Sodium channels enable fast electrical signaling and regulate phagocytosis in the retinal pigment epithelium

Background Voltage-gated sodium (Nav) channels have traditionally been considered a trademark of excitable cells. However, recent studies have shown the presence of Nav channels in several non-excitable cells, such as astrocytes and macrophages, demonstrating that the roles of these channels are more diverse than was previously thought. Despite the earlier discoveries, the presence of Nav channel-mediated currents in the cells of retinal pigment epithelium (RPE) has been dismissed as a cell culture artifact. We challenge this notion by investigating the presence and possible role of Nav channels in RPE both ex vivo and in vitro. Results Our work demonstrates that several subtypes of Nav channels are found in human embryonic stem cell (hESC)-derived and mouse RPE, most prominently subtypes Nav1.4, Nav1.6, and Nav1.8. Whole cell patch clamp recordings from the hESC-derived RPE monolayers showed that the current was inhibited by TTX and QX-314 and was sensitive to the selective blockers of the main Nav subtypes. Importantly, we show that the Nav channels are involved in photoreceptor outer segment phagocytosis since blocking their activity significantly reduces the efficiency of particle internalization. Consistent with this role, our electron microscopy results and immunocytochemical analysis show that Nav1.4 and Nav1.8 accumulate on phagosomes and that pharmacological inhibition of Nav channels as well as silencing the expression of Nav1.4 with shRNA impairs the phagocytosis process. Conclusions Taken together, our study shows that Nav channels are present in RPE, giving this tissue the capacity of fast electrical signaling. The channels are critical for the physiology of RPE with an important role in photoreceptor outer segment phagocytosis. Electronic supplementary material The online version of this article (10.1186/s12915-019-0681-1) contains supplementary material, which is available to authorized users.


Introduction
In the vertebrate eye, the retinal pigment epithelium (RPE) forms a barrier between the retina and the choroid [1][2][3]. Its cells are associated closely with photoreceptors: their apical sides surround the outer segments with long microvilli, and the basolateral sides are attached to Bruch's membrane, an extracellular matrix separating the RPE from the choroid [3,4]. The RPE has many functions that are vital to retinal maintenance and vision, such as maintaining the visual cycle, secreting important growth factors, delivering nutrients to the photoreceptors from the bloodstream while removing metabolic end products, and absorbing scattered light [1,3]. Additionally, RPE maintains ionic homeostasis in the subretinal space [5] and sustains photoreceptor renewal by phagocytosing their shed outer segments [1,6]. Phagocytosis is highly essential for vision, and it is under strict diurnal control, initiated at light onset for rods and typically at light offset for cones [7,8]. This evolutionarily conserved molecular pathway is receptor mediated and precisely regulated; however, the exact signaling cascades are still not completely understood [9]. Recent studies imply the importance of specific ion channels in this process including the L-type calcium channels as well as calcium-dependent potassium and chloride channels [10][11][12].
Since the first single-cell recordings from RPE in 1988 [13], a large variety of different ion channels have been identified in them [5]. Among these are several voltagegated calcium, potassium, and chloride channels. However, the identity of sodium conductive ion channels in RPE has remained elusive [5], even though the importance of sodium homeostasis to normal RPE function is acknowledged. Of the two main families of sodium channels, there is evidence of both epithelial Na + channels and voltage-gated Na + (Na v ) channels in RPE [5,14,15,18,19]. However, electrophysiological data demonstrating their functionality is missing in mature RPE. More importantly, Na v channels that are characteristic of excitable cells have to date only been detected from cultured RPE. This has resulted in the interpretation that their expression is due to neuroepithelial differentiation that can occur in culture [5,20,21].
Here, we shed light on this crucial issue by demonstrating the presence of Na v channels both in cultured human embryonic stem cell (hESC)-derived RPE and freshly isolated mouse RPE. We show that Na v channels co-regulate photoreceptor outer segment (POS) phagocytosis. Our hypothesis is supported by a recent demonstration of the involvement of Na v channels in phagocytosis of mycobacteria by macrophages [22]. Our work provides evidence that Na v 1.8 accumulates with the phagosomal particles. Na v 1.4 also accumulates to phagosomes but displays localization to cell-cell junctions outside phagocytosis. Interestingly, selective Na v channel blockers significantly reduced this phagosomal translocation. Moreover, the selective blockers combined with the universal Na v blocker tetrodotoxin (TTX) reduced the total number of ingested POS particles by up to 41% while not affecting their binding. Reduction was also observed when the expression of Na v 1.4 was silenced with short hairpin RNA (shRNA). More generally, our observations add to the growing body of evidence that Na v channels play diverse roles in a variety of classically non-excitable cell types ranging from astrocytes and microglia to macrophages and cancer cells (for review, see [23]). Collectively, our results show that this epithelium is electrically more complex than was previously thought.

