Fast voltage sensitivity in retinal pigment epithelium: sodium channels and their novel role in phagocytosis

Despite the discoveries of voltage-gated sodium channels (Nav) from a number of non-excitable cell types, the presence of Nav-mediated currents in cells of the retinal pigment epithelium (RPE) has been dismissed as a cell culture artifact. Here, we challenge this notion by demonstrating functional Nav1.4-Nav1.6 and Nav1.8 channels in human embryonic stem cell derived and mouse RPE. Importantly, we show that Navs are involved in photoreceptor outer segment phagocytosis: blocking their activity significantly reduces the efficiency of this process. Consistent with this role, Nav1.8 co-localizes with the endosomal marker Rab7 and, during phagocytosis, with opsin. Nav1.4 localizes strongly to the cell-cell junctions together with the gap junction protein Connexin 43. During phagocytosis, both are localized to the phagosomes with a concurrent decrease in the junctional localization. Our study demonstrates that Navs give the capacity of fast electrical signaling to RPE and that Navs play a novel role in photoreceptor outer segment phagocytosis.


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 blood stream while removing metabolic end products, and absorbing scattered light 1,3 . Most importantly, though, the 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 at light offset for cones 7,8 . This evolutionary 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 currents linked to calcium-dependent potassium and chloride channels 10 .
Since the first single-cell recordings from RPE cells in 1988 11 , a large variety of different ion channels have been identified in them 5 . Much of this work was done on RPE cultured from various species, and some channels have been found only in cultured RPE [12][13][14] .
Thus, there appear to be important differences between freshly isolated and cultured RPE 15 .
One of the channel types identified in cultured RPE is the voltage-gated Na + channel (Nav) 16,17 ; Nav-mediated currents have not been recorded from freshly-isolated RPE 18,19 . Thus, it is widely thought that the expression of Nav in RPE cells is the result of the neuroepithelial differentiation in culture 5,15,20 . 3 Here, we challenge this conventional wisdom by demonstrating functional Nav channels in cultured human embryonic stem cell (hESC) -derived RPE in vitro and freshly-isolated mouse RPE in vivo. Our work reveals that Nav channels in RPE are primarily localized to the tight junctions that connect the cells. We conclude that this specific localization of Navs to cellcell contacts accounts for previous findings that dissociated RPE cells lack Nav-mediated currents. Specifically, we hypothesize that Nav channels could co-regulate photoreceptor outer segment (POS) phagocytosis. Our hypothesis is supported by a recent demonstration of the involvement of Nav channels in phagocytosis of mycobacteria by macrophages 21 . Our work provides evidence that Nav1.8 localizes with the endosomal marker Rab7, and that they both accumulate with the phagosomal particles. Further, Nav1.4 co-localizes with connexin 43 (Cx43), a constituent of the gap junctions connecting RPE cells 22 . During phagocytosis, however, they both are localized to the phagosomes with a concomitant decrease in junctional localization. Interestingly, this phagosomal translocation was significantly reduced by selective Nav channel blockers. Moreover, the selective blockers combined with the universal Nav blocker tetrodotoxin (TTX) reduced the total number of POS particles by up to 34%. More generally, our observations add to the growing body of evidence that Navs 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 ).

