We BLAST-searched the genome of A. pullulans EXF-150 with the amino acid sequence of different fungal rhodopsins including CarO, LR, and nop-1. The amino acid sequences from BR (gb| AAA72504), CarO (gb| CAD97459), LR (gb| AAG01180), ApOps1 (gb|KEQ89910.1), ApOps2 (gb|KEQ89333.1), and ApOps3 (gb|KEQ87154.1) were aligned using PSI/TM-Coffee alignment (Di Tommaso et al., 2011). ApOps1, ApOps2, and ApOps3 were modeled with the Swissmodel (Biasini et al., 2014) based on the crystal structure of Coccomyxa subellipsoidea rhodopsin CsR (6gyh.1.A; ApOps1; Fudim et al., 2019) and LR (7bmh.1.A ApOp2, ApOps3; Zabelskii et al., 2021).

A. pullulans EXF-150 cDNA, kindly provided by Cene Gostinčar, was synthesized using SuperScript III first-strand cDNA synthesis kits (Invitrogen) and random hexamer primers (Promega) according to the manufacturer instructions. Coding sequences of ApOps1 and ApOps2 were amplified from cDNA by PCR. ApOps3 was synthesized by ProteoGenix (France). ApOps1 and ApOps2 were cloned into pcDNA5/FRT/TO-CarO::eYFP (García-Martínez et al., 2015) via BamH1 restriction sites as described before (Panzer et al., 2019).

The PCR-amplified DNA fragments of ApOps1-3 and CsR [Coccomyxa subellipsoidea rhodopsin; the template DNA fragment for CsR was synthesized by GeneArt Strings DNA Fragments (LifeTechnologies, Thermo Fisher Scientific)] were inserted via BglII and XhoI restriction sites into the pGEMHE vectors already containing the cleavable N-terminal signal peptide Lucy-Rho (LR), eYFP, the plasma membrane trafficking signal (T), and the ER export signal (E). Mutations were made by QuikChange site-directed mutagenesis. The generated plasmid sequences were confirmed by DNA sequencing. The linearized plasmids (by NheI digestion) were used for in vitro generation of cRNA with an AmpliCap-MaxT7 high-yield message maker kit (Epicentre Biotechnologies). Afterward, the DNA fragments coding for proton pumps or channels together with fused signal peptides and eYFP were transferred to plasmid pcDNA3.1 (-) via BamHI and HindIII restriction sites for mammalian cell expression.

For the construction of the fungal-rhodopsin-channelrhodopsin-tandem plasmids, the SmaI restriction site in the backbone of pcDNA3.1(-)-hChR2(H134R)-mKate-hbetabR (BL191) (Kleinlogel et al., 2011) was removed by site-directed mutagenesis. In-frame replacement of BR with ApOps1 was enabled by means of overlap extension fusion PCR, yielding pcDNA3.1(-)-hChR2(H134R)-mKate-hbeta-ApOps1. The latter plasmid was used as a template in classical cloning with SmaI/HindIII restriction sites yielding pcDNA3.1(-)-hChR2(H134R)-mKate-hbeta-CarO, pcDNA3.1(-)-hChR2(H134R)-mKate-hbeta-LR, pcDNA3.1(-)-hChR2(H134R)-mKate-hbeta-ApOps1, and pcDNA3.1(-)-hChR2(H134R)-mKate-hbeta-ApOps2. Fungal LR was amplified from plasmid FCK-Mac-GFP, which was a gift from Edward Boyden (Addgene plasmid # 22223) (Chow et al., 2010).

Initial experiments of nonmodified ApOpsins without 2.0 cassette were performed with human embryonic kidney (HEK) Flp-InTM T-REx™-293 cells (Thermo Fisher Scientific) that were stably transfected with pcDNA™⁵/FRT/TO-ApOps1-eyfp or pcDNA™⁵/FRT/TO-ApOps2-eyfp to obtain the respective HEK293 cell lines following the manufacturer’s instructions. pcDNA3.1(-)-ApOps1-3 2.0 constructs (and for comparison pcDNA™⁵/FRT/TO ApOps1-eyfp and pcDNA™⁵/FRT/TO ApOps2-eyfp) were transiently transfected in NG108-15 cells using Lipofectamine 2000 as described recently (Feldbauer et al., 2016).

