An inner activation gate controls TMEM16F phospholipid scrambling

Transmembrane protein 16F (TMEM16F) is an enigmatic Ca2+-activated phospholipid scramblase (CaPLSase) that passively transports phospholipids down their chemical gradients and mediates blood coagulation, bone development and viral infection. Despite recent advances in the structure and function understanding of TMEM16 proteins, how mammalian TMEM16 CaPLSases open and close, or gate their phospholipid permeation pathways remains unclear. Here we identify an inner activation gate, which is established by three hydrophobic residues, F518, Y563 and I612, in the middle of the phospholipid permeation pathway of TMEM16F-CaPLSase. Disrupting the inner gate profoundly alters TMEM16F phospholipid permeation. Lysine substitutions of F518 and Y563 even lead to constitutively active CaPLSases that bypass Ca2+-dependent activation. Strikingly, an analogous lysine mutation to TMEM16F-F518 in TMEM16A (L543K) is sufficient to confer CaPLSase activity to the Ca2+-activated Cl− channel (CaCC). The identification of an inner activation gate can help elucidate the gating and permeation mechanism of TMEM16 CaPLSases and channels.

C ell surface exposure of phosphatidylserine (PS), a major phospholipid species that resides exclusively in the inner leaflet of the plasma membrane, serves as an important physiological signal that triggers various cellular processes, including phagocytosis, blood coagulation, bone mineralization, cell-cell fusion, and viral infection 1 . The CaPLSases from the TMEM16/ANO family are major phospholipid transporters that catalyze PS externalization in response to intracellular Ca 2+ elevation [1][2][3][4][5][6][7][8] . TMEM16F, one of the most well characterized CaPL-Sases, has been shown to be involved in blood coagulation 2,9 , bone development 10 , and HIV infection 11 . Loss-of-function mutations in TMEM16F cause Scott syndrome, a hemostatic disorder due to deficient PS externalization in platelets 2,12 . In addition, single-nucleotide polymorphisms (SNPs) of TMEM16F/ ANO6 are associated with the etiology of ankylosing spondylitis 13 . Besides TMEM16F, mutations in TMEM16E, another CaPLSase of the TMEM16 family, have also been implicated in a number of inherited diseases including gnathodiaphyseal dysplasia 14 , limbgirdle muscular dystrophy 7,[15][16][17] , and Miyoshi myopathy [15][16][17] . Considering their importance in health and disease, a comprehensive understanding of the structure and function of the mammalian TMEM16 CaPLSases would facilitate drug discovery for these therapeutic targets.
An X-ray structure of a TMEM16 homolog from the fungus Nectria haematococca (nhTMEM16) captured in an activated Ca 2+ -bound state illuminates a homodimeric assembly of TMEM16 with each monomer containing ten transmembrane (TM) helices 4 (Fig. 1a). Recent structural studies on TMEM16A-CaCC 18,19 and TMEM16F-CaPLase 20 further indicate that mammalian TMEM16 proteins also adopt a double-barreled dimeric architecture. Ca 2+ binding to a highly conserved Ca 2+ -binding site in TM helices 6-8 (refs. 4,21,22 ) triggers ion or phospholipid permeation through a hydrophilic pathway/groove comprising 20,[23][24][25]. Interestingly, the hydrophilic groove in the Ca 2+ -bound nhTMEM16 structure is exposed to the lipid environment owning to the physical separation of TM4 and TM6, both of which are located at the periphery of the protein. This intricate architecture thus suggests that phospholipid headgroups can enter and move along the permeation pathway while maintaining their hydrophobic acyl tails in the hydrocarbon core of the membrane 4,26-28 , consistent with the widely accepted credit-card reader model for phospholipid permeation 29 (Fig. 1b). Nevertheless, the credit-card reader model does not implicate a gating control mechanism, which is required to regulate passive phospholipid permeation through TMEM16 CaPLSases in response to Ca 2+ binding.
Recent computational and functional studies guided by the Ca 2+ -bound nhTMEM16 structure have proposed three critical sites for controlling phospholipid scrambling [26][27][28] : an extracellular S E site formed by charged residues from TM3 and TM6, a central constriction site in the middle of the hydrophilic groove and a cytosolic S C site composed of charged residues in TM4 (Fig. 1a). Although the exact roles of the S C site and the central constriction site in nhTMEM16 gating are unclear, molecular dynamic (MD) simulations and mutagenesis studies have suggested that the network of charged amino acids within the S E site (Fig. 1a) can serve as a putative extracellular gate to control phospholipid permeation 26,28 . In this model, Ca 2+ binding to nhTMEM16 triggers rearrangements of these charged amino acids at the S E gate allowing the hydrophilic pathway to accommodate and permeate phospholipids. While the nhTMEM16 structure has provided an invaluable structural template for computational and functional studies to unravel the molecular mechanism of phospholipid scrambling and regulation 4,26-28 , lack of structural information of the mammalian TMEM16 scramblases has precluded a comprehensive mechanistic understanding of phospholipid scrambling, and casted questions on how mammalian TMEM16 CaPLSases gate their phospholipid permeation pathways in a tightly controlled fashion.
Combining structure-guided mutagenesis, quantitative measurements of CaPLSase activity, and atomistic MD simulations, we identify F518, Y563, and I612 residues, which are located in the middle of the TMEM16F phospholipid permeation pathway, form an inner activation gate to control phospholipid translocation. Amino acid substitutions at these inner gate residues profoundly affect TMEM16F scrambling activity. Remarkably, we find that F518K and Y563K mutations lead to constitutively active CaPLSases independent of Ca 2+ activation. Moreover, TMEM16A-L543K, a corresponding mutation of TMEM16F-F518K, confers the CaCC the capability to constitutively permeate phospholipids. Our identification of an inner activation gate in controlling TMEM16F-CaPLSase phospholipid permeation thus provides an important framework for the mechanistic understanding of the gating mechanism of mammalian TMEM16 CaPLSases.