Functional voltage-gated sodium channels are present in RPE monolayer cultures derived from human embryonic stem cells
We used whole-cell recordings from mature hESC-derived RPE in K + free extra-and intracellular solutions to observe transient inward currents elicited by a series of depolarizing 4 voltage pulses after strong hyperpolarization to -170 mV (Fig. 1c, n=33). These recordings were done from an intact monolayer (Fig. 1a) in the presence and absence of a gap-junction antagonist (18α-glycyrrhetinic acid). Similar currents were also identified, albeit infrequently and with reduced amplitudes, in cells from freshly dissociated mature hESC-derived RPE (Fig.   1b, 1d, n=6). In all experiments, the current resembled the Nav current characteristic of excitable cells: it had the typical currentvoltage relationship (Fig. 1e) and showed fast activation and inactivation (Fig. 1c). The current was activated at about -50 mV and peaked at about -13 mV with a maximum amplitude of 294 ± 79 pA (mean ± SEM, n=18). Taking into account the average membrane capacitance 39 ± 5 pF (mean ± SEM), the average current density was 8.7 ± 3.4 pA/pF (mean ± SEM, n=18). To investigate the time dependency of recovery from inactivation, we used a paired-pulse protocol (Fig. 1f). The current was recorded after a second depolarizing pulse given at increasing time intervals until it finally recovered to its full size (Fig. 1f). The second peak currents were subsequently normalized to the prepulse peak current and plotted against the time between the two voltage pulses (Fig. 1g). Our data was fitted with an exponential function and the best fit yielded to  = 47 ms (n=3).
The presence of Nav currents was confirmed using the universal extracellular Nav channel blocker tetrodotoxin (TTX). 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).
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). On the other hand, 2mM QX-314, an intracellular Nav channel blocker added to the internal solution of the patch pipette, removed the current rapidly after breaking into the whole-cell configuration (Fig. 2a). 5 Voltage-gated sodium channels localize to the apical side of the tight junctions in hESC-

derived as well as mouse RPE
Our patch-clamp data indicated that functional Nav channels are present in the hESC-derived RPE cells. We also wanted to verify the expression in situ and study the precise cellular localization of the channels by performing immunofluorescence studies: the cellular retinaldehyde binding protein (CRALBP), a marker for mature RPE cells 24,25 , was labelled together with the universal Nav channel marker. These hESC-derived RPE samples were then imaged with a laser scanning confocal microscope (LSCM) by acquiring 3D image stacks (Fig.   2b) and the data were denoised by deconvolution. This showed that Nav was present in fully differentiated RPE. Furthermore, the Nav label concentrated primarily on the cellular borders while the CRALBP label was more uniformly localized to the apical side of the hESC-derived RPE (Fig. 2b).
Since the Nav label seemed to localize to the cell-cell junctions, immunolabelling was next carried out with the tight junction marker ZO-1. The labels showed highly overlapping distributions, suggesting that Nav localizes mostly into the tight junctions (Fig. 2b). The expression of Nav channels in RPE has previously been thought to be induced in vitro by the cell culturing 18,19 , therefore we wanted to confirm the presence of the Nav channels in vivo by using freshly isolated and non-cultured mouse RPE (Fig. 2c). The same antibodies gave highly similar labeling distributions in mouse RPE as in hESC-derived RPE: the CRALBP label was cytoplasmic on the apical side of the cells while ZO-1 and Nav concentrated on the cell-cell junctions (Fig. 2c).
We investigated the mechanism underlying the disappearance of Nav currents from acutely isolated RPE cells (Fig. 1d) and discovered evidence for internalization following destruction of the tight junctions. In dissociated hESC-derived RPE cells seeded for 30 min on glass and immunolabeled with the universal Nav marker, CRALPB and ZO-1, the Nav label 6 was primarily concentrated in the narrow region separating the apical and basolateral sides of the cell. ZO-1 and Nav formed a clear ring-like structure between the apical and basal membranes following relaxation of junctional tension, suggesting that cytoskeletal contractility is involved in the internalization process (Fig. 2d). Due to the junctional disruption, Navs might not be accessible to pass ionic currents.
RPE cells express voltage-gated sodium channel subtypes Nav1.4, Nav1.5, Nav1.6 and Nav1.8 Our previous experiments indicated that functional voltage-gated sodium channels are present in RPE. Since ten different Nav channel subtypes have been identified, Nav1.1 -Nav1.9 and Nax, 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 shown expression of several Nav channels in donated human RPE, specifically Nav subtypes 1.2-1.6 and Nav1.9 26,27 . We performed immunolabeling experiments with mouse and hESC-derived RPE using specific antibodies against channel subtypes Nav1.1 -Nav1.9 (Fig. 3a, 3b, Fig. S1). Confocal microscopy showed that Nav1.4 localizes as beadson-a-string to the cell-cell junctions and strongly co-localizes with the gap junction marker connexin 43 (Cx43) (Fig. S3). Nav1.8, on the other hand, localized overall to the apical side of the RPE cells (Fig. 3a, 3b). These data indicated that especially the Nav1.4 and Nav1.8 channels which are usually expressed in skeletal muscle and dorsal root ganglia 28,29 , respectively, are also present in RPE cells. Moreover, subtypes Nav1.5 and Nav1.6, the predominant channels of cardiac muscle, and the adult central nervous system, respectively 30 , were also identified.
However, the expression of Nav1.5 was punctate and not detectable in all cells whereas Nav1.6 showed a more homogenous labeling pattern and uniform expression between the cells (Fig.   3a, 3b). Subtypes 1.1-1.3, 1.7 and 1.9 were not detected (Fig. S1). Additionally, we investigated 7 the changes in channel subtype localization patterns during maturation of hESC-derived RPE (Fig. S2). The immunolabeling experiments indicated that the subtypes Nav1.4, Nav1.5 and Nav1.8 changed from homogeneous cellular distribution to more specific localization either to cell-cell junctions (Nav1.4) or to the apical side of the epithelium (Nav1.5 and Nav1.8) during the first 9 days of maturation (Fig. S2).
To verify further the functional expression of the most prominent channel subtypes by electrophysiology, we repeated our patch clamp recordings using highly selective blockers for the channels Nav1.4 and Nav1.8 and a less specific blocker for Nav1.6. At the time of our study, to our knowledge there was no selective blocker available for Nav1.5 without potential crossreactivity. 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 (Nav1.6 blocker), 1 µM A-803467 (Nav1.8 blocker) and 600 nM µ-conotoxin GIIB (Nav1.4 blocker) (Fig. 3c) and the effect of inhibition was more potent with each 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 10µM TTX (Fig. 3d).

