A convenient protocol for establishing a human cell culture model of the outer retina.

The retinal pigment epithelium (RPE) plays a key role in the pathogenesis of several blinding retinopathies. Alterations to RPE structure and function are reported in Age-related Macular Degeneration, Stargardt and Best disease as well as pattern dystrophies. However, the precise role of RPE cells in disease aetiology remains incompletely understood. Many studies into RPE pathobiology have utilised animal models, which only recapitulate limited disease features. Some studies are also difficult to carry out in animals as the ocular space remains largely inaccessible to powerful microscopes. In contrast, in-vitro models provide an attractive alternative to investigating pathogenic RPE changes associated with age and disease. In this article we describe the step-by-step approach required to establish an experimentally versatile in-vitro culture model of the outer retina incorporating the RPE monolayer and supportive Bruch’s membrane (BrM). We show that confluent monolayers of the spontaneously arisen human ARPE-19 cell-line cultured under optimal conditions reproduce key features of native RPE. These models can be used to study dynamic, intracellular and extracellular pathogenic changes using the latest developments in microscopy and imaging technology. We also discuss how RPE cells from human foetal and stem-cell derived sources can be incorporated alongside sophisticated BrM substitutes to replicate the aged/diseased outer retina in a dish. The work presented here will enable users to rapidly establish a realistic in-vitro model of the outer retina that is amenable to a high degree of experimental manipulation which will also serve as an attractive alternative to using animals. This in-vitro model therefore has the benefit of achieving the 3Rs objective of reducing and replacing the use of animals in research. As well as recapitulating salient structural and physiological features of native RPE, other advantages of this model include its simplicity, rapid set-up time and unlimited scope for detailed single-cell resolution and matrix studies.


Abstract
The retinal pigment epithelium (RPE) plays a key role in the pathogenesis of several blinding retinopathies. Alterations to RPE structure and function are reported in Age-related Macular Degeneration, Stargardt and Best disease as well as pattern dystrophies. However, the precise role of RPE cells in disease aetiology remains incompletely understood. Many studies into RPE pathobiology have utilised animal models, which only recapitulate limited disease features. Some studies are also difficult to carry out in animals as the ocular space remains largely inaccessible to powerful microscopes. In contrast, models provide an attractive alternative to investigating pathogenic RPE in-vitro changes associated with age and disease. In this article we describe the step-by-step approach required to establish an experimentally versatile in-vitro culture model of the outer retina incorporating the RPE monolayer and supportive Bruch's membrane (BrM). We show that confluent monolayers of the spontaneously arisen human ARPE-19 cell-line cultured under optimal conditions reproduce key features of native RPE. These models can be used to study dynamic, intracellular and extracellular pathogenic changes using the latest developments in microscopy and imaging technology. We also discuss how RPE cells from human foetal and stem-cell derived sources can be incorporated alongside sophisticated BrM substitutes to replicate the aged/diseased outer retina in a dish. The work presented here will enable users to rapidly establish a realistic model of the outer retina that is amenable in-vitro to a high degree of experimental manipulation which will also serve as an attractive alternative to using animals. This model therefore has the in-vitro benefit of achieving the 3Rs objective of reducing and replacing the use of animals in research. As well as recapitulating salient structural and physiological features of native RPE, other advantages of this model include its simplicity, rapid set-up time and unlimited scope for detailed single-cell resolution and matrix studies. 1 1 1 2 2 1 3 1

Research highlights
Scientific benefits: • We provide a step-by-step protocol to rapidly establish an in-vitro model of the outer retina incorporating the Retinal Pigment Epithelium (RPE) and the supportive Bruch's membrane.
• We discuss the advantages and limitations of RPE cells (the ARPE-19 cell-line) used in this work.
• This in-vitro model allows the use of powerful confocal microscopes (fast, high-resolution imaging) and new platforms such as 3View and Lightsheet.
• Allows a high degree of experimental manipulation.
3Rs benefits: • This in-vitro culture model can be used as an alternative to in-vivo experiments in spontaneously arising, acutely-induced or transgenic mouse models of retinal degeneration, or be used in parallel with animal studies.

