins involved in this process or constitutively present in the RPE were investigated. The AQP1 mRNA levels were detected only in AQP1 CT N− and AQP1 TNT N− (Fig. 5A). In addition, the AQP1 mRNA levels were significantly increased in both AQP1 TNT N− and MB compared to CT CT N + and RPE, respectively (Fig. 5A). AQP1 protein was expressed in the cytoplasm and maybe also the membrane of AQP1 CT N− and AQP1 TNT N− (Fig. 5B). Expression of AQP1 and RPE cell markers. (A) mRNA expression are expressed as median ± interquartile range of counts per 20 million transcripts (cp20m) (cell culture: n = 4; RPE: n = 39; MB: n = 44). When the FDR value is significant (< 0.05), the -log10FDR value is indicated on the graph. (B) Immunofluorescence images in the four cell culture conditions (taken with a 20x magnification with Zeiss LSM 900 confocal microscope). Negative controls (Neg CT) were performed for each culture condition, but only one is shown in the figure for illustration purposes. Protein expression (in red) and cell nuclei (in blue; DAPI staining). AQP1: aquaporine-1; MB: human membranes from proliferative vitreoretinal diseases; RPE: human retinal pigment epithelium. The mRNA levels of constitutive markers of RPE cells, Retinal pigment epithelium-specific 65 kDa protein (RPE65) and Microphtalmia-associated Transcription Factor (MITF), were significantly decreased in both AQP1 TNT N− and MB conditions compared to CT CT N + and RPE, respectively (Fig. 5A). By opposition, the mRNA levels of EMT markers, such as vimentin (VIM) and N-cadherin (CDH2), were significantly increased in both AQP1 TNT N− and MB conditions compared to CT CT N + and RPE, respectively (Fig. 6A). Expression of EMT markers. (A) mRNA expression are expressed as median ± interquartile range of counts per 20 million transcripts (cp20m) (cell culture: n = 4; RPE: n = 39; MB: n = 44). When the FDR value is significant (< 0.05), the -log10FDR value is indicated on the graph. (B) Immunofluorescence images in the four cell culture conditions (taken with a 20x magnification with Zeiss LSM 900 confocal microscope). Negative controls (Neg CT) were performed for each culture condition, but only one is shown in the figure for illustration purposes. Protein expression (in red) and cell nuclei (in blue; DAPI staining). CDH 2: N-Cadherin; MB: human membranes from proliferative vitreoretinal diseases; RPE: retinal pigment epithelium; VIM: Vimentin. At the protein level, a strong cytoplasmic and possibly membranous expression of RPE65 was observed in CT CT N + and AQP1 CT N−, while RPE65 expression was nearly absent in CT TNT N− and AQP1 TNT N−. MITF shows nuclear staining accompanied by limited cytoplasmic staining in CT CT N+ (Fig. 5B). In contrast, in AQP1 CT N−, nuclear staining was weaker, suggesting a less diffuse distribution, while cytoplasmic staining appeared unchanged. In CT TNT N− and AQP1 TNT N−, no staining is observed. These results were consistent with a strong decrease in MITF mRNA levels in AQP1 TNT N− (Fig. 5B). VIM is exclusively expressed in the cytoplasm across all four conditions (Fig. 6B). Its localization follows a fusiform pattern in TNT-treated conditions and in AQP1 CT N−, with a more pronounced effect in TNT-treated samples. These observations concur with the observed elongated cell morphology (Fig. 2). N-cadherin appeared to be primarily located in the membranous area in CT CT N+ (Fig. 6B). However, in AQP1 CT N−, its distribution was slightly decreased in the cytoplasmic membrane, whereas in CT TNT N− and AQP1 TNT N−, it was predominantly cytoplasmic (Fig. 6B). The mRNA levels of Zinc finger E-box-binding homeobox-1 (ZEB1) is significantly increased in both AQP1 TNT N− and MB conditions compared to CT CT N + and RPE, respectively whereas CTNNB1 (β-catenin) mRNA expression is not significantly altered (Fig. 7A). The expression of ZEB1 was, as expected, almost absent in CT CT N+. However, its expression, both cytoplasmic and nuclear, was increased in AQP1 CT N− cells, as well as in all cell groups treated with TNT (Fig. 7B). β-catenin is located in the cytoplasm in all 4 conditions (Fig. 7B). However, in the CT CT N + condition, it also appears to be present at the membrane level, although absent in the nuclei. In the AQP1 CT N− condition, its membrane presence seems reduced, and some nuclei appear to express it. With TNT treatment, nuclear localization of β-catenin increases (Fig. 7B). Expression of transcription factors involved in the EMT process. (A) mRNA expression are expressed as median ± interquartile range of counts per 20 million transcripts (cp20m) (cell culture: n = 4; RPE: n = 39; MB: n = 44). When the FDR value is significant (< 0.05), the -log10FDR value is indicated on the graph. (B) Immunofluorescence images in the 4 cell culture conditions (taken with a 20x magnification with Zeiss LSM 900 confocal microscope). Negative controls (Neg CT) were performed for each culture condition, but only one is shown in the figure for illustration purposes. Protein expression (in red) and cell nuclei (in blue; DAPI staining). CTNNB1: β-catenin; MB: human membranes from proliferative vitreoretinal diseases; RPE: retinal pigment epithelium; ZEB1: Zinc finger E-box-binding homeobox-1. Given the role of Matrix Metallopeptidase-2 (MMP2) and collagen type-1 alpha-1 (COL1A1) as markers of extracellular remodelling occurring in PVDs, their gene expression and protein localization were also studied. The MMP2 and COL1A1 mRNA levels were significantly higher in both AQP1 TNT N− and MB compared to CT CT N + and RPE, respectively (Fig. 8A). The protein expression of MMP2 was almost absent, with very weak cytoplasmic staining in CT CT N+ (Fig. 8A). In AQP1 CT N−, cytoplasmic staining was significantly more intense than in CT CT N+, whereas in CT TNT N−, its staining appeared comparable to AQP1 CT N−. However, AQP1 TNT N− showed a strong cytoplasmic staining when compared to control conditions (Fig. 8B), consistently with the strong overexpression of MMP2 mRNA in AQP1 TNT N− compared to CT CT N+ (Fig. 8A). The protein expression of COL1A1 was not detected in the CT CT N + condition but was weakly found in the cytoplasm of AQP1 CT N− and strongly in CT TNT N− and AQP1 TNT N− (although without a notable difference between these conditions) (Fig. 8B). These data align with the stronger expression of COL1A1 mRNA in AQP1 TNT N− (Fig. 8B). Expression of various proteins involved in extracellular matrix remodelling occurring in PVD. (A) mRNA expression levels are expressed as median ± interquartile range of counts per 20 million transcripts (cp20m) (cell culture: n = 4; RPE: n = 39; MB: n = 44). When the FDR value is significant (< 0.05), the -log10FDR value is indicated on the graph. (B) Immunofluorescence images in the 4 cell culture conditions (taken with a 20x magnification with Zeiss LSM 900 confocal microscope). Negative controls (Neg CT) were performed for each culture condition, but only one is shown in the figure for illustration purposes. Protein expression (in red) and cell nuclei (in blue; DAPI staining). COL1A1: collagen type-1 alpha; MB: human membranes from proliferative vitreoretinal diseases; MMP2: matrix metallopeptidase-2; RPE: retinal pigment epithelium. Discussion This study aimed to characterize an in vitro PVD model based on ARPE-19 cells and to determine the impact of AQP1 in this context. Our results indicate that both CT TNT N− and AQP1 TNT N− ARPE-19 cells exhibit morphological characteristics of EMT, confirmed by increased cell migration and the modified expression of various RPE, EMT, and PVD markers (compared to control). These findings align with previous studies showing that the addition of TNT to ARPE-19 cells induced a more elongated cell morphology, enhanced migration and cytoskeletal rearrangements30. The use of TNT on ARPE-19 cells was reported to be more efficient to induce EMT than TGF-β treatment alone, which was insufficient to effectively disrupt cell-cell contacts and initiate EMT31. While other methods, such as alternative cytokines or mechanical stimulation, have been explored, they have shown limited efficacy and have rarely been used8. Indeed, transcriptomic analysis indicates that the AQP1 TNT N− condition shares the highest number of DEGs with human PVD (49%), which is 5% and 21% more than CT TNT N− and AQP1 CT N−, respectively. However, AQP1 TNT N− and CT TNT N− exhibit similar and significantly higher cell migration rates compared to other culture conditions. These data suggest that AQP1 does not further enhance the cell migration induced by TNT. Regarding the role of AQP1 in PVDs, AQP1 CT N− cells display a mesenchymal morphology, share 28% of DEGs with MB, and exhibit significantly higher migration rates than CT CT N + cells after only 6 h. However, both AQP1 TNT N− and CT TNT N− migrate faster than AQP1 CT N−, suggesting that adding TNT to the culture medium further enhances the EMT process in the presence of AQP1 expression alone. While the combined effect of AQP1 and TNT appears synergistic, not all experimental results concur. Based on morphological and migration analyses, it seems that TNT alone is sufficient to establish an effective in vitro PVD model. However, the transcriptomic analysis shows that AQP1 expression combined with TNT treatment provides a model that more closely resembles human pathological