Inhibition of voltage-gated sodium channels significantly reduces the number of POS particles in hESC-derived RPE
Our previous experiments showed that Nav1.4, Nav1.5, Nav1.6 and Nav1.8 are expressed both in 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 Nav channels, as it requires rapid activation and high synchronization 31 . Thus, next we wanted to investigate the potential importance of Nav channels for POS phagocytosis in vitro. Our phagocytosis assays and subsequent labeling with opsin and phalloidin indicated 8 that the hESC-derived RPE cells have the ability to bind and take up purified POS particles isolated from porcine eyes (Fig. 4a). To study the involvement of Nav channels in this pathway, the assay was repeated in the presence of the highly selective Nav 1.4 and 1.8 blockers and TTX. Interestingly, the immunolabeling with opsin and ZO-1, a tight junction protein and cell border marker, showed a reduction in the total number of bound and internalized POS particles ( Fig. 4b). To quantify the effect, large fields were imaged from each blocker condition and the number of POS particles was compared with controls. The results showed that the selective blockers caused a 13% (Nav 1.4 blocker, n= 10), 27% (Nav 1.8 blocker, n=14) or 34% (Nav 1.4 and Nav 1.8 blockers together with TTX, n=18) reduction in the total number of POS particles labeled with opsin ( Fig. 4c).

