SARS-CoV-2 ORF3a accessory protein is a water-permeable channel that induces lysosome swelling

Abstract

ORF3a, the most abundantly expressed accessory protein of SARS-CoV-2, plays an essential role in virus egress by inactivating lysosomes through their deacidification. However, the mechanism underlying this process remains unclear. While seminal studies suggested ORF3a being a cation-selective channel (i.e., viroporin), recent works disproved this conclusion. To unravel the potential function of ORF3a, here we employed a multidisciplinary approach including patch-clamp electrophysiology, videoimaging, molecular dynamics (MD) simulations, and electron microscopy. Preliminary structural analyses and patch-clamp recordings in HEK293 cells rule out ORF3a functioning as either viroporin or proton (H⁺) channel. Conversely, videoimaging experiments demonstrate that ORF3a mediates the transmembrane transport of water. MD simulations identify the tetrameric assembly of ORF3a as the functional water transporter, with a putative selectivity filter for water permeation that includes two essential asparagines, N82 and N119. Consistent with this, N82L and N82W mutations abolish ORF3a-mediated water permeation. Finally, ORF3a expression in HEK293 cells leads to lysosomal volume increase, mitochondrial damage, and accumulation of intracellular membranes, all alterations reverted by the N82W mutation. We propose a novel function for ORF3a as a lysosomal water-permeable channel, essential for lysosome deacidification and inactivation, key steps to promote virus egress.

ORF3a, an accessory protein of SARS-CoV-2, mediates the transmembrane transport of water, leading to lysosome swelling. This function could, in principle, result in lysosome inactivation, a crucial step that facilitates virus egress.

Introduction

Severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2) is a positive-strand RNA β-coronavirus that causes coronavirus disease-19 (COVID-19), responsible for the latest pandemic. Despite numerous antiviral strategies and vaccines in development, no fully effective anti-COVID-19 therapies have been establieshed yet. SARS-CoV-2 infection begins with the spike (S) protein interacting with the angiotensin-converting enzyme 2 (ACE2) on the host cell plasma membrane. This interaction determines the subsequent entry of the virus into the cell. Once inside, the virus exploits the host cell translation machinery and intracellular organelles such as the endoplasmic reticulum (ER) and Golgi apparatus, to synthesize new viral particles before escaping the cell^1,2. Since the processes of entry and egress are crucial steps in the viral infective cycle, the underlying molecular mechanisms may represent potential targets for the development of new vaccines or therapeutic strategies.

The SARS-CoV-2 genome encodes 29 proteins essential for virus replication, maturation, and egress, classified as structural, non-structural, and accessories³. Two overlapping open reading frame (ORF) regions of the viral genome, ORF1a and ORF1b, encode non-structural proteins (NSPs), which constitute the majority of the proteins synthesized by the virus and are responsible for viral RNA replication. Additional genomic regions encode four structural proteins—S, envelope (E), membrane (M) and nucleocapsid (N), that are components of the virion^4,5. In addition, the SARS-CoV-2 genome contains sequences that encode a series of viral accessory proteins involved in the entry and egress processes. Among these, ORF3a is one of the largest (274 amino acid residues) and the most expressed. It is prevalently localized in endosomes/lysosomes as an integral membrane protein, and consists of an N-terminus luminal domain, three transmembrane segments and a short C-terminus cytoplasmic domain⁶. Both dimeric and tetrameric assemblies of the protein have been described when expressed in insect cells, with the tetramer formed by the association of two dimers⁶. However, a recent paper reported that the majority of ORF3a at the plasma membrane of HEK293 cells exists as monomers and dimers⁷.

In recent years, growing evidence has suggested that ORF3a plays a central role in virus egress from the host cell by hijacking lysosomal pathways to facilitate exocytosis of viral components through the “non-conventional” pathway^8–11. Normally, lysosomes maintain an acidic environment essential for their functions, including the degradation of cellular waste and pathogens. Therefore, for the virus to exit via lysosomes, they must be inactivated and repurposed as exocytotic vesicles. Inactivation likely occurs through deacidification, which reduces lysosomal luminal acidity and impairs enzyme activity. Gosh and collaborators first reported this reduction in lysosomal acidity and activity in cells infected with β-coronaviruses⁸. Subsequent studies demonstrated that ORF3a abolishes lysosomal proteolytic and digestive functions by promoting their swelling, deacidification, and consequent inactivation^8,9,12, enabling them to serve as vesicles for virus egress through the lysosomal exocytotic pathway. In addition, ORF3a blocks autophagy by inhibiting the fusion of autophagosomes with lysosomes through its interaction with the VPS39 protein, a component of the homotypic fusion and protein sorting (HOPS) complex^10,13,14, thus preventing the digestion/destruction of viral particles.

Although several studies have confirmed these roles, the exact function of ORF3a and its mechanism for inactivating the proteolytic/digestive power of lysosomes through their deacidification, remain uncertain and elusive, and a great effort is being made to unveil its functional role. In 2021, two research groups independently proposed that ORF3a from SARS-CoV-2 functions as a viroporin (i.e., viral ion channel) that mediates the passage of cations, such as Ca²⁺, Na⁺, and K^+6,15,16, a function that could potentially alter the luminal content of lysosomes and lead to their inactivation. This conclusion is fully consistent with previous reports demonstrating that ORF3a from SARS-CoV-1 is a non-selective cation channel^17–19. However, a recent study has overturned this notion by demonstrating that ORF3a from both SARS-CoV-1 and SARS-CoV-2 is not an ion channel¹⁴.

To investigate the potential function of SARS-CoV-2 ORF3a, in the present study we employed a multidisciplinary approach including molecular biology, patch-clamp electrophysiology, videoimaging for the assessment of cell volume changes, molecular dynamics (MD) simulations, and electron microscopy (EM). Collectively, our results demonstrate that ORF3a from SARS-CoV-2 is neither an ion channel nor a proton (H⁺) channel. We also propose a novel function for ORF3a as a water-permeable lysosomal channel. The ORF3a-mediated water transport across lysosomal membrane would cause lysosomal swelling and deacidification, key steps in promoting virus egress from the cell.

Materials and methods

HEK293 cell culture

HEK293 cells (ATCC, CRL-1573™) were grown in Dulbecco’s Modified Eagle Medium (DMEM, high glucose, with L-glutamine, and Phenol red, Invitrogen, Life Technologies) supplemented with 10% FBS (Gibco, Thermo Fisher Scientific) and 1% Pen-Strep (100 U/mL, Sigma-Aldrich). The cell culture was maintained at 37 °C and 5% CO₂.

Plasmid constructions

The plasmid expressing ORF3a fused at its C-terminus to the EGFP was generated by amplifying the ORF3a coding region from cDNA of SARS-CoV-2-infected Vero cells using primers P1 and P2 (Supplementary Information, Table S1, reports the nucleotide sequence of all primers used in this study). ORF3a cDNA was digested with the restriction enzymes HindIII and BamHI and cloned into the vector pcDNA3-EGFP, upstream of the EGFP in the same restriction sites. The plasmid expressing the truncated form of ORF3a, lacking the first 41 amino acids, and fused to the N-terminus of the EGFP, was obtained by amplifying ORF3a with primers P3 and P2 and cloned in the pcDNA3-EGFP vector as the previous plasmid. Plasmids carrying the mutation of the asparagine 82 (N82) with either tryptophan (N82W) or leucine (N82L) were generated by overlap PCR using primers P5 and P6 or P7 and P8, respectively. All constructs were sequenced to rule out the occurrence of nucleotide substitutions in the coding sequences.

