Structure–Function of TMEM16 Ion Channels and Lipid Scramblases

Abstract

The TMEM16 protein family comprises two novel classes of structurally conserved but functionally distinct membrane transporters that function as Ca²⁺-dependent Cl^– channels (CaCCs) or dual functional Ca²⁺-dependent ion channels and phospholipid scramblases. Extensive functional and structural studies have advanced our understanding of TMEM16 molecular mechanisms and physiological functions. TMEM16A and TMEM16B CaCCs control transepithelial fluid transport, smooth muscle contraction, and neuronal excitability, whereas TMEM16 phospholipid scramblases mediate the flip-flop of phospholipids across the membrane to allow phosphatidylserine externalization, which is essential in a plethora of important processes such as blood coagulation, bone development, and viral and cell fusion. In this chapter, we summarize the major methods in studying TMEM16 ion channels and scramblases and then focus on the current mechanistic understanding of TMEM16 Ca²⁺- and voltage-dependent channel gating as well as their ion and phospholipid permeation.

6.1. Introduction

Cell membranes are a bilayer structure made up of amphipathic phospholipid molecules. The fatty acid acyl chains of phospholipids form a hydrophobic core that creates a huge energy barrier for the transport of ions, glucose, and amino acids as well as spontaneous flip-flopping of phospholipids. To maintain cellular homeostasis, communication, and survival, numerous membrane proteins have evolved to overcome this energy barrier to facilitate membrane transport [1] (Fig. 6.1a). TMEM16 transmembrane proteins are a recently discovered family of membrane transport proteins that passively permeate ions, phospholipids, or both (Fig. 6.1b). Over the past decade, tremendous advances have been made to understand these mysterious TMEM16 proteins and their roles in human health and diseases. In this chapter, we briefly discuss the concise history of the molecular identification of TMEM16 proteins and then primarily focus on the current mechanistic understanding of how TMEM16 proteins work in response to Ca²⁺ and voltage to catalyze the permeation of two structurally distinct substrates: ions and phospholipids. This chapter by no means can cover all of the elegant works in the field and the physiological aspects of TMEM16 biology. We hope that our summary can give the readers an overview of the TMEM16 proteins and an up-to-date summary of the current understanding of TMEM16 molecular mechanisms. Interested readers can refer to several excellent reviews for more detailed insights into TMEM16 proteins and their biology [3–19].

Fig. 6.1.

TMEM16 protein family. (a) TMEM16 proteins can function as Ca²⁺-activated chloride channels (CaCCs) or Ca²⁺-dependent phospholipid scramblases (CaPLSases). (b) TMEM16 protein family with diverse functionality. Note that the yeast IST2 and C. elegans ANOH-1 and −2 have not been assigned as CaCCs or CaPLSases. (c) TMEM16A CaCC and TMEM16F small-conductance, Ca²⁺-activated nonselective channel (SCAN) at different Ca²⁺ concentrations [2]. (d) A fluorescence assay to monitor CaPLSase activity

6.2. Molecular Identifications of TMEM16 Proteins

6.2.1. TMEM16A and TMEM16B Form the Canonical Ca²⁺-Activated Chloride Channels

TMEM16A/DOG1 with the unknown function was first found to be highly expressed in gastrointestinal stromal tumor (GIST) in 2004 [20]. However, systematic investigations of the TMEM16 family of proteins were not reported until 2008, when three groups independently identified TMEM16A as well as its closely related member TMEM16B as the long sought after Ca²⁺-activated chloride channels (CaCCs) [21–23].

The Oh group searched for membrane proteins having multiple TMs with unknown functions and came across the TMEM16 family. They showed that co-expressing of TMEM16A with G_q-coupled receptors (G_qPCRs) gave rise to robust CaCC conductance following the application of receptor agonists. TMEM16A’s biophysical properties strikingly resemble those of the canonical CaCCs [24] exemplified by the outward-rectifying current at low Ca²⁺ and linear current–voltage relationship at high Ca²⁺ (Fig. 6.1c) and especially with an anion selectivity sequence of NO₃⁻ > I⁻ > Br⁻ > Cl⁻ > F⁻ [23]. In addition, TMEM16A was found to be highly expressed in the epithelial cells of pulmonary bronchioles, pancreatic acinar cells, epithelia of the renal tubules of the kidney, the outer nuclear layer of the retina, small-diameter sensory neurons of dorsal root ganglia (DRG), acinar cells of submandibular glands, and Leydig cells of testes. These patterns of tissue expression of TMEM16A are consistent with the reported expression of endogenous CaCCs [24].