Voltage-gated sodium channels Nav1.4 and Nav1.8 are involved in POS phagocytosis in mouse RPE in vivo
The experiments with pharmacological inhibitors of Nav channel subtypes Nav1.4, and Nav1.8 indicated that these channels are involved in POS particle uptake. To investigate further their role in the phagocytosis process in vivo, we performed immunolabeling experiments with mouse eyes that had been prepared either at light onset near the diurnal peak of phagocytosis or 10 h after light onset. The role of the channels in POS uptake was studied by comparing the immunolabeling of Nav1.4, Nav1.8 and opsin. At light onset both channels localized to the bound POS particles (Fig. 5a). Since previous experiments had indicated that Nav1.4 localized strongly to gap junctions (Fig. S3c), we labeled it together with Cx43 as well as opsin. Colocalization of these three markers was most evident at light onset (Fig. 5b), yet we observed a reduction in Nav1.4 and opsin co-localization at 2 h time point (Fig. 5c). Interestingly, after 10 h, when phagocytosis process is over 32 , both Nav1.4 and Cx43 stains were again detected at cell-cell junctions (Fig. 5d). 9 We next wanted to investigate further the role of Nav1.8 in the POS phagocytosis. For this purpose, we combined the labeling of Nav1.8 with Rab7, a GTPase which is generally recognized in the regulation of vesicle docking and fusion, specifically with late endosomes and lysosomes 33,34 . In macrophages Nav1.5 has previously been demonstrated to preferentially localize with Rab7 rather than the endosomal markers early endosome antigen 1 (EEA1) or lysosomal-associated membrane protein 1 (LAMP-1) 35 . To date, controversy persists as to whether Rab7 is additionally involved in the earlier steps of endosomal trafficking 36 . We investigated how Nav1.8 labeling correlated with Rab7 and EEA1 as well as the phagocytosis markers MerTK and integrin αvβ5 (Fig. S4, Fig. 6). Nav1.8 was found to partially co-localize with EEA1 and integrin αvβ5, however the distribution similarity was most evident with Rab7 ( Fig. S4, Fig 6). Moreover, both Nav1.8 and Rab7 also localized with opsin throughout the 2 h follow-up of phagocytosis (Fig. 6a, 6b, Fig. 5a). When the peak of phagocytosis was over (10 h time point), the overall labelling pattern of Nav1.8 appeared significantly more homogeneous (Fig. 6c).
The redistribution of Nav channels occurring during phagocytosis (Fig. 7a) was studied ex vivo with the channel blockers (Fig. 7b). 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 Nav1.4 as well as the combination of all Nav blockers significantly prevented the redistribution of Nav1.4 compared to the control (Fig. 7b). The inhibition effect was similarly observed in hESC-derived RPE in vitro when the POS challenge was carried out with blocker solutions (Fig. 7c). We did not observe significant differences in the labeling pattern of Nav1.8 or Rab7 after the blocker incubation. Taken together, these experiments demonstrate the participation of Nav channels in the phagocytotic processes of RPE cells in vitro and in vivo. 10