Transfection of HEK293 cells

HEK293 cells were plated in 6-well plates with cover slips, aiming to reach 70-90% confluency by the time of transfection, which was performed the day after (0.25 × 10⁶ cells for each well). On the day of transfection Lipofectamine^TM 3000 (Invitrogen, Life Technologies) was diluted in DMEM, at a concentration of 5 µl in 125 µl of medium. The master mix of DNA was prepared by diluting 5 µg of plasmid DNA and 10 µl of P3000^TM reagent in 125 µl of medium. The DNA master mix was then added to the Lipofectamine mixture and gently mixed by pipetting several times. After incubating for 10 min, the DNA-lipid complex was added to the cells. HEK293 cells were incubated overnight with the transfection master mix which was then replaced with DMEM supplemented with 10% FBS. Cells were then incubated for 2–4 days at 37 °C and 5% CO₂. Starting the day after the transfection, fluorescence signal from the transfected plasmid was detectable using a fluorescence light microscope.

MD simulations

ORF3a protein models were based on the Cryo-EM structure, Protein Data Bank entry 6XDC⁶. In few instances, we used a second available structure, 8EQJ¹⁴. The setup of the simulation box was performed with CHARMM-GUI²⁰. The protein was inserted in the lipid membrane and the systems were solvated using TIP3P water molecules and K⁺ and Cl^- ions to reach 200 mM KCl. Van der Waals interactions were truncated at 9 Å. Standard scaling of 1−4 interactions was applied. Long-range electrostatic interactions were calculated with the particle mesh Ewald method using a grid spacing of 1.0 Å. The SETTLE algorithm was used to constrain bonds with the hydrogen atom. The temperature was maintained at 300 K by coupling to a Langevin thermostat with a damping coefficient of 1 ps⁻¹. We chose to run our simulations at 300 K since our primary goal was to obtain results comparable to our experimental (patch-clamp and videoimaging) results, which were obtained at room temperature. A pressure of 1 atm was maintained by coupling the system to a Nose−Hoover Langevin piston, with a damping constant of 25 ps and a period of 50 ps. The simulation protocol consisted of 2000 energy minimization steps, followed by an annealing of the system to a temperature of 300 K over 0.3 ns simulation time, and then 50 ns constrained dynamics with a timestep of 2 fs. During the annealing and constrained dynamics, restraints were applied on the C_α atoms of the backbone protein. Thereafter a 150 ns simulation was performed without restraints using a timestep of 2 fs. Additional 150 ns replicas were generated by reinitializing the atoms velocity. NAMD2.11 and CHARMM36 force field were used, although in some instances the AMBER force field was used (Supplementary Information, Table S2). The N82W and N82L mutants were generated using the CHARMM-GUI and the resulting structures were minimized and equilibrated with the same protocol used for the WT protein. A summary Table of all the simulations performed is presented in the Supplementary Material (Supplementary Information, Table S2). PDB files containing the initial and final coordinates from the CHARMM36 simulation on the tetrameric structure are provided as Supplementary Data.

Immunofluorescence assay

For immunofluorescence assay, HEK293 cells were grown overnight, after transfection, on cover slips. Initially, they were washed twice with PBS, and then fixed for 20 min with 4% (w/v) paraformaldehyde in PBS, permeabilized with 0.1% Triton X-100 in PBS for 20 min and blocked with 10% (v/v) FBS in PBS for 1 h. For primary antibody staining, mouse anti-GFP antibody (monoclonal derived from hybridoma in culture; 1:200) was diluted in 2% (v/v) FBS in PBS. Cover slips were washed three times for 5 min in 2% (v/v) FBS in PBS. A goat anti-mouse Alexa Fluor 488 secondary antibody (Invitrogen, cat. # A11001) was used for 1 h, diluted in 2% (v/v) FBS in PBS at 1:1000. Cover slips were then washed three times for 5 min in 2% (v/v) FBS in PBS and then mounted on slides with 25 ml of Vectashield® antifade mounting medium with DAPI (Vector Laboratories). Images were taken on a fluorescence light microscope (Nikon-Eclipse-TE2000-U) at 100× in oil immersion and analyzed using Fiji software (ImageJ).

Electrophysiology

The patch-clamp whole-cell dialyzed configuration was used for electrophysiological recordings. Currents were amplified with a HEKA EPC‐10 amplifier (List Medical), digitized with a 12‐bit A/D converter (TL‐1, DMA interface; Axon Instruments), and analyzed with the Patch‐Master software package (version 2_60, ELEKTRONIK) and Microcal Origin 8.0. For on‐line data collection, macroscopic currents were filtered at 3 kHz and sampled at 50 μs/point. Patch electrodes with resistances 3–5 MΩ were fabricated on PUL-100 Micropipette Puller (World Precision Instruments) from borosilicate glass (OD1.5 mm, ID 0.84 mm, World Precision Instruments). Access resistances ranging between 5 and 15 MΩ were achieved by membrane disruption following seal formation and were actively compensated to ~50%. For ion current recordings, the internal solution contained: KCl 155 mM, EGTA-K 1 mM, HEPES 5 mM, MgCl₂ 1 mM, adjusted to pH 7.2 with KOH. The external standard Ringer’s solution contained: NaCl 140 mM, KCl 5 mM, CaCl₂ 2 mM, MgCl₂ 2 mM, HEPES 5 mM, glucose 10 mM, adjusted to pH 7.4 with NaOH. To assess ORF3a mechanosensitivity, a hypotonic solution, prepared by adding 30% distilled water to the extracellular standard Ringer’s solution, was used to promote cell swelling and the resulting increase of membrane stretch/tension. To prevent the activation of the swelling-activated Cl^- current mediated by the volume-regulated anion channel (VRAC), which is endogenously expressed in HEK293 cells, the external solution was supplemented with the VRAC inhibitor DCPIB at the saturating concentration of 10 µM, as previously reported²¹. For H⁺ current measurements, both the internal and external solutions contained NMDG⁺ and MeSO³⁻ as the major cations and anions, respectively, as described in previous studies²². The internal solution contained: NMDG⁺ 100 mM, MeSO³⁻ 70 mM, MgCl₂ 2 mM, EGTA 10 mM, HEPES 100 mM, adjusted to pH 7.2 with 28 mM NMDG, with a theoretical calculated osmolarity of 314 mOsm. The external solutions contained: NMDG⁺ 100 mM, MeSO₃^- 70 mM, CaCl₂ 2 mM, MES 100 mM (pH 5.5) or HEPES 100 mM (pH 7.4). The osmolarity was adjusted to a theoretical value of 314 mOsm by adding NMDG-MeSO₃ 14 mM. All reagents were freshly prepared daily at the specified concentrations and applied to the bath using a gravity perfusion system. Experiments were carried out at room temperature (18–22 °C).