Results
Functional voltage-gated sodium channels are present in RPE derived from human embryonic stem cells We used whole-cell recordings from mature hESCderived RPE in K + free intracellular solution to observe transient inward currents elicited by a series of depolarizing voltage pulses after strong hyperpolarization to − 170 mV (Fig. 1c, n = 19). These recordings were performed from an intact monolayer (Fig. 1a, results summarized in Fig. 1j) in the presence and absence of a gap-junction antagonist (18α-glycyrrhetinic acid). Resembling currents, but with only a fraction of the amplitude, were occasionally identified in cells from freshly dissociated mature hESC-derived RPE (Fig. 1b, d, n = 6), that is the conventional configuration for RPE patch clamp recordings. The current resembled the Na v current characteristic of excitable cells: it had the typical currentvoltage relationship (Fig. 1e) and showed fast activation and inactivation (Fig. 1i). The current was activated at about − 50 mV and peaked at about − 13 mV with a maximum amplitude of 330 ± 50 pA (mean ± SEM, n = 19). The average membrane capacitance was 28 ± 2 pF (n = 19), and the average current density was 13 ± 3 pA/ pF (n = 19). The average resting membrane potential, measured in the presence of K + -based intracellular solution was − 47 ± 1 mV (mean ± SEM, n = 15). The inactivation time constant decayed exponentially with increasing command voltages, while the decay of the activation time constant was more shallow (Fig. 1i). The steady-state inactivation curve was determined by measuring the amplitude of a response to − 10 mV test pulse following a series of prepulses (from − 140 mV to − 40 mV at 10 mV intervals). The normalized current amplitude was plotted against the prepulse voltage and fitted with the Boltzmann equation resulting in the half-inactivation voltage V 1/2 = − 94 ± 1 mV (n = 7) (Fig. 1f). To investigate the time dependency of recovery from inactivation, we used a paired-pulse protocol (Fig. 1g). The current was recorded after a second depolarizing pulse given at increasing time intervals until it finally recovered to its full size. The second peak currents were subsequently normalized to the prepulse peak current and plotted against the time between the two voltage pulses (Fig. 1h). Our data was fitted with an exponential function, and the best fit yielded to τ = 54 ± 3 ms (n = 5). The presence of Na v currents was confirmed using the universal extracellular Na v channel blocker TTX. By comparing the responses elicited with a voltage step from − 170 to − 10 mV, it was evident that addition of 1 μM TTX to the bath reduced the amplitude of the current to roughly one half of that recorded in the control extracellular solution (Fig. 2a, left). Thus, the recorded current was sensitive to TTX but required reasonably high concentrations. Furthermore, the sensitivity to TTX varied between the cells and in some cases even 10 μM TTX was not enough to block the current (Fig. 2a, left). The current was also sensitive to 2 mM QX-314, an intracellular Na v channel blocker added to the internal solution of the patch pipette that typically removed the current rapidly after breaking into the whole-cell configuration (Fig. 2a, right).  1 Patch clamp recordings of Na + currents from hESC-derived RPE. a, b Brightfield light microscopy images of hESC-derived RPE cells. a Mature hESC-derived RPE grown on insert for 2 months showing strongly pigmented cells and characteristic epithelial morphology. b Mature hESC-derived RPE was dissociated yielding single cells with typical morphology showing pigmented apical and non-pigmented basal sides. Scale bars 10 μm. Whole-cell patch clamp recordings as responses to a series of depolarizing voltage pulses (− 80 to + 60 mV, 10 mV steps) after strong hyperpolarization (− 170 mV) either c from mature monolayer of hESC-derived RPE or d from single hESC-derived RPE cells. Patch clamp pipette is visible in the center of the a and b images. e-i Analysis of the monolayer recordings. e The average current-voltage relationship (I vs V m , mean ± SEM, n = 12). f Steady-state inactivation curve was analyzed by plotting the normalized peak current at − 10 mV test pulse against the prepulse voltage (− 140 to − 40 mV, 10 mV steps) and fitting the data with the Boltzmann equation. The best fit was obtained with V 1/2 = − 94 ± 1 mV and k = 10 (n = 7). Data points indicate mean ± SEM. g, h The time dependency of recovery from inactivation. The second peak currents were normalized and plotted against the voltage pulse interval (10-270 ms). The best fit to an exponential function was obtained with τ = 54 ± 3 ms (n = 5) (individual datapoints for h available in Additional file 7: Table S2). i The activation (squares) and inactivation (circles) time constants were obtained from single exponential fits to the rising and decaying phases of the current responses shown in c and plotted against the command voltage (n = 7). j Summary of the patch clamp results

Voltage-gated sodium channels localize near cell-cell junctions in RPE
Our patch clamp data indicated that functional Na v channels are present in the hESC-derived RPE. The cellular localization of the channels was investigated by performing immunofluorescence studies where the cellular retinaldehyde-binding protein (CRALBP), a marker for RPE cells [16,17], was labeled together with the universal Na v channel marker. These hESC-derived RPE samples were then imaged with a laser scanning confocal Fig. 2 Blocker sensitivity and distribution of Na v channels. Patch clamp recordings were performed on mature hESC-derived RPE monolayers. a Applying TTX extracellularly (either 1 μM or 10 μM) did not entirely block the current (left). The current was completely removed by intracellular QX-314 (2 mM) (right). Laser scanning confocal microscopy (LSCM) images on Na v distribution in RPE cells. LSCM data inverted greyscale Z-maximum intensity projections of b hESC-derived and c mouse RPE stained against Na v channels (green) and RPE marker CRALBP (red). Scale bars 10 μm. d Immunogold labeling and transmission electron microscopy images showing Na v distribution at the apical membrane in the vicinity of the cell-cell junctions (black arrows). Scale bars 250 nm. e Dissociated hESC-derived RPE cells were let to adhere to poly-l-lysine coated coverslips for 30 min, fixed and immunolabeled against Na v together with CRALBP (up) or tight junction marker ZO-1 (down). The Na v label concentrated on the belt-like region in the middle of the cell, between the basal and apical sides. Scale bars 5 μm microscope (LSCM) by acquiring 3D image stacks (Fig. 2b), and the data were denoised by deconvolution. This showed that Na v channels were present in fully differentiated RPE. Furthermore, the Na v label concentrated primarily on the cellular borders with low expression elsewhere on the cell membrane while the CRALBP label was more uniformly localized to the apical side of the hESC-derived RPE (Fig. 2b).
Since the expression of Na v channels in RPE has previously been thought to be induced in vitro by the cell culturing [18,19] and since cells derived from ESCs might not fully replicate the pattern of ion channel expression in vivo [11,20,[24][25][26][27][28], we wanted to confirm their presence by using freshly isolated and non-cultured mouse RPE (Fig. 2c). The same labeling showed highly similar distributions in mouse RPE as in hESC-derived RPE: the CRALBP label was cytoplasmic on the apical side of the cells while Na v concentrated more on the cellular borders. Furthermore, the immunogold labeling for electron microscopy (immuno-EM) demonstrated the presence of Na v channels in the cell-cell junctions (Fig. 2d) and our immunolabeling with the tight junction marker ZO-1 showed highly overlapping distributions, strongly suggesting the primary Na v localization near the tight junctions (Additional file 1: Figure S1).
We investigated the mechanism underlying the previously reported lack of Na v currents from acutely isolated RPE cells (Fig. 1d). The hESC-derived RPE cells were seeded on glass coverslips for 30 min and immunolabeled with the universal Na v marker, CRALPB and ZO-1. Surprisingly, the Na v label was primarily concentrated in the narrow region separating the apical and basolateral sides of the cell. Together with ZO-1, Na v channels formed a clear ring-like structure between the apical and basal membranes following relaxation of junctional tension (Fig. 2e). Due to this junctional disruption, Na v channels might not be accessible to pass ionic currents in acutely dissociated RPE cells.

