Main

The epithelial sodium channel (ENaC) is a constitutively open sodium-selective ion channel that lacks voltage sensitivity and mediates sodium absorption across various epithelia1. Classical ENaCs are heterotrimers composed of three subunits, α, β and γ, which assemble as a sodium-selective ion channel in the plasma membrane (Fig. 1)2,3. In renal epithelia, ENaCs mediate Na+ reabsorption by the kidney and are involved in maintaining extracellular volume and blood pressure. Gain-of-function mutations cause a hereditary form of hypertension (Liddle syndrome), whereas loss-of-function mutations give rise to the salt-wasting syndrome pseudohypoaldosteronism type 14,5. ENaC-mediated sodium absorption by the pulmonary epithelium drives lung liquid clearance and is thus a critical regulator of pulmonary water content. Increased ENaC activity in the airways is the cause of cystic-fibrosis-like lung disease, whereas decreased ENaC activity in the distal lung leads to impaired alveolar fluid clearance and the development of pulmonary oedema6,7,8,9. As such, ENaCs are a primary target for drugs that affect electrolyte and water homeostasis, such as diuretics or oedema medications10,11,12,13.

Figure 1: ENaCs and their blockers.
figure 1

a, Schematic drawing of heterotrimeric αβγENaC and δβγENaC. Each subunit contains two transmembrane helices TM1 and TM2, relatively short C and N termini at the intracellular (in) site and a large extracellular (ex) loop. Under physiological conditions, ENaCs allow for a constant influx of sodium into the cell (indicated by blue arrows). b, Chemical structure of amiloride and its more potent derivatives phenamil, benzamil and compound 1. c, Structures of photoswitchable amiloride derivatives that contain an azobenzene functional group. d, Photoisomerization of PA1. Illumination with 400 nm induces isomerization to cis-PA1, which is the thermodynamically less-stable form. The cis isomer can relax back to the trans state thermally, or, more rapidly, by illumination using 500 nm radiation.

Full size image

An additional ENaC-subunit, δ, was identified in 199514. There is only scarce knowledge about its physiological function; however, it can assemble with the β and γ subunits to form channels with a high open probability14,15,16. Expression of δENaC (and other ENaCs) has been confirmed in various non-epithelial tissues, most prominently the central nervous system of primates17,18,19,20. It has been speculated that δ-containing ENaCs play a role in communicating ischaemic and hypoxic signals in neurons, as well as in adjusting nervous excitability21,22. Complicating the analysis, rodents do not appear to have a gene for the δ subunit. Furthermore, no pharmacological tools are available for functional differentiation between α- and δ-containing ENaCs. Thus, in contrast to αβγENaC, little is known about the physiological function of δβγENaC and, in particular, the function of δENaC in the brain22.

ENaCs belong to a large group of ion channels, the ENaC/degenerin (ENaC/deg) protein family, which also includes neuronal acid-sensing ion channels23. Recently, the latter were elucidated with X-ray crystallography, which confirmed a trimeric architecture that probably applies to all channels of the family (Fig. 1)24. Every individual ENaC subunit comprises only two transmembrane helices (TM1 and TM2) with relatively short intracellular C and N termini and a large, cysteine-rich extracellular domain.

Although ENaC/deg channels respond to a variety of input signals, none are inherently sensitive towards light. In recent years, our group succeeded in developing a range of synthetic photoswitches that can confer light sensitivity to native or slightly modified receptor proteins. These can either function as freely diffusible photochromic ligands (PCLs) or as photoswitchable tethered ligands, which are covalently attached to the target protein25. Using either strategy, effectively we were able to turn a variety of ion channels and G-protein-coupled receptors into photoreceptors26,27,28,29,30.

We now show that an entirely new class of transmembrane proteins, namely trimeric ENaC/deg channels, are amenable to the PCL approach. To this end we have designed, synthesized and characterized a small series of photoswitchable azobenzene derivatives of amiloride and identified one candidate that can be used primarily to control δβγENaC activity with light. Our photochromic ligands provide insights into the structure–activity relationships of ENaC blockers and could be used to discriminate functionally between isoforms.