membranes (MB). Given the overexpression of AQP1 in PVDs previously described14 and confirmed within the present study, the AQP1 TNT N− model may represent the best in vitro PVD model to date compared to CT TNT N−. Since some studies have described both the expression of AQP1 and AQP11 in human and stem cell-derived RPE cells, we might have expected to observe a modification of the expression of other aquaporins to compensate for the lack of AQP1 expression in non-transfected ARPE cells23,32,33. However, besides AQP1 and AQP11, no other aquaporin was expressed in ARPE-19 cells, either transfected with the AQP1-vector or not. Furthermore, treatment with TNT seemed to decrease AQP11 expression, while AQP1 transfection had no effect on any of the other aquaporins’ expression. However, it is interesting to note that human PVDs also showed a decreased expression of AQP11 when compared to human RPE cells while showing an increase of AQP1, underlining the similarities with the in vitro PVD model using differentiated RPE cells. To further characterize the in vitro PVD model, the expression of selected proteins related to RPE and EMT function, and PVDs was evaluated by transcriptomic analysis and IF. The findings highlight the modification of key RPE, EMT and PVD markers. Specifically, VIM (which is an essential protein for cell adhesion and migration), that is significantly altered during ARPE-19 dedifferentiation induced by TNT and AQP1. In AQP1 CT N− cells, EMT is initiated, suggesting that AQP1 expression alone, even in the absence of TNT, promotes migration and cellular dedifferentiation. Furthermore, our results support a previous study indicating that tight junction disruption contributes to the morphological and functional changes associated with EMT in RPE34. Indeed, during EMT induction, we observed an increase in mRNA expression and cytoplasmic protein localization of N-cadherin, suggesting a loss of its cell junction function. Increased N-cadherin expression is associated with reduced E-cadherin levels, leading to the dissociation of the β-catenin-cadherin complex, followed by nuclear activation of ZEB1, which further suppresses E-cadherin expression and promotes VIM expression. This ZEB1-driven signaling cascade ultimately induces EMT and the formation of PVDs35. Our findings support this mechanism, as evidenced by the increased ZEB1 protein staining in cells undergoing EMT. The increased protein expression of ZEB1 observed in AQP1 CT N−, CT TNT N−, and AQP1 TNT N− conditions, as well as the elevated ZEB1 mRNA levels in AQP1 TNT N−, may be attributed to the intrinsic role of AQP1 and TNT in inducing ZEB1 expression. Furthermore, with its increased expression, ZEB1 is a key player in PVD pathogenesis. Therapeutic targeting of ZEB1 with miR-194 has been proposed as a strategy to inhibit EMT of RPE cells36. Following EMT induction in our in vitro PVD model, COL1A1 and MMP2 expression increases, whereas MITF expression decreases. It has been suggested that COL1A1 overproduction contributes to collagen accumulation in epiretinal membranes, leading to vitreoretinal traction in PVD. MMP2, a metalloproteinase involved in extracellular matrix degradation, is known to promote cell migration and tissue invasion37. Our results suggest that MMP2 expression may play a role in PVD development, despite conflicting findings in the litterature38,39. However, this hypothesis requires further validation. AQP1 has already been described to promote cell migration by increasing the levels of β-catenin and focal adhesion kinase (FAK), and to activate the Wnt/β-catenin signaling pathway (a pathway known for its role in EMT)20,40. In addition, the Wnt/β-catenin signaling pathway is involved in EMT of RPE cells and PVD41–43. Therefore, given the increased AQP1 expression in human PVD membranes and the β-catenin redistribution from the plasma membrane to the nucleus and cytoplasm, we hypothesize that overexpression of AQP1 in ARPE-19 cells may participate to EMT through this mechanism, that is further enhanced with the addition of TNT in the medium. This hypothesis may be comforted by previous in vitro models of PVD which were based on mechanical disruption of cell adhesions such as wound healing assay44. This is further supported by evidence showing that EMT in RPE cells is prevented by adequate cell-cell adhesions, and that disruption of these adhesions is considered necessary for EMT to happen45. Therefore, increased AQP1 expression may facilitate the EMT process in RPE cells and prevent adequate differentiation when maintained in N + medium. Furthermore, since it has been described that TGF-β alone is not sufficient to induce EMT in RPE