Cell volume measurements

Cells were perfused with an extracellular Ringer’s solution identical to that used in the electrophysiological experiments. To study the effect of hypotonic‐induced cell swelling, a 30% hypotonic solution, identical to the one used for assessing ORF3a mechanosensitivity, was applied. Relative projection area of cells was measured using a videoimaging technique^23,24. The cells were observed through a 40× objective lens of a microscope connected to a video camera (Axiocam/Cm1, Zeiss). Images of the cells were saved as TIFF files and the area of each image was measured using ImageJ. Images were acquired every 10 s over a total recording period of 6 min. Relative cell area (A_rel) was calculated by dividing the area of the cell at each time point (A_t) upon exposure to hypotonic solution by the initial area (A₀) taken under isotonic conditions (more precisely the average of the three time points recorded in the first 20 s). Finally, the relative cell area was converted to relative cell volume by assuming a spherical cell shape.

Fluorescence-activated cell sorting (FACS)

Cells were harvested from cell culture plates with trypsin, resuspended in complete DMEM medium, and centrifuged at 900 × g for 5 min. Afterward, cells were washed with PBS at 37 °C and centrifuged again. For sorting, cells were resuspended in Accumax - Cell Aggregate Dissociation Medium (Invitrogen, Thermo Fisher) and filtered with BD Filcon 70 μm (Fisher Scientific) to prevent clumping. GFP-positive cells, at least 100.000, were isolated using a FACSAria III cell sorter (BD Biosciences, San Jose, CA, USA). For gain strategy, please refer to the Supplementary Fig. 5 and legend.

Electron microscopy (EM)

For EM preparation, sorted transfected HEK293 cells were centrifuged at 10,000 × g for 5 min, washed three times with PBS, fixed with 3.5% glutaraldehyde in 0.1 M sodium cacodylate buffer for 1 h, and stored at 4 °C. For embedding, cells were post-fixed in 2% osmium tetroxide (OsO₄) in the same buffer for 2 h and block-stained in saturated uranyl acetate replacement. After dehydration, specimens were embedded in epoxy resin (Epon 812). Ultrathin sections (about 40 nm) were cut using a Leica Ultracut R microtome (Leica Microsystem, Austria) with a Diatome knife (DiatomeLtd. CH-2501 Biel, Switzerland) and double-stained with uranyl acetate replacement and lead citrate. Sections were viewed and photographed in a 120 kV JEM-1400 Flash Transmission Electron Microscope (Jeol Ltd, Tokyo, Japan) equipped with CMOS camera Matataki and TEM Center software (Jeol Ltd, Tokyo, Japan).

Quantitative analyses of EM images

Quantitative analysis was done on random micrographs of HEK293 cell cultures. The average size of lysosomes was measured by drawing the outline of each lysosome using the Analy-SIS software of the EM digital camera (Olympus Soft Imaging Solutions). Only clearly membrane-limited lysosomes were measured. The analyses were conducted on approximately 40 cells per sample.

Statistics and reproducibility

To ensure the reproducibility of results, at least three independent experiments were conducted. To evaluate the significance of differences, the two‐sample unpaired t-test was used to compare two distinct normally distributed populations, and the non-parametric Kruskal-Wallis test followed by post-hoc Dunn’s test was used to compare three independent, non-normally distributed populations. A p-value of < 0.05 was considered statistically significant, and all data are shown as mean ± SEM. The sample size and number of replicates for each experiment are indicated in corresponding figure legends. All experiments and data analysis were performed blinded.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Results

The structure and electrostatic properties suggest that ORF3a is not an ion channel

Since its discovery following the 2003 SARS epidemic, studies on ORF3a of SARS-CoV-1 have shown that this protein can form a non-selective cation channel (i.e., viroporin)^17–19. This view aligns with the cryo-EM structure of ORF3a from SARS-CoV-2, first reported by Kern and colleagues in 2021 and is further supported by electrophysiological data from both their group and Toft-Bertelsen and coworkers, suggesting that SARS-CoV-2 ORF3a functions as a non-selective channel that allows the passage of Na⁺, K⁺, and Ca^2+6,16. However, these results were challenged in the same year by Grant and Lester²⁵ and two years later by the Clapham's group¹⁴, who showed that ORF3a from both SARS-CoV-1 and SARS-CoV-2 does not form a cation-permeable viroporin, thereby questioning the exact function of this protein. To explore its possible function, we examined the ORF3a structure provided by Kern et al.⁶ to determine whether it possessed the necessary properties to function as a cation-selective channel.

Figure 1A shows the backbone of ORF3a in cyan, with positively (lysine and arginine) and negatively (glutamate and aspartate) charged amino acid residues highlighted in yellow and red, respectively. The distribution of positive charges at the intracellular entrance of the putative cation permeation pore (indicated), would make the passage of positive ions energetically unfavorable (Fig. 1A). This conclusion was further supported by the electrostatic potential profile of ORF3a, calculated from the equilibrated structure prepared for MD simulations (see “Methods”), and shown along a plane perpendicular to the plasma membrane in Fig. 1B. A positive electrostatic potential was clearly evident at the entrance of the putative permeation pore. Consistent with the findings of Clapham’s group¹⁴, these preliminary structural observations suggested that ORF3a is unlikely to function as a cation channel.

Fig. 1. Structure and electrostatic potential evaluation of ORF3a.

A 3-D structure of 6XDC ORF3a, provided by Kern and colleagues, in which the protein backbone is represented in cyan while atoms belonging to both positively and negatively charged amino acid residues are shown as yellow and red spheres, respectively. The intracellular (bottom view) and extracellular/luminal (top view) sides of the protein are also shown. Also indicated is the putative cation permeation pore suggested by Kern and colleagues⁶. B Profile of the electrostatic potential of ORF3a along a plane perpendicular to the plasma membrane. The electrostatic potential was calculated using the PMEPot Plugin of the VMD software, by averaging 100 frames taken every 0.1 ns of MD simulation. It is represented by the blue and red colors, corresponding to the positive and negative potentials, respectively, and quantified in volts using the palette in the top right corner.

ORF3a is not an ion channel

We then attempted to confirm the prediction of the structural analysis by directly testing, through patch-clamp experiments, whether ORF3a behaves as an ion channel. To this end, we used HEK293 cells and generated two experimental groups: i) cells transfected with a plasmid encoding only the enhanced green fluorescent protein (EGFP), serving as the control group (cEGFP) and ii) cells transfected with a plasmid encoding for the ORF3a fused to the EGFP at its C-terminus (ORF3a). We first examined ORF3a expression and localization through immunofluorescence analyses, as reported in Fig. 2A. Consistent with the known localization of ORF3a in the ER/Golgi/endolysosomal compartments, the EGFP fluorescence signal displayed a large intracellular distribution in ORF3a-transfected cells (Fig. 2A, left).

Fig. 2. Assessment of both ORF3a and tORF3a localization and patch-clamp recordings of ion currents in HEK293 cells.

A Representative immunofluorescence images of cells transfected with both the full-length (ORF3a, left) and truncated (tORF3a, right) ORF3a protein. The EGFP protein, fused to the ORF3a at the C-terminus, is shown in green and excited with a wavelength of 488 nm. The nuclei, stained with DAPI, are shown in blue. Scale bars: 20 µm. B Representative current ramps obtained by applying 1 s voltage ramps from −100 to 100 mV, starting from a holding potential of −40 mV, as shown in the upper part. The traces are from native non-transfected cells (black trace) and from cEGFP (gray trace), ORF3a (red trace), and tORF3a (blue trace) transfected cells. C Mean current-voltage (I-V) relationships of steady-state current obtained in native non-transfected cells (black circles, n = 5 cells) and in EGFP (gray circles, n = 5 cells), ORF3a (red circles, n = 6 cells), and tORF3a (blue circle, n = 6 cells) transfected cells. Inset: Representative families of current traces obtained by applying voltage steps from −100 to +100 mV, starting from a holding potential of −40 mV, with each step lasting 1 s and increasing by 20 mV as shown in the upper part. Data are shown as mean ± SEM. Individual data points for each experimental group are reported.