Results and discussion

Design and synthesis of photoamilorides

Amiloride was introduced as a potassium-sparing diuretic in the 1960s without particular knowledge of its molecular target31. Subsequently, ENaC in kidney epithelia was established as the ion channel primarily responsible for the potassium-sparing diuretic effect3. To date, countless amiloride analogues have been synthesized and tested, which has provided insights into its structure–activity relationships32,33. Most importantly, it has been found that lipophilic substitution at the terminal guanidine nitrogen boosts both potency and selectivity for ENaC. Indeed, the potency gained by an aromatic substituent increases with linker length (Fig. 1b)34. Accordingly, we designed photochromic amiloride derivatives that have an azobenzene unit attached directly to the guanidine, which gave photoamiloride-1 (PA1) and photoamiloride-2 (PA2), both of which resemble the drug phenamil (Fig. 1). Phenamil is a long-lasting blocker of ENaCs and has been used to determine channel density35. In addition, we designed the compounds photoamiloride-3 (PA3) and photoamiloride-4 (PA4), which bear an azobenzene moiety linked through an ethylene spacer and resemble the known amiloride analogue 136. In PA2 and PA4, the azobenzenes are substituted further in the 4′-position with a diethylamino substituent, which is known to red shift the action spectra of azobenzenes37. The photoswitching of PA1 between its trans and cis configuration is shown in Fig. 1d.

The synthesis of the shorter photoamilorides PA1 and PA2 started with aminoazobenzenes 2 and 3, respectively, which were converted into aryl guanidines using cyanamide (Fig. 2a and Supplementary Information). The resulting guanidine hydrochlorides were deprotonated and condensed with an active ester of pyrazinic acid 4, which yielded PA1 and PA2, respectively.

Figure 2: PA synthesis and characterization.
figure 2

a, Synthesis of PA1 and PA2 starts with 4-aminoazobenzenes 2 and 3, respectively. Reaction with cyanamide installs guanidine functional groups, which can be coupled with an active ester of 4 to yield PA1 and PA2. b, Synthesis of PA3 and PA4 starts from the protected phenethylamine 5, which can be either condensed with nitrosobenzene or oxidized to a nitrosobenzene and condensed with p-(diethylamino)aniline. After acidic deprotection, primary amine building blocks 6 and 7 are functionalized to guanidines, which can react with ester 9 to yield PA3 and PA4, respectively. c, X-ray structure of PA1 indicates an extended hydrogen-bond network (dotted lines). d, Optimal wavelengths for photoswitching were determined using UV/vis spectroscopy. Generally, the more conjugated candidates PA1 and PA2 are more red-shifted than their corresponding homologues PA3 and PA4. Substitution with the electron-donating diethylaniline functionality leads to a red shift compared to the unsubstituted candidates (PA2 and PA4 versus PA1 and PA3). e, IC50 values of PAs in their dark-adapted trans form. PA4 is the most-potent and PA2 is the least-potent ENaC blocker. All PA molecules are more potent on the αβγ isoform. PA2 discriminates the most between αβγENaC and δβγENaC.

Full size image

To generate the longer photoamilorides, aniline 5 was condensed with nitrosobenzene and subsequently deprotected to yield 6 (Fig. 2b and Supplementary Information). Alternatively, oxidation of 5 and condensation with p-(diethylamino)aniline, followed by deprotection, gave azobenzene 7. With these two building blocks in hand, the guanidine moiety was installed using carboxamidine 8. The resulting guanidinium salts were deprotonated via a basic work-up and added directly to an ethanolic solution of ester 9. Heating to reflux then yielded PA3 and PA4, respectively. Interestingly, attempts to furnish PA3 or PA4 under the same coupling conditions as PA1 and PA2 (and vice versa) failed. Similarly, attempts to condense 4-aminoazobenzenes 2 and 3 with carboxamidine 8 did not yield the corresponding guanidines.

A single-crystal X-ray structure of PA1 is shown in Fig. 2c. It provides the first structural data of an amiloride derivative substituted at the terminal guanidine nitrogen and reveals an extended hydrogen-bonding network that involves an amine on the pyrazine, the adjacent carbonyl and the guanidine (dotted lines).

Photophysical characterization of photoamilorides

To determine absorption maxima and the optimal wavelengths for photoswitching, solutions of PA1-4 (50 µM, dissolved in dimethylsulfoxide (DMSO) were irradiated at different wavelengths using a monochromator (Till Photonics Polychrome 5000, Supplementary Fig. 1). DMSO was chosen as the solvent to slow down the thermal relaxation to resolve the absorption spectra of cis and trans isomers for red-shifted candidates38. Thus, it was possible to identify wavelengths that led to the highest cis and trans content for each molecule (Fig. 2d).