Staining procedures and imaging of HEK293 and NG108-15 cells were performed in poly-D-lysine covered in eight-well tissue culture chambers (eight-well tissue culture chamber, Sarstedt, or eight-well Lab-Tek^®II Chambered # 1.5 German Coverglass System, Nunc™, Thermo Fisher Scientific). NG108-15 cells were seeded in a density of about 10⁴ cells per well and grown for 12–24 h. A confocal laser scanning microscope (SP700, Zeiss, Germany) equipped with three laser lines (488 nm: 10 mW, 555 nm: 10 mW, 639 nm: 5 mW) was used. Images were recorded using a plan-apochromat 63x/1.40 oil M27 objective. The frame size for imaging was set to 1024 × 1024 pixels except otherwise specified, with a bit depth of 16 bits. The PMT detector gain for all channels was 500–700, and laser powers between 1 and 2.8% and pixel dwell times from 1.27 to 3.15 μs were used. Every line was averaged from two recordings with the laser scanning unidirectionally. The pinhole was adjusted to 1 airy unit (AU). Images were processed with ZEN software (ZEN 2012, Zeiss) or Fiji, Version ImageJ 1.50f (Schindelin et al., 2012).

Two days after cRNA injection, the expression level of optogenetic tools was evaluated by the fluorescence from the fused YFP. The images of Xenopus oocytes were obtained using a Leica DMi8 inverted microscope. An HC FL PLAN 2.5 × 0.07 DRY objective and a DFC3000G camera were used to capture images. The frame size was set to 1296 × 966 pixels, and the digitization was 8 bits. The wavelength of excitation light was 490–510 nm, and the emission channel used was 520–550 nm. The imaging process was performed with Leica Application Suite X software.

cRNA-injected oocytes were incubated in ND96 solution (in mM, 96 NaCl, 5 KCl, 1 MgCl₂, 1 CaCl₂, 5 HEPES, pH 7.4) containing 10 µM all-trans-retinal at 16°C. Thirty nanograms of cRNA were injected into the Xenopus oocyte for all the constructs. Photocurrents were measured 2 days after injection. Illuminations were performed through a 532 nm solid-state laser (Changchun New Industries, China). Electrophysiological measurements with Xenopus oocytes were performed at room temperature (20–23°C) with a two-electrode-voltage clamp amplifier (TURBO TEC-03X, npi electronic GmbH, Tamm, Germany). The bath solutions for electrophysiological recording are indicated in each figure legend. All buffers contained 1 mM MgCl₂, 2 mM CaCl₂, and 5 mM HEPES. Electrode capillaries (Ф = 1.5 99 mm, wall thickness 0.178 mm, Hilgenberg) were filled with 3 M KCl, with tip openings of 0.4–1 MΩ. A USB-6221 DAQ device (National Instruments) and WinWCP (v5.5.3, Strathclyde University, United Kingdom) were used for data acquisition. Origin2020 Pro was used for data analysis. The laparotomy to obtain oocytes from Xenopus laevis was carried out in accordance with the principles of the Basel Declaration and recommendations of Landratsamt Wuerzburg, Veterinaeramt. The protocol under license #70/14 from Landratsamt Wuerzburg, Veterinaeramt, was approved by the responsible veterinarian.

Patch-clamp experiments were performed using a setup described in detail before (García-Martínez et al., 2015; Adam et al., 2018; Panzer et al., 2019). Measurements of the tandem proteins were performed with the lasers used previously (García-Martínez et al., 2015). For ApOps 2.0 constructs, the beam of a 532 nm DPSS laser (Changchun New Industries Optoelectronics) was coupled into a 400 µm light fiber that was coupled to a fiber-optic cannula (Thor labs, CFM14L10, Ø400 µm Core, 0.39 NA). Cells were illuminated directly in front of the fiber. For the measurement of dose dependency (0.02–84 mW mm⁻²), the laser intensity was set by means of neutral density (ND) filters. The composition of the standard patch-clamp solutions was described recently (Adam et al., 2018). For the identification of the transported ion species in the extracellular solution, NaCl was replaced with an equimolar concentration of sodium gluconate (absence of chloride). Similarly, in the pipette solution, NaCl was replaced with an equimolar concentration of CsCl. A voltage-step protocol was used for recording the I–V plots. The time between start of sweeps was set to 2 s (time without illumination: 1.9 s). Data were analyzed with ClampFit 10.7 software, Excel, and Origin Pro 2016 64Bit. The mean and standard deviation of the values obtained from different cells were calculated. Current values were normalized to the value obtained at a 0 mV clamp voltage in bath solution NaCl pH 7.4 and plotted against the applied membrane voltage. Action spectra were recorded and analyzed as described previously (García-Martínez et al., 2015).