Introduction
The retinal pigment epithelium (RPE) consists of a monolayer of largely cuboidal-shaped pigmented cells found beneath the neuroretina and overlying the vascular blood supply of the choriocapillaris. Occupying this strategic position in the outer retina the RPE performs multiple functions which are essential for retinal homeostasis and maintenance of life-long vision. This includes the daily phagocytosis of shed Photoreceptor Outer Segments (POS), re-isomerization of all-trans-retinal to 11-cis-retinal in the visual cycle, protection against effects of photo-oxidation, trans-epithelial transport as well as the polarised secretion of molecules towards the overlying neuroretina and the underlying choroid. The RPE also forms part of the outer blood-retinal barrier (BRB) which functions to confer an immune privileged state within the ocular environment 1 . Dysfunction or abnormalities of the RPE monolayer is correlated with early stages of pathology linked to a range of ocular conditions such as Age-related Macular Degeneration (AMD), Sorsbys fundus dystrophy, Stargardt disease and Best disease, diabetic retinopathy as well as pattern dystrophies [1][2][3] . However, the origins of RPE dysfunction and how they contribute to such diverse ocular conditions remains incompletely understood.
Numerous in-vivo models including non-human primates, pigs, sheep, rabbits and rodents have been used to study retinal pathobiology 4 . Of these, the most widely used are mice, which show regional differences in Bruch's membrane (BrM) thickness and photoreceptor density, a similar rod to cone ratio at locations comparable to the peripheral human macula as well as a similar RPE monolayer to humans 5 . Mice also offer advantages in terms of costs compared to the use of larger animals and the possibility of studying salient disease features in a matter of months. This has led to the use of spontaneously arising 6 , acutely-induced 7 and transgenic mouse models 8 , or indeed combined models where genetics and diet has been manipulated and mice aged for long periods to bring about disease features 9,10 . However, given the lack of anatomical specialisation equivalent to humans, rodent models are of limited value for studies into macular conditions. Moreover, no single mouse model is capable of replicating the full disease spectrum observed in human retinopathies. This has often led to the unnecessary and over-use of poorly characterised rodent models, many of which show only limited disease features and/or have to be aged for long periods before any obvious retinal pathology is detected 4,11 . The arrangement of ocular tissues such as the RPE also makes them difficult to image, particularly for studies requiring dynamic, real-time imaging or data at single-cell resolution. Welfare concerns and severity limits of mouse models, other than basic information on animal husbandry, are also poorly reported in the literature. For instance, there is limited data on how a particular genetic alteration or mice maintained over long periods (>18 month) might affect their behaviour and quality of life. In contrast, in-vitro models, although simplistic by comparison, are not limited by these issues and boast distinct advantages over mouse models for delineating cellular pathways of damage, or for drug screens to identify effects on a given cell type. Cells cultured under in-vitro conditions that recapitulate their in-situ environment have been shown to reproduce a phenotype that closely resemble native tissues. These cells not only adopt native-like structural and physiological characteristics, but also a genetic profile closely matching their in-situ counterpart. RPE cells were initially grown on plastic substrates and did not exhibit a fully differentiated phenotype. Investigators therefore started culturing RPE monolayers on commercially-sourced transwell inserts with varying pore sizes which mimics important features of the underlying BrM 2,12 . The culture of RPE cells on 0.4μm pore-size inserts is now widely regarded to produce the most desirable RPE phenotype 13  16% reduction in the number of reports using in-vivo models encompassing rodents, rabbits, porcine, bovine and non-human primates over a similar period. Based on these findings and an average annual increase of 4.3% in citations with RPE studies, we estimate that at least 45 publications reporting in-vivo work will be replaced by in-vitro RPE modelling studies. Given the rapid rate at which RPE modelling work is progressing, this will yield at the very least 188 annual citations by 2023. In this article we provide a detailed step-by-step approach for optimising an in-vitro RPE cell model using the spontaneously arisen ARPE-19 cell-line. We also provide steps used to characterise this model, discuss its advantages and limitations as well as how it fulfils the 3Rs objectives.

Methods
Establishing the RPE cell model This was subsequently corrected for the growth area using the following formula (Equation 1). Measurements were performed at room temperature within 6 minutes of removing cultures from the incubator. A full media change was also performed after weekly measurements to minimise the risk of contamination.

Preparation of POS-FITC.
Porcine eyes (maximum of 2 days post mortem) were sourced from a butcher. An incision was made proximate to the ora serata after which the anterior ocular portion was removed and retinae detached gently from the underlying RPE. These were subsequently pooled in KCl buffer (0.3M KCl, 10mM HEPES, 0.5mM CaCl 2 , 1mM MgCl 2 ; pH 7.0) with 48% sucrose (w/v), agitated for 2 minutes and centrifuged at 5000g for 5 minutes to facilitate POS detachment. The resultant supernatant containing isolated POS was filtered through a sterile gauze into 1.5ml Eppendorf tube containing an equal volume of KCl buffer without sucrose. POS were pelleted by centrifugation at 4,000g for 7 minutes, washed three times in 1× PBS and re-suspended in DMEM with 2.5% sucrose (w/v). POS were covalently tagged to Fluorescein isothiocyanate (FITC) by incubation with 5ml labelling buffer (20Mm phosphate buffer pH 7.2, 5mM taurine with 10% w/v sucrose) and 1.5ml FITC stock solution (2mg/mL FITC isomer I in 0.1M Na 2 CO 3 buffer; pH 9.5) at room temperature for 1 hour in the dark on a Stuart SB2 Rotator (Camlam Ltd, UK). POS-FITC conjugates were pelleted by centrifugation at 3,000g for 5 minutes and re-suspended in DMEM with 2.5% sucrose (w/v). Isolated POS can be stored for up to 6 months at -80°C. The total protein content in preparations was quantified using a BCA assay (23225, Thermo Fisher, UK) following the manufacturer's instructions.

POS feeding assay and assessment of trafficking dynamics.
ARPE-19 monolayers on transwell inserts were incubated at 17°C for 30 minutes prior to exposure with 4mg/cm 2 POS-FITC for a further 30 minutes at 17°C. This facilitates maximal POS binding with minimal internalisation 39 to initiate a pulse-chase assay. The POS-FITC solution was aspirated to remove unbound POS. Cultures were supplemented with fresh media and returned to a humidified 37°C incubator with 5% CO 2 and 95% air. Cells were subsequently fixed at 2, 4, 6, 12, 24 and 48 hours with 1×PBS containing 4% formaldehyde for 30 minutes at 4°C, after which they were incubated with 1% BSA in PBS-Tween to block/permeabilise cells for 30 minutes. Cultures were probed overnight at 4°C with the following primary antibodies prepared in blocking buffer. Statistical Analysis. Statistical analyses were conducted using the GraphPad Prism 7 Software (GraphPad, US). Values were first assessed to ensure data met assumptions of the selected statistical test. Tests for each experiment appear in figure legends. Briefly, ELISA quantification was assessed using the unpaired student's t-test, whilst TEER were evaluated using a one way ANOVA and Tukey's multiple comparisons tests. In both cases, a single well corresponded to an experimental unit. Data is presented as means ± standard error of the mean (SEM) where n represents independent experiments. Statistical significance is denoted as * p ≤ 0.05, ** p ≤ 0.01, *** p ≤ 0.001 and **** p ≤ 0.0001.