Electrophysiological characterization of photoamilorides in the dark

To identify the amiloride derivative that would be the best photochromic ENaC blocker, initially we performed microelectrode recordings on Xenopus oocytes that express human αβγENaC and δβγENaCs for all the molecules (Supplementary Fig. 2). In the presence of a dim room light to ensure that PAs were in their trans form, a series of increasing ligand concentrations was applied, followed by full block using a 100 µM amiloride solution. This enabled both the determination of half-maximum inhibitory concentration (IC50) values and a relative block compared to the full blocker amiloride (Fig. 2e and Supplementary Fig. 2). It was found that all the photochromic ligands synthesized were ENaC blockers. In general, they have a higher affinity for α-containing than δ-containing ENaCs. PA4 is the most potent molecule with IC50 values of 4 and 14 nM for αβγENaC and δβγENaC, respectively. The weakest blocker is PA2 with IC50 values of 346 and 8,000 nM for αβγENaC and δβγENaC, respectively. For comparison, amiloride has IC50 values of 100 nM and 2,600 nM on αβγENaC and δβγENaC channels, respectively, in a similar experimental setting.22

The structure–activity relationships of our small molecular series provided additional interesting findings (Fig. 2e). Matching the pharmacology of previously known ENaC blockers, the addition of a C-2 linker markedly increased affinity, as PA3 and PA4 are substantially more potent than their shorter analogues. The addition of a substituent in the 4′-position in PA4 further increased potency compared to the unsubstituted PA3. However, this trend is reversed when there is no linker between the guanidine and the azobenzene moiety.

Photocurrents in Xenopus oocytes

To investigate which compound elicits the most-pronounced photoeffect, we performed electrophysiological recordings on Xenopus oocytes that expressed either αβγENaC or δβγENaC in combination with monochromatic illumination. PA1–PA4 were applied at varying concentrations and the illumination wavelengths were switched in intervals of one minute as the perfusion was stopped (Supplementary Fig. 3). Among our four amiloride derivatives, PA1 emerged as the most-efficient photochromic blocker, especially when applied to the δβγ isoform. Our further investigations therefore focused on this compound. As shown in Fig. 3, illumination of PA1 with 400 nm radiation immediately triggered a strong reduction in current. This effect could be reversed by switching back to 500 nm (Fig. 3 and Supplementary Fig. 4a). Thus, PA1 functions as a cis blocker of δβγENaC. We found that the overall switching amplitude evoked by PA1 is consistent between 1 and 10 µM and decided to settle on 10 µM as a working concentration, which provides maximal block in the trans state, close to a full amiloride block (Supplementary Fig. 4b). Control experiments using oocytes that expressed ENaC channels showed no innate sensitivity to light (Supplementary Fig. 5). Similarly, ENaC-expressing oocytes blocked by a 100 µM solution of amiloride did not respond to illumination39.

Figure 3: PA1 photoswitching in Xenopus oocytes.
figure 3

a, Reversible block of δβγENaC currents evoked by photoisomerization of PA1 (10 µM) in 10 s and 5 s intervals. b, Action spectrum of PA1. Wavelengths were switched between 500 nm and 400 + x nm (x = 0, 20, 40, 60). The largest photoswitchable currents are observed when 500 and 400 nm are used as illumination wavelengths for cis and trans isomerization. c, Photoswitching of PA1 tested on an amiloride-sensitive current (IA) of αβγENaC-expressing (left) and δβγENaC-expressing (right) Xenopus oocytes by two-electrode voltage clamp recordings. Perfusion was stopped and illumination wavelengths were switched between 500 and 400 nm in one minute intervals. Current values were taken at the end of each interval and averaged (n = 12 and 10, respectively). Error bars indicate s.e.m., and significance refers to switching from 400 nm to 500 nm (*p < 0.05). There is a reversible photoswitching effect on both αβγENaC and δβγENaC. However, the amplitude of PA1-evoked photoswitching is higher on δβγENaC.

Full size image

Figure 3a shows the effect of PA1 on a δβγENaC-mediated current. Wavelengths, and consequentially currents, were switched alternately in intervals of ten and five seconds. With light-mediated activation, perfusion-induced mechanical effects can be avoided, as recently demonstrated for ENaCs as well as for other members of the ENaC/deg family40,41,42,43.