Since we needed to measure the ion currents on the plasma membrane, to enhance ORF3a localization at the cell surface, we generated a third group of HEK293 cells transfected with a version of ORF3a lacking the first N-terminus 41 amino acids, referred to as truncated ORF3a (tORF3a). Consistent with previous studies^6,7, removal of this sequence substantially increased ORF3a localization on the plasma membrane, as clearly shown by the enhanced EGFP fluorescence signal on the cell surface in tORF3a-transfected cells (Fig. 2A, right), compared to ORF3a-transfected cells (Fig. 2A, left).

To evaluate whether the expression of the ORF3a protein correlated with the presence of additional ion currents on the plasma membrane, we performed patch-clamp experiments in the whole-cell dialyzed configuration. Ion currents were recorded at varying potentials using different protocols: either 1 s voltage ramps from −100 to +100 mV (Fig. 2B) or discrete 1 s steps of potential from −100 to +100 mV with increments of 20 mV (Fig. 2C). In both protocols, the holding potential was set at −40 mV. Recordings were conducted on all experimental groups, cEGFP, ORF3a and tORF3a cells. Moreover, to assess if transfection per se could affect the functional expression of endogenous currents, few experiments were conducted in native non-transfected HEK293 cells. As shown by both the representative current traces (Fig. 2B, C) and the quantitative analysis of the current-voltage (I-V) relationship of the steady state current (Fig. 2C), the expression of the full-length ORF3a protein did not result in any additional current compared to the control group (cEGFP). To ensure that the absence of a difference in current density was not due to a low level of ORF3a on the plasma membrane, we repeated the same electrophysiological recordings on tORF3a-transfected cells. Again, the current density did not differ from that recorded in cEGFP cells. The absence of ion currents when the 41 residues at the N-terminus were deleted, was consistent with a recent study showing that the same construct, which exhibited robust cell surface expression, did not produce any additional currents compared to control cells⁷. It is worth noting that all the three groups of transfected cells exhibited similarly higher basal membrane currents than native non-transfected cells, strongly suggesting that transfection per se may alter somehow the plasma membrane electrical properties. In any case, the similarly higher currents observed in all transfected groups cannot be attributed to ion currents through ORF3a, as essentially identical currents were observed in cEGFP cells, which did not express ORF3a, compared to both ORF3a and tORF3a cells. Therefore, these results tell us that native non-transfected cells are not suitable as a control group, a place that should be taken by the cEGFP.

It has been previously suggested that ORF3a may have mechanosensitive properties²⁶. We tested this possibility by performing electrophysiological experiments in which cells were exposed to a 30% hypotonic solution that we previously showed to induce cell swelling (Supplementary Fig. 2) and activate the mechanosensitive Piezo1 channel^23,24,27. Exposure to the hypotonic stimulus was not able to activate any currents in all four groups tested. Further details of these experiments are reported in the Supplementary Information.

In conclusion, the combination of structural analysis of ORF3a (Fig. 1) and electrophysiological data obtained from ORF3a-tranfected HEK293 cells (Fig. 2), rules out the notion that ORF3a is an ion channel.

ORF3a mediates water transport across the membrane

Results published in recent years have demonstrated that ORF3a increases lysosomal pH. There are two possible ways by which ORF3a can do this: i) ORF3a may act as a H⁺ channel, directly mediating H⁺ flux across lysosomal membrane; ii) alternatively, ORF3a may function as a water-permeable channel that increases water permeability of the lysosomal membrane, thus affecting lysosomal pH. In particular, the increased water permeability may lead to the swelling of lysosomes and, by consequence, promote their deacidification. This notion is supported by three lines of evidence: i) the lysosomal membrane is generally considered to be relatively impermeable to water, and the expression of water-permeable proteins induces rapid swelling²⁸; ii) exposure of lysosomes to a hypotonic solution leads to lysosomal swelling and alkalinization (i.e., deacidification)²⁹; iii) expression of ORF3a results in significant enlargement (likely swelling) of lysososmes¹². Therefore, another possibility for promoting lysosomal swelling and deacidification would be that ORF3a mediates water flux across the lysosomal membrane.

To verify if ORF3a functions as a H⁺ channel, we conducted patch-clamp experiments to record H⁺ currents in cEGFP and tORF3a cells at two different pH (7.4 and 5.5, see Methods for details) Results from these experiments clearly showed that changing the pH from 7.4 to 5.5 did not elicit any detectable current both in cEGFP and tORF3a cells, which showed virtually identical basal currents at the two pH levels analyzed (Fig. 3A, B). The absence of H⁺ currents in cells expressing ORF3a on the plasma membrane suggests that ORF3a is not a H⁺ channel.

Fig. 3. Patch-clamp recordings of H⁺ currents and cell volume assessment for water permeability in HEK293 cells.

Average current-voltage (I–V) relationships in cEGFP (gray circles, n = 4 cells) and tORF3a cells (blue circles, n = 5 cells), obtained with an extracellular solution at pH 7.4 (A) and pH 5.5 (B). Insets: Representative families of current traces obtained by applying 1 s voltage steps from −100 to +100 mV, starting from a holding potential of −40 mV, with increments of 20 mV. C Time course of cell swelling, evaluated as relative cell volume changes, in control isotonic conditions (first three circles) and during application of a 30% hypotonic solution (30% Hypo, light gray bar), in control cEGFP cells (grey circles, n = 11 cells) and in ORF3a (red circles, n = 12 cells) and tORF3a (blue circles, n = 10 cells) transfected cells. The dashed box indicates the experimental range for the analysis of water permeability, as described in (D). D Analysis of the slope of the cell volume increment as evaluated during the first 100 s of recording upon application of 30% hypotonic solution, for the same experimental groups shown in (C). Inset: bar graph of the slope. Data are shown as mean ± SEM; *p = 1.35e⁻⁶ for cEGFP vs. ORF3a; *p = 4.54e⁻⁸ for ORF3a vs. tORF3a; *p = 2.22 × e⁻¹⁶ for cEGFP vs. tORF3a, as evaluated with two-sample unpaired t-test. Individual data points for each experimental group are reported.

To experimentally demonstrate that ORF3a can function as a water-permeable channel, we quantified the rate of water flux across the plasma membrane by measuring the rate of cell volume increase (i.e., cell swelling). To induce cell swelling, we perfused cells with a 30% hypotonic solution to create the osmotic gradient required for net water influx. Cell volume changes were monitored using a phase contrast microscope, and images were acquired at 10 s intervals, for a total of 6 min, sufficient to reach the peak of swelling. Based on the experiments in Fig. 2B, C, which show that transfection per se markedly influences endogenous currents, thereby eliminating the use of native non-transfected cells as a control group, we performed these experiments on all three transfected groups i.e., cEGFP, ORF3a, and tORF3a, with cEGFP cells serving as the control group. As shown by the average time courses of cell volume changes, all cells analyzed exhibited an initial rapid and linear increase in volume upon application of the hypotonic solution, which lasted about 100–110 s from the onset of the hypotonic shock, followed by a slower, less steep, increment that reached a plateau (Fig. 3C). As described in Supplementary Information, Note 1, the volume of a cell is expected to change linearly with time during the initial phase of cell swelling, with the slope being proportional to the water permeability of the cell membrane. Therefore, the slope of the linear portion of the curves serves as an indicator of the rate at which the water passes through the plasma membrane (Fig. 3D). The quantitative analysis of the relative cell volume during the first 100 s after the exposure to the hypotonic solution revealed that the rate of increase in cell volume in cell volume strictly correlated with ORF3a expression on the plasma membrane. In fact, fitting the experimental data to a linear function, revealed that the slope of cell volume increase was 1.3 and 1.6 times significantly greater in ORF3a and tORF3a, respectively, compared to cEGFP cells (Fig. 3D).