Protocol
Here we describe the step-by-step procedure required for establishing and validating long-term cultures of ARPE-19 monolayers on transwell inserts. A schematic highlighting the sequence of steps and timelines are summarised in Figure 1.
Protocol for establishing the culture model Step 1: Preparation of cell culture media ARPE-19 cells (CRL-2302™ ATCC®, USA) require growth in optimised culture media, which can be prepared according to Table 1. Freshly prepared cell culture media is passed through a vacuum filter to maximise sterility and used within 2 weeks. A volume of 250ml is sufficient for the culture of 6 transwell plates of 12mm diameter inserts or 4 plates of 24mm diameter inserts (Table 4) for a period of approximately 2 weeks. Volumes should be scaled according to the size and number of desired transwells as storage of media for longer periods is not recommended.
Step 2: Culture of ARPE-19 cells ARPE-19 cells should be maintained in a humidified incubator set at 37°C with an atmosphere of 5% CO 2 . Cells should be cultured in a T25cm 2 flask containing 5ml of freshly prepared media and passaged at a 1:3 ratio when confluent or approximately every 3 weeks. A complete media change should be performed every 2-3 days to retain physiological glucose levels 41 . Although it is possible to maintain ARPE-19 cells in T25cm 2 flasks for longer periods, cells becoming increasingly difficult to passage following the formation of an underlying extracellular matrix. Passaging involves removal of conditioned medium and washing cells with 1xHBSS followed by exposure to 1.5ml 0.25% trypsin/EDTA for 6 minutes in an incubator. The trypsin/EDTA solution is neutralised using 7ml of complete media and triturated to obtain single cells after which the resulting suspension is centrifuged at 300g for 5 minutes. The pellet is re-suspended in a volume of freshly prepared culture media and split at a 1:3 ratio between T25cm 2 flasks. The importance of correct cell passaging to maintaining an epithelial phenotype is discussed elsewhere 42 . For the long term storage in liquid nitrogen (-195.8 o C) cells should be suspended in the desired concentration in freezing medium comprising 75% complete culture medium and 25% dimethylsulfoxide (DMSO; S-002-M, Sigma Aldrich, UK). For ARPE-19 cells we recommend 1ml aliquots of 1x10 6 cells. These must first be frozen in a suitable container (Mr Frosty™, Thermo Fisher Scientific UK) or equivalent with 100% isopropanol at -80°C overnight prior to storage in liquid nitrogen.
Step 3: Coating transwell inserts with fibronectin Prepare lyophilised fibronectin (F2006, Sigma Aldrich, UK) to a final concentration of 50μg/ml by adding 20ml of sterile ddH 2 O. We recommend preparing an initial 5ml solution using sterile ddH 2 O followed by transfer to a 50ml falcon containing 15ml sterile ddH 2 O. The fibronectin solution should be used immediately and is sufficient to coat six 12mm diameter or five 24mm diameter transwell plates, or should be divided into aliquots for storage at -20 o C. Users should apply the stock solution to the apical transwell compartment as indicated in Table 2. Ensure that the entire surface of the membrane   is covered after which transwells are left partially covered in a laminar flow hood overnight. The following day any residual fibronectin should be aspirated and wells washed with 1x sterile PBS for cell seeding (Table 3). We observed that ARPE-19 cells readily proliferate and mature to form in-situ RPE-like monolayers on an underlying fibronectin matrix although others had used a laminin substrate 26 . Consequently, work presented here are carried out on transwells coated with fibronectin. It is also possible to culture cells in the absence of an underlying coating on transwell membranes (Figure 2). ARPE-19 cultured without an extracellular matrix (ECM) substrate develop pigmentation, establish a trans-epithelial barrier and secrete proteins directionally 25 . The culture of cells without an underlying coating is particularly useful for studies in which de-novo synthesis/deposition of extra cellular components and their turnover can be monitored without any influence of artificial substrates.
Technical tip: Rapid thawing of soluble fibronectin can result in the irreversible precipitation of proteins. To avoid this we suggest that the stock solution should be gradually thawed at 4 o C.
Step 4: Seeding and culture of ARPE-19 cells on transwell inserts Cells are passaged as described in step 2. A confluent flask of T25cm 2 ARPE-19 cells yield between 3-5 million cells. Consequently, one T25cm 2 flask is sufficient to seed 10 plates of 24mm diameter transwell inserts or 20 plates of 12mm diameter inserts. Cells are seeded on fibronectin coated transwells (Table 3), although the benefit of using uncoated transwell membranes have also been discussed. Culture media are applied to the apical and basal transwell compartment at least one hour prior to cell seeding (Table 4). Following seeding, we recommend leaving cultures undisturbed for approximately 4 days prior to the first media change. Cultures are maintained in a 37°C incubator with an atmosphere of 5% CO 2 and media changed every 2-3 days (Table 4). After approximately 1 week cultures appear confluent and by 2 weeks exhibit characteristic cobblestone morphology. Obvious signs of pigmentation will develop after 3-4 months, although some evidence of pigmentation is apparent under light microscopy after 2 months. The size of pores in transwell membranes are known to influence RPE morphology 13 hence we suggest users adhere to these recommendations. Cultures are maintained for a minimum of 2 months prior to validation studies. For functional experiments, such as POS feeding assays, we recommend maintaining cultures for approximately 4 months.
Technical tip: Pipette solutions along plastic walls of the transwell chamber at a steady state during media changes to avoid inducing cell stress or cell detachment. When removing media we recommend first tilting the transwell plate to a 45° angle to avoid disturbing the RPE monolayer.
Protocol for characterising and validating the culture model In this section we describe the steps used to validate and characterise ARPE-19 monolayers on transwell inserts. These approaches however may be adopted for the culture of RPE from different sources, although the time taken to obtain monolayers displaying physiological and structural features akin to the native RPE may vary depending on the specific type/source of RPE cells.  Step 5: Confocal immunofluorescence studies of ARPE-19 monolayers Cells that have been in culture for at least 2 month are used to ensure RPE monolayers had adopted structural and physiological features of native RPE. Transwell inserts are washed with 1x sterile PBS prior to fixation in 4% PFA for 30 minutes. Each transwell membrane is removed from its insert by running a blade along the circumferential ring ( Figure 3). The amount of material (RPE monolayers) required to carry out experiments may be maximised by sectioning transwell membranes into multiple sections, although caution must be exercised to prevent disturbing the delicate cell layer. We recommend using a sharp razor blade to guillotine sections of the membrane outright as cutting or slicing generates sheer forces which rucks membranes leading to cell detachment (Figure 3). Transwell membranes are washed three times in 1x PBS and blocked/permeabilised in 5% NGS in 0.1% PBST for 1 hour. A battery of antibodies are used to probe for components of the BRB (ZO-1 and Occludin), to assess RPE polarisation (Na + /K + ATPase) and to detect expression of the cell-specific marker (RPE65), although other proteins such as CRALBP may also be included. Readers are also referred to studies described in Ahmado et al., 2011 25 . Primary Antibodies are diluted 1:100 in blocking buffer for incubation at 4°C overnight (Table 5). The following day membranes are washed three times in 0.05% PBST followed by incubation with the appropriate Alexa Fluor® labelled secondary antibody (RRID: AB_2534115, RRID: AB_2534085, RRID: AB_2534116, RRID: AB_2534087, RRID: AB_2534114, RRID: AB_2534064, Life Technologies, UK) prepared in 0.05% PBST for 1 hour. Membranes are washed three times in ddH 2 O to remove any unbound secondary antibody and mounted between two glass coverslips with Mowiol® mounting medium (Harco Chemical Company Ltd., UK). We recommend sandwiching membranes between 2 glass coverslips as opposed to pairing a single coverslip with a thicker glass slide as this allows either side of the sample to be imaged without potential optical interference from the porous membrane. Z-stack image are captured with a laser-scanning confocal microscope (Supplementary Figure S1 and Figure 4). We recommend a minimum optical slice thickness of 1μm through z-stacks to help assess the polarised expression of RPE markers.
Technical tip: Pores within the membrane may be used as a reference point to help orient the position of apical and basolateral RPE surfaces. However, users may have to account for small undulations in the membrane which will alter the focal plane across the sample.
Step   to prevent the cells drying out which would render them useless for microscopy. Samples are passed successively through a graded series of ethanol concentrations (30%, 50%, 70% and 95% ethanol) for 10 minutes each, followed by two successive incubation periods in absolute ethanol for 20 minutes to achieve optimal dehydration. The link reagent acetonitrile (Fisher Scientific, UK) is applied for 10 minutes and membranes incubated in a mixture containing an equal ratio of acetonitrile to Spurr resin overnight. In our experience, Spurr resin appears to be the best medium to effectively bond filters, although sections often split along the resin/filter interface during sectioning and when viewing under the microscope. The following day, samples are incubated for an additional 6 hours in Spurr resin and embed in fresh Spurr resin for polymerisation at 60°C for 24 hours. Samples should be embedded in Spurr resin as triangular slices with the apex of the triangle positioned towards the bottom of the embedding capsule ( Figure 5), which facilitates ease of cutting. This procedure is carried out without a rotator as cells could otherwise detach from the underlying membrane. Ultrathin/silver TEM sections are prepared using a Reichert Ultracut E ultramicrotome and collected on 200 mesh carbon and formvar coated copper palladium grids. We advise against the use of chloroform to stretch sections which exacerbates potential separation of the resin/membrane interface during sectioning. Sections are subsequently stained with Reynold's lead stain and visualised using a Hitachi 7000 transmission electron microscope fitted with a SIS Megaview III plate EMSIS camera ( Figure 6).
Technical tip: We recommend viewing samples starting at a lower magnification with the electron beam spread widely in order to minimise shrinkage or movement across sections, and to protect against the possibility of splitting at the resin/membrane interface.
Step 7: Trans-epithelial electrical resistance measurement of ARPE-19 cultures TEER studies are carried out after a minimum of 6 weeks in culture as ARPE-19 cells do not form an effective barrier before this time ( Figure 7A-B). ARPE-19 cells also generate relatively poor barriers compared to hfRPE or PSC-RPE. However, the method describe herein can be adopted to test barriers created by RPE cells from different sources. Electrical recordings are obtained using an EVOM 2 epithelial voltohmmeter and a 4mm STX2 chopstick electrode (EVOM 2 ; 300523, World Precision Instruments Inc., USA). As importance is given to maintaining sterility in RPE cultured for long periods, electrodes are first sterilised in 70% ethanol, rinsed in ddH 2 O and equilibrated in pre-warmed culture medium prior to use. For this reason we also recommend performing a complete media change in both transwell compartments after measurements. Electrodes are inserted perpendicularly into the apical and basal compartments so that the tip of each arm is immersed in media. Five recordings are taken from each transwell at set time intervals (10 seconds) to calculate the average TEER value. Measurements are recorded from at least three separate transwell inserts. The reference value from a fibronectin coated transwell without cells is subtracted from initial measurements (Equation 2) and the net recording corrected for area of cell growth to yield a final TEER value (Equation 3, Table 2). All measurements are performed at room temperature within 6 minutes of removing cells from the incubator.