The photostationary state of azobenzenes, that is their cis/trans ratio, is a function of the illumination wavelength, which allows for controlling their effective concentrations with light. As a consequence, the degree to which δβγENaC channels are blocked by PA1 also depends on the wavelength applied (Fig. 3b). Switching in 20 nm steps between a given wavelength and 500 nm results in a graded effect on the current. In line with our previous ultraviolet/visible (UV/vis) experiments, 400 nm radiation emerged as the most effective blocking wavelength. As shown in Fig. 3c, photoswitching mediated by 10 µM PA1 could be detected on both αβγENaC- and δβγENaC-expressing oocytes. However, switching events were generally more robust and amplitudes were higher on δβγENaC (also see Supplementary Fig. 6).

PA1 photoswitching in human embryonic kidney cells

We next turned to human embryonic kidney (HEK293t) cells as a mammalian expression system. Given that these cells are smaller and transparent, we expected faster and more pronounced effects compared to those of the pigmented oocytes, for which only a fraction of the membrane surface can be illuminated. As depicted in Fig. 4, PA1 is an excellent photoswitchable blocker of δβγENaC-expressing HEK cells. In line with the data from Xenopus oocytes, 400 nm radiation translates into a rapid block of ENaC currents, which can be released with 500 nm radiation (Fig. 4a). In the current-clamp mode, the action of PA1 on δβγENaC translates into a pronounced and fast light-dependent change in membrane potential. Photoeffects are generally higher on the δβγ isoform over the tested range of 10 nM to 25 µM PA1 and the difference between αβγENaC and δβγENaC increases with higher concentrations (Supplementary Fig. 6c,d). A comparison of photoswitching using HEK cells expressing either δβγENaC or αβγENaC and 10 µM PA1 revealed that the average amplitude of the photocurrents is ~180 pA on the former and ~10 pA on the latter isoform. Similarly, the average change in membrane potential was ~57 mV and ~4 mV, respectively (Fig. 4b; overall expression was identical (Supplementary Fig. 8)). Thus, the general trends and effects that were found in Xenopus oocytes could also be observed in HEK cells, but much more pronounced and with a faster onset (Supplementary Fig. 7). We also found that cis-PA1 thermally relaxes back to trans-PA1 on a fast time scale (Fig. 4c). Thus, our novel optical tool can be operated by turning on and off a simple light source. As in the case of oocytes, intermediate wavelengths between 400 and 500 nm translate into graded effects (Fig. 4d). We were thus able to induce current and voltage steps in δβγENaC-expressing HEK cells by switching between 400, 433, 466 and 500 nm.

Figure 4: Switching δβγENaC in HEK cells using 10 μM PA1.
figure 4

a, By switching between 400 and 500 nm both large currents (upwards deflection = inhibition) and potentials (downwards deflection = hyperpolarization) can be controlled by light. b, The average current amplitude of photoswitching (ΔIP) is 182 pA for δβγENaC and 9 pA for αβγENaC. The average amplitude of the light-mediated membrane potential (ΔVP) is 57 mV for δβγENaC and 4 mV for αβγENaC (currents and potentials were recorded from the same cell, n = 10 biological repeats from three independent transfections and holding potential was –60 mV). c, Comparison between light-induced and thermal relaxation. d, Application of light with wavelengths between 400 nm and 500 nm reveals the physiological action spectrum of PA1 on δβγENaC-expressing HEK cells. Graded effects can be installed for both current (top) and membrane potential (bottom). e, Light-induced hyperpolarization and depolarization in the current-clamp mode. When the starting potential is adjusted to –70 mV under green light (left), photoswitching leads to hyperpolarization. When the starting potential is adjusted to –70 mV under blue light (right), photoswitching leads to depolarization. f, PA1 can be loaded into HEK cells by a short incubation. After one minute incubation at room temperature (r.t.) in 10 µM PA1 (yellow Petri dish) and subsequent wash (blue Petri dish), cells are ready for photoexperiments using the patch–clamp technique.

Full size image

To determine further the potential of PA1 as a tool for adjusting membrane potential, we set the starting potential in HEK cells in the current–clamp mode at –70 mV, initially using either blue or green illumination (Fig. 4e). Starting from these conditions, PA1 can be used not only to hyperpolarize (switching from 500 to 400 nm), but also to depolarize (switching from 400 to 500 nm) a cell. As such, the δβγENaC–PA1 system could be used as an optogenetic tool.