RPE cells express various voltage-gated sodium channel subtypes
Since ten different Na v channel subtypes, Na v 1.1-Na v 1. 9 and Na x , have been identified with drastically different expression profiles in diverse cell types, we wanted to investigate which specific channel subtypes are functionally expressed in the RPE cells. At the mRNA level, previous work has detected all of the Na v channels in donated human RPE-choroid preparations, specifically Na v subtypes 1.2-1.6 and Na v 1.9 [29,30]. We performed immunolabeling experiments with mouse and hESCderived RPE using specific antibodies against channel subtypes Na v 1.1-Na v 1.9 (Fig. 3a, b, Additional file 2: Figure S2). Confocal microscopy showed that Na v 1.4 localizes as beads-on-a-string to the cell-cell junctions ( Fig. 3a, b). Na v 1.8, on the other hand, localized overall to the apical side of the RPE cells (Fig. 3a, b). These data suggested that especially the Na v 1.4 and Na v 1.8 channels, which are usually expressed in skeletal muscle and dorsal root ganglia [31,32], respectively, are also present in RPE cells. Na v 1.6 the predominant channel of the adult central nervous system [33] showed a more homogenous labeling pattern in hESC-derived RPE and foci-like pattern in mouse RPE (Fig. 3a, b).
Subtypes Na v 1.1, Na v 1.3, Na v 1.5, Na v 1.7, and Na v 1.9 were detected in cell-cell junctions and apical membrane but their labeling was more prominent after fixation with lower concentration of paraformaldehyde (Additional file 2: Figure S2). The subtype Na v 1.2 was only weakly detected in both hESC-derived and mouse RPE. Additionally, we investigated the changes in channel subtype localization patterns during maturation of hESC-derived RPE (Additional file 3: Figure S3). The immunolabeling experiments indicated that the subtypes Na v 1.4, Na v 1.5, and Na v 1.8 changed from homogeneous cellular distribution to more specific localization either to cell-cell junctions (Na v 1.4) or to the apical side of the epithelium (Na v 1.5 and Na v 1.8) during the first 9 days of maturation.
To further verify the functional expression of the most prominent channel subtypes by electrophysiology, we repeated our patch clamp recordings using highly selective blockers for the channels Na v 1.4, Na v 1.6, and Na v 1.8. The average current-voltage relationship (I-V curve) was determined from all these recordings (n = 7) (Fig. 3c). The current was sensitive to the combination of 30 nM 4,9-anhydro-TTX (Na v 1.6 blocker), 1 μM A-803467 (Na v 1.8 blocker), and 600 nM μ-conotoxin GIIB (Na v 1.4 blocker), and the effect of inhibition was more potent with each added blocker thus confirming the expression and functionality of these channel subtypes in the hESC-derived RPE. However, the effect of inhibition was more significant when the blockers were combined with 10 μM TTX indicating the presence of Na v subtypes additional to 1.4, 1.6, and 1.8 (n = 11) (Fig. 3d).
Finally, the channel subtype composition was verified by carrying out mass spectrometry (MS) analysis of gel bands obtained from hESC-derived RPE protein lysates that had been tested to show the main Na v subtypes by Western blot (Additional file 4: Figure  S4). Here, we followed the "two-peptide rule" [34], considering a hit positive if two or more specific peptides were identified. Intriguingly, all of the nine types, except subtype Na v 1.2, were identified. This analysis thus further confirmed the expression of the three major subtypes (Na v 1.4, Na v 1.6, Na v 1.8) in RPE and was also positive for the Na x channel expression (Fig. 3e).