Final TEER (Ω/cm 2 ) = Net TEER (Ω) × Area of transwell membrane (cm 2 ) [Equation 3]
Technical tip: Care should be taken to prevent electrodes from touching chamber walls as this results in inconsistent TEER values.
Step 8: ELISA studies of ARPE-19 cultures The capacity to secrete proteins directionally can be assessed by performing an ELISA on conditioned media harvested from apical and basal transwell compartments ( Figure 1). A Novex® human VEGF solid-phase sandwich ELISA kit (Life Technologies, UK) is used to measure secreted levels of VEGF, whilst a Biovendor Human PEDF solid-phase sandwich ELISA kit (Biovendor, UK) is used to measure secreted levels of PEDF. A complete media change is performed prior to quantifying soluble protein levels during a 2-3 day period. Collected samples are kept at 4°C or on ice before quantification to prevent protein degradation or stored at -80°C for future use. ELISA quantification is carried out in triplicate on a minimum of three separate wells ( Figure 7C-D). The volume of media lost due to sampling is restored afterwards by the addition of freshly prepared media into apical and basal transwell compartments. Assays are carried out following the manufacturers' guidelines (Table 6).
Step 9: Photoreceptor outer segment phagocytosis assay Post-confluent ARPE-19 cells are reported to exhibit phagocytic activity after 2 weeks in culture 43 , although we recommend using cultures of approximately 4 months so cells exhibit a gene profile comparable to native RPE 36 .