Pharmacokinetics of PA1 in HEK cells

In the course of our experiments using PA1 on HEK cells expressing δβγENaC, we found that photoswitching is persistent and channel block remains after washout of PA1 from the external recoding solution (Supplementary Fig. 9a,b). This interesting observation hints at a possible cell uptake or a reservoir formation in the cell membrane. Interestingly, Xenopus oocytes did not take up PA1 as readily as HEK cells, which might be another reason for the difference in overall photoeffects (Supplementary Fig. 9c). To investigate this finding further, we preincubated coverslips with HEK cells expressing δβγENaC for one minute at room temperature in 10 µM PA1. Coverslips were then washed three times and investigated using whole-cell patch–clamp experiments. We obtained a fully functional photoswitchable system with strong, persistent ENaC photocurrents (Fig. 4f). Figure 4f further demonstrates the rapid robust and reversible action of PA1 on δβγENaC. Intriguingly, cells prepared in such a manner could be blocked and unblocked by amiloride without affecting photoswitching (Supplementary Fig. 10). However, photoswitching was reduced when phenamil was used as a blocker with similar lipophilicity, which is also taken up by cells (Supplementary Fig. 10c). In addition, we loaded PA1 through the patch pipette (Supplementary Fig. 11). Photoswitching was possible instantaneously and a reversible amiloride block was also observed. Additional PA1 delivered by perfusion did not enhance the photoeffect. These observations suggest that PA1 finds its binding site from a membrane reservoir and has similar pharmacokinetics to long-lasting, use-dependent open channel blockers44.

Use of PA1 in a lung-tissue model

With the ability of PA1 to switch ENaC currents and to differentiate between αβγENaC and δβγENaC in HEK cells, we next wanted to employ PA1 in a mammalian tissue model that endogenously expresses ENaC subunits. Human H441 bronchiolar epithelial cells were the model of choice because these cells (1) express all four ENaC subunits, including δENaC16, and (2) form a polarized epithelium, which allows the electrophysiological assessment of sodium transport across an intact cell monolayer. Furthermore, despite the molecular evidence for δENaC expression in pulmonary epithelia, a functional contribution of this isoform to transepithelial sodium transport has not been demonstrated22.

To address the functional role of δβγENaC in these cells, we measured the transepithelial potential of H441 monolayers, which were cultured on permeable membrane supports (Fig. 5a). First, we determined the magnitude of the overall amiloride-sensitive transepithelial potential (VE) (on average ~4 mV, Fig. 5c). We then exchanged the apical bath solution for 10 µM PA1 and switched between 400 and 500 nm, and 400 nm and 694 nm for three cycles, respectively (Fig. 5b). The deep-red wavelength functionally corresponds to darkness because our azobenzenes do not absorb in this region. Results were identical for the different wavelength protocols and the overall magnitude was a photosensitive transepithelial potential VE of ~0.4 mV (Fig. 5b). This finding hints towards a small amount of functional δβγENaC. To investigate whether the photoeffect comes from δβγENaC or is a small effect of αβγENaC, we addressed the lithium conductivity of each isoform. The ratio between lithium and sodium conductivity is 0.64 for δβγENaC and 2.0 for αβγENaC22. Thus, we expected decreased photoeffects for the δ-containing isoform and increased effects for the α-containing isoform when sodium is replaced by lithium. We confirmed this hypothesis using Xenopus oocytes (Fig. 5d), which had shown photoeffects for either isoform before. Similarly, δβγENaC photocurrents were significantly reduced in HEK cells by extracellular lithium (Supplementary Fig. 12). When sodium was substituted by lithium at the apical side of the H441 monolayer, PA1-mediated photoeffects decreased from 0.4 to 0.15 mV for both wavelength protocols (Fig. 5c). Thus, the photosensitive fraction of VE results from δβγENaC rather than from αβγENaC. These data strongly indicate a functional expression of δβγENaC in H441 cells and underline the ability of PA1 for functional differentiation between αβγENaC and δβγENaC.

Figure 5: Photocontrol of transepithelial potential in H441 monolayers.
figure 5

a, Experimental configuration with an H441 monolayer (ML) on a filter inlet. The apical compartment contained 10 µM PA1 in sodium- or lithium-based buffer and was illuminated with a monochromator. The transepithelial potential VE between the apical and basolateral compartment was measured by a voltmeter (apical electrode is the reference). b, VE can be switched by toggling between 500 nm and 400 nm radiation (top), as well as between 694 nm and 400 nm (bottom). The 400 nm radiation was applied for ten seconds, the longer wavelengths for 20 seconds and the voltage was recorded at the end of each illumination period. Photoeffects are robust and reversible for both protocols. c, Comparison of the entire amiloride-sensitive VE and PA1-mediated photosensitive VE with apical sodium or lithium using both 400/500 nm and 400 nm/dark switching (n = 7). d, Comparison of the effect of extracellular lithium on αβγENaC and δβγENaC photocurrents in Xenopus oocytes (n ≥ 7). Error bars indicate s.e.m., *p-value < 0.05.