We BLAST-searched the genome of Aureobasidium pullulans (isolate EXF-150, CBS 100280, from the hypersaline waters of the Sečovlje solar saltern in Slovenia) (Gostinčar et al., 2014) and found three genes coding for putative opsins. The putative (rhod)opsins ApOps1 (gb|KEQ89910.1), ApOps2 (gb|KEQ89333.1), and ApOps3 (gb|KEQ87154.1) were identified by conserved residues that are typically present in microbial rhodopsins including the lysine in the seventh transmembrane domain providing the retinal binding site (Figure 1; complete sequence in Supplementary Figure S1). ApOps1 and ApOps2 share 59% sequence identities, and both are assigned to the clade of CarO-like rhodopsins, while ApOps3 is a representative of the LR-like rhodopsins (Adam et al., 2018). All three rhodopsins also show aspartate residues in the proton-acceptor position, homologous to BR(D85), as well as in the proton-donor position (a homologue of BR-D96). The proton-releasing group (BR: E194, E204, D212) consists of either two glutamate and one aspartate residues (ApOps1 and ApOps2) or one glutamate and two aspartate residues (ApOps3). While ApOps1 and ApOps2 show glutamate in the BR-G116 analogue position, which was considered as a hallmark of fungal rhodopsins (Fan et al., 2011), we find a valine in ApOps3 in this position. Overall, from the conserved residues, we would expect proton pump functions from all three A. fumigatus rhodopsins.

Using Swissmodel, we modeled the 3D structure of all three rhodopsins. All three rhodopsins consist of seven transmembrane domains. As observed in other fungal rhodopsins like CarO from F. fujikuroi or the rhodopsins from U. maydis also, all three ApOpsins show an extended extracellular loop between TM2 and TM3 which might play a role in the regulation of pump activity (Zabelskii et al., 2021). Modeling ApOps2 and ApOps3 to the crystal structure of LR (7bmh.1.A) revealed solid QMEANDisCo values of 0.71 ± 0.05 and 0.86 ± 0.05, respectively. Modeling of ApOps1 to LR (0.69 ± 0.05) revealed slightly smaller values than to Coccomyxa subellipsoidea Rhodopsin CsR (6gyh.1.A; 0.72 ± 0.06). The major difference observed in the two models of ApOps1 is the folding of the extracellular loop between TM2 and TM3, which was shown to interact with the N-terminus in LR (Zabelskii et al., 2021).

The expression of fungal rhodopsins in mammalian cells tends to be challenged by the fact that the trafficking is not optimally addressing the plasma membrane and some of the protein is sticking in the endoplasmic reticulum (Supplementary Figure S2).

We therefore added to our construct membrane trafficking factors from other membrane proteins, which were already successfully used in combination with other microbial rhodopsins for improved expression in plants (Zhou et al., 2021). The cleavable N-terminal signal peptide Lucy-Rho was linked to the N-terminus, the Golgi apparatus trafficking signal from Kir2.1 was used twice as a linker between the fungal rhodopsin and the eYFP tag, and the endoplasmic reticulum export signal from Kir2.1 was attached at the C-terminus (Figure 2A).

This modification of the membrane proteins led to improved expression and trafficking toward the plasma membrane (Figures 2B,C; Supplementary Figure S3A) and increased pump currents (Supplementary Figures S3B,C), whereas the physiological properties were unchanged (Supplementary Figure S3D). Also, the Lucy-Rho signal peptide improved the photocurrent of eArch3.0 [an improved version of Arch with maximal absorption around 530–550 nm (Mattis et al., 2011)] and Coccomyxa subellipsoidea rhodopsin (CsR), which was yet smaller than ApOps2 (Supplementary Figures S3B,C). Therefore, in the following, we present the electrophysiological data that were obtained with the expression-enhanced versions, called ApOps1 2.0, ApOps2 2.0, and ApOps3 2.0 hereafter.

According to the structural models and the conserved residues, we expected all rhodopsins from A. pullulans to be proton pumps. To investigate this, we performed whole-cell patch-clamp experiments in the voltage-clamp mode at NG108-15 cells heterologously expressing the respective rhodopsin. All experiments were done at pH7.4 intra- and extracellular in the presence of sodium chloride and hence under conditions similar to those in previous investigations with other microbial rhodopsins (García-Martínez et al., 2015; Adam et al., 2018; Panzer et al., 2019).