Figure 5. Preparation of polyethylene terephthalate (PET) membranes for transmission electron microscopy (TEM).
Schematic outlining steps carried out to embed segmented transwell membranes into capsules containing fresh Spurr resin. The apex is positioned downwards, which greatly assists with cutting sections for TEM.  Fluctuations between average weekly TEER were observed prior to week 6 (p=0.009 at 3 weeks, p=0.013 at 4 weeks and p=0.001 at 6 weeks, one-way ANOVA with Tukey's multiple comparisons) after which a stable value of 40.72 Ω.cm 2 was achieved. Data is presented as mean ± SEM. Next, we quantified polarised secretion of Vascular Endothelial Growth Factor (VEGF) and Pigment Epithelium Derived Factor (PEDF) by ARPE-19 cells. Conditioned media was collected (n=3) after 72 hours and proteins quantified by ELISA.
[C] The apical compartment was found to contain 0.942 ± 0.035ng/ml of VEGF compared to 2.852 ± 0.145ng/ml in the basal chamber, which was statistically significant (p=0.0002).
[D] PEDF concentrations in the apical compartment was 16.95 ± 0.72ng/ml compared to 25.05 ± 3.93ng/ml in the basal chamber. There were no significant differences (p= 0.112) although more PEDF was secreted via the basolateral RPE surface. Data is presented as mean ± SEM with statistical comparisons made using the unpaired student's t-test and sourced in part from material published previously 18 under the Creative Commons licence. Isolation of photoreceptor outer segments Porcine eyes are obtained from a butcher or abattoir within 2 days of post mortem and POS isolated on the same day. An incision is made at the ora serata to remove the anterior eye portion after which the retina can be gently detached. We find this is best achieved by teasing the retina away from the RPE in a circular fashion. The optic nerve is severed at the nerve head to detach the retina. The retinae are pooled in KCl buffer (0.3M KCl, 10mM HEPES, 0.5mM CaCl2, 1mM MgCl2; pH 7.0) with 48% sucrose (w/v) and agitated vigorously for 2 minutes on a rotation mixer after which the solution is centrifuged for 5 minutes at 5,000g. At this point isolated POS appear in the supernatant and the pellet can be discarded. Filter the POS containing supernatant through a sterile surgical gauze positioned on a 1.5ml Eppendorf tube into an equal volume of KCl buffer without sucrose and incubate at room temperature for 5 minutes. Centrifuge the suspension at 4,000g for 7 minutes to pellet isolated POS and discard the supernatant. Wash POS pellets three times in 1xPBS and re-suspend in DMEM with 2.5% (w/v) sucrose 44 . POS is covalently conjugated to fluorescein isothiocyanate (FITC). This is achieved by incubating pooled POS in 5ml labelling buffer (20Mm phosphate buffer pH 7.2, 5mM taurine with 10% w/v sucrose) and 1.5ml FITC stock solution (2mg/mL FITC isomer I in 0.1M Na2CO3 buffer; pH 9.5) on a rotator mixer at room temperature for 1 hour in the dark. Pellet the POS-FITC by centrifugation at 3,000g for 5 minutes, suspend in DMEM with 2.5% sucrose (w/v) and store for a maximum of 6 months at -80°C. Once thawed isolated POS should not be refrozen. The total protein content of POS preparations can be quantified using a BCA assay prior to use.