Discussion
Recent studies in diverse cell types have revolutionized our understanding of the roles that Nav channels have in cellular functions; no longer are these proteins considered important only in "classically" electrically excitable tissues. Here, we provide the first evidence that Nav channels are localized to the tight junctions between the epithelial cells (Fig. 2) and that their activity co-regulates phagocytosis in RPE. Our observations of Nav channels and Nav-mediated currents in intact RPE preparations (mature hESC-derived RPE monolayers and freshly isolated mouse RPE) demonstrate that previous observations of Nav-mediated currents in cultured RPE cells are not preparation-dependent artifacts 19 . Rather, the absence of Navmediated currents in acutely isolated RPE cells (Fig. 1d) likely results from the destruction of tight junction complexes during dissociation (Fig. 2d). Internalization of Nav 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 . The observation of Nav currents in recordings from hESC-derived RPE monolayers, we believe, is strong evidence that cells in RPE with intact tight junctions usually express functional Navs in their plasma membranes.
Unfortunately, the denser and longer apical microvilli in mouse RPE prevented us from performing recordings from intact mouse tissue.
The properties of Nav-mediated currents in hESC-derived RPE cells are consistent with observations in other non-neuronal cells. Specifically, the relative insensitivity of the channels to TTX (block by 10 µM TTX) is similar to that of Navs in macrophages 37 and in cultured cancer cells 38 . Additionally, it is consistent with our observation that RPE cells were labeled strongly by an anti-Nav1.8 antibody; Nav1.8 is the least sensitive of Nav channels to TTX 39 .
Our pharmacological analysis of Nav-mediated currents using Nav subtype-specific blockers and immunohistochemical analysis (Fig. 3) provided additional evidence for Nav1.8 expression along with lesser expression of Nav1.5. 11 A role for Nav channels in POS phagocytosis is consistent with previous studies showing that Nav1.5 and Nav1.6 channels enhanced mycobacteria processing in human macrophages and that TTX attenuated phagocytosis in microglia 21,37,40,41 . The mechanisms by which voltage-gated channels with rapid activation and inactivation kinetics regulate phagocytosis are unclear. Although inhibition of Nav channels did not abolish phagocytosis here-the observed 34% attenuation (Fig. 4c) was less dramatic than the previously reported effect of TTX in microglia 42 -Nav channel activity clearly modulated phagocytosis as indicated by observations of bound POS particles. Furthermore, the channels could be involved directly in the circadian control of the pathway, as has been recently shown for other ion channels 10 .
Our co-labeling studies give a strong indication that Nav1.4 consistently interacts with Cx43 (Fig. 5). Similarly, an interaction between Nav channels and Cx43 has been suggested for Nav1.5 in cardiac myocytes (for review, see 43 ). Both Nav1.4 and Cx43 have previously been implicated in phagocytosis in macrophages, although the findings have been controversial 44,45 .
We observed a redistribution of Nav1.4 and Cx43 from tight junctions during phagocytosis and for Cx43 the localization with opsin was still evident 2 h after light onset, near the peak expression level of a phagocyte cell surface receptor tyrosine-phosphorylated MerTK 32 . The involvement of Nav1.4 was supported by the fact that following Nav blocker incubation, we observed a decrease in its translocation (Fig. 7) with a concurrent reduction in the number of POS particles (Fig. 4c). As with Nav1.4 and Cx43, Nav1.8 and Rab7 showed consistent colocalization (Fig. 6), and their cellular localization varied throughout phagocytosis during which both also localized with opsin. It is possible that the role of Nav1.8 and Rab7 is related to the regulation of endosomal processing or acidification of the endosomal pH, as has been suggested for Nav1.5 of macrophages in multiple sclerosis lesions 35 . 12 The fact that RPE expresses such a versatile array of Nav channels suggests that besides phagocytosis, these channels also have other roles in the physiology of the RPE. Epithelial cells, including RPE, show strong calcium waves in response to mechanical stimulation 46-48 , and it is possible that Nav channels serve to amplify or accelerate voltage changes. Moreover, it is well established that the Ca 2+ binding protein calmodulin (Cam) interacts directly with the C-terminal domain of Nav 49 , and it was recently shown that the Ca 2+ -free form of Cam, ApoCam, enhances the Nav channel opening by several-fold 50 . Thus, Na + -and Ca 2+ -dependent signaling pathways can interact in epithelia as has been reported in the case of astrocytes 51 In this study, we have demonstrated that Nav channels can be identified in both hESCderived and mature mouse RPE. This strongly indicates that expression of these channels is not associated with neuroepithelial differentiation nor due to specific culturing conditions. The channels are in fact implicated in POS phagocytosis, one of the key functions of RPE, which is essential for the survival of the retina. Further studies are required to elucidate the additional roles these channels have in the tissue, but it is evident that they are a vital part of the physiology of RPE and may thus contribute to the development of retinal diseases.
Briefly, the hESC-derived RPE were spontaneously differentiated in floating cell clusters. The

Phagocytosis assay for hESC-derived and mouse RPE
The porcine POS particles were isolated and purified as previously described 52,53 . 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 x g for 10 min, +4ºC and stored in 73 mM sucrose phosphate buffer at -80°C.
The purified POS particles were fed to the hESC-derived cells in a KO-DMEM medium supplemented with 10% fetal bovine serum (FBS) and incubated for 2 h at 37° in 5 % CO 2 . In the blockers experiments, selective blockers for NaV1.4, NaV1.6 and NaV1.8 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 15 min, 2 h and 10 h after it. The mice were reared in normal 12-hour 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° with the retina left intact.

Quantification of bound POS particles in hESC-derived RPE
To detect and quantify bound POS particles, large random fields were imaged with Zeiss        The role of Nav1.8 and its interaction with endosomal marker Rab7 was studied by dissecting mouse eyes at various time points during the circadian cycle. Filamentous actin was stained with phalloidin (gray in the merged image) to highlight epithelial cell-cell junctions and lower panels show a high contrast blow up of selected regions. Laser scanning confocal microscopy Z-maximum intensity projections of mouse RPE prepared (a) at 15 min after the light onset 28 indicated clustering of Rab7 (red) and Nav1.8 (green) together with opsin (blue) labeled POS particles. (b) 2 h after the light onset, Nav1.8 (green) and Rab7 (red) still co-localize with POS particles but start to show more homogeneous localization pattern. (c) 10 h after light onset a diffuse labeling was shown for both and Rab7 (red) and Nav 1.8 (green). Scale bars 10 µm.