Voltage-gated sodium channels Na v 1.4 and Na v 1.8 are involved in POS phagocytosis in RPE
Our previous experiments showed that several Na v subtypes are present in both mouse and mature hESC-derived RPE. However, their physiological relevance remained unknown. Phagocytosis of POS is one of the major roles of RPE [3], and a plausible candidate function for the Na v channels, as it requires rapid activation and high synchronization [35]. We therefore next investigated the potential importance of Na v channels for POS phagocytosis.
To study their role in the phagocytosis process, we performed immunolabeling experiments with mouse eyes that had been prepared at light onset near the diurnal peak of phagocytosis. The role of the channels in POS uptake was studied by comparing the immunolabeling of the three major subtypes (Na v 1.4, Na v 1.6 and Na v 1.8) and opsin. Interestingly, at light onset, Na v 1.4 and Na v 1.8 localized to the bound POS particles (Fig. 4a). To confirm this redistribution of Na v channels, we next performed immuno-EM experiments (Fig. 4b, c), where we labeled the subtypes with gold nanoparticles in hESC-derived RPE. When the cells had not been exposed to POS particles, the localization of both channel subtypes was junction adjacent. This labeling pattern was particularly evident for Na v 1.4 ( Fig. 4b) that formed clusters at the apical part of the cell-cell junctions. After 2 h or 4 h of phagocytosis, however, we could again observe the change in labeling distribution as the channels interacted directly with the phagocytic cups or recently ingested phagosomes (Fig. 4b, c).
The redistribution of Na v channels occurring during phagocytosis ( Fig. 5a) was studied ex vivo with the channel blockers (Fig. 5b). For this purpose, we developed an assay where freshly opened mouse eyecups were incubated in physiological conditions with blocker solutions for 1 h starting at 15 min prior to light onset. The blocker for Fig. 3 Immunolabeling of different Na v subtypes in hESC-derived and mouse RPE, mass-spectrometry studies of Na v expression, and patch clamp recordings with selective Na v blockers. a, b The specific pattern of Na v subtypes was studied by immunolabeling. Laser scanning confocal microscopy Z-maximum intensity projections (xy-MIP) and yz cross-sections of a mature hESC-derived or b mouse RPE. Na v subtypes 1.4, 1.6, and 1.8 (green) were immunolabeled together with filamentous actin (phalloidin stain, red). Scale bars 10 μm. Right side panels show a higher magnification of the highlighted regions. Patch clamp recordings were performed on mature hESC-derived RPE using selective blockers for channel subtypes. c Na v subtypes were sequentially blocked by extracellularly applied 4,9-AnhydroTTX (30 nM, Na v 1.6 blocker), A-803467 (1 μM, Na v 1.8 blocker) and μ-Conotoxin GIIB (600 nM, Na v 1.4 blocker). The average normalized peak current-voltage relationship (I/I max vs V m ) was determined from all recordings (mean ± SEM, n = 7). d Applying the selective blockers in combination with TTX (10 μM) removed most of the Na v currents (n = 11). e Mass spectrometry analysis of Na v channel expression in hESC-derived RPE. Specific peptides were identified for all Na v subtypes, excluding Na v 1.2 Na v 1.4 as well as the combination of all Na v blockers significantly prevented the disappearance of Na v 1.4 from cellcell junctions when compared to the control (Fig. 5b). The inhibition effect was similarly observed in hESC-derived RPE in vitro when the cells were incubated for 2 h with POS mixed with blocker solutions (Fig. 5c). We did not observe significant differences in the overall labeling pattern of Na v 1.8 after the blocker incubation. Taken together, these experiments indicate the participation of Na v channels in the phagocytic processes of RPE cells in vitro and in vivo.
Na v 1.4 knockdown and the inhibition of Na v channels significantly reduces the number of ingested POS particles in hESC-derived RPE Our LSCM and immuno-EM imaging of POS phagocytosis in RPE indicated a close interaction between Na v channels and phagocytosed POS particles. Therefore, we hypothesized that reducing the Na v channel activity could affect the rate of phagocytosis. After observing the dramatic change in the localization of Na v 1.4, we decided to study its effect further by silencing the channel subtype expression by shRNAs (Fig. 6). Due to the challenges associated with passaging of hESC-derived RPE cells, such as loss of the cobblestone morphology and poor cell proliferation, we opted for the lentivirus shRNA constructs. The transduction of the RPE cells had to be conducted several days after the cell seeding yielding a monolayer with sparse distribution of single GFP-positive cells (Fig. 6e). Since it was not possible to confirm the knockdown efficiency in hESC-derived RPE, the constructs were first validated with ARPE-19 cells (Additional file 5: Figure S5). The cells were transduced Laser scanning confocal microscopy Z-maximum intensity projections of mouse RPE prepared at light onset showed localization of opsin labeled POS particles (blue) and Na v 1.4 (green) together with Na v 1.8 (red). Lower panels show a high contrast blowup of highlighted regions. Scale bars 10 μm. To study phagocytosis in vitro, mature hESC-derived RPE were labeled with 1.4 nm nanogold-conjugated antibodies against b Na v 1.4 and c Na v 1.8 during phagocytosis of purified porcine POS particles and in control conditions. Without POS exposure, both channels showed localization near the cell-cell junctions (black arrows) but by incubating the monolayers with POS particles for b 2 h or c 4 h, the localization (black arrows) was also evident around the phagocytic cups and recently ingested phagosomes. Scale bars 250 nm with the shRNA constructs and collected for Western blot. Next, the knockdown effect of the verified construct was confirmed in hESC-derived RPE by conducting single cell patch clamp recordings and applying μ-conotoxin GIIB extracellularly (Fig. 6a-d).
The cells expressing the target shRNA had both highly reduced Na v currents and minimal reactivity to the blocker (Fig. 6d) when compared to EGFPexpressing (Fig. 6c) or wildtype hESC-derived RPE cells (Fig. 6b). Intriguingly, when the cells were used in the phagocytosis assay (Fig. 6e), the silencing of Na v 1.4 caused a drastic reduction in the total number of POS particles found in individual GFP-positive cells on the monolayer (Fig. 6f, g). Laser scanning confocal microscopy Z-maximum intensity projections of a Na v 1.4 localization in mouse RPE at light onset and 2 h after it showed strong reduction of the beads-on-a-string type labeling from cell-cell junctions. Different assays were used to investigate Na v 1.4 distribution during phagocytosis and the effect of selective blockers for Na v 1.4 (600 nM μ-Conotoxin GIIB) and Na v 1.8 (1 μM A-803467) in combination with 10 μM TTX, or only of the selective blocker for Na v 1.4. b The redistribution of Na v 1.4 was studied ex vivo by incubating opened mouse eyecups in control solution or with the selective blockers. In both of the blocker samples, the redistribution was inhibited and the beads-on-a-string type labeling remained visible (white arrows) in the cell-cell junctions. c The hESC-derived RPE phagocytosis assay in vitro showed a highly similar redistribution of Na v 1.4 and the blockers had the same effect as in the ex vivo mouse eyecup assay. Scale bars 10 μm To then study the effect of all Na v channels on a larger population of cells, we performed the in vitro phagocytosis assay (Fig. 7a) in the presence of Na v 1.4 and 1.8 blockers and TTX. The effect was first quantified by counting the number of particles from the immuno-EM images that had been tagged with gold nanoparticle labeled opsin (Fig. 7b). This revealed a drastic reduction in the total number of bound and internalized POS particles. To better analyze the effect, the assay was carried out by imaging large fields of immunolabeled opsin and ZO-1 and by comparing the number of POS particles in Na v blocker and control conditions after 2 h at +37°C (Fig. 7c). The results showed that the blocker combination caused a 34% (n = 18) reduction in the total number of POS particles labeled with opsin (Fig. 7d).  Table S3). e The level of POS phagocytosis was analyzed with the EGFP expressing hESC-derived RPE cells. Filamentous actin was stained with phalloidin (blue) to highlight epithelial cell-cell junctions, EGFP (red) was used to identify the transduced cells and POS were labeled with opsin (green). f The average distribution of POS particles was analyzed from several images that had a single shRNA expressing cell placed in the middle. g The relative intensity of POS labeling in each square of the 3 × 3 grid was analyzed from Na v 1.4 shRNA cells (n = 22 images) and control EGFP cells (n = 18 images). Scale bars 10 μm Next, we wanted to independently quantify the binding and internalization stages of the phagocytosis pathway. The binding was investigated by incubating the POS-supplemented hESC-derived RPE cells at room temperature (RT) [35,36] with 5% CO 2 . Most of the unbound POS were then removed by gentle washing, and the monolayers were placed in the incubator for either 2 h or 5 h to investigate the internalization stage of phagocytosis. This was carried out also with the combination of Na v blockers and the total number of POS particles was compared to control cells at each timepoint (Fig. 7e, f ). The results showed no statistically significant changes (p = 0.1) in the number of POS particles due to the Na v inhibition in the particle binding phase (2 h RT, Control n = 15, Na v blockers n = 10). However, Na v blocking caused a 41% reduction of POS particle Fig. 7 POS phagocytosis assay of hESC-derived RPE with selective Na v blockers. POS phagocytosis assays were performed on mature hESCderived RPE by incubating the monolayers with purified porcine POS particles with or without Na v blockers (600 nM μ-Conotoxin GIIB, 1 μM A-803467 and 10 μM TTX). a Laser scanning confocal microscopy (LSCM) Z-maximum intensity projections (Z-MIP) and yz and xz cross-sections of the RPE and POS particles (green) with filamentous actin staining (red) after 2 h of phagocytic challenge. Scale bar 10 μm. b Quantification of POS particles from TEM images with immunogold labeled opsin showed a 60% decrease of POS particles/cell in the presence of the blockers. Scale bar 250 nm. c LSCM Z-MIP images of ZO-1 (gray) together with opsin (green). Scale bars 20 μm. d Quantification of control (n = 15) and Na v blocker samples (n = 18) during phagocytosis (2 h +37°C) showed a 34% reduction in the total number of POS particles in the presence of Na v blockers. e, f Na v channel role in POS binding, internalization and further processing. e LSCM Z-MIP images of phalloidin (red) together with opsin (green) during POS binding and internalization. Scale bars 10 μm. f Quantification of the binding phase (2 h RT, Control n = 15, Na v blockers n = 10) showed no significant reduction in POS numbers due to Na v channel blockers, but a 41% decrease in the internalization phase (2 h RT + 2 h at +37°C, Control n = 15, Na v blocker n = 15). In control cells, the POS numbers decreased in the processing phase, but not in the presence of Na v channel inhibitors (2 h RT + 5 h at +37°C, 25% decrease, Control n = 15, Na v blocker n = 15). Center lines show the medians; box limits indicate the 25th and 75th percentiles as determined by R software and whiskers extend to minimum and maximum values numbers in the internalization phase (2 h RT + 2 h at +37°C, Control n = 15, Na v blocker n = 15) (p < 0.001). In control cells, the internalized particles were degraded, which was detected as a 25% reduction in particle numbers in the processing phase (2 h RT + 5 h at +37°C, n = 15, p < 0.01) (Fig. 7f). Na v blocking substantially reduced further processing of POS particles in RPE cells as there were no significant changes in particle numbers between internalization and processing phases (n = 15, p = 0.16). Together, our data indicate that functional Na v channels directly interact with phagosomes (Fig. 4) and that they are involved in the POS particle internalization and further processing (Fig. 7e, f ).