Results
Homology models of TMEM16F predict an inner activation gate. In order to study TMEM16F-CaPLSase gating, a closed state structural model is necessary. Given the high sequence conservation between mammalian TMEM16F and TMEM16A, and low sequence conservation between mammalian TMEM16F and fungal nhTMEM16 ( Supplementary Fig. 1a), we hypothesized that the Ca 2+ -free structure of a murine TMEM16A (PDB: 5OYG) 23 would be a more suitable template for modeling the closed state of TMEM16F ( Supplementary Fig. 1b). The Ca 2+ -free TMEM16A-derived homology model of TMEM16F clearly stays in a closed conformation with its hydrophilic groove constricted, which results in high free energy barriers for both water and phosphate permeation during atomistic simulations (Fig. 1d).
To build an open state TMEM16F homology model, we first attempted to build a structural model (Fig. 1c) based on the Ca 2+ -bound structure of TMEM16A (PDB: 5OYB) 23 , which shows that Ca 2+ binding straightens TM6 around the glycine hinge to trigger TMEM16A activation ( Supplementary Fig. 1b). Due to this Ca 2+ -induced conformational change, the pore in the Ca 2+ -bound TMEM16F model becomes more accessible to water permeation in comparison to the Ca 2+ -free closed state (Fig. 1d, e). However, there remains a major constriction for phosphates near the center of the groove (Fig. 1d, f) 4 , whose hydrophilic groove is widely opened (Fig. 1a and Supplementary Fig. 1c). However, owing to the low sequence homology between nhTMEM16 and TMEM16F, particularly outside of the TM region ( Supplementary Fig. 1a), we started with the 5OYB-derived TMEM16F intermediate model and used steered MD to gradually move its TMs 3, 4, and 6 ( Fig. 1c), whose conformations are most distinct between the TMEM16A and the nhTMEM16 structures 23 Fig. 2a, b). More importantly, this hydrophilic groove becomes readily accessible to both water and especially phosphates ( Fig. 1d-f), supporting the notion that this open state model could represent an active TMEM16F-CaPLSase. Interestingly, despite the limited simulation timeframe of 400 ns, we observed a total of three phosphatidylcholine externalization events (one in each run) via the widened hydrophilic groove, closely recapitulating the wellknown credit-card reader transport mechanism 29 (Supplementary Fig. 2c and Movie 2).
After carefully scrutinizing the closed, intermediate and fully open models of TMEM16F and their capabilities to permeate water and phospholipids ( Fig. 1d-f), we find that the constriction impeding the permeation of phospholipid headgroups mainly originates from three hydrophobic residues, F518, Y563, and I612, whose side chains point toward the central axis of the hydrophilic groove (Fig. 1d, f). Indeed, in the closed and intermediate states, the side chains of these residues stay in close proximity to create a physical barrier restricting phospholipid permeation, whereas in the Ca 2+ -bound fully open state, they separate to widen the hydrophilic groove, thus permitting phospholipid permeation (Fig. 1d-f and Supplementary Fig. 2c). Based on these observations, we hypothesize that F518, Y563 and I612 constitute an inner activation gate of the mammalian TMEM16F-CaPLSase.
An imaging assay to quantify CaPLSase activity. A microscopy assay utilizing fluorescently conjugated Annexin V (AnV), a PSbinding protein, to detect CaPLSase-mediated PS surface exposure enables quantification of phospholipid scrambling at a single-cell resolution 30 . In a recent study, we optimized this assay by genetically ablating endogenous TMEM16F expression, which eliminates endogenous CaPLSase interference in HEK293T cells 31 . We used the TMEM16F-null cell line throughout this study to quantify the scrambling activities of the heterologously expressed wild type (WT) and mutant murine TMEM16F (mTMEM16F) CaPLSases in real time. Heterologous expression of WT mTMEM16F into the TMEM16F-null HEK293T cells via transient transfection induced strong CaPL-Sase activity with minimal sign of apoptosis 31 (Fig. 2a, t = 0 min and Supplementary Fig. 3a). We limited our CaPLSase measurements to the first 10 min after ionomycin stimulation, during  Fig. 3b), further suggesting that transient transfection and overexpression of TMEM16F does not induce apoptosis-dependent phospholipid scrambling. To circumvent the variations in scrambling activity due to the differences in TMEM16F expression levels and ionophore-mobilized intracellular Ca 2+ levels, we use t 1/2(Imax) , the time at which each cell reaches 50% of its maximum AnV fluorescent intensity, instead of the maximum AnV fluorescent intensity to quantify CaPLSase activity 31 (Fig. 2b, c). Using this assay, we confirmed that D409G, a TMEM16F gain-of-function mutation 2 , displayed an accelerated phospholipid scrambling as seen by its significantly shorter t 1/2(Imax) in comparison to that of WT mTMEM16F (Fig. 2b, c). Without ionomycin treatment, D409G expression did not induce spontaneous PS exposure (Fig. 2a, 0 min and Supplementary Fig. 3c), suggesting that this mutant CaPLSase is not constitutively activated in HEK293T cells. On the other hand, the Ca 2+ -binding site mutation D703R, whose equivalent mutation in TMEM16A completely eliminated Ca 2+ -dependent activation of the CaCC 21 , abolished phospholipid scrambling activity (t 1/2(Imax) value assigned as infinity and denoted as non-scrambling) (Fig. 2a-c). Taken together, our optimized CaPLSase imaging assay provides us a faithful method to characterize the putative inner gate residues of TMEM16F.
Ala mutations at the inner gate promote TMEM16F-CaPLSase.