Photoreceptor outer segment feeding assay
Cultures are incubated at 17°C for 30 minutes after which 4mg/cm 2 POS-FITC is applied to RPE cultures for 30 minutes to maximise binding with minimal internalisation 39 . This concentration is sufficient to challenge each RPE cell with approximately 10 isolated POS molecules 44 . Alternatively, if cultures cannot be chilled to 17°C they may be incubated with isolated POS for 2 hours at 37°C to achieve a similar effect 45 . Following the feeding assay, wash inserts once in fresh medium and return to an incubator set at 37°C and 5% CO 2 . Transwells are removed at desired time points after which they are washed once in 1xHBSS followed by fixation in 1xPBS containing 4% formaldehyde for 30 minutes at 4°C. Wash cells three times in 1x PBS and store at 4°C until use. Immunostaining is performed by blocking/permeabilising cells in PBS-Tween containing 1% BSA for 30 minutes followed by incubation at 4°C overnight with the desired antibody (Table 5) prepared in the same solution. The following day, wash cells three times with 1xPBS to remove any unbound primary antibodies and incubate with the appropriate secondary antibody (step 5) for 1 hour at room temperature. Wash samples as before and incubate with 1μg/ml DAPI (prepared in ddH 2 O) for 10 minutes before performing three final washes in 1xPBS. Mount the sample between two glass coverslips using Mowiol ® mounting medium for confocal microscopy studies (step 5).
For co-localisation studies we use an unbiased statistical algorithm described by Costes et al. 40 and performed using Volocity Software (Perkin Elmer, UK). Considerations prior to undertaking co-localisation studies include careful selection of suitable fluorophores to avoid bleed through and chromatic aberration as well as pixel saturation (Figure 8). We also suggest selecting non-overlapping and non-adjacent fluorophores and refer to several excellent articles on co-localisation studies 40,46-48 .

Results
Characterisation studies ARPE-19 monolayers on transwell inserts can be easily maintained in long term culture (Figure 1). This allows them to mature and express structural and physiological features of native RPE. Our experiments were carried out on monolayers that had been in culture for 2-4 months. We also tested the ability of ARPE-19 cells to attach and spread on transwell membranes with or without the presence of fibronectin; the preferred substrate for these cells in our experience 18,49,50 . Our findings show that cells were capable of attachment and growth to confluence on PET membranes irrespective of the presence/absence of an underlying fibronectin matrix (Figure 2). Prior to carrying out imaging studies transwell membranes were carefully removed from their plastic wells. We show a convenient method by which even a small transwell insert can be sectioned into several segments so that the investigator is able to probe for multiple markers and thus maximise the possibility of obtaining data from each transwell (Figure 3).
In-vitro RPE monolayers were studied for physiological and structural features characteristic of RPE cells. We first probed for junctional complexes zonula occludens (ZO-1) ( Figure 4A-D) and occludin ( Figure 4E-H). We also looked for evidence of apically expressed Na +/ K + ATPase ( Figure 4I-L) and the cell-specific marker RPE65 ( Figure 4M-P). After 2 months in culture ARPE-19 monolayers expressed the early tightjunction protein ZO-1 with a border demarcating cell-to-cell contact, and 3D imaging revealing polarisation towards the apical cellular region. ZO-1 staining was also observed in the cytoplasm and prominently in the nucleus, which is consistent with reported literature 51 . Expression of occludin during mid-late stages of barrier formation was also observed. Next, we probed for expression of the Na + /K + ATPase transporter to assess the presence of a polarised plasma membrane. Na +/ K + ATPase is predominantly expressed on the apical RPE surface where it facilitates the process of photo-transduction. Apically expressed Na + /K + ATPase is also linked with a highly differentiated, polarised RPE phenotype 52 . We detected Na + /K + ATPase in some but not all cells, although expression appeared to be limited to the apical RPE surface ( Figure 4L). We also probed for the cell-specific retinoid isomerohydrolase RPE65 marker to confirm identity of RPE cells in the monolayer ( Figure 4M-P). RPE65 was observed as punctate, cytoplasmic staining as reported by others 53 and confirmed the identity of RPE cells in longterm culture. Next, we assessed the extent to which ARPE-19 monolayers on transwells adopt ultrastructural features of native RPE. We describe a convenient technique by which transwell membranes can be sectioned into smaller segments to be embedded in resin blocks for TEM studies ( Figure 5). We observed evidence of numerous microvilli on the apical RPE surface ( Figure 6A) and infolded/convolutions of the basolateral cell membrane (Figure 6B), characteristic of native RPE. Micrographs also showed details of intracellular organelles including mitochondria, compartments in the endocytic-lysosomal pathway and pigment molecules ( Figure 6C-E). Mitochondria, for instance, appear in cross-section as a double membranebound structures with luminal cristae, whilst vesicles contained cargos of varying electron densities. The arrangements of these organelles conformed to the apical-basolateral axis of native RPE. Junctional complexes between RPE, detected previously by immunofluorescence studies (Figure 4A-H), were also observed at ultrastructural resolution as tight junctions and adherens junctions along membranes at the apical region of RPE cells ( Figure 6F-G). These were observed as electron-dense regions indicating points of cell-to-cell contact and associated with desmosomes in some instances.