Discussion
Recent studies show revolution in our understanding of the roles that Na v channels have in cellular functions; no longer are these proteins considered important only in "classically" electrically excitable tissues. Here, we provide the first evidence, to our knowledge, that Na v channels are expressed in the epithelial cells (Figs. 1, 2, and 3, Additional file 1: Figure S1, Additional file 2: Figure  S2) and that their activity co-regulates phagocytosis in RPE. Our observations of Na v channels and Na v -mediated currents in intact RPE preparations (mature hESCderived RPE monolayers and freshly isolated mouse RPE) demonstrate that previous observations of Na v -mediated currents in cultured RPE cells are not preparation-dependent artifacts [18,19]. Rather, the absence of Na v -mediated currents in acutely isolated RPE cells (Fig. 1d) likely results from the destruction of tight junction complexes during dissociation (Fig. 2e). Internalization of Na v channels, of course, would result in diminution or absence of membrane currents mediated by these channels as observed by us (Fig. 1d) and others [18,19]. Reductions in Na v currents have also been reported in dorsal root ganglion neurons following peripheral axotomy [37,38] but this effect was found to be reversible by exposure to growth factors [37,38]. The observation of Na v currents in recordings from hESCderived RPE monolayers, we believe, is strong evidence that cells in RPE with intact tight junctions usually express functional Na v channels in their plasma membranes.
The properties of Na v -mediated currents in hESCderived RPE cells are consistent with the earlier recordings from other non-neuronal cells [23]. Furthermore, non-excitable cells have been shown to display varying sensitivity to TTX based on their Na v subtype composition [23], and our recordings from RPE align with this observation. The high TTX concentration needed for a full inhibition of the current is supported by the finding that RPE cells were labeled strongly by an anti-Na v 1.8 antibody; Na v 1.8 is the least sensitive of Na v channels to TTX [39]. Our pharmacological analysis of Na v -mediated currents using Na v subtype-specific blockers, immunofluorescence microscopy and MS-based analysis (Fig. 3, Additional file 2: Figure S2) indicated that Na v 1.1 as well as Na v 1.3-Na v 1.9 subtypes are present in the RPE together with the sodium concentration sensitive Na x channel [40]. This data was further supported by the Na v current characteristics: the relatively slow recovery from inactivation [41,42] (Fig. 1h) and the large variation visible in the early phase of the recovery, indicate the presence of several different Na v subtypes in the RPE. Compared to neurons, other non-excitable cells, such as non-process bearing astrocytes and supporting cells of the porcine vomeronasal organ [43][44][45], have been shown to exhibit more negative steady-state halfinactivation voltages, and similar V 1/2 values were found for hESC-derived RPE (Fig. 1f ). Accordingly, and similarly to astrocytes [46], we never observed spontaneous action potentials in these cells. Earlier recordings from RPE have shown V 1/2 values much closer to neurons; however, these results were obtained after only a short period in culture [15,19]. As our data demonstrates that the Na v channel localization and subtype composition is dynamically regulated during development (Additional file 3: Figure S3), we believe that RPE maturation stage including the junctional localization of the channels could greatly influence their electrophysiological properties. This is supported by the previous recordings that identified Na v channels in astrocytes in hippocampal slices [46].
For the MS-based identification of Na v proteins, we chose to perform the analysis from gel bands [47][48][49] due to the high sensitivity and possibility for absolute protein identification [48] provided by this approach. However, it was challenging to purify the Na v channel proteins from the RPE cells, most probably due to their several transmembrane domains and high hydrophobicity [50]. Therefore, the number of detected Na v channel-specific peptides does not necessarily directly correlate to their relative abundance. Na v 1.4 and Na v 1.8 of the identified Na v subtypes showed the strongest staining in the immunohistochemical analysis and carried over one third of the total current in patch clamp recordings. Therefore, we focused on these two channel subtypes while investigating the physiological roles of Na v channels. It is noteworthy, however, that the subtypes Na v 1.1, Na v 1.3, Na v 1.5, and Na v 1.7 only showed strong junctional labeling after fixation with 1% PFA. Our antibodies labeling these subtypes are directed against different residues of the same intracellular loop between the domains III and IV. Importantly, this region might be sequestered in the dense cell-cell junctions thus hampering their detection through conventional immunocytochemistry [51]. On the other hand, cells derived from ESCs can bear differences compared to native cells regarding subtype composition. Earlier studies have demonstrated high similarity but also certain deviations in terms of ion channel distribution and biophysical characteristics as well as channel pharmacology and transcriptional profile [11,20,[24][25][26][27][28].
Previous studies have shown that in macrophages, Na v channels have important roles in phagocytosis [22,[52][53][54]. We observed an accumulation of Na v 1.4 and Na v 1.8 towards POS particles during phagocytosis. The translocation of Nav1.4 from tight junctions was still evident 2 h after light onset (Fig. 5a), near the peak expression level of a phagocyte cell surface tyrosine kinase receptor MerTK [55]. The involvement of Na v 1.4 in phagocytosis was supported by the fact that its silencing decreased the amount of POS particles (Fig. 6). Furthermore, following Na v blocker incubation, we observed a decrease in its translocation (Fig. 5) with a concurrent reduction in the number of POS particles (Fig. 7). Na v channel redistribution has been observed in demyelinated axonal membrane [56] suggesting that these ion channels can display dynamic regulation of distribution and have important implications in various pathologies.
Participation of Na v channel activity to POS phagocytosis was further indicated by their direct association with both the forming phagocytic cups and ingested phagosomes (Fig. 4). Although inhibition of Na v channels did not abolish phagocytosis, the observed 41% attenuation (Fig. 7) was similar to the previously reported effect of TTX in microglia [57]. Our assays suggest that the Na v channels are involved in the engulfment and further processing of the phagosomes, since inhibiting Na v activity did not impair the binding of POS particles (Fig. 7). Interestingly, changes in intracellular free calcium concentration regulate phagocytosis in RPE [58] and, more specifically, particle engulfment in other phagocytes [59]. Na v mediated sodium influx could result in increased calcium concentration via reversed functioning of sodiumcalcium exchangers expressed in the RPE apical membrane [60,61] thus affecting phagocytosis. Alternatively, Na v channels could regulate endosomal acidification by providing a sodium efflux pathway to enhance the entry of protons, similarly to macrophages [22,53]. In these cells, phagocytosis has been associated with membrane potential hyperpolarization due to the activation of a Ca 2+dependent K + conductance [62,63]. Such change in potential could relieve the Na v channel inactivation in the early phase of phagocytosis, and the channels could subsequently be activated in the phagosomes and endosomes with membrane potential in the Na v activation range [64][65][66]. Furthermore, the channels could be involved directly in the circadian control of the pathway, as has been recently shown for other ion channels [10].
The fact that RPE expresses such a versatile array of Na v channels suggests that besides phagocytosis, these channels also have other roles in the physiology of the RPE. Overall, sodium homeostasis is critical to epithelial transport mechanisms, and our observation of Na v channels, including the non-voltage-gated Na x channel, brings a new piece in the ongoing identification of the sodium conducting proteins in RPE. In excitable cells, Na v and Ca 2+ channels form local signaling complexes that are essential for various intracellular processes [67]. Similar roles for these channels could be possible in RPE as well. Moreover, epithelial cells, including RPE, show strong calcium waves in response to mechanical stimulation [68][69][70], and it is likely that Na v channels are involved in the process. It is well established that the Ca 2+ binding protein calmodulin (Cam) interacts directly with the C-terminal domain of Na v [71], and it was recently shown that the Ca 2+ -free form of Cam, ApoCam, enhances the Na v channel opening by several-fold [72]. Thus, Na + -and Ca 2+ -dependent signaling pathways can interact in epithelia as has been reported in the case of astrocytes [73]. Lastly, it has been suggested that retinal Müller cells that display similar low Na v channel densities could be activated by the adjacent neurons and it is possible that RPE cells could also serve as voltage sensors reacting to signals arising from photoreceptors [44,74].