To examine the roles of the putative gate residues in regulating phospholipid movement across the hydrophilic groove, we first replaced these bulky residues with Ala. We hypothesized that the smaller and less hydrophobic sidechain from Ala would widen the inner gate to alleviate the constrictive effect, thereby facilitating phospholipid permeation. Using the optimized CaPLSase assay, we found that all three Ala mutations of F518, Y563, and I612 exhibited strong gain-of-function phenotypes (Fig. 3a, b). When quantified with t 1/2(Imax) , F518A and I612A showed significantly enhanced CaPLSase activity compared to WT-TMEM16F. Consistent with our observation, a recent report also shows an enhanced TMEM16F scrambling activity by the I612A mutation 32 . c Ionomycin-induced CaPLSase activity for TMEM16F-WT, D409G, and D703R as quantified by t 1/2(Imax) . Each data point represents one cell and n is the total number of cells analyzed. Cells that did not exhibit ionomycin-induced CaPLSase activity were denoted as non-scr (non-scrambling) and are excluded from statistical analysis. As a majority of D703R expressing cells does not scramble, mean and SEM are not assigned for this mutation. The pie charts illustrate the percentage of scrambling cells in response to ionomycin stimulation. Statistical analysis was performed using unpaired two-sided Student's t-test for WT and D409G. ****: p < 0.0001. Error bars indicate SEM. All of the experiments based on AnV labeling were done in Ca 2+ -containing buffer. Source data are provided in the Source Data file Strikingly, we found that Y563A expression in regular Ca 2+ -containing medium resulted in constitutively exposed PS even in the absence of ionomycin stimulation (Fig. 3c). Rounded cell morphology and strong caspase 3/7 activity staining (TF3-DEVD-FMK signal) further suggest that overexpressing Y563A likely induces cytotoxicity (Fig. 3c). Interestingly, a rounded cell morphology has also been observed when a gain-of-function mutation of TMEM16E is heterologously expressed in HEK293 cells, suggesting that gain-of-function CaPLSases tend to induce cytotoxicity 33 . On the other hand, overexpression of TMEM16F D409G gain-of-function mutation 2 in the TMEM16F deficient HEK293T cells only resulted in accelerated ionomycin-induced phospholipid scrambling (Fig. 2) without apparent changes in cell morphology and constitutive PS exposure in regular cell culture medium ( Supplementary Fig. 3c). This result is different from the observation of constitutive PS exposure when D409G was expressed in lymphoma cells 34 . We speculate that this discrepancy may be due to the differences in basal Ca 2+ activities and cellular regulations in the two different cell types. Thus, Y563Ainduced cytotoxicity in Ca 2+ containing medium suggests that Y563A might be a more potent gain-of-function mutation compared with D409G.
To test whether this cytotoxicity was due to the excessive Ca 2+ -dependent activation of TMEM16F, we cultured Y563Aexpressing cells in a Ca 2+ -free medium to suppress the basal Ca 2+ activity. In this condition, we observed about 42% of the Y563A-expressing cells still exhibited spontaneous PS surface exposure (Fig. 3c, d). Notably, these PS-positive cells possessed healthy, polarized morphology and displayed minimal caspase activity, all of which indicate that they were viable and that Alanine mutations of the inner gate residues promote TMEM16F activation. a Representative images of ionomycin-induced scrambling activity of TMEM16F-knockout HEK293T cells expressing eGFP-tagged TMEM16F-F518A, Y563A, or I612A. Y563A was expressed in Ca 2+ -free medium to minimize cytotoxicity, and its CaPLSase activity was measured in Ca 2+ -containing buffer. AnV-CF 594 signal representing phospholipid scrambling was recorded by time-lapsed imaging for 10-min post 5 µM ionomycin application (0 min) at a 5-s acquisition interval. Scale bars, 25 µm. b Ionomycin-induced CaPLSase activity for wild type (WT) and the mutant TMEM16F quantified by t 1/2(Imax) . Each data point represents one cell and n denotes the total number of cells measured. The pie charts illustrate the percentage of ionomycin-induced scrambling cells. The cells that did not exhibit ionomycin-induced CaPLSase activity were denoted as non-scr (non-scrambling) and were excluded from statistical analysis. Statistical analysis was performed using one-way ANOVA with Tukey's multiple comparisons test. ****: p < 0.0001. Error bars indicate SEM. # denotes mutations with spontaneous CaPLSase activity. TMEM16F-Y563A expression promotes spontaneous PS externalization. While the remaining (58%) Y563A-expressing cells did not show spontaneous CaPLSase activity, likely due to lower expression levels, they did exhibit gain-of-function phenotype upon ionomycin stimulation as evidenced by a smaller t 1/2(Imax) value in comparison to that of TMEM16F-WT (Fig. 3a, b). These observations suggest that, distinct from D409G, which is not spontaneously active (Fig. 2a), Y563A is a potent gain-of-function mutation that can facilitate strong CaPLSase activity under resting intracellular Ca 2+ . Consistent with its increased CaPLSase activity, we found that Y563A also significantly enhanced its ion channel activation by Ca 2+ (Supplementary Fig. 4a-d). In addition, Y563A completely abolishes TMEM16F channel desensitization or rundown in addition to altering ion selectivity ( Supplementary Fig. 4e, f). All taken together, these results further support that Y563 plays a critical role in gating TMEM16F substrate permeation. As F518A, Y563A, and I612A all exhibited pronounced gainof-function phenotypes in their scrambling activity, we test whether their effects on phospholipid permeation can be additive. We therefore generated a triple Ala mutation, F518A-Y563A-I612A, and interrogated its Ca 2+ -dependent phospholipid scrambling. Interestingly, we found that this triple mutation displayed more enhanced gain-of-function phenotype than the single Ala mutations (Fig. 3c, d). Specifically, more than 90% of the triple mutation-expressing cells exhibited constitutive CaPL-Sase activity in the absence of ionomycin treatment. The observed additive effect of the triple Ala mutation suggests a possibility that the three putative inner gate residues work synergistically to control phospholipid movement across the hydrophilic groove (Fig. 1d, f).
