Introduction

Ion channels and phospholipid scramblases catalyze the passive flux of ions and phospholipids down their respective chemical gradients. Compared to ion channels, both our understanding of scramblases and the evolutionary origins of phospholipid scrambling are underdeveloped. Discoveries of two scramblase families—the TMEM16 calcium-activated phospholipid scramblases (CaPLSases) and XKR caspase-dependent phospholipid scramblases—have revealed roles for scramblases in processes such as blood coagulation¹, cell death signaling², phagocytosis³, cell migration⁴, membrane repair^5,6, microparticle release⁷, cell-cell fusion^8–10, and viral infection^11,12. Among identified scramblases, TMEM16 CaPLSases are the most extensively studied at the molecular level¹³. The TMEM16 family (Extended Data Fig. 1a) was originally classified as calcium-activated chloride channels (CaCCs) based on the first-discovered members, TMEM16A and TMEM16B^14–16. The remaining members, however, exhibit CaPLSase activity, with TMEM16F representing the canonical CaPLSase¹⁷ (Fig. 1a). Uniquely, CaPLSases also function as non-selective ion channels along with their non-specific permeability to phospholipids. Substrate permeation is facilitated by calcium-induced conformational changes causing the clamshell-like separation of TMs 4 and 6 to facilitate ion and phospholipid permeation^18–20 (Extended Data Fig. 1b-e). These conformational changes further catalyzed phospholipid permeation by thinning the membrane near the permeation pathway^21,22.

Figure 1:

The TMEM16 family together with the TMEM63/OSCA family and TMC family^23–26 collectively form the Transmembrane Channel/Scramblase (TCS) superfamily¹³. Despite the conserved 10-TM architectural core^27–31, phospholipid permeability has not been experimentally demonstrated in mechanosensitive TMEM63/OSCA or TMC proteins, raising the intriguing question of how TCS proteins discriminate between ion and ion/phospholipid substrates. Interestingly, studies have shown that the TMEM16A CaCC can be genetically modified to enable phospholipid permeability, either by substituting TMEM16F domains as small as 15 amino acids³², or through a single lysine mutation in TM 4 at the hydrophobic gate¹⁸. These findings suggest a modest energetic barrier for scramblase activity and lead us to hypothesize that evolutionary relatives of the TMEM16 family may maintain a conserved potential for both ion and phospholipid permeation.

To test this hypothesis, we introduced single lysine mutations along TM 4 of TMEM16F, TMEM16A, OSCA1.2, and TMEM63A at sites near the corresponding hydrophobic gate of TMEM16F¹⁸. TM 4 lysine mutations in TMEM16F and TMEM16A resulted in constitutive phospholipid scramblase activity. Strikingly, the equivalent TM 4 lysine mutations of OSCA1.2 and TMEM63A also converted these ion channels into phospholipid scramblases, which were either constitutively open or further activated by osmotic stimulation. All mutants also resulted in commensurate gain-of-function ion channel activity, suggesting that the charge in TM 4 disrupts the activation gate of these evolutionarily related proteins. Together, our findings advance the mechanistic understanding of gating and substrate permeation in the TMEM16 and OSCA/TMEM63 families, underscore a key design principle for phospholipid scramblases, and support our hypothesis that the TCS superfamily maintains a conserved potential for both ion and phospholipid permeation.