MD simulations confirm ORF3a functioning as water transporters

To verify whether the ORF3a structure has properties consistent with a water transporter, we performed MD simulations to investigate potential water conductance. Since ORF3a has been shown to exist both as homodimer and homotetramer⁶, our simulations were conducted using both these structural forms. As shown in Fig. 4A, B (upper panels), ORF3a dimers did not exhibit any significant water columns (Fig. 4A), nor did they show water density across the protein connecting the two aqueous compartments (Fig. 4B). As detailed in Supplementary Information, Table S2, MD simulations like those shown in Fig. 4A, B were repeated using both available ORF3a structures from the PDB (6XDC of Kern et al.⁶; 8EQJ of Miller et al.¹⁴), along with CHARMM36 and AMBER force fields as starting points, to verify whether our negative results were influenced by the model structure or the parameters used in the simulation. In addition, since a potential influence of membrane lipids on the dynamics of ORF3a has been suggested³⁰, we also performed simulations using both a relatively simple phospholipid bilayer (either 100% POPC or POPC:POPA 3:1) and a lipid mixture more representative of the plasma/lysosomal membrane composition (POPC 15%, POPE 20%, sphingomyelin 15%, monoacylglycerol phosphate 20%, cholesterol 30%) (PC:PE:PI:PS:SM:CHOL:BMP 35:25:8:3:6:18:7). Finally, we also tested whether water conductance could be observed in presence of a thinner phospholipid bilayer (DDPC:DLPA 3:1). In all these simulations we were unable to detect significant water conductance in ORF3a dimers (Supplementary Information, Table S2). By contrast, significantly different results were obtained when MD simulations were performed using the tetrameric structure of ORF3a. While no water conductance was detected when the tetramer was simulated in the POPC:POPA 3:1 bilayer, we observed the presence of water columns connecting the two aqueous compartments through the TM2 and TM3 transmembrane helices of the cis monomers (where we defined cis as the closest monomers within the two different dimers, and trans as the most distant ones), when the tetramer was simulated in the DDPC:DLPA 3:1 (Fig. 4A, B). Conversely, no water columns were observed for the trans monomers of the tetrameric structure, suggesting that the cis monomer/cis monomer interface could favor the passage of water. This also implies that there may be differences in the conformational properties of the two subunits in each dimer. We examined the trajectory of water as it moved from one vestibule to the other one (Fig. 4C, D) and counted the number of water molecules crossing the protein. The quantitative analysis revealed an average water conductance of 4.1 ± 1.0 molecules*ns⁻¹ (Fig. 4E).

Fig. 4. ORF3a water permeability as assessed through MD simulations.

A Snapshot of ORF3a protein after the equilibration period, for both the dimeric (top) and tetrameric (bottom) structures of the 6XDC ORF3a. Note the water columns crossing the protein between transmembrane helices of the cis monomers in the tetrameric structure, linking the intra- and extracellular environments. B Water density plots assessed for both the dimeric (top) and tetrameric (bottom) ORF3a structures. The densities were averaged from frames taken during the production period of the simulations. C Snapshot of ORF3a structure showing two water molecules passing in the opposite direction through the two cis monomers of the tetramer. D Graph showing the z trajectory of water molecules during their passage through the protein. E Average rate of water transport assessed in two replicas for the dimeric structure (dark orange bar) and in three replicas (n = 3 independent simulations) for the tetrameric structure (light green bar). Data are shown as mean ± SEM. Individual data points for each experimental group are reported.

Identification of the relevant residues for the passage of water through ORF3a

The MD simulations presented in Fig. 4 reveal the presence of a preferential route for water passage through the protein. This route is represented by the wide intracellular vestibule, a central route between the TM2 and TM3 helices, and the lipid-protein interface at the outermost part of the membrane. To identify the permeation pore and its sorrounding residues, we used the HOLE program³¹. Figure 5A reports the ORF3a tetramer, where the two dimers of the protein are indicated as dimer 1 and dimer 2. The permeation pores identified using HOLE are color-coded based on their size: blue for a radius wider than 2.8 Å (indicating space for two or more water molecules), green for a radius between 1.4 and 2.8 Å (indicating space for a single water molecule), and red for a radius lower than 1.4 Å (indicating the pore was too narrow for even a water molecule). The estimated pore sizes showed that only the cis monomers allowed the transmembrane passage of water, as evidenced by the absence of red regions along the permeation pore. Conversely, the trans monomers displayed restriction zones at the outer part of the protein, at the interface with the lipid membrane, which hindered the proper transit of water molecules. As shown in Fig. 5B only the pore radius of the cis monomers remained large enough to accomodate the passage of a water molecule (whose radius is indicated by the gray region), while trans monomers of each dimer displayed a narrower pore radius at the transition zone between the intra- and extracellular environments. To identify residues of ORF3a involved in water permeation, we analyzed the trajectories of the permeating water molecules and examined the protein residues in close contact (i.e., with a center-to-center distance of less than 3 Å) with these waters during permeation. Fig. 5C shows a plot of the number of contacts detected for all the protein residues throughout the entire production run. The most prominent residues are shown in different colors, and their positions are indicated in the structure in Fig. 5D. As expected, most of these residues are polar, enabling them to form electrostatic interactions with water. Several permanently charged residues, such as K71, K75, R126 and D142 are located in the large intracellular vestibule, where they likely contribute to the positive electrostatic potential that hinders the entry and permeation of cations (Fig. 1). We found particularly interesting a pair of asparagines (N82 and N119) that are in close proximity to each other, positioned where the intracellular vestibule restricts enough to allow the passage of only one water at a time (i.e., the green part of the pore in Fig. 5A). Given the localization of these asparagines in the transition zone between the internal vestibule and a more restricted passage through the protein, and since at this level the permeation pore coincides with the dimension of a single water molecule, we hypothesized that the two asparagines, along with the nearby histidine (H78) and isoleucine (I123), might constitute a potential selectivity filter for water (Fig. 5D). This notion is supported by the fact that a pair of asparagines are critical for water permeation and selectivity in aquaporins^32,33. To test this hypothesis, we performed MD simulations of a mutated ORF3a tetramer where the asparagine at position 82 was substituted with either leucine or tryptophan, creating the N82L and N82W mutant proteins (Fig. 5E). Our MD simulation results showed that both mutations disrupted the ability of the ORF3a tetramer to transport water (Fig. 5F and Supplementary Table 2). These results were confirmed by water permeability experiments (similar to those shown in Fig. 3C, D) performed in cells transfected either with the wild type truncated ORF3a (tORF3a-WT) or with truncated ORF3a carrying the N82W (tORF3a-N82W) or N82L (tORF3a-N82L) mutations. Consistent with the MD predictions, these experiments demonstrated that both mutations completely abolished the ORF3a ability to mediate the passage of water. In fact, the slope of hypotonic-induced cell swelling was significantly reduced in cells expressing the mutated tORF3a protein (N82L or N82W), with values nearly identical to those observed in cEGFP cells (Fig. 5G).