Bulky mutations at the inner gate hinder TMEM16F-CaPLSase. Having established the critical roles of the putative inner gate residues in phospholipid gating, we next tested whether replacing these residues with a bulkier amino acid could disrupt the physical constriction and affect CaPLSase activity. A recent study identifies residues that are important for lipid scrambling in nhTMEM16 via Trp substitutions 28 . Taking a similar approach, we found that Trp substitutions of Y563 and I612, Y563W and I612W, markedly impaired phospholipid scrambling of TMEM16F, giving rise to loss-of-function scramblases (Fig. 4a, b). Almost all Y563W-expressing cells and~52% of I612Wexpressing cells showed no ionomycin-induced CaPLSase activity. Paradoxically, F518W slightly enhanced TMEM16F phospholipid permeation (Fig. 4a, b). We suggest that this enhancement in scrambling of F518W is likely due to the higher hydrophilicity of Trp compared with that of Phe (the hydropathy indexes for Trp and Phe are −0.9 and 2.8, respectively 35 ). This mild gain-of-function phenotype of F518W is further corroborated by a similar observation that an equivalent mutation in nhTMEM16 also enhanced phospholipid scrambling 28 . To examine the significance of the hydrophobicity at F518 for phospholipid permeation, we substituted F518 with Leu, a more hydrophobic residue with the hydropathy index of 3.8 (ref. 35 ). In contrast to F518W, F518L exhibited a loss-of-function phenotype with a reduced scrambling activity (Fig. 4a, b). Furthermore, we found that a large population of F518L-expressing cells showed no CaPLSase activity whereas cells that did scramble exhibited significantly prolonged t 1/2(Imax) values compared with that of WT-TMEM16F (Fig. 4b). Taken together, our findings highlight the importance of both size and hydrophobicity at the inner gate, both of which are critical in controlling phospholipid permeation.
Hydrophilic and charged mutations at the inner gate. To further dissect the role of the inner gate in controlling phospholipid permeation, we disrupted the hydrophobicity at the inner gate by substituting the inner gate residues with either hydrophilic or charged amino acids. Surprisingly, all of these mutations turned out to be gain-of-function ( Fig. 5 and Supplementary Fig. 5). Furthermore, with the exception of I612Q, which showed the modest gain-of-function phenotype, all other tested mutations exhibited constitutive scrambling activities ( Fig. 5a and Supplementary Fig. 5a). Similar to Y563A and the triple Ala mutation (Fig. 3), all of the spontaneous scrambling mutations from the hydrophilic/charged substitutions, except I612Q, induced cell stress and cytotoxicity when cultured in regular Ca 2+ medium.
To verify that the cytotoxicity was due to Ca 2+ -dependent activation of TMEM16F, we cultured the transfected cells in the Ca 2+ -free medium. Removal of extracellular Ca 2+ alleviated the cytotoxicity from most of the spontaneous gain-of-function mutations, except for the two most potent ones, F518K and Y563K ( Fig. 5a and Supplementary Fig. 5a). In fact, these two mutations were highly active that all of the expressing cells were PS-positive with apparent signs of apoptosis (rounded morphology and strong caspase 3/7 activities) even when cultured in Ca 2+ -free medium ( Supplementary Fig. 5a). Cells expressing other spontaneous gain-of-function mutations were healthy in the Ca 2+ -free medium and exhibited various levels of spontaneous and ionomycin-induced CaPLSase activities (Fig. 5a, b and Supplementary Fig. 5b). Interestingly, negatively charged substitution to the inner gate residues, especially F518E and I612E, also caused a large population of transfected cells to display spontaneous PS externalization ( Fig. 5a and Supplementary  Fig. 5a). Thus, these results suggest that increasing hydrophilicity or introducing charges at the inner activation gate promotes phospholipid permeation. Besides lowering the permeation energy barrier by facilitating phospholipid headgroups interaction, the hydrophilic and charged residues introduced at the inner gate may also shift the equilibrium between the open and closed states of TMEM16F-CaPLSase by destabilizing the closed state and/or stabilizing the open state. Other than directly affecting the inner gate itself, the mutations at the putative inner gate may allosterically enhance Ca 2+ -binding to give rise to the gain-of-function phenotype. To examine this possibility, we couple our polar and charged inner gate mutations with the Ca 2+ -binding site mutation D703R (Fig. 2). Interestingly, abrogating Ca 2+ binding successfully suppresses the gain-of-function phenotypes for most of the polar and charged inner gate mutations (except F518K and Y653K) even in the regular Ca 2+ -containing medium ( Fig. 5c and Supplementary Fig. 6). Since all of the double mutations with D703R have suffice cell surface expression ( Supplementary Fig. 6), membrane targeting is unlikely to be the reason why these double mutations lost their spontaneous CaPLSase activities. We speculate that the resting, basal Ca 2+ level (below 100 nM) is sufficient to trigger the opening of the activation gate in the corresponding single gain-of-function mutations to induce spontaneous CaPLSase activity (Fig. 4a).
Notably, we found that disrupting Ca 2+ binding by the D703R mutation did not prevent spontaneous scrambling of F518K and Y563K (Fig. 5c, d). In fact, about 34% of F518K-D703R and nearly 93% Y563K-D703R-expressing cells still exhibit spontaneous PS exposure in Ca 2+ -free medium. The constitutively active F518K-D703R and Y563K-D703R mutations further support that the positive charges at these two inner gate residues destabilize the closed state of the phospholipid pathway independent of Ca 2+ binding. I612K in TM6, however, when coupled with D703R, failed to spontaneously expose PS (Fig. 5c), suggesting that the charge on this residue imposes a weaker effect in destabilizing the activation gate than those of F518K and Y563K in TM4 and TM5, respectively. The results also illustrate that positive charges introduced at these two positions allow TMEM16F phospholipid pathway to stay constitutively open without Ca 2+ binding. Although we could not investigate the Ca 2+ binding affinity of TMEM16F mutations to completely rule out the possibility that these gain-of-function mutations allosterically enhance Ca 2+ binding, the constitutive scrambling observed in F518K-D703R and Y563K-D703R mutations, whose Ca 2+ binding is severely impaired, strongly indicates that these charged residues introduced at the inner activation gate directly open the phospholipid pathway without much facilitation from Ca 2+ binding.