Results

TM 4 lysine mutations result in a constitutively active TMEM16F scramblase

Our previous study identified the activation gate of TMEM16F, consisting of hydrophobic residues from TMs 4, 5, and 6, which collectively govern ion and phospholipid permeation¹⁸. Introducing polar or charged residues to the activation gate results in gain-of-function (GOF) CaPLSases, some of which are constitutively activated without requiring calcium. Structural and computational studies on the fungal NhTMEM16³³ and AfTMEM16²¹ orthologs, as well as human TMEM16K³⁴ further support a clamshell like gating model¹⁸ whereby scrambling activity is promoted by separation of the TM 4/6 interface¹³ (Extended Data Fig. 1b-d). The conformational transition involves the N-terminal half of TM 4, which bends away from TM 6, and the C-terminal half of TM 6, which collapses onto the calcium binding sites formed together with TMs 7 and 8¹³. Given that the N-termini of TM 4 are largely hydrophobic among TMEM16 CaPLSases (Extended Data Fig. 1f) and a lysine mutation at F518 leads to a constitutively open TMEM16F activation gate¹⁸, we hypothesized that introducing a charged side chain along the TM 4/6 interface might disrupt the hydrophobic gate and result in TMEM16 scramblases with GOF scramblase activity. To test this hypothesis, we overexpressed eGFP-tagged mutant constructs in TMEM16F knockout (KO) HEK293T cells and used an established scramblase assay^18,35,36 that detects phosphatidylserine (PS) exposure using fluorophore-tagged Annexin V (AnV) as a reporter (Fig. 1b). In the absence of calcium stimulation, PS is predominantly in the inner leaflet of the plasma membrane of the TMEM16F wildtype (WT) expressing cells and is therefore not labelled by extracellular AnV. Similar to TMEM16F F518K¹⁸, overexpressing the single lysine mutations TMEM16F I521K and M522K led to spontaneous, global exposure of PS on the plasma membrane without requiring calcium stimulation (Fig. 1b, c). We further tested whether I521K and M522K have GOF ion channel activity using whole-cell patch clamp. In the absence of calcium, the mutants, but not WT TMEM16F, showed robust depolarization-activated outward rectifying currents (Fig. 1d-f). Our imaging and patch clamp experiments thus demonstrate that I521K and M522K are functionally expressed on the plasma membrane and their GOF lipid scramblase activity results in spontaneous, global PS exposure on the cell surface. I521 and M522 are approximately one helical turn below the hydrophobic gate residue, F518,¹⁸ (Fig. 1a) supporting the idea that a lysine mutation on TM 4 along the TM 4/6 interface promotes gate opening and substrate permeation (Fig. 1g).

TM 4 lysine mutations convert TMEM16A into a constitutively active scramblase

TMEM16A is a CaCC^14–16 without scramblase activity (Fig. 2a-c)^18,32. We previously reported that a single TM 4 lysine mutation (L543K) at the hydrophobic gate of TMEM16A (Fig. 2a) allows the CaCC to permeate phospholipids spontaneously, analogous to the F518K mutation in TMEM16F¹⁸. Thus, we reasoned that lysine mutations equivalent to TMEM16F I521K and M522K might also result in spontaneous phospholipid permeability. To test this hypothesis, we overexpressed eGFP-tagged TMEM16A I546K and I547K mutants in TMEM16F KO HEK293T cells, and assessed their ability to expose PS on the plasma membrane with confocal microscopy. Similar to TMEM16A L543K¹⁸, I546K- and I547K-expressing cells exhibit spontaneous, global PS exposure (Fig. 2b, c), analogous to the equivalent TMEM16F mutations I521K and M522K (Extended Data Fig. 1f; Fig. 1b, c). Similarly, whole-cell patch clamp revealed GOF ion channel activity at depolarizing potentials, even in the absence of calcium stimulation (Fig. 2d-f). Together, these results suggest that lysine mutations in this region strongly destabilize the hydrophobic gate to endow the TMEM16A CaCC with constitutive phospholipid permeability and GOF ion channel activity (Fig. 2g).