Fig. 5. MD analysis of ORF3a tetramers and cell volume assessment for water permeability in HEK293 cells expressing ORF3a mutants.

A Representative snapshot of ORF3a tetramer, showing the permeation pore analyzed with the HOLE program, for all four monomers. The permeation pore shows blue regions with a pore radius of ~3 Å, green regions with a pore radius of ~1.5 Å, and red regions with a pore radius of ~1 Å. These regions are wider (blue), comparable to (green), and narrower (red) than the radius of a water molecule (~1.4 Å), respectively. B Plot showing the profile of the pore radius (Å) as a function of the protein z coordinate (Å), with the cis monomer represented by the black line and trans monomer represented by the red line. The gray region indicates a pore radius that allows the passage of water. C Plot showing the number of detections of the indicated residue during water permeation through the protein. The plot was generated by analyzing a total of 543 water molecules in 72 ns, using 1 frame every 0.5 ns from a single simulation (identifier 9, as reported in Supplementary Information, Table S2). D Snapshot of the putative selectivity filter for water permeation, constituted by a pair of asparagines (N82 and N119), a frequently detected histidine (H78) and an isoleucine (I123). E Snapshot of the mutant selectivity filters where the asparagine at position 82 was either replaced with leucine (N82L, upper) or tryptophane (N82W, bottom). F Average number of water molecules passing through the selectivity filter in MD simulations with the WT and mutant structures. G Analysis of the slope of cell volume increase during the first 100 s of recording upon application of 30% hypotonic solution, in HEK293 cells transfected with EGFP (cEGFP, gray circles, n = 11 cells), the wild type tORF3a (tORF3a-WT, blue circles, n = 10 cells) or tORF3a with asparagine at the position 82 substituted either with tryptophane (tORF3a-N82W, green circles, n = 10 cells) or leucine (tORF3a-N82L, red circles, n = 9 cells). cEGFP and tORF3a cells are the same shown in Fig. 3D. Inset: bar graph of the slope. Data are shown as mean ± SEM; *p = 1.26 × 10⁻¹² for tORF3a-WT vs. tORF3a-N82W; *p = 4.84 × 10⁻¹² for tORF3a-WT vs. tORF3a-N82L, as evaluated with two-sample unpaired t-test. Individual data points for each experimental group are reported.

ORF3a-WT, but not ORF3a-N82W transfected cells exhibit enlarged lysosomes

ORF3a localizes to lysosomes, where it induces their increase in diameter and deacidification, leading to inactivation of proteolytic enzymes^9,12. Lysosomal inactivation, together with the inhibition of autophagic pathways, is the prerequisite for virus egress through the lysosomal-mediated exocytotic pathway^8–10,34. Given the strong evidence that ORF3a mediates the passage of water, and considering that lysosomal swelling can increase lysosomal diameter and reduce or suppress lytic power¹², we investigated potential ultrastructural alterations/modifications of lysosomes caused by ORF3a as a water transporter. To this aim, we first examined lysosome morphology through ultrastructural EM analyses in cEGFP cells and in cells transfected with the full-length ORF3a-WT protein. In electron micrographs, lysosomes are tipically identified as spherical organelles, appearing as single membrane-limited, rounded vesicles with diameters ranging from 0.05 to 0.5 µm. These lysosomes may contain a granular, electron-dense material due to the presence of digested content. Conversely, autolysosomes are double-membrane limited organelles of variable shape derived from the fusion of lysosomes with autophagosomes/amphisomes when digestion begins^35,36. To assess whether transfection per se could alter the shape or size of lysosomes and/or cause ultrastructural modifications, we performed EM analysis by comparing native non-transfected and cEGFP-transfected cells (Supplementary Fig. 3A). No significant differences in lysosome size were observed between the two cell groups (Supplementary Fig. 3B), nor were any other alterations such as mitochondrial damage, swelling or intracellular membrane accumulation (see below).

Interestingly, as shown in Fig. 6, in contrast to cEGFP cells, where lysosomes exhibited the standard morphology, i.e., circular vesicles usually containing spots of electron-dense granular material (Fig. 6A and inset), lysosomes in cells transfected with the full-length ORF3a-WT often appeared as enlarged, white/clear circular vesicles, lacking electron-dense material inside (Fig. 6B and inset). EM quantitative analysis fully supported visual observations. Indeed, the average size of lysosomes (area, µm²) found in ORF3a-WT cells (0.28 ± 0.06 µm²) was almost three times greater than that found in cEGFP-transfected cells (0.11 ± 0.01 µm²) (Fig. 6D). We also evaluated the ORF3a-N82W mutant and found that lysosomes were abundantly present and showed the standard morphology (Fig. 6C). Interestingly, in ORF3a-N82W cells, lysosomes were significantly smaller (0.06 ± 0.008 µm²) compared to those observed in cEGFP cells.

Fig. 6. EM analysis of lysosomes in HEK293 cells.

Representative EM images of cEGFP (A), ORF3a-WT (B) and ORF3a-N82W cells (C). Lysosomes were falsely labeled in light blue for easier visualization. N as for nucleus. Scale bars: A–C, 1 µm. Insets: 0.5 µm. D Bar graph showing average size of primary lysosomes (in µm²) in cEGFP (gray bar, n = 27 cells from 3 biologically independent samples), ORF3a-WT (cyan bar, n = 30 cells from 4 biologically independent samples), and ORF3a-N82W (green bar, n = 28 cells from 3 biologically independent samples). Data are shown as mean ± SEM; *p = 0.00004199 for cEGFP vs. ORF3a-WT; *p = 6.37e⁻¹¹ for ORF3a-WT vs. ORF3a-N82W; *p = 0.0194 for cEGFP vs. ORF3a-N82W, as evaluated with Kruskal-Wallis ANOVA test followed by post-hoc Dunn’s test. Individual data points for each experimental group are reported.

In the process of examining lysosomes, we also observed additional intracellular alterations in cells transfected with ORF3a-WT protein. These included large double-membrane structures with materials inside, as well as a striking remodeling of intracellular membranes (likely ER/Golgi) to form a labyrinth of a quite regular pattern resembling a honey-comb structure. In addition, mitochondria appeared structurally abnormal, exhibiting an irregular bean-shaped morphology, and an aberrant internal matrix. Notably, none of these structural alterations were found in either control cEGFP or ORF3a-N82W cells (Supplementary Fig. 4).