Conclusion
The results of this study demonstrate that functional Na v channels are present in mouse and hESC-derived RPE cells with intact tight junctions. Specifically, we confirm the presence of Na v 1.1 as well as Na v 1.3-Na v 1.9 subtypes and the sodium concentration sensitive Na x channel, showing that their expression is not due to specific culturing conditions. Our data shows that the most prominent subtypes Na v 1.4 and Na v 1.8 are involved in photoreceptor outer segment renewal by directly interacting with phagosomes. Inhibiting the activity of these channels by either pharmacological blockers or shRNAmediated silencing impairs the phagocytosis process, particularly at the engulfment or further processing stages. Collectively, we demonstrate that Na v channels yield RPE cells the capacity for fast voltage sensitivity and that the channels are a vital part of its physiology.

Antibodies and reagents
Catalog and batch numbers as well as other information for the chemicals and antibodies used in this study can be found from the Additional file 6: Table S1.
Generation of Na v 1.4 shRNA cultures ARPE-19 cells (ATCC, USA) were maintained in DMEM/F12 medium containing 10% FBS, 1% Gluta-MAX and 1% penicillin/streptomycin at 37°C with 5% CO 2 . The medium was changed 3 times a week. The confluent cells were dissociated with trypsin-EDTA (25200-056, Thermo Fisher Scientific) and transfected the following day with shRNA expression vectors containing the reporter vector pLKO.1-CMV-tGFP or pLKO.1-puro-CMV-TurboGFP (Sigma-Aldrich). The expression of shRNA was then investigated by Western blot analysis for Na v 1.4 for clones TRCN0000416043, TRCN0000425151, and TRCN000044419. This was carried out by comparing the labeling intensity against βactin that was used as the loading control (n = 3). The normalization was carried out by subtracting the background intensity (I Bgrnd Nav ; I Bgrnd Actin ) from the intensity of Na v and β-actin bands ( I Band Nav ; I Band Actin ). The bands were then normalized to the maximum intensity (I Max Nav ; I Max Actin ), yielding normalized Western blot band intensities between values 0 and 1 as follows: : These normalized intensities were then used to calculate the relative knockdown as After Western blot analysis, 2 μl of the verified clone TRCN000044419 (8.1 × 10 6 TU/ml) and 8 mg/ml polybrene were added on hESC-derived RPE cells grown on insert. Transduction was made 5-23 days after the cell seeding and virus particles were incubated for 1 day before changing medium. The silencing of Na v 1.4 current was verified by patch clamp from maturated, 8-10 weeks old, hESC-derived RPE cells.

Sample preparation
For monolayer patch clamp recordings and immunolabeling, the membrane of the culture insert was removed from the insert holder and cut into smaller pieces. The cells were rinsed three times either with PBS (for immunolabeling) or with Ames' solution (for patch clamp recordings). For the experiments on dissociated cells, the hESC-derived RPE monolayers were treated with TrypLE Select for 10 min in +37°C, gently mechanically triturated with a pipette and centrifuged for 5 min at 1000 rpm. Dissociated cells were resuspended in culture medium, seeded on glass coverslips coated with poly-Llysine (Sigma-Aldrich) and allowed to settle down for 10 min for patch clamp recordings and 30 min for immunolabeling.
Mouse RPE was prepared for immunolabeling as follows. C57BL/6 mice were euthanized by CO 2 inhalation and cervical dislocation. The eyes were enucleated and bisected along the equator, and the eyecups were sectioned in Ames' solution buffered with 10 mM HEPES and supplemented with 10 mM NaCl, pH was adjusted to 7.4 with NaOH (Sigma-Aldrich). The retina was gently removed from the eyecup leaving the RPE firmly attached to the eyecup preparation.