Figure 2:

TM 4 lysine mutation L438K converts OSCA1.2 into a constitutively active scramblase

We next applied our approach to a TMEM16 relative from the OSCA/TMEM63 family (Extended Data Fig. 1a), which was first discovered in plants as a family of mechanosensitive and osmolarity-activated cation non-selective ion channels³⁷ (Fig. 3a). We hypothesized that analogous lysine mutations on TM 4 along the TM 4/6 interface of OSCA/TMEM63 proteins would result in GOF channels that may also permeate phospholipids. Within the family, we chose OSCA1.2 from Arabidopsis thaliana due to previous structural²⁷ and biophysical³⁸ characterization demonstrating that the channel is activated directly by membrane tension. Overexpressing eGFP-tagged OSCA1.2 WT in TMEM16F KO HEK293T cells did not induce PS exposure, demonstrating that OSCA1.2 WT lacks spontaneous lipid scrambling activity (Fig. 3b). We then introduced TM 4 lysine mutations near the putative gating residues identified by structural (Fig. 3a) and sequence alignment (Extended Data Fig. 1e). Strikingly, a single point mutation, L438K, causes cells overexpressing the OSCA1.2 mutant to exhibit spontaneous and global PS exposure (Fig. 3b, c). This mirrors our results with both TMEM16F (Fig. 1) and TMEM16A (Fig. 2), suggesting that the L438K mutation allows the OSCA1.2 channel to permeate phospholipids. Next, we used inside-out patch clamp to examine if L438K also enhances OSCA1.2 channel activity. We found that the mutant significantly left-shifts the conductance-voltage (G-V) relationship (Fig. 3e, f) and accelerates channel activation kinetics (Fig. 3g) compared to WT. Under −50 mmHg of pressure, L438K has half-maximal voltages (V_0.5) of 66.7±3.7 mV, while the WT V_0.5 is nearly 108.7±5.6 mV (Fig. 3f), underscoring that this mutation also disrupts channel gating. Together, these experiments show that, like TMEM16A, a single lysine mutation near the putative gate allows the mechanosensitive and osmolarity-activated OSCA1.2 channel to become spontaneously permeable to phospholipids (Fig. 3h).

Figure 3:

TM 4 lysine mutation A439K converts OSCA1.2 into an osmolarity-sensing scramblase

Interestingly, mutating the neighboring amino acid (Fig. 3a), A439, to lysine resulted in minimal spontaneous PS exposure (Fig. 3b, c). We thus reasoned that A439K scramblase activity may require additional stimulation. Given that OSCA1.2 is an osmolarity-sensitive ion channel³⁸ and reduction of extracellular osmolarity does not induce PS exposure in OSCA1.2 WT expressing cells (Fig. 4a, c), we acutely treated A439K-expressing cells with a hypotonic solution (120 mOsm/kg). Indeed, PS exposure was robustly induced (Fig. 4b, c). Inside-out patch clamp further demonstrated that A439K enhances OSCA1.2 ion channel activity as evidenced by the accelerated activation kinetics (Fig. 3g) and left-shifted G-V relationship (Fig. 3f). Our experiments thus indicate that A439K disrupts OSCA1.2 gating and converts the osmolarity-activated ion channel into an osmolarity-activated phospholipid scramblase (Fig. 4d).

Figure 4:

TM 4 lysine mutations convert TMEM63A into a constitutively active scramblase

We next turned our attention to TMEM63A to further investigate the evolutionary conservation of our observation that TM 4 lysine mutations convert TMEM16 and OSCA members into scramblases. TMEM63s represent the mammalian members of the OSCA/TMEM63 family with three members present in humans (TMEM63A-C). Recent structural and functional characterizations indicate that TMEM63s function as mechanosensitive ion channels gated by high-threshold membrane tension and, in notable contrast to all other structurally resolved TCS members, they likely function as monomers^39–41. Given their structural homology to TMEM16s and OSCA1.2³⁹, we hypothesized that TM 4 mutations in TMEM63A would also result in GOF activity. We again identified residues near the putative gate by structural (Fig. 5a) and sequence alignment (Extended Data Fig. 1f), selecting W472 (equivalent to F518 in TMEM16F and L543 in TMEM16A) and S475 (equivalent to L438 in OSCA1.2). Indeed, overexpressing eGFP-tagged mouse TMEM63A with single lysine mutations at either W472 and S475 led to spontaneous PS exposure (Fig. 5b, c), though the AnV staining revealed punctate rather than global patterns of PS exposure. To confirm membrane localization and further probe mutant effects, we exploited TMEM63A ion channel function using cell-attached patch clamp. As TMEM63A exhibits voltage-dependent activity under high pressures, we compared mutant and WT I-V relationships at −80 mmHg. W472K and S475K caused marked reductions in V_0.5 from 122.3±3.5 mV for WT to 95.9±6.3 mV and 92.1±8.6 mV, respectively (Fig. 5d-f). Together, these results are consistent with our observations in TMEM16F, TMEM16A, and OSCA1.2, indicating that single lysine mutations along TM 4 of TMEM63A can facilitate both ion and phospholipid permeation (Fig. 5g).