An equivalent inner gate controls TMEM16A gating. As TMEM16F-CaPLSase and TMEM16A-CaCC share a high degree of sequence similarity in their TMs ( Supplementary Fig. 1a), we tested whether equivalent residues in TMEM16A could also serve as an inner activation gate. Interestingly, unlike WT-TMEM16A 21 , our inside-out patch clamp recordings show that both TMEM16A-L543K and I637K mutant channels could be activated by membrane depolarization in the absence of intracellular Ca 2+ (Fig. 6a, b), suggesting that introducing positive charges to these two hydrophobic residues in TM4 and TM6 could unlatch the gate and promote TMEM16A channel opening. Among all three mutations, S588K showed the weakest effect on TMEM16A voltage-dependent activation despite being activated by Ca 2+ (Supplementary Fig. 7a).
More strikingly, similar to TMEM16F-F518K (Fig. 5a), the corresponding mutation TMEM16A-L543K induced cytotoxicity when the transfected cells were cultured in regular Ca 2+ -containing medium (Supplementary Fig. 7b). However, culturing these cells in a Ca 2+ -free medium alleviated cytotoxicity; yet about 44% of the L543K-TMEM16A-expressing cells still exhibited constitutive PS exposure as indicated by the positive AnV signal (Fig. 6c, d). Their morphology as well as activated caspase 3/7 level were similar to the healthy non-transfected cells (Fig. 6d). For the L543K-expressing cells that did not show spontaneous PS exposure, application of a lower concentration of ionomycin (2.5 µM) rapidly induced PS exposure ( Fig. 5e and Supplementary Fig. 7c), consistent with TMEM16A having a higher Ca 2+ sensitivity than TMEM16F 9 . These results clearly demonstrate that L543K, a single point mutation at the inner activation gate, is capable of converting TMEM16A-CaCC into a CaPLSase. In addition, a potent TMEM16A-CaCC blocker Ani9 (ref. 36 ) effectively blocked ionomycin-induced phospholipid scrambling through L543K (Fig. 6c), confirming that this mutation rendered TMEM16A the capability to permeate phospholipids. Interestingly, TMEM16A-L543K also has altered ion selectivity ( Supplementary Fig. 7b-d), further supporting the critical role of L543 residue in controlling ion permeation. In contrast to L543K, neither S588K nor I637K expression caused spontaneous or ionomycin-induced PS exposure through TMEM16A (Fig. 6c, e and Supplementary Fig. 7c). As I637K shows a gain-of-function in its channel activation (Fig. 5a, b), the lack of CaPLSase activity indicates that while the positive charge at I637 can enhance ion permeation, its activation gate might not be opened widely enough to permeate phospholipids. This is also consistent with the milder gain-of-function phenotype observed in the equivalent I612K-TMEM16F-CaPLSase (Fig. 5a, c). In summary, our observation that the conversion of TMEM16A-CaCC to a constitutively activated CaPLSase by L543K further strengthens the importance of positive charges at the inner activation gate in promoting gating of TMEM16 proteins.

Discussion
In this study, we identified an inner activation gate formed by three hydrophobic residues, F518, Y563, and I612, near the middle of the phospholipid permeation pathway of TMEM16F-CaPLSase. Side chain properties of the inner gate residues have profound effects on TMEM16F-CaPLSase activities. In particular, introducing positive charges to F518 and Y563 constitutively opens the activation gate to allow spontaneous phospholipid permeation even in the absence of Ca 2+ binding. Moreover, the positive charge at the putative inner activation gate not only promotes phospholipid scrambling in TMEM16F but also enhances channel activation in TMEM16A, consistent with a possibility that they share a similar mechanism of activation. More strikingly, we found that a single point mutation, L543K, at the inner activation gate can convert the non-scrambling CaCC TMEM16A into a constitutively activated CaPLSase. This observation stands in contrast to a previous study, which shows that substitution of a long stretch of 34 amino acids at TMs 4-5 from TMEM16F is required to transform CaCC TMEM16A into a scramblase 30 .
Based on our functional and computational studies, here we propose a clam shell model to describe the gating mechanism of TMEM16F-CaPLSase (Fig. 7). We speculate that the interface between TM4 and TM6 can open and close like a clam shell to control the accessibility of phospholipids to the interior of the hydrophilic groove, where phospholipid permeation is catalyzed ( Supplementary Fig. 2c and Movie 2). F518 in TM4 and I612 in TM6 likely serve as gate-keepers for the opening of this interface. In the absence of Ca 2+ , they stay close to each other to block the accessibility of phospholipid headgroups to the hydrophilic groove. Upon Ca 2+ binding, large-scale conformational changes, likely including the movement of TM6 (Supplementary Fig. 1b) around a conserved glycine hinge 23,37 and the rearrangement of TMs 3-5, separate TM4 and TM6 (Fig. 1c). These conformational rearrangements likely culminate in the opening of the TM4-TM6 interface as well as the inner gate residues. Subsequently, the interior of the hydrophilic groove becomes exposed to the surrounding phospholipids such that their phospholipid headgroups can enter and translocate via the credit-card reader mechanism. By projecting its bulky sidechain into the center of the hydrophilic groove, Y563 in TM5 likely serves as a cap that stabilizes the inner gate and further obstructs phospholipid permeation in the closed state. Dilation of the inner gate following TM4 and TM6 separation allows the Y563 hydroxyl group to interact with phospholipid headgroups and facilitate phospholipid permeation. Introducing smaller, polar or charged residues to these critical locations tends to remove steric hindrance, likely via destabilizing the TM4-TM6 interface to promote spontaneous opening transitions, resulting in enhanced phospholipid permeation. In the case of F518K and Y563K, their inner activation gates become so severely disrupted that phospholipids can freely While this work is under review, four independent structural characterizations of TMEM16 CaPLSases were released, reporting the structures of the fungal nhTMEM16, afTMEM16, murine TMEM16F (mTMEM16F), and human TMEM16K (hTMEM16K) 20,24,25,38 . Interestingly, the Ca 2+ -free mTMEM16F structure exhibits high similarities to the murine TMEM16A structures 18,23 , especially in the TM region including TM4 and TM6. This is consistent with our hypothesis that mammalian TMEM16F and TMEM16A share more structural similarities than fungal nhTMEM16 (Supplementary Fig. 1a). While the structures of mTMEM16F provide valuable insights into the mammalian CaPLSases, minimal conformational changes were observed between the Ca 2+ -bound and the Ca 2+ -free mTMEM16F or between mTMEM16F and the non-scrambling mTMEM16A. It thus remains unclear how TMEM16F-CaPLSase regulates phospholipid permeation based on these structural studies. Moreover, the extracellular segments of TM4 and TM6 in the Ca 2+ -bound mTMEM16F structure are packed against each other preventing phospholipids from accessing the interior of the hydrophilic groove 20 , suggesting that this structure may represent an intermediate conformation rather than a fully open and active conformation (Fig. 1e, f). On the other hand, widening of TM4-TM6 interface was observed in the Ca 2+ -bound structures of fungal afTMEM16 (ref. 24 ), nhTMEM16 (refs. 4,25 ), and hTMEM16K 38 , which is consistent with our proposed clam shell gating model in TMEM16F.