Patch clamp recordings
Ionic currents were recorded from mature hESC-derived RPE monolayers or freshly dissociated cells using the standard patch clamp technique in whole-cell configuration. Patch pipettes (resistance 5-6 MΩ) were filled with an internal solution containing (in mM) 83 CsCH 3 SO 3 , 25 CsCl, 10 TEA-Cl, 5.5 EGTA, 0.5 CaCl 2 , 4 ATP-Mg, 0.1 GTP-Na, 10 HEPES, and 5 NaCl; pH was adjusted to~7.2 with CsOH and osmolarity was~290 mOsm (Gonotec, Osmomat 030, Labo Line Oy, Helsinki, Finland). For the recordings with K + -based internal solution, CsCl was replaced with KCl and CsCH 3 SO 3 was replaced with K-gluconate. In some experiments, the internal solution also contained 2 mM QX-314-Cl (from Sigma-Aldrich). During all recordings, the tissue was perfused at 2.5 ml min −1 with Ames' solution (Sigma-Aldrich) buffered with 10 mM HEPES and supplemented with 10 mM NaCl and 5 mM TEA-Cl. The pH was adjusted to 7.4 with NaOH and the osmolarity set to3 05 mOsm. The bath solution contained 10 nM-10 μM TTX citrate (from Tocris Bioscience) when the effect of TTX on the recorded currents was investigated, and 30 μM 18α-glycyrrhetinic acid (from Sigma-Aldrich) when the effect of gap junctional coupling was tested.
For the channel subtype recordings, the bath solution was supplemented with 30 nM 4,9-AnhydroTTX, 1 μM A-803467, or 600 nM μ-Conotoxin GIIB. All recordings were made in voltage clamp mode with pClamp 10.2 software using the Axopatch 200B patch clamp amplifier connected to an acquisition computer via AD/DA Digidata 1440 (Molecular Devices, USA). The access resistance was below 30 MΩ and the membrane resistance above 150 MΩ. Series resistance was 15-30 MΩ and was not compensated. Holding potentials were corrected for a 3 mV liquid junction potential during the data analysis. All recordings were performed at room temperature.

Immunolabeling
Prior to immunolabeling, samples were washed three times with PBS and fixed for 15 min with 4% paraformaldehyde or 10 min with 1% paraformaldehyde (pH 7.4; Sigma-Aldrich). After repeated washes with PBS, samples were permeabilized by incubating in 0.1% Triton X-100 in PBS (Sigma-Aldrich) for 15 min and subsequently blocked with 3% BSA (BSA; Sigma-Aldrich) for 1 h. All immunolabeling incubations were done at room temperature. Primary antibodies against the following proteins were used in this study: cellular retinaldehyde-binding protein (CRALBP) 1:400 (ab15051, Abcam), Na v 1.

Pre-embedding immunogold labeling
The hESC-derived RPE monolayers were washed three times with phosphate-buffered saline (PBS) and then fixed for 2 h at RT in periodate-lysine-paraformaldehyde (PLP) fixative. Fixed cells were prepared for preembedding EM as described previously [76,77]. Cells were treated with 0.01% saponin and 0.1% BSA in 0.1 M phosphate buffer, pH 7.4 (Buffer A) before adding the primary antibodies diluted in Buffer A. The concentration of all primary antibodies was doubled for the experiment compared to immunolabeling. After 1 h incubation at RT and washes with Buffer A, 1.4 nm nanogold-conjugated polyclonal Fab' fragment of goat anti-rabbit IgG or of goat anti-mouse IgG (Nanoprobes. com, Yaphank, NY, USA) diluted to 1:50 in Buffer A was applied for 1 h, followed by washes with Buffer A and 0.1 M phosphate buffer (pH 7.4). Cells were post-fixed with 1% glutaraldehyde in phosphate buffer for 10 min at RT, quenched with 50 mM NH 4 Cl in phosphate buffer for 5 min at RT, and then washed with phosphate buffer and water.
The samples were treated in dark with HQ-silver (Nanoprobes.com) for 5 min followed by washes with water and gold toning (2% sodium acetate 3 × 5 min at RT, 0.05% gold chloride 10 min at +4°C, 0.3% sodium thiosulphate 2 × 10 min at +4°C). After washes with water, the cells were reduced in 1% osmium tetroxide in 0.1 M phosphate buffer for 1 h at +4°C and dehydrated with graded series of ethanol (70%, 96%, 100%), then stained with 2% uranyl acetate. Finally, the monolayers were embedded in Epon (TAAB Embedding resin, medium, TAAB Laboratories Equipment Ltd., Berks, UK) and after polymerization, sections perpendicular to the membrane were cut with an ultramicrotome (Leica ultracut UCT ultramicrotome, Leica Mikrosysteme GmbH, Austria). The thin sections (200 nm) were placed on carbon-coated single-slot grids and were imaged with JEOL JEM-1400 transmission electron microscope (JEOL Ltd., Tokyo, Japan) equipped with bottom-mounted Quemesa CCD camera (4008 × 2664 pixels). High voltage of 80 kV was used for imaging.

Western blotting
The hESC-derived RPE and ARPE-19 protein lysates were obtained by incubating 1 × 10 6 cell pellets in RIPA buffer supplemented with Halt protease inhibitor cocktail (87786, Thermo Fisher Scientific) for 30 min at +4°C on constant agitation. The lysate was then centrifuged at +4°C for 20 min at 12,000×g, mixed with Novex sample buffer (NP0007, Thermo Fisher Scientific) and heated at +70°C for 10 min. The protein lysates were then loaded onto 3-8% NuPage gel (EA0375, Thermo Fisher Scientific) or Bolt™ 4-12% Bis-Tris Plus Gels (NW04120, Thermo Fisher Scientific), fractionated by SDS-PAGE and then either processed for MS analysis or transferred to nitrocellulose membrane via Trans Blot Turbo Transfer system according to the manufacturer's protocols (BioRad).

Sample preparation of mass spectrometry
The SDS-page gels were labeled overnight at RT with coomassie blue dye to identify the bands.