Figure 5:

Discussion

Mechanistically, our study improves models of TMEM16 substrate permeation and gating. We identified multiple lysine mutations in TMEM16F and TMEM16A that promote phospholipid permeation and cause commensurate changes in ion channel activities. Our findings complement functional characterizations of TMEM16F gating mutants where gate destabilization is inversely correlated with side chain hydropathy^18,42. Lysine is above only arginine on the hydropathy index and thus likely explains why it readily destabilizes the gate. For instance, F518K exhibits spontaneous PS exposure, even when the calcium binding site is destroyed¹⁸, whereas F518H does not exhibit spontaneous PS exposure⁴². Interestingly, the recent TMEM16F F518H structure (Extended Data Fig. 1d) shows local membrane thinning due in part to unexpected conformational changes in TM 3⁴². Future structural studies are needed to assess whether TMEM16A or OSCA/TMEM63 mutant scramblases also promote membrane thinning and/or conformational rearrangements in TM 3. More broadly, our results highlight an increasingly appreciated design principle of scramblases where polar and charged residues often line a membrane-spanning groove. This observation has been noted for TMEM16^43,34,44, Xkr^45,46, and opsin⁴⁷ scramblases and should be a key criterion for identifying and characterizing new scramblases.

Our findings also advance our understanding of evolutionary relatives of the TCS superfamily¹³ and help uncover their roles in human diseases. Although OSCA/TMEM63 proteins are not known to scramble phospholipids, we show that TM 4 lysine mutations in OSCA1.2 and TMEM63A convert these osmolarity-activated and mechanosensitive ion channels into phospholipid scramblases, analogous to our findings with the equivalent TMEM16A mutations. We thus speculate that the conserved structural architecture in the transmembrane region endows TCS proteins with a potential to scramble phospholipids, though this capability may have been lost during evolution. It will be interesting to test whether equivalent TM 4 lysine mutations can convert transmembrane channel-like (TMC) proteins— the third TCS relative of TMEM16 and OSCA/TMEM63¹³—into phospholipid scramblases. TMC1 and TMC2 are best known for their roles in auditory sensation, and thus far have mostly been characterized in vivo due to expression difficulties in heterologous systems⁴⁸. However, recent in vivo characterization of mouse TMC1 M412K, known as the Beethoven mutation, provided an important insight⁴⁹. The deafness-associated mutation is located in TM 4 (Extended Data Fig. 1f; equivalent of M522 in TMEM16F, I547 in TMEM16A, and A439 in OSCA1.2) and results in constitutive PS exposure when expressed in the hair cell membranes of both heterozygous and homozygous mice⁴⁹. This raises the intriguing possibility that the Beethoven mutation may enable TMC1 to spontaneously permeate phospholipids, leading to a loss of membrane homeostasis and ultimately, auditory sensation⁵⁰. The possibility of converting TMC proteins into phospholipid scramblases should be thoroughly investigated. Additionally, disease-associated mutations in TMEM63 proteins are present along the TM 4/6 interface, such as TMEM63B T481N⁴⁹. We speculate that introducing more hydropathic side chains along this interface may lead to spontaneous ion and or phospholipid permeability, perhaps contributing to underlying pathophysiology.