Discussion

The present study investigates the possible functional role of the SARS-CoV-2 ORF3a accessory protein. First, structural analysis of the dimer reveals that ORF3a does not possess the typical architecture of ion channels expressed by eukaryotic cells, nor does it have the electrostatic properties required to allow the permeation of cations (i.e., Na⁺, K⁺ and Ca²⁺). Consistent with this, patch-clamp recordings show no additional basal current in cells expressing ORF3a compared to control cEGFP cells, a finding reported by other research groups^7,14. Given that ORF3a could be also gated by mechanical forces acting on the plasma membrane²⁶, we investigated its potential mechanosensitivity by measuring ion currents in native non-transfected cells and in cEGFP, ORF3a, and tORF3a cells exposed to a 30% hypotonic solution, a stimulus known to activate mechanosensitive channels like Piezo1 in eukaryotic cells^23,24,27,37. However, no detectable hypotonic-induced current is observed in any of the groups, strengthening the notion that ORF3a does not function as an ion channel. Our electrophysiological data also rule out the possibility that ORF3a operates as a H⁺ channel, thereby excluding the notion that ORF3a promotes deacidification of lysosomes by directly mediating the efflux of H⁺. Remarkably, our videoimaging results, which assess water permeability by measuring the rate of cell volume increase during osmotic shock, reveal that ORF3a can function as a water transporter. This finding is fully supported by MD simulations of the tetrameric ORF3a structure, which demonstrate the formation of water columns connecting the intra- and extracellular environments, resulting in a flux of water molecules across the membrane at a rate of ~4.1 ± 1.0 molecules*ns⁻¹. Water molecules pass between vestibules through a dedicated permeation pore, mediated by four specific amino acid residues that form a putative selectivity filter, including two asparagines (N82 and N119). Notably, mutation of residue N82, disrupts water transit, excluding the possibility that the increased water permeability in ORF3a-expressing cells results from modulation of endogenous water-permeable channels in the host cell. This strongly supports the idea that ORF3a itself functions as a water transporter. A limitation of these conclusions is that both patch-clamp recordings of ion and H⁺ currents and videoimaging experiments on water permeability were conducted in cells expressing the protein on the plasma membrane. It is well established that ORF3a mainly localizes to the endomembrane system, particularly to the ER and endosomes/lysosomes, where it plays its pathophysiological roles during the virus life cycle. Although ORF3a has also been observed on the plasma membrane, this may result from lysosomal exocytosis, suggesting that ORF3a does not play any functional role in this compartment. Therefore, direct measurements of H⁺ currents and water transport on lysosomes are necessary to provide the required information for establishing the ORF3a function. However, our EM results offer a strong evidence that ORF3a may act as water transporter in lysosomes, as they show that ORF3a expression leads to an increase in lysosomal volume (swelling), an effect that is reverted by a mutation that inhibits water transport. Collectively, these findings reveal a new function of ORF3a as a water-permeable channel, which could be relevant for the SARS-CoV-2 viral cycle.

ORF3a is not an ion channel

To date, the exact function of the ORF3a accessory protein is still unclear and under debate. In recent years, conflicting reports have emerged regarding the ability of ORF3a to function as a viroporin. Seminal studies postulated that ORF3a from both SARS-CoV-1 and SARS-CoV-2, is a highly hydrophobic, virus-encoded protein forming a nonselective cation channel with a preferential permeability for Ca²⁺ ions^6,15–18. Initial studies by Kern et al.⁶ and Toft-Bertelsen et al.¹⁶ showed that ORF3a from SARS-CoV-2 forms a viroporin in Xenopus oocytes and liposome membranes. However, Grant and Lester, using the same experimental preparation, were unable to see any ORF3a-dependent conductance when the protein was expressed in the plasma membrane²⁵. A recent comprehensive study by Clapham’s group, unequivocally demonstrates that ORF3a from both SARS-CoV-1 and SARS-CoV-2 is not an ion channel¹⁴. Using three high-resolution cryo-EM structures of the SARS-CoV-2 ORF3a under different conditions, the authors observed a constriction within the transmembrane region of the protein and the presence of a positively charged aqueous vestibule, both of which would strongly discourage cation permeation. This finding is further confirmed by a recent report from Marquez-Miranda and coworkers, who conducted full-atom MD simulations on the dimeric ORF3a structure from Kern et al.⁶. Their simulations reveal that the polar cavity within the inner half of the putative permeation pathway exhibits high-affinity sites for anions (i.e., Cl^- ions), whereas no high-affinity regions for cations (i.e., K⁺) were found³⁸. Furthermore, Clapham and colleagues were unable to record ORF3a-dependent currents in several mammalian cell types, as well as in other experimental preparations, including Xenopus oocytes, reconstituted systems using purified ORF3a and endo/lysosomes from HEK293 cells¹⁴. The same authors downgraded the ion currents through ORF3a reported by Kern et al.⁶ and by Toft-Bertelsen et al.⁶ to technical artefacts, like an unusually high protein-to-lipid ratio in reconstituted systems⁶ or contamination with endogenous channels¹⁶.

In line with the findings of Clapham and colleagues, we also conclude that ORF3a is not an ion channel. This conclusion is based on the following lines of evidence. First, structural analysis of ORF3a shows that the putative ion permeation pore is too narrow to accommodate a dehydrated cation. Second, the distribution of positively charged amino acid residues at the intracellular entrance of the putative cation permeation pore, would make cation passage thermodynamically unlikely. Third, patch-clamp recordings under control conditions show no ion current from the ORF3a-enriched plasma membrane of HEK293 cells (obtained using the truncated construct)¹⁴. This last result is fully consistent with another study showing no additional currents in cells expressing either the full-length or truncated ORF3a⁷. Moreover, no additional currents are observed when cells are exposed to an extracellular hypotonic solution, which induces cell swelling and membrane stretch/tension, ruling out the possibility of ORF3a having mechanosensitive gating as previously hypothesized²⁶. Therefore, our data, along with the results from Grant and Lester, 2021²⁵ and Miller et al.¹⁴, strongly indicate that ORF3a is not a viroporin.

We note that all studies using HEK293 cells as an experimental model to test whether ORF3a is a viroporin, have failed to record ion currents through ORF3a^7,14, in agreement with our findings. Therefore, the discrepancies in the literature regarding the ORF3a function as a viroporin may arise from differences in experimental preparations (i.e., Xenopus oocytes, artificial nanodiscs, and animal cells), which are challenging to compare.

ORF3a as a water transporter: a possible new function

To clarify the function of ORF3a, we focused on several pieces of evidence previously reported in the literature: i) β-coronaviruses, including SARS-CoV-2, use the lysosomal exocytotic pathway to escape the host cell; ii) the use of lysosomes as exocytotic vesicles requires their deactivation through a deacidification process; iii) the expression of ORF3a in mammalian cells promotes lysosomal deacidification and swelling^8,12. In principle, ORF3a may dissipate the H⁺ gradient across the lysosomal membrane in two ways. The first possibility is that ORF3a acts as an H⁺ channel, allowing direct passage of H⁺ from the lysosomal lumen to the cytosol. Alternatively, ORF3a could function as a water transporter, promoting lysosomal swelling and subsequent alkalinization. These assumptions are supported by previous observations demonstrating that increasing lysosomal membrane permeability to water (which is normally very low) leads to rapid lysosomal swelling²⁸ and deacidification²⁹.

However, our failure to record H⁺ currents, despite of the rigorous experimental conditions set up for this purpose, rejects the hypothesis of ORF3a being a H⁺ channel. Instead, we show that ORF3a mediates water flux across the plasma membrane, as shown by videoimaging experiments. Namely, slope analysis of the initial increase in cell volume following the application of hypotonic stress reveals that the rate of volume change correlates with the tendency of the transfected ORF3a construct (cEGFP vs ORF3a vs tORF3a) to localize to the cell membrane.