Identification of the inner activation gate for TMEM16F-CaPLSase also offers insights into our understanding of the gating mechanism of TMEM16-CaCCs. We suggest that, instead of utilizing the Ca 2+ -dependent clam shell activation mechanism for TMEM16 CaPLSases, Ca 2+ binding to TMEM16-CaCCs may induce coordinated movement of TM4 and TM6 to dilate the ion permeation pathway, thereby allowing ions to go through while still maintaining a tight barrier at the TM4-6 interface, which excludes phospholipid permeation. This gating mechanism is supported by our observation that I637K-TMEM16A, which is a gain-of-function CaCC, does not facilitate phospholipid permeation ( Fig. 6a-d). On the other hand, when a positive charge is introduced to L543 in TM4, the interaction between TM4 and TM6 might be dramatically weakened so that they may open like a clam shell to allow spontaneous permeation of phospholipids. We therefore postulate that TMEM16 CaPLSases and CaCCs may share evolutionarily conserved Ca 2+ -dependent gating mechanism and overall similar design in their inner activation gates. Their distinct substrate selectivity and permeation may be partially derived from the differences on how widely the putative inner activation gate can open. We hope that our current findings can inspire future structural, computational, and functional studies to further establish the molecular mechanisms underlying Ca 2+ -dependent substrate permeation of TMEM16 proteins.

Methods
Cell lines and cell culture. HEK293T cells are authenticated and tested negative for mycoplasma by Duke Cell Culture Facility. The Cas9 control HEK293T and TMEM16F-KO HEK293T cell lines were generated by Duke Functional Genomics Core and were characterized in a recent report 31 . Briefly, the sgRNA (5′-AATAGTACTCACAAACTCCG-3′; see also Supplementary Table 1) targeting exon 2 of ANO6/TMEM16F was first cloned into lentiguide-puro (Addgene #52962). LentiCAS9-blast (Addgene #52962) was used to transduce HEK293T cells (Duke Functional Genomics Facility) to generate stable Cas9-expressing cells. All transductions were done at a multiplicity of infection (MOI) of <1 in the presence of 4 µg ml −1 polybrene. Twenty-four hours post infection, cells were selected with 10 µg ml −1 blasticidin or 2 µg ml −1 puromycin for 48-72 h and then expanded. Genomic DNA was harvested from cells 1 week after transduction with sgRNA. The ANO6 exon 2 locus was PCR amplified from genomic DNA and subsequently analyzed using the Surveyor assay (Integrated DNA Technologies) to confirm the presence of insertions and deletions (Supplementary Table 2). The TMEM16F-KO HEK293T cells were further purified using single-cell colonies from serial-dilution. The single-cell colonies were selected and expanded. Western blots and phospholipid scrambling assay were used to screen for the colonies without TMEM16F protein expression and function. The purified TMEM16F-KO HEK293T cells were used throughout to characterize phospholipid scrambling activity. HEK293T cell line with stable expression of C-terminally eGFP-tagged mTMEM16F was a kind gift from Dr. Min Li. All cells were maintained in Dulbecco's modified Eagle's medium (DMEM; Gibco, #11995-065) supplemented with 10% fetal bovine serum and 1% penicillin-streptomycin and supplied with 5% CO 2 at 37°C. The same medium with 100 µg ml −1 of hygromycin was used for culturing mTMEM16F stable HEK293T cells.
Microscope-based phospholipid scrambling assay. Cells were seeded and transfected on poly-L-lysine (Sigma)-coated coverslips prior to the imaging experiments. Fluorescence-tagged (CF 594) Annexin V (Annexin V-CF 594, Biotium # 29011) was diluted at 0.5 µg ml −1 with Annexin V-binding solution (10 mM HEPES, 140 mM NaCl, 2.5 mM CaCl 2 , pH 7.4). The coverslips were placed into our homemade imaging chambers that contain the diluted Annexin V-CF 594. After focusing on the cell surface eGFP signal from the expressed TMEM16F-eGFP, an equal volume of Annexin V-CF 594 with 10 μM ionomycin, which was diluted from a fresh aliquot of 5 mM ionomycin (Alomone) stock solution, was added into the chamber to reach the final concentration of 5 µM to activate TMEM16F phospholipid scramblase. For Q-VD-OPh treatment, after TMEM16F-null HEK293T cells are transfected with eGFP-tagged mTMEM16F for 5-h, culture medium was replaced with the fresh medium containing 10 µM Q-VD-OPh (Sigma). After 24-h of transfection, the cells were tested for their scrambling activity as described above in the presence of 10 µM Q-VD-OPh.
In case of TMEM16A and its mutants, the final concentration of 2.5 µM of ionomycin was used. For Ani9 treatment, TMEM16A-L543K-expressing cells were treated with 10 µM Ani9, and then stimulated with 2.5 µM of ionomycin. The PS exposure of the cells was detected by the accumulation of Annexin V-CF 594 on the cells membrane. Time-lapse imaging of lipid scrambling activity was captured at 5-s intervals by a Prime 95B Scientific CMOS Camera (Photometrics) connected to an Olympus IX71 inverted epi-fluorescent microscope (Olympus IX73) for 10min upon ionomycin application (0-min). A ×60 oil objective with NA of 1.35 was used for imaging. Image acquisition was controlled by Metafluo software (Molecular Devices).