Full size image

Conclusion

In summary, the photoswitchable ENaC blocker PA1 has been designed, synthesized and functionally characterized using electrophysiology with Xenopus oocytes, HEK293t cells and H441 cell monolayers. Our results provide new insights into the structure–activity relationships of the amiloride class of diuretics and outline a strategy to address selectively the αβγ and δβγ isoforms of ENaC channels. PA1 can be used to control δβγENaCs optically by switching between blue and green light, or by turning on and off blue light. This therefore allows ENaC to be investigated without mechanical stress. We showed that PA1 is long lasting and can be applied by perfusion, cell incubation or loading through the patch pipette. We used PA1 to address the significance of different ENaC isoforms in H441 monolayers and identified functional δβγENaC. The combination of photoswitchable PA1 and the voltage-insensitive δβγENaC is a powerful tool to control precisely the membrane potential with light, allowing for both hyper- and depolarization. With its unique functional profile, PA1 could enable new insights into the δβγENaC function, for instance in the primate brain.

Methods

The Supplementary Information gives details of chemical syntheses and Xenopus oocytes cultures.

Cell culture and electrophysiology

HEK293t cells were cultured in DMEM buffer with 10% fetal bovine serum in a 10% CO2 37 °C incubator. At 90–95% confluency, cells were split and used for transient transfection using the jetPRIME polyplus transfection reagent in accordance with the supplier's protocol. Poly-L-lysine-coated acid-etched glass coverslips were placed in a 24-well plate and consecutively treated with the DNA mix (150 ng of each ENaC subunit and 50 ng yellow fluorescent protein (YFP)) and 30,000 to 50,000 cells in growth medium. After 2–4 hours, the supernatant was removed and growth medium supplemented with 10 µM amiloride was added. Cells were used for electrophysiological recordings 12–48 hours post-transfection. Only YFP-positive cells with an amiloride-sensitive current were used for the experiments. For each repeat, a new cover slip was used.

Whole-cell patch–clamp experiments were performed using a standard electrophysiology set-up equipped with a HEKA Patch Clamp EPC10 USB amplifier and PATCHMASTER software. Micropipettes were generated from Science Products GB200-F-8P with filament pipettes using a vertical puller. Resistance varied between 5 and 8 MΩ. The bath solution contained (in mM) 140 NaCl (or LiCl), 3 KCl, 2 CaCl2, 1 MgCl2, 10 D-glucose, 20 HEPES (NaOH to pH 7.4). The pipette solution contained (in mM): 90 K gluconate, 10 NaCl, 10 KCl, 1 MgCl2, 10 EGTA, 60 HEPES (KOH to pH 7.3). Photoswitchable ligands and reference agonist were dissolved in bath solution from a 1,000× DMSO stock.

H441 cells were purchased from American Type Culture Collection in the 65th passage and cultured under liquid/air conditions in the presence of 200 nM dexamethasone, as previously described45. For the experiments, H441-containing cell culture inserts were placed in six-well plates containing H441 saline (2 ml in the basolateral compartment and 500 µl in the apical compartment). H441 saline contained (in mM): 130 NaCl (or LiCl), 2.7 KCl, 1.5 KH2PO4, 1 CaCl2, 0.5 MgCl2, 10 HEPES, 10 glucose (37 °C, pH 7.35 (TRIZMA)). VE was measured with an epithelial voltohmmeter (EVOM, World Precision Instruments). The shorter electrode was placed in the apical compartment and served as reference. When VE was stable, amiloride was added to the apical compartment to a final concentration of 100 µM to investigate the amiloride-sensitive fraction of VE. Monolayers were washed three times with H441 saline and afterwards 10 µM PA1 was added to the apical compartment. When VE stabilized again, the EVOM was zeroed (for better comparability between cell preparations) and monolayers were exposed to different wavelengths. Afterwards, cells were rinsed with lithium-containing saline and the same measurement protocol was employed. Experiments were performed at 37 °C on seven H441 monolayers from two independent cell preparations.

Irradiation was performed on a TILL Photonics Polychrome 5000 monochromator. In patch–clamp experiments, the light beam was guided through the microscope objective and operated by the amplifier and PATCHMASTER software. In all other experiments, a light fibre was positioned in proximity to the object of interest and the monochromator was operated using the POLYCON 3.1 software.