Indeed, when illuminated with green light (DPSS-laser, 532 nm), all rhodopsins responded with a typical pump signal, similar to those observed with other fungal rhodopsins like CarO (García-Martínez et al., 2015; Adam et al., 2018) or UmOps1 and UmOps2 (Panzer et al., 2019) (Figures 3A–C). Hence, ApOps1, ApOps2, and ApOps3 are light-gated ion pumps. We also analyzed the voltage dependency of the rhodopsins, revealing that all rhodopsins are still active at negative membrane potentials. Nevertheless, as expected from the increasing gradient, the pump activity decreased with more negative membrane potentials. Notably, the characteristic initial peak that is well pronounced in many microbial proton pumps like BR and CarO was only present in ApOps1 and ApOps3, especially at more negative membrane potentials, whereas at more positive voltages, the stationary currents were dominating. Interestingly, at depolarized membranes, ApOps1 reached its maximal activity late, after about 20–30 ms (Figure 3A).

All three rhodopsins show a biexponential current decay upon light-off, exhibiting a fast and a slow time constant. The fast time constants fall into the lower single-digit range, and the slow ones fall in the double and lower three-digit range. The slow time constants show a clear voltage dependency. Concerning the mean slow time constant, ApOps1 exhibited much higher values than ApOps2 and ApOps3, suggesting a slow photocycle [at 0 mV 72 ± 16 ms (ApOps1) vs 19 ± 2.9 ms (ApOps2) and 20.2 ± 2.6 ms (ApOps3); Supplementary Figure S4].

Fungal rhodopsins typically are supposed to perceive green light. To investigate if this is also the case for the Aureobasidium rhodopsins, we analyzed the spectral range promoting maximal pump activity by recording the action spectrum. All rhodopsins yielded bell-shaped action spectra ranging from 400 to 640 nm (Figures 3D–F) with the maximal pump activity in the green spectral range (excitation with 532 nm).

Of note, the pump activity of these rhodopsins, especially of ApOps2 and ApOps3, was relatively high in comparison to other fungal rhodopsins, reaching mean current densities of 4.5 ± 1.2 pA/pF (n = 16, ApOps1 2.0), 5.4 ± 4.8 pA/pF (n = 13, ApOps2 2.0), and 7.3 ± 3.2 pA/pF (n = 14, ApOps3 2.0) at 15 mW mm⁻². However, the highest current densities were observed with ApOps2 2.0, sometimes reaching values above 15 pA/pF.

The number of activated pumps can be calculated from the amplitude of the light-induced pump-current I_stat and the slow time constant of current decay τ_slow under saturating conditions and assuming that one charge is transported per photon. The products of I_stat and τ_slow divided by the elementary charge (e = 1.6 * 10–¹⁹ As) and the cell surface area [obtained from the membrane capacitance divided by the NG108-15-specific membrane capacitance of 1.72 ± 0.14 μF/cm² (Feldbauer et al., 2016)] yield 10.8 ± 7.2 * 10³ pumps/µm² (ApOps1 2.0, n = 5), 7.4 ± 5.7 * 10³ pumps/µm² (n = 5, ApOps2 2.0), and 4.9 ± 2.9 * 10³ pumps/µm² (n = 8, ApOps3 3.0), suggesting that ApOps2 and ApOps3 are more potent pumps reaching higher pump activities with a lower expression rate.

For the comparison of protein functions independent of the expression level, we used a genetic tandem cassette combining two rhodopsins in one functional protein (Kleinlogel et al., 2011; Feldbauer et al., 2016; Vierock et al., 2020). This protein combines an N-terminal rhodopsin (here, ChR2) via a fluorescent protein and the beta-subunit of the H⁺/K⁺-ATPase with the second rhodopsin (the respective fungal rhodopsin) in the C-terminal position (Supplementary Figure S5A). We measured the fungal rhodopsin currents (561 nm = low activation of ChR2) at 0 mV and normalized it to the ChR2-currents (473 nm) at –120 mV (Supplementary Figures S5B,C). With this stoichiometric comparison in the tandem protein, the pump activity of ApOps2 was about 4 times higher than that of ApOps1 (Supplementary Figure S5D). This is in accordance with the faster time constants observed in ApOps2 (Supplementary Figure S4) and shows once more that this relatively high pump activity was not (only) due to an expression effect but an intrinsic feature of this fungal rhodopsin. Indeed, in this comparison, MAC, the fungal rhodopsin currently used most in optogenetic applications, showed 35% less activity than ApOps2 under the same conditions (Supplementary Figure S5D). High pump activity is directly coupled to high energy consumption. In accordance, we found that ApOpsins are saturating at very high light intensities higher than the maximal solar radiation (Supplementary Figure S6).