Validation studies
Establishment of an effective trans-epithelial barrier is a key feature of native RPE, and one that can be readily measured in transwell cultures. We carried out TEERs of ARPE-19 cultures over an approximately 3 month period. A stable electrical gradient was established following 6 weeks in culture ( Figure 7A), after which there were no appreciable changes to the barrier ( Figure 7B). An average TEER value of 40.72 Ω/cm 2 was noted once cultures had established a stable barrier in-line with previous reports 22,25 . Polarised secretion of molecules towards the overlying neuroretina and the underlying choroid is an important feature of RPE cells 15 . Proteins such as VEGF and PEDF that are synthesised/secreted by RPE are known to possess pro-angiogenic and neuroprotective effects, respectively 1 .
Directional secretion of such molecules can easily be quantified in transwell compartments once cells establish an effective trans-epithelial barrier. To assess if this was achieved in culture we measured VEGF and PEDF levels in conditioned media using two different ELISAs. ARPE-19 cells secreted VEGF through both apical and basolateral surfaces at concentrations of 0.942 ± 0.035ng/ml and 2.852 ± 0.145ng/ml, respectively. VEGF secretion towards the choroid was therefore significantly higher compared to amounts released towards the neuroretina ( Figure 7C). PEDF levels were also secreted via both surfaces at concentrations of 16.95 ± 0.72ng/ml (apical) and 25.05 ± 3.93ng/ml (basal). Statistically, there were no differences in amounts of PEDF secreted towards the choroid or neuroretina ( Figure 7D). Next, we assessed the ability of cultured ARPE-19 cells to bind and internalise POS cargo. In-situ RPE daily internalises and proteolytically degrade POS from overlying photoreceptors, the impairment of which plays a key role in retinopathy 1 . POS-FITC cargos were fed to 4 month old monolayers using a pulse-chase method described previously 39 . Each compartment in the endosome/phagosomelysosomal and autophagy pathway was assessed at 2, 4, 6, 12, 24 and 48 hours for the extent of co-localisation with fluorescently-labelled POS. Cargos initially appeared in early Rab 5 compartments ( Figure 8A-B), which by ~6 hours had trafficked to Rab 7 late vesicles ( Figure 8C-D). Between 6-12 hours, a large proportion of cargo had co-localised to early LAMP1 ( Figure 8E-F) and mature LAMP2A lysosomes ( Figure 8G-H). 48 hours after the pulse-chase assay was initiated, a large proportion of cargo appeared in LC3B-positive autophagy bodies ( Figure 8I-J).

Discussion
In this article, we describe a convenient protocol by which users can rapidly establish and study RPE cells in long-term culture. We used the human ARPE-19 cell-line, although the approaches described herein may be adopted for studying RPE from a variety of sources. We provide examples from our own laboratory as well as other groups to highlight the type of questions which investigators could realistically address using this model. In summary, this article provides an in-depth set-up and validation protocol for establishing a culture model of the outer retina using the widely utilized ARPE-19 cell-line. We also discussed advantages and limitations of transwell models in general and ARPE-19 cells in particular, so that users may best exploit this versatile system for their studies. Advantages over mouse models such as (1) its use as a viable alternative, (2) ability to rapidly generate functional RPE monolayers akin to native tissues, and (3) ability to reproduce disease features that could only be previously studied in mice 18,55,56 , makes in-vitro models of the outer retina especially attractive. Their versatility is further demonstrated by studies in which the surface of transwell membranes are directly modified to mimic effects of aging 23 . Investigators are also well-placed to take advantage of new developments in stem-cell technology and refinements to in-vitro cultures described herein 60-62 as well as a plethora of artificial BrM substrates on offer. The latter may be set-up and assembled similarly to transwells by using commercially available products such as CellCrown TM inserts (Sigma, UK). The growing interest in microfluidic devices allow laboratories to model relationships between the RPE vs. choroidal endothelial cells and blood flow by incorporating the latter into transwell devices. These advances combined with the development of fast/high-resolution imaging and new 3D imaging platforms such as serial block face scanning electron microscopy and Lightsheet are likely to usher in further opportunities to exploit in-vitro models. Consequently, investigators may wish to consider these culture models as attractive alternatives to using animals, or at least as powerful new tools to be exploited in parallel that will also have the benefit of reducing and replacing animals used in research.

Data availability
stimuli/factors should also be elaborated upon including use of hydrogels, co-culture, surface topography, surface geometry/curvature, growth factors-see literature. The coating exploits surface adsorption-why do you let the Transwell "dry" overnight and how can you aspirate residual fibronectin if so-rather to "coat" overnight? Confirmation of the presence of fibronectin coating was achieved by? Agitation during the coating procedure? Is the ELISA a single timepoint at 2-months or cumulative collection of media-how is this achieved when there are frequent media changes required? FBS contains exogenous (ECM) proteins as well as VEGF and PEDF? Please state "room temperature" for TEER measurement (and POS)-I assume that this was performed in a hood rather than "on the bench" hence airflow/room temperature appropriate? Capital "T" for Transwell throughout.-registered product.