Quantifying phospholipid scrambling activity. We used two ways to quantify TMEM16F lipid scrambling activity. For TMEM16F-expressing cells without spontaneous lipid scrambling activities prior to ionomycin application, t 1/2(Imax) was analyzed. t 1/2(Imax) indicates the amount of time each individual transfected cell takes to reach 50% of maximum AnV fluorescent intensity by the course of 10-min post ionomycin treatment. To determine t 1/2(Imax) , a customized MATLAB program (Mathworks) was written to analyze the AnV fluorescent intensity change overtime of each cell. Briefly, a region of interest (ROI) was manually chosen around the scrambling cells, and the AnV fluorescent intensity was calculated using the following equation for each frame: where i equals the fluorescent intensity of each pixel and N is the number of the pixels in the ROI. The time reaching half of I max was defined as t 1/2 . For the mutations that induce constitutive scrambling activity prior to ionomycin treatment, we quantify the percentage of spontaneous PS-positive cells by manually counting the number of AnV-positive expressing cells over all expressing cells, which were detected by GFP signals, in 10-20 random fields of view per coverslip. Each data point represents the percentage of spontaneous PSpositive expressing cells in one coverslip.
Live cell Caspase 3/7-binding assay. To detect cleaved Caspase 3/7 level in live cells, we employed Live cell Caspase 3/7 binding assay kit (AAT Bioquest, # 20101). After 24-h post transfection, culture media were removed and the cells were washed once with fresh media. The stock solution of TF3-DEVD-FMK (150×) was prepared and stored as described in the manual complemented with the assay kit (AAT Bioquest). The TF3-DEVD-FMK dye was diluted to 1× in either normal or Ca 2+ -free medium depending on the mutations that were transfected to the cells.
After incubating with 1× dye for 1 h at 37 o C, 5% CO 2 , the cells were washed twice with the kit's washing buffer. 1:100 dilution of CF 640R-tagged AnV in AnVbinding buffer (Biotium # 29014) was also added into the imaging chamber to monitor the cleaved caspase 3/7 and spontaneous PS exposure simultaneously. The results were collected with a ×63/1.4 NA Oil Plan-Apochromat DIC in Zeiss 780 inverted Confocal microscope. The results were analyzed with ImageJ and Zeiss.
Electrophysiology. TMEM16F-KO HEK293T cells were used for expression and recordings of mTMEM16F current. Normal HEK293T cells were used for expression and recordings of mTMEM16A current. HEK293T cells grown on poly-L-lysine (PLL, Sigma) and laminin (Sigma)-coated coverslips placed in a 24-well plate (Eppendorf) reaching 40-60% confluency were transiently transfected using X-tremeGENE 9 DNA transfection reagent (Sigma). Two hundred nanograms of DNA was used for each well to yield sufficient channel expression for recordings. All recordings were carried out at room temperature after 24-48 h following transfection. Inside-out recordings were performed on patches excised from cells expressing eGFP-tagged TMEM16A or TMEM16F constructs. Pipette electrodes were made from borosilicate glass capillaries (1.5 mm × 0.86 mm) with a puller (Sutter Instruments), fire-polished with a microforge (Narishge), and had a resistance of~2-3 MΩ when filled with recording solutions. The pipette (external) solutions contain 140 mM NaCl, 5 mM EGTA, 10 mM HEPES, 2 mM MgCl 2 , adjusted to pH 7.3 (with NaOH). The perfusion (internal) solution contain 140 mM NaCl, 10 mM HEPES, 5 mM EGTA, adjusted to pH 7.3 (with NaOH). The internal solution with 100 µM Ca 2+ was made by directly adding CaCl 2 into a solution containing 140 mM NaCl and 10 mM HEPES, adjusted to pH 7.3 (with NaOH). The internal solution with 2.26 µM Ca 2+ contains 140 mM NaCl, 10 mM HEPES, 5 mM EGTA, and 4.64 mM CaCl 2 , adjusted to pH 7.3 (with NaOH). The amount of added CaCl 2 was calculated using WEBMAXC (https://somapp.ucdmc. ucdavis.edu/pharmacology/bers/maxchelator/). Free Ca 2+ concentration of EGTAbuffered solution was further verified using the ratiometric Ca 2+ dye Fura-2 (ATT Bioquest) and plotted against a standard curve from a calibration kit (Biotium).
For the current-voltage relationship (I-V) recordings, the membrane was held at −60 mV and 250-ms test voltage steps ranging from −120 mV to +140 mV were applied at a 20 mV increment. Due to the small amplitudes and rapid deactivation of the tail currents at the −60 mV repolarization step of TMEM16A in the absence of Ca 2+ , steady-state peak currents from the test voltage steps were used to construct the I-V plot. For TMEM16F's I-V recordings in the presence of 2.26 µM and 100 µM Ca 2+ , peak tail currents measured at −60 mV repolarization steps were normalized to the max tail current to generate the G-V curves.
For the reversal potential (E rev ) measurements, TMEM16A and TMEM16F channels were first activated by perfusion of internal solution containing 100 µM Ca 2+ . Both the pipette (external) and bath (internal) are symmetric and contain 140 mM NaCl, 5 mM EGTA, 10 mM HEPES, adjusted to pH 7.3 (NaOH). For TMEM16A, a ramp protocol from −100 mV to +100 mV was used to elicit currents. For TMEM16F, an inverted V-shaped protocol in which the membrane was ramped from −120 mV to +120 mV and back to −120 mV was used. The second phase of the ramp (from −120 mV to −120 mV) was used to construct the I-V plots for measuring the E rev . For both TMEM16A and TMEM16F, the changes in E rev were triggered by switching internal solution with a low NaCl solution containing 14 mM NaCl, 5 mM EGTA, 10 mM HEPES, and 100 µM Ca 2+ adjusted to pH 7.3. The E rev was determined as the membrane potential at which the current was zero, and the shifts in the E rev (ΔE rev ) are the differences between the E rev measured in symmetric NaCl and low internal NaCl. The permeability ratio P Cl / P Na was calculated using the Goldman-Hodgkin-Katz equation: For time-course monitoring of TMEM16F channel activity, a voltage step protocol in which the membrane was stepped to +80 mV and then −80 mV lasting 200 ms each. The membrane was held at 0 mV. Channel opening was elicited by perfusion of internal solution containing 100 µM Ca 2+ . Peak currents measured at +80 mV steps were used for time-course monitoring of channel opening.