To investigate if ApOps1-3 are really proton pumps, we first analyzed the pump currents in different buffers of the same pH by TEVC with Xenopus oocytes. The photocurrents of ApOps1-3 were similar in different buffers of the same pH 7.5 (Figure 4A). We then further tested the pH dependency of the pump activity in NG108-15 cells, where the intracellular pH can be controlled. An outward-directed proton pump should be challenged by an acid extracellular environment and negative membrane potentials (Figure 4B). Indeed, all three rhodopsins showed this expected behavior, with the lowest pump activity at negative voltages in pH 5. In contrast, the highest pump activity was observed at pH 9 and positive voltages. This is in accordance with the hypothesis that these proteins act as proton pumps. Interestingly, for ApOps1 and ApOps3, we observe an I-V curve non-linear at positive membrane potentials (Figure 4B).

To further confirm the proton pump activity in NG108-15 cells, we replaced either extracellular chloride with gluconate or intracellular sodium with cesium in the whole-cell patch-clamp measurements as besides protons, potentially, chloride or sodium is a potential target for the charge transfer by the ApOps proteins as well. The resulting I-V relationship was similar to the I-V curve observed with NaCl (Supplementary Figure S7; Figure 4B), with the only exception that the saturation effect observed from ApOps1 and ApOps3 in sodium chloride was absent under these conditions, suggesting a modulating role of sodium under acidic conditions. Furthermore, when using gluconate, a slightly enhanced pump current could be measured. This issue derives from the “weak organic acid (WOA) effect,” which could be shown previously for other fungal rhodopsins (Garcìa-Martìnez et al., 2015; Adam et al., 2018; Panzer et al., 2019) dealing with the fact that rhodopsins show increased pump activity in the presence of extracellular weak organic acids.

While for ApOps1 2.0 and ApOps2 2.0 the current densities were similar, in contrast, for ApOps3 2.0, we observed a significant decrease in pump activity in the presence of cesium/absence of sodium (Supplementary Figure S7).

ApOps2 and ApOps3 showed very intense pump activity, making them potentially superior optogenetic tools for membrane voltage and pH regulation. Besides proton pumps, proton-specific channels are of interest for the regulation of pH by light as well. We therefore wanted to test if the Aureobasidium proton pumps could be altered by mutagenesis to proton channels. In our previous studies with a Fragilariopsis proton pump rhodopsin, a random mutation of arginine (R) to histidine (H) happened during the cloning process. Since the R position is mostly conserved in the transmembrane helix 3 of different rhodopsins and it is known that it is of importance for the proton transfer, we decided to keep the wrong clone/mutant for a measurement. We found it to be a light-gated proton channel (unpublished, PhD thesis of S. Gao). We used site-directed mutagenesis to introduce the similar mutation R112H to ApOps2 and also the corresponding mutations in ApOps1 and ApOps3 (Figure 5A). We found that this single mutation turns ApOps1-3 into light-gated proton channels (Figure 5A). Interestingly, ApOps1 with the least proton pump current showed the largest channel current after mutation (Figure 5B). From the modeled structure, we thought that Y78 or F79, corresponding to Y57 in CsR, might participate in conformational modification of the proton-transport pathway. Introducing an additional mutation F79A indeed further increased the proton channel current of ApOps2 R112H (Figure 5A). A single F79A mutation will not change the proton pump character but only influence the photocurrent slightly (Supplementary Figure S8). ApOps2 2.0 F79A/R112H yielded the largest proton channel current among the tested ApOps1-3 mutants (Figure 5B). The ApOps2 2.0 F79A/R112H mutant showed about a 3-fold higher proton channel current than the corresponding CsR mutants and the previously published CsR Y14E mutant, whose reversal potential did not meet requirements for a pure proton channel (Fudim et al., 2019). In general, protein expression and membrane insertion in the mutants were similar to the wild type (Supplementary Figure S9A).

Due to the promising intense pump activity, we further investigated the ApOps2 2.0 F79A/R112H mutant (Figure 6). In NG108-15 cells, this mutant showed good expression (Supplementary Figure S9B) and light-gated proton channel activity (Figure 6A) that was strongly dependent on the extracellular pH (Figures 6A,B). In Xenopus oocytes, we showed that the reverse potentials measured for ApOps2 2.0 F79A/R112H are similar to the reversal potentials calculated for protons using the Nernst equation with E_s = 58.17*(pHi-pHe), and pHi and pHe represented intracellular and extracellular pH, respectively (Figure 6C). pHi was regarded as 7.43 (Cicirelli et al., 1983). In accordance, testing the ApOps2 2.0 F79A/R112H double mutant in different extracellular buffers confirmed the current specificity for protons (Figure 6D).