These findings are confirmed by MD simulations of the tetrameric structure of ORF3a, where we observed continuous water columns connecting the two aqueous compartments through the TM2 and TM3 transmembrane helices of the cis monomers in each dimer. Interestingly, quantitative analysis of the number of water molecules moving between the two vestibules reveals an average conductance of 4.1 ± 1.0 molecules*ns⁻¹, a value comparable to that observed for aquaporins in animal cells³⁹. Regarding water passage through the protein, we identified a preferential route and a putative selectivity filter formed by four amino acid residues located in the transition zone between the intra- and extracellular environments. Notably, mutations of one of these residues—asparagine 82 (N82)—with tryptophan (N82W) or leucine (N82L) disrupt the ORF3a ability to transport water, as confirmed by both MD simulation analyses and videoimaging experiments.

An intriguing observation from our MD simulations is that the water conductance occurs only for the tetrameric structure, but not for the dimeric one. As it has been postulated that ORF3a can exist both as a homodimer and a homotetramer⁶ and since it remains unclear which of the two structures is functionally relevant, our simulations could help clarify this aspect. However, a recent study reported that in mammalian cells, ORF3a is primarily present at the plasma membrane as a dimer or monomer, with the monomer being the predominant form⁷. Another key point to emphasize is that our simulations show water conductance through the tetramer only when the protein is embedded in a phospholipid bilayer made of short fatty acid alkyl chains (i.e., 12 carbon atoms, DDPC:DLPA 3:1). A potential limitation of this result is that alkyl chains with 12 carbon atoms are not typically present in most biological membranes, which primarily consist of fatty acid alkyl chains ranging from 14 to 24 carbon atoms⁴⁰. Further studies will be required to explore these points in greater detail.

Finally, the positive electrostatic potential observed at the intracellular vestibule of ORF3a might possibly contribute to its selectivity against H⁺, as demonstrated in our patch-clamp experiments. Indeed, this mechanism has been convincingly shown in aquaporin water channels, where a highly conserved arginine along the water permeation pathway generates a positive electrostatic potential that plays a crucial role in destabilizing positively charged hydronium ions, thus hampering H⁺ conductance⁴¹. The positive electrostatic potential may also help exclude small organic osmolytes, as pointed out by Oliva and colleagues⁴².

Possible physio-pathological implications of ORF3a-mediated water transport

Generally, the virus lifecycle comprises different major stages, ranging from virus attachment and entry into the host cell to its release. In the case of SARS-CoV-2, while the early stages of the infectious cycle are relatively well understood, the mechanisms underlying the later stages (i.e., virus egress) are instead less clear. A recent paper by Ghosh and colleagues reveals that the egress process of β-coronaviruses, including SARS-CoV-2, relies on a lysosomal, Arl8b-dependent exocytosis pathway and requires lysosomal inactivation through a deacidification (alkalinization) process⁸. In SARS-CoV-2 lysosomal alkalinization, which is essential for their inactivation and subsequent fusion with the plasma membrane, is known to be mediated by ORF3a^9,10,12. However, the mechanism(s) through which ORF3a promotes lysosome deacidification and exocytosis remains debated and poorly understood.

Our electrophysiological recordings discard the possibility that ORF3a directly mediates H⁺ extrusion, as no H⁺ currents are obsereved in HEK293 cells expressing truncated ORF3a. Rather, both videoimaging experiments and MD simulations suggest that ORF3a mediates the passage of water across biological membranes. This rises the question of how a water-permeable channel can dissipate a H⁺ gradient and deacidify a lysosome. Lysosomes are intracellular organelles with inherently low basal water permeability, as indicated by a study showing that the introduction of synthetic water channels to lysosomal membrane significantly increases osmotic tension in lysosomes²⁸, and by the finding that VRAC-mediated water flux, triggered by hypotonic conditions, leads to lysosomal swelling^43,44. Our EM data, which show that cells transfected with ORF3a exhibit enlarged lysosomes, suggest that ORF3a expression may promote an osmotic flux of water into the hyperosmotic lysosomal lumen, causing lysosomal swelling. In smooth muscle cells, Busch and colleagues reported that hypotonic-induced lysosome swelling leads to lysosomal alkalinization²⁹. Although not fully proven experimentally, the authors proposed that the rapid increase in lysosomal pH following osmotic shock, may result from increased H⁺ leakage across lysosomal membrane or from lysosomal fusion with other organelles with a higher pH²⁹. This idea supports our hypothesis that increased ORF3a-dependent lysosomal water permeability leads to lysosomal swelling and deacidification.

Another potential role of ORF3a-mediated water transport in the SARS-CoV-2 lifecycle may involve exocytosis-driven virus egress. Lysosomes express Ca²⁺ permeable TRPML2 channels, which can be activated by lysosomal swelling⁴⁵. Thus, TRPML2-dependent Ca²⁺ release from lysosomes may contribute to their fusion with the plasma membrane (i.e., lysosomal exocytosis) in ORF3a-expressing cells, with ORF3a facilitating the Ca²⁺-dependent process. Increased lysosome fusion with the plasma membrane during osmotic shock-induced swelling has been well-documented^43,44.

Conclusion

This study confirms that ORF3a is not a cation channel (a viroporin) and, for the first time, shows that its tetrameric assembly functions as a water-permeable channel. This new feature may offer insights into how ORF3a promotes deacidification and inactivation of lysosomes, crucial steps for virus egress through the lysosomal exocytotic pathway. Understanding the molecular mechanisms by which ORF3a promotes viral egress could lead to new antiviral therapies.

Author contributions

A.M. and L.C. conceived and designed the study and wrote the manuscript. A.M. and L.S. performed all the electrophysiological and videoimaging experiments, analyzed the results, and prepared the figures. R.F., M.V.L., and A.DiB. performed cell culture preparations and transfections and obtained immunofluorescence images; L.C. performed MD simulations and analysis; G.R. prepared samples for EM analysis; S.V. sorted EGFP positive cells used for EM preparations and edited the manuscript; S.B. performed EM experiments, analyzed results, prepared figures, and edited the manuscript; M.DiC. designed all the plasmids used in the study, supervised all molecular biology preparations, and edited the manuscript.

Peer review

Peer review information

Communications Biology thanks Omid Bavi, Han-Gang Yu and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. Primary Handling Editors: Laura Rodríguez Pérez and Ophelia Bu. [A peer review file is available].

Data availability

All relevant information supporting the findings of this study are presented in the manuscript and in Supplementary Information. Source data underlying graphs and structure and topology files resulting from MD simulations can be found in Supplementary Data. Additional supporting data is available from the corresponding authors upon request. Plasmids used in this study are deposited in the Addgene repository. Plasmid ID numbers: pCDNA3sars-cov2-ORF3A-EGFP 7021 bp (Plasmid #231697); pMDC-656 (pCDNA3sars-cov2-myc-41-ORF3A-EGFP) 6931 bp (Plasmid #231698); pMDC-722 (pCDNA3sars-cov2-ORF3AmutN-W-EGFP) 7021 bp (Plasmid #231699); pMDC-726 (pCDNA3sars-cov2-c-myc-41-ORF3AmutN-W-EGFP) 6931 bp (Plasmid #231700); pMDC-727 (pCDNA3sars-cov2-c-myc-41ORF3AmutN-L-EGFP) 6931 bp (Plasmid #231703).

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s42003-024-07442-5.