Homology modeling of TMEM16F in open and closed states. Homology models of the closed (without Ca 2+ ) and intermediate (with Ca 2+ ) states of mTMEM16F were derived directly using Ca 2+ -free (PDB: 5OYG) and Ca 2+ -bound (PDB: 5OYB) mTMEM16A structures respectively with the sequence alignment shown in Supplementary Fig. 1a using the Swiss-pdb server 39 . Even though the Ca 2+ -bound structure of nhTMEM16 (PDB: 4WIS) is believed to provide a good model of the open conformation of the TM region of mTMEM16F, the sequence conservation in the extracellular domain and loops is poor. To derive a model of mTMEM16F in the open state, steered MD was initiated from the intermediate state model derived from 5OYB to reposition TMs 3-6 to the configuration observed in the Ca 2+bound nhTMEM16 structure (PDB: 4WIS) using the CHARMM software 40,41 . We note that, even though TM5 is similarly positioned between PDBs 5OYB and 4WIS, it lines the putative phospholipid permeation pathway and thus was included in the steered MD to ensure the structural integrity. For this, a homology model of mTMEM16F TM region alone was first derived from nhTMEM16 using the sequence alignment derived from Clustal Omega (see above). This model was then used as the target (open state TM conformation) during the steered MD simulation, where root mean square deviation (RMSD) restraint was applied on TMs 3-6 with increasing force constants ranging from 5 to 75 kcal mol −1 Å −2 . In addition, all heavy atoms of adjacent five residues of loops before and after TMs 3-6 were harmonically restrained using a weak force constant of 0.1 kcal mol −1 Å −2 . This allows appropriate, well-controlled loop reconfiguration to accommodate the movement of TMs 3-6. The rest of the whole protein were restrained by harmonic positional restrains with a force constant of 100 kcal mol −1 Å −2 to avoid unnecessary structure disruption. The simulation was performed in vacuum with a distance-dependent dielectric constant. Timestep was set to 1 fs. The final model was only~0.14 Å backbone RMSD from 4WIS for TMs 3-6 (Supplementary Movie 1).
MD simulations of mTMEM16F. The mTMEM16F structures (in open, intermediate, or closed state) were first inserted in model lipid bilayers and then solvated in TIP3P water 42 using the CHARMM-GUI web server 43 . To mimic mammal plasma membranes, pure POPC and 2:1 POPC/POPS mixture were used for upper and lower leaflets, respectively. The solvated systems were then neutralized and 150 mM KCl was added. The final simulation boxes contain about 660 lipid molecules and~73,000 water molecules, with a total of~335,000 atoms and dimensions of~155 × 155 × 140 Å 3 . The CHARMM36m all-atom force field 44 and the CHARMM36 lipid force field 45 were used. All simulations were performed using CUDA-enabled Amber14 46 . Long-range electrostatic interactions were described by the Particle Mesh Ewald (PME) algorithm 47 with a cutoff of 12 Å. Van der Waals interactions were cutoff at 12 Å with a smooth switching function starting at 10 Å. The lengths of hydrogen-containing covalent bonds were constrained using SHAKE 48 and the MD time step was set at 2 fs. The temperature was maintained at 298 K using the Langevin dynamics with a friction coefficient of 1 ps −1 . The pressure was maintained semi-isotopically at 1 bar at both x and y (membrane lateral) directions using the Monte Carlo (MC) barostat method. Several segments were absent in the cryo-EM structures of TMEM16A; they were considered dynamic and thus not included in simulations. These include the Nand C-terminal segments (M1-N86 and S882-E911) and a long loop in the cytosolic domain (Y101-V134). The residues before and after the missing part were capped with either an acetyl group (for N-termini) or a N-methyl amide (for Ctermini). To minimize the effects of missing residues on the cytosolic domain, the backbone of structured parts of the cytosolic domain (E92-A100, L135-F182, S190-D194, P206-V220, G232-L248, R420-V431, K862-E881) were harmonically restrained with a modest force constant of 1 kcal mol −1 Å −2 during all simulations.
All systems were first minimized for 5000 steps using the steepest descent algorithm, followed by a series of equilibration steps where the positions of heavy atoms of the protein/lipid were harmonically restrained with restrained force constants gradually decreased from 10 to 0.1 kcal mol −1 Å −2 . In the last equilibration step, only protein heavy atoms were harmonically restrained and the system was equilibrated 10 ns in under NPT (constant particle number, pressure, and temperature) conditions. All production simulations were performed under NPT conditions. For each state (open, intermediate, and closed), three independent 400-ns simulations were performed. Snapshots were saved every 50 ps. Only snapshots from 100 to 400 ns of each trajectory were included for analysis. The PMFs of water and phosphate group of lipids were calculated directly from the corresponding probability distributions along the membrane normal. The standard errors were estimated from the calculated averages over three independent simulations. The 3D phosphate density distribution near the channel was calculated using a cubic grid with a resolution of 0.5 Å.
Statistical analysis. All statistical analyses were performed in Prism software (GraphPad). Unpaired two-tailed Student's t-test was used for single comparisons between two groups, and one-way ANOVA (with Tukey's multiple comparisons test) was used for multiple comparisons. Unless otherwise described, the data are representative of mean ± standard error of the mean (SEM). P values less than 0.05 were considered statistically significant.
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
Data supporting the findings of this manuscript are available from the corresponding author upon reasonable request. A reporting summary for this Article is available as a Supplementary Information file. The source data underlying Figs. 2b, 2c, 3b, 3d, 4b, 5a-c, 6b, 6c, 6e and Supplementary Figs. 3b, 4c, 4d, 4e, 4f, 4i, 4j, 7f are provided as a Source Data file.