Identification of proteins
Identification of the proteins was done using Protein Pilot® 4.5 (Sciex, Redwood City, USA) and all data dependent analysis (DDA) runs MS/MS spectra were identified against respective Na v channel protein data retrieved from UniprotKB/SwissProt library. FDR 1% and 99% peptide confidence level was used in the library creation and only distinctive peptides were used in the identification. Mass accuracy was set to 5 ppm for each peptide.
The flush system was then switched to line with analytical column. The peptide samples were analyzed with 120 min 6 step gradient using eluent A: 0.1% FA in 1% ACN and eluent B: 0.1% FA in ACN (eluent B from 5 to 7% over 2 min; 7 to 24% over 55 min; 24 to 40% over 29 min; 40 to 60% over 6 min; 60 to 90% over 2 min and kept at 90% for 15 min; 90 to 5% over 0.1 min and kept at 5% for 13 min) at 300 nl/min. The following key parameters were applied for Triple-TOF mass spectrometer in shotgun identification analysis: ion spray voltage floating (ISVF) 2300 V, curtain gas (CUR) 30, interface heater temperature (IHT) +125°C, ion source gas 1 13, declustering potential (DP) 100 V. Methods were run by Analyst TF 1.5 software (Ab Sciex, USA). For IDA parameters, 0.25 s MS survey scan in the mass range 350-1250 mz was followed by 60 MS/MS scans in the mass range of 100-1500 Da (total cycle time 3.302 s). Switching criteria were set to ions greater than mass to charge ratio (m/z) 350 and smaller than 1250 (m/z) with charge state 2-5 and an abundance threshold of more than 120 counts. Former target ions were excluded for 12 S. IDA rolling collision energy (CE) parameters script was used for automatically controlling CE.

Phagocytosis assay for hESC-derived and mouse RPE
The porcine POS particles were isolated and purified as previously described [75,78]. Briefly, the eyecups obtained from a slaughterhouse were opened and retinas were removed using forceps under dim red light. The retinas were shaken gently in 0.73 M sucrose phosphate buffer and separated after filtering in sucrose gradient using an ultracentrifuge (Optima ultracentrifuge, Beckman Coulter, Inc., Brea, CA) at 112,400 x g for 1 h at +4°C. The collected POS layer was centrifuged 3000×g for 10 min at +4°C and stored in 73 mM sucrose phosphate buffer at −80°C.
The purified POS particles were fed to the hESCderived RPE cells in a KO-DMEM medium supplemented with 10% fetal bovine serum (FBS) and incubated for either 2 h at RT or 2 h, 4 h or 5 h at +37°C in 5% CO 2 . In the blocker experiments, selective blockers for Na v 1.4, Na v 1.8 and TTX were also added to the medium for the incubation. Then the monolayers were washed twice briefly with PBS and fixed with PFA according to the immunostaining protocol. Phagocytosis was studied in vivo by preparing the mouse eyes under dim red light either at light onset or 2 h and 10 h after it. The mice were reared in normal 12-h light/dark cycle. When blockers were used, the eyecup was opened and then incubated in blocker solutions diluted in Ames' as described above, for 1 h at +37°C with the retina left intact.

Quantification of POS particles in hESC-derived RPE
To detect and quantify POS particles, large random fields were imaged from 3 different samples in each condition with Zeiss LSM780 LSCM (the total number of images in each case is included in the figure legends as "n"). The images were first blurred with a Gaussian function after which a Z-maximum intensity projection was binarized using a global threshold. The number of POS particles was then analyzed from the images converted to mask. In the hESC-derived RPE where subtype Na v 1.4 had been silenced with lentiviral vectors encoding shRNAs, phagocytosis was analyzed by capturing several fields with a GFP-positive cell in the center of the image. The MIP images were then combined and the average distribution of POS particle labeling was compared between control-GFP (EGFP) construct and the clone TRCN000044419. The image was split into a 3 × 3 grid and the relative intensity of POS labeling was analyzed for each individual square of the grid.

Statistical analysis of the POS phagocytosis quantification
Each phagocytosis experiment was repeated three times and the images were pooled together. The normality of the data was tested by using Shapiro-Wilk normality test and the differences were first analyzed using ANOVA. Finally, pairwise comparison was conducted by using Kruskal-Wallis test to confirm the possible statistical significance between the experimental conditions.

Confocal microscopy and image processing
Confocal microscopy was performed with Zeiss LSM780 LSCM on inverted Zeiss Cell Observer microscope (Zeiss, Jena, Germany) by using Plan-Apochromat 63x/ 1.4 oil immersion objective. Voxel size was set to x = y = 66 nm and z = 200 nm and 1024 × 1024 pixel stacks of 70-120 slices were acquired with line average of 2. The Alexa Fluor 405 was excited with 405 nm diode laser; Alexa Fluor 488 with 488 nm laserline from Argon laser; Alexa Fluor 568 and TRITC with 561 nm DPSS or 562 nm InTune laser; Atto 633 and Alexa Fluor 647 with 633 nm HeNe and with 628 nm InTune laser. Emission was detected with windows of (in nm) 410-495 (DAPI, Alexa Fluor 405), 499-579 (Alexa Fluor 488), 579-642 (Alexa Fluor 568), and 642-755 (Alexa Fluor 647). Laser powers were minimized to avoid bleaching and photomultiplier tube sensitivities were adjusted to obtain optimal signal-to-noise ratio of the signal. The data was saved in .czi format and deconvolved using Huygens Essential (SVI, Hilversum, Netherlands) software. The deconvolution was performed with theoretical PSF, signalto-noise ratio of 5 and quality threshold of 0.01. Information regarding the refractive index of the sample was provided by the manufacturer of the ProLong Gold antifade mounting media. Images were further processed with ImageJ [79] and only linear brightness and contrast adjustments were performed for the pixel intensities.

Additional files
Additional file 1: Figure S1. Immunolabeling of Na v in hESC-derived and mouse RPE. Z-maximum intensity projections (Z-MIP) of (a) hESCderived and (b) mouse RPE stained against Na v channels (green) and tight junction marker ZO-1 (red), together with cross-sectional X-MIPs from the highlighted regions. (PNG 729 kb) Additional file 2: Figure S2. Immunolabeling of different Na v subtypes in hESC-derived and mouse RPE. Different Na v channel subtypes were immunolabeled in (a) mature hESC-derived and (b) mouse RPE that had been fixed with 1% PFA. Laser scanning confocal microscopy Zmaximum intensity projections of Na v subtypes (green) labeled together with filamentous actin (phalloidin, red). In both samples, the subtypes Na v 1.1, Na v 1.3, Na v 1.5, Na v 1.7 and Na v 1.9 showed labeling in cell-cell junctions and apical membrane. The subtype Na v 1.2 gave extremely weak signals in both samples. Scale bars 10 μm. (PNG 4721 kb)