Methods

Cloning and mutagenesis

All constructs use a peGFP-N1 vector backbone. Wild type sequence and mutation numbers correspond to NCBI: NP_780553.2 (Mus musculus TMEM16F) with a three amino acid (MQM) N-terminal truncation, NCBI: NP_001229278 (Mus musculus TMEM16A), GenBank: AIU34614.1 (Arabidopsis thaliana OSCA1.2), and NCBI: NP_001404481.1 (Mus musculus TMEM63A). AtOSCA1.2 and MmTMEM63A cDNAs were subcloned using In-Fusion Snap Assembly (Takara, #638947). Point mutants were generated by PCR site-directed mutagenesis with primers from IDT DNA Technologies. Sequences were confirmed by Sanger sequencing (Azenta).

Bioinformatics

The following sequences were obtained from UniProt and aligned using Clustal Omega: Q8BHY3-2 (MmTMEM16A), Q9NQ90 (HsTMEM16B), Q9BYT9 (HsTMEM16C), Q32M45 (HsTMEM16D), Q75V66 (HsTMEM16E), Q6P9J9 (MmTMEM16F), Q6IWH7 (HsTMEM16G), Q9HCE9 (HsTMEM16H), A1A5B4 (HsTMEM16J), Q9NW15 (HsTMEM16K), C7Z7K1 (NhTMEM16), Q4WA18 (AfTMEM16), Q9XEA1 (AtOSCA1.1), Q5XEZ5 (AtOSC1.2), B5TYT3 (AtOSCA1.3), A0A097NUQ0 (AtOSCA1.4), A0A097NUS0 (AtOSCA1.5), A0A097NUP1 (AtOSCA1.6), A0A097NUP8 (AtOSCA1.7), A0A097NUQ2 (AtOSCA1.8), A0A097NUQ5 (AtOSCA2.1), A0A097NUS5 (AtOSCA2.2), A0A097NUP6 (AtOSCA2.3), A0A097NUQ3 (AtOSCA2.4), A0A097NUQ7 (AtOSCA2.5), Q9C8G5 (AtOSCA3.1), A0A097NUT0 (AtOSCA4.1), Q91YT8 (MmTMEM63A), Q5T3F8 (HsTMEM63B), and Q9P1W3 (HsTMEM63C), Q8R4P5 (MmTMC1), Q8R4P4 (MmTMC2), Q7TQ69 (MmTMC3), Q7TQ65 (MmTMC4), Q32NZ6 (MmTMC5), Q7TN60 (MmTMC6), Q8C428 (MmTMC7), and Q7TN58 (MmTMC8). A subset of the alignment was selected for Extended Data Fig. 1. Structural models were obtained from the PDB, aligned, and visualized using Pymol (Schrödinger).

Cell culture

The HEK293T cell line was authenticated by the Duke Cell Culture Facility. The TMEM16F KO HEK293T cell line was generated by the Duke Functional Genomics Core and characterized in previous studies^18,35. All cells were cultured with DMEM (Gibco, #11995-065) supplemented with 10% fetal bovine serum (FBS) (Sigma-Aldrich, # F2442) and 1% penicillin/streptomycin (Gibco, #15-140-122) at 37°C in 5% CO₂-95% air.

Transfection

Plasmids were transiently transfected into TMEM16F KO HEK 293T cells by using X-tremeGENE9 or X-tremeGENE360 transfection reagents (MilliporeSigma). Media was replaced with calcium free DMEM (Gibco, 21068-028) 3-4 hours after transfection. The transfected cells were imaged or patched 18-24 or 18-48 hours after transfection, respectively.

Fluorescence imaging of scramblase-mediated PS Exposure

A Zeiss 780 inverted confocal microscope was used to monitor scramblase activity in live cells using the methods described in previous publications^9,18,35. Briefly, 18-24 hours after transfection, the cells were incubated in AnV buffer (1:175 dilution of the fluorescently tagged AnV (Biotium) in Hank’s balanced salt solution) before imaging. Spontaneous PS positive cells were labeled by fluorescently tagged AnV and imaged. Results were quantified as a percentage of PS positive cells among all GFP positive cells. For osmolarity activation, 2 mM CaCl₂ in ddH₂O was added to the AnV buffer at a 2:1 ratio. The final osmolarity was ~120 mOsm/kg as measured by a micro-osmometer (Advanced Instrument). Cells overexpressing WT or mutant OSCA1.2 were treated with low osmolarity AnV buffer, and the scramblase activity was measured by recording fluorescent AnV surface accumulation at 5 second intervals. A custom MATLAB code was used to quantify AnV signal and is available at Github (https://github.com/yanghuanghe/scrambling_activity).