Protocol:
: Additionally filter step may be seen as "overkill"-an additional step could actually increase Step 1 infection risk-reagents can simply be added using aseptic technique? Please mention storage condition and "warming-up" protocol for media. Discrepancy to name of "FBS" mentioned in the M&M and Table 1: FBS = foetal calf serum; N4762 is newborn calf serum (NBCS)-difference in composition especially antibodies and exogenous protein profile.
Amend to ensure that the FN-coating step is mentioned before the seeding of cells-the former is Step 2: required before the latter and should also tally with Figure 1. What % confluency is reached before media change? The 6-minute trypsin protocol is very exact-did you confirm visually under a microscope? Over/under-incubating can be problematic as it may lead to selective population isolation-also, as the author stated, longer-term cells generate their own ECM which will enhance the cells' adhesiveness to the substrate. Was a simple viability and cell count i.e. trypan blue performed-needs to be mentioned in Step 4-determine initial seeding densities? any agitation used during the coating process?
Step 3: fibronectin solution should never be vortexed/centrifuged either-also results in Step 3 (technical tip): "crashing out" of solution. Note that "steps" in Figure 1 and main text are out of synch-please amend e.g.
Step 3 in Figure 1 is "seeding" whereas in the text, Step 4 is seeding the Transwell.
: State volume applied to apical and basal compartments prior to seeding? What volume was the Step 4 cell suspension when adding to the Transwells-possible sink conditions? Clarify what "obvious signs of pigmentation is".
: was there a slight agitation when cells were first added to ensure uniform coating Step 4 (technical tip) of substrate?
Steps out of sync with Figure legend. Can anything about morphology be discussed from the Step 5: images? The study emphasises the long-term ability of the culture conditions i.e. 2-4 months-do you have any associated images for the extended timepoints to indicate viability and/or correct phenotype? Previous researchers have exploited wax embedding as a way to preserve/protect the samples before imaging-similar to your TEM preparation-any reason/justification/mention of the suitability? Capital A in antibodies. What confocal was used and associated operating conditions required. Could Figure 4, panel antibodies. What confocal was used and associated operating conditions required. Could Figure 4, panel I, J and K be recomposed with the same magnification/scale as the other panels-allow easier comparison. State timepoint of image acquisition-assuming 2 months based on RPE65 sentence? Full name for NGS required-normal goat serum?
State the dimensions/manufacturer of the embedding capsule used. Details for Reynold's Lead Step 6: stain required i.e. citrate? concentration, time.
Please state pore size/manufacturer for the surgical gauze used for filtration.

Isolation of POS:
how were the samples chilled/incubated to 17°C-state instrument used and/or POS feeding assay: humidified/gaseous incubator? Optimal internalisation occurs at 37°C-so why 2 hours at 37°C same as 30 minutes at 17°C?
Authors comment on attachment and spreading characteristics but have not alluded to what is Results: ideal/typical of RPE cells-expansion needed regarding morphology of cells. Growth, per se, has not really been documented-absolute cell count, mitochondrial activity assessment i.e. MTT/XTT/MTS assays required alongside a complementary LDH-release profile would fully confirm that the cells are active, alive and not subject to (induced) death i.e. apoptosis or necrosis. No cell numbers are mentioned throughout the long-term culture aside from the initial cell seeding density and justification of contact inhibition. Appropriate representative markers have been selected and results suggest that the cells are of the correct phenotype, barrier formation and correct functionality.
A good descriptive and critical overview is provided by the authors. However, there is actually Discussion: no direct incorporation of a Bruch's membrane e.g. ex vivo tissue or biomimetic collagen-elastin substitute within this study which contrasts the statement(s) made in the Abstract and Research Highlights. The authors also point out the shortfall of their model e.g. ATPase expression, TEER/barrier function, POS binding. I would like a bit more discussion with regards to how this technique/protocol could be applied to mimic diseased states-is it simply culturing cells from (diseased) donor tissues/cells? Would a full (re)validation of culture conditions be required or is the assumption that Transwell plus culture media optimisation would suffice? Could this be incorporated within a 3D model? Multilayer/co-culture format? Slight expansion on future integration in ocular toxicity assessment and drug discovery would be beneficial.
All appropriate-suitable and up-to-date References: All appropriate and relevant Supplementary material:

If applicable, is the statistical analysis and its interpretation appropriate? Yes
Are the 3Rs implications of the work described accurately? Yes Is the rationale for developing the new method (or application) clearly explained?

Is the description of the method technically sound? Yes
Are sufficient details provided to allow replication of the method development and its use by others? Yes If any results are presented, are all the source data underlying the results available to ensure full reproducibility? Yes Are the conclusions about the method and its performance adequately supported by the findings presented in the article? Yes Currently a NC3R Training Fellowship Assessment Panel Board member Competing Interests: (https://www.nc3rs.org.uk/funding-panel-membership) and have previously been awarded multiple summer studentship grants from the Animal Free Research UK (formerly known as the Dr Hadwen Trust) charity (https://www.animalfreeresearchuk.org/2017-summer-studentships/).

Referee Expertise:
Referee suggested by the NC3Rs for their scientific expertise and experience in assessing 3Rs impact.
I have read this submission. I believe that I have an appropriate level of expertise to confirm that it is of an acceptable scientific standard.
The benefits of publishing with F1000Research: Your article is published within days, with no editorial bias You can publish traditional articles, null/negative results, case reports, data notes and more The peer review process is transparent and collaborative Your article is indexed in PubMed after passing peer review Dedicated customer support at every stage For pre-submission enquiries, contact research@f1000.com