cells, and that it requires a loss of cell-cell contacts, it may be interesting to investigate the use of induced constitutive AQP1 expression associated with TNF-α adjunction alone, to better understand the signaling pathways involved in the EMT of RPE cells8,31,46. Additionally, AQP1 expression slightly enhances MMP2 expression, even in the presence of TNT, which may increase the cell migration ability. The precise role of AQP1 in diseases involving EMT remains to be precisely understood, but its expression has been found to be increased in different tumors, and to increase cell motility and invasiveness in such cases15–17,47. Taken together, all our results suggest that ARPE-19 cells differentiated using N + medium (CT CT N+) subjected to TNT treatment in N− medium (CT TNT N−) is an accessible in vitro PVD model. Data comparison between AQP1 CT N− and CT CT N− and between AQP1 CT N + and CT CT N + indicated an effect of AQP1 on transcriptomic modifications related to cell migration, morphogenesis and EMT, as well as an increase in cell migration. However, comparison between AQP1 TNT N− and CT TNT N− showed transcriptomic modifications without any additional effect on cell migration. Besides, the comparison of CT CT N− to CT CT N + showed an effect of nicotinamide removal on transcriptomic modifications related to cell adhesion, morphogenesis, extracellular matrix, as well as an increase cell migration. The AQP1 TNT N− model shares the highest degree of similarities with human PVD, and might therefore be a new reference in vitro PVD model. Notably, the incorporation of AQP1 overexpression in the TNT-treated ARPE-19 cells marks a significant advancement over the previously reported ARPE-19 TNT model. While TNT alone induces EMT-related changes, the AQP1 TNT N⁻ model more closely mimics the pathological environment of PVD by further amplifying transcriptomic reprogramming and functional alterations such as increased migration. Furthermore, the use of differentiated cells prior to inducing EMT allows a better mimicking of the in vivo situation, in which normal RPE cells tend to undergo EMT following trauma, retinal detachment or inflammation for PVD to develop. The marked increase in DEGs and stronger phenotypic changes validate the added pathophysiological relevance of this enhanced model, thereby improving its utility for mechanistic studies and therapeutic screening. However, further studies will be needed to evaluate the exact mechanism through which it induces such loss of differentiation, as well as to identify key pathways that are mainly regulated by TNT, AQP1 or both, and these will be addressed in future research projects. The AQP1 TNT N− model has some limitations considering that the expression of AQP1 appears to impede the ARPE-19 differentiation induced with N + medium. Therefore, the use of another model in which an inducible AQP1 expression could be triggered after the differentiation process in the ARPE-19 cells may offer a useful solution in the future to clarify the role of AQP1 alone or in combination with TNT on the EMT process. Further studies will also be necessary to validate the findings of our study at the protein level in PVD samples. Nevertheless, the AQP1 TNT N− ARPE-19 cell model developed herein offers added value and great promise in studying the mechanisms underlying PVD, while awaiting potentially more refined in vitro PVD models in the future. Conclusions This study successfully developed an in vitro PVD model based on differentiated ARPE-19 cells, with the AQP1 TNT N− condition proving to be most representative of human PVD membranes. Our findings highlight the synergistic role of AQP1 and TNT in enhancing EMT progression, cell migration and transcriptomic pattern, reinforcing AQP1’s potential role in PVD pathogenesis. Given its ability to promote β-catenin translocation and slightly increase MMP2 expression, AQP1 emerges as a promising therapeutic target. Future research involving inducible AQP1 expression will be crucial to understanding its role in PVD progression and improving in vitro modeling strategies. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 Acknowledgements The authors express their gratitude to all members of the vitreoretinal surgery team from the Department of Ophthalmology of the Hôpital Universitaire de Bruxelles (HUB) - Erasme Hospital (Brussels, Belgium), in particular Dr Demols, for their precious help in collecting PVD samples during surgery. The authors also thank Professor Caroline Verhoeven for statistical advice (Laboratoire d’Enseignement des Mathématiques, Université Libre de Bruxelles, Belgium), Dr Audrey Penning for providing advices regarding cell migration assay (Laboratory of Cancer Epigenetics, Faculty of Medicine, ULB-Cancer Research Center (U-CRC), Université libre de Bruxelles, Institut Jules Bordet, Brussels, Belgium), Françoise Gregoire for logistical help (Laboratory of Pathophysiological and Nutritional Biochemistry, Université libre de Bruxelles) and both Dr Frederick Libert and Dr Anne Lefort for performing RNA sequencing (BRIGHTcore, Brussels, Belgium). Use of artificial intelligence tools: Not applicable. Some figures were partly created using Servier Medical Art (https://smart.servier.com/), licensed under a Creative Commons Attribution 4.0 Unported License (CC BY 4.0). Abbreviations One-way analysis of variance Aquaporin-1 ARPE-19 + AQP1 cells differentiated for 28 days in N + medium and then treated for 5 days with either N− medium ARPE-19 + AQP1 cells differentiated for 28 days in N + medium and then treated for 5 days with N− medium containing 10ng/ml TNF-α and 10ng/ml TGF-β Spontaneously arising retinal pigment epithelia cell line derived from the normal eyes of a 19-year-old male who died from head trauma in a motor vehicle accident ARPE-19 cells stably transfected with aquaporin-1 N-cadherin Collagen type-1 alpha-1 ARPE-19 cells differentiated for 28 days in N + medium and then treated for 5 days with N + medium ARPE-19 cells differentiated for 28 days in N + medium and then treated for 5 days with N− medium containing both 10 ng/mL TNF-α and 10 ng/mL TGF-β β-Catenin Differentially expressed genes Deoxyribonucleic acid 2,2’,2’’,2’’’-(Ethane-1,2-diyldinitrilo)tetraacetic acid Epithelial-mesenchymal transition Epiretinal membranes Fetal bovine serum Fold change False discovery rate Gene ontology Membranes from proliferative vitreoretinal diseases Microphthalmia-associated transcription factor Matrix Metallopeptidase 2 Nicotinamide-enriched medium Medium devoid of nicotinamide Ribonucleic acid Human retinal pigment epithelium Phosphate buffered saline Proliferative diabetic retinopathy Proliferative vitreoretinal disease Retinal pigment epithelium Retinal pigment epithelium-specific 65 kDa protein 10 ng/ml TNF-α and 10ng/ml TGF-β Vimentin Zinc finger E-box-binding homeobox-1 Author contributions C.Delporte, A.D., E.M., F.W., J.P. designed and supervised experiments. A.D., C.Duveau, A.Z.E.D. and N.B., carried out experiments. A.D., C.Duveau, A.Z.E.D., J.P. and C.Delporte analyzed experimental results. A.D. and C.Duveau wrote the manuscript. C.Delporte. E.M., F.W., A.D. and C.Duveau obtained funding. V.L. provided materials. C.Delporte, A.D. and C.Duveau confirm the authenticity of all the raw data. All authors have reviewed and approved the final manuscript. Funding This work was supported by the Université Libre de Bruxelles, the Fonds de la Recherche Scientifique - FNRS (Fund for Scientific Research) (Research Credit grant J.0128.23; E.M.); Fund Iris Research from the King Baudouin Foundation (2022-J1820690-226884; F.W.). A.D. benefited from a research fellowship of the Fonds Erasme pour la recherche médicale, and fundings for Individual Research Projects in Ophthalmology (IRPO) from the Fonds pour la Recherche en Ophtalmologie (FRO) and from Vocatio. C. Duveau. is a Research Fellow of the Fonds de la Recherche Scientifique - FNRS (Fund for Scientific Research). Data availability Raw RNA-sequencing data of ARPE-19 cells and AQP1-transfected ARPE-19 cells have been deposited in the NCBI Sequence Read Archive (SRA) under the BioProject accession number PRJNA1304145 (temporary link for reviewers: https://dataview.ncbi.nlm.nih.gov/object/PRJNA1304145?reviewer=sclfsbvc2dssf223crde63r880). RNA-sequencing data of human PVD samples and RPE cells isolated from post-mortem ocular globes have been deposited in the European Genome-phenome Archive (EGA) and may be accessed upon approval of the Data Access Committee EGAC50000000721. Declarations Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Azine Datlibagi, Clément Duveau, Christine Delporte and Elie Motulsky contributed equally to this work. References Associated Data Supplementary Materials Supplementary Material 1 Data Availability Statement Raw RNA-sequencing data of ARPE-19 cells and AQP1-transfected ARPE-19 cells have been deposited in the NCBI Sequence Read Archive (SRA) under the BioProject accession number PRJNA1304145 (temporary link for reviewers: https://dataview.ncbi.nlm.nih.gov/object/PRJNA1304145?reviewer=sclfsbvc2dssf223crde63r880). RNA-sequencing data of human PVD samples and RPE cells isolated from post-mortem ocular globes have been deposited in the European Genome-phenome Archive (EGA) and may be accessed upon approval of the Data Access Committee EGAC50000000721.