Electrophysiology

All electrophysiology recordings were conducted using an Axopatch 200B amplifier with the signal digitally sampled at 10 kHz using an Axon Digidata 1550A (Molecular Devices, Inc.). All electrophysiology recordings were carried out at room temperature 18-48 hours after transfection. Glass pipettes were pulled from borosilicate capillaries (Sutter Instruments) and fire-polished using a microforge (Narishige). Pipettes had resistances of 2–3 MΩ in bath solution.

Inside-out patch clamp recordings

The pipette solution (external) contained 140 mM NaCl, 10 mM HEPES, 2 mM MgCl₂, adjusted to pH 7.3 (NaOH), and the bath solution (internal) contained 140 mM NaCl, 10 mM HEPES, 5 mM EGTA, adjusted to pH 7.3 (NaOH). OSCA1.2 WT and mutations were held at constant pressure administrated using a syringe calibrated with a manometer, similar to a previous study⁵⁴. Patches were held at a membrane potential of −60 mV and at the indicated pressure, then stimulated using the indicated voltage protocol.

Cell-attached patch clamp recordings

For TMEM63 recordings, cell-attach mode was used. The bath solution contained (in mM): 140 KCl, 10 HEPES, 2 MgCl₂, 10 glucose, pH 7.3 adjusted with KOH. The pipette solution contained (in mM): 130 NaCl, 5 KCl, 10 HEPES, 10 TEA-Cl, 1 CaCl₂, 1 MgCl₂, pH 7.3 (with NaOH). The mechano-activated current was evoked with a 200 ms pressure pulse at −80 mmHg using a high-speed pressure clamp system (HSPC-1, ALA Scientific Instruments, Farmingdale, NY). The membrane potential inside the patch was held at −60 mV. The voltage pulse alone was run first followed by voltage pulse with pressure. The mechanosensitive current was obtained by subtracting the voltage pulse from the voltage with pressure pulse.

Whole cell patch clamp recordings

The pipette solution (internal) contained 140 mM CsCl, 10 mM HEPES, 5 mM EGTA, adjusted to pH 7.3 (CsOH), and the bath solution (external) contained 140 mM NaCl, 10 mM HEPES, 5 mM EGTA, adjusted to pH 7.3 (NaOH). Patches were held at a membrane potential of −60 mV, then stimulated using the indicated voltage protocol.

Data analysis for electrophysiology

All data analysis was performed using Clampfit (Molecular Devices), Excel (Microsoft), MATLAB (MathWorks), and Prism softwares (GraphPad). Individual G-V curves were fitted with a Boltzmann function,

where G_max denotes the fitted value for maximal conductance, V_0.5 denotes the voltage of half maximal activation of conductance, Z denotes the net charge moved across the membrane during the transition from the closed to the open state, and F denotes the Faraday constant.

Acknowledgements

This work was supported by the National Institute of Health (DP2GM126898 and R21GM146152 to H.Y.) and the National Science Foundation Graduate Research Fellowship Program (DGE 2139754 to A.J.L.). Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the National Science Foundation.

Author contributions

H.Y. and Y.Z. conceived and H.Y. supervised the project. Y.Z. and A.J.L. imaging. P.L. electrophysiology. A.J.L. sequence and structure alignments. A.J.L. and Y.C.S.W. cloning and mutagenesis. Z.P. plasmids. Y.Z. MATLAB codes. A.J.L. and H.Y. drafted the manuscript with input from all authors.

Competing interests

The authors declare no competing interests.