The Galietta group discovered TMEM16A based on the previous observations that prolonged stimulation of human bronchial epithelial cells with interleukin-4 or −13 (IL-4 or IL-13) upregulated the expression of CaCCs [25, 26]. Using gene expression microarrays, the authors identified a list of genes that are upregulated following IL-4 and IL-13 stimulation. Among these candidates, only siRNA knockdown of TMEM16A gene significantly attenuated CaCC activity as measured via iodide influx, short-circuit current, and patch-clamp recordings of several cell lines known to highly express CaCCs such as bronchial CFBE41o, pancreatic CFPAC-1 as well as primary human bronchial epithelial cultures [21].

The Jan group took advantage of the Axolotl oocytes, which do not express endogenous CaCCs for an expression cloning strategy [22]. Size-fractionated mRNAs extracted from Xenopus oocytes, which highly express CaCCs, were injected into Axolotl oocytes for measurement of Ca²⁺-activated Cl^– currents. After cycles of functional screening, the TMEM16A gene was identified to confer robust CaCC currents in Axolotl oocytes. Heterologous expression of TMEM16A as well as TMEM16B in HEK293 cells further confirmed that these two proteins are the bona fide CaCCs. The identification of TMEM16B as a CaCC was subsequently confirmed by other groups [27–29].

6.2.2. TMEM16F Encodes a Dual Functional CaPLSase and Nonselective Ion Channel

It was surprising and puzzling that TMEM16F does not function as a CaCC despite the fact that it shares about 45% of sequence identity with TMEM16A/B CaCCs. Through a series of elegant experiments, the Nagata group made a surprising discovery that TMEM16F (ANO6) plays an indispensable role in mediating Ca²⁺-activated phospholipid scrambling to catalyze the flip-flop of phospholipids across the membrane [30]. The authors utilized the mouse Ba/F3 cell line and carried multiple rounds of annexin V (AnV)-based fluorescence-activated cell sorting (FACS) to identify a subpopulation of cells displaying enhanced phosphatidylserine (PS) externalization upon Ca²⁺ ionophore treatment (Fig. 6.1d). A gain-of-function mutation of TMEM16F, D409G, located in the first intracellular loop between TM2 and TM3 was identified to give rise to this enhanced PS exposure by this subpopulation. Interestingly, loss-of-function mutations of TMEM16F are responsible for Scott syndrome, a rare inherited bleeding disorder caused by a defect in Ca²⁺-activated phospholipid scramblase (CaPLSase)-mediated PS exposure in platelets both in humans [30–32] and in dogs [33]. The Jan group subsequently demonstrated that TMEM16F plays a key role in blood coagulation and the TMEM16F-null mice recapitulate the human Scott syndrome phenotype of prolonged bleeding [2]. Remarkably, the TMEM16F-null mice resist strong thrombotic challenges induced by FeCl₃, suggesting that TMEM16F could serve as a novel anticoagulant target to prevent thrombotic disorders such as stroke, heart attack, and venous thromboembolism. These early findings suggested that TMEM16F may serve as a CaPLSase or function as a key element for the CaPLSase.

One of the reasons why TMEM16F was not immediately considered as a bona fide CaPLSase was that it also has an ion channel function (Fig. 6.1c). Different studies from several groups suggested that TMEM16F may function as a small conductive nonselective cation channel (SCAN) [2, 3], an outward-rectifying chloride channel [34], volume-gated anion channel [35], a CaCC [36–40], or a CaCC of delayed activation [41]. Despite the controversial aspect of TMEM16F’s ion selectivity (see discussion in a later section), the proposal that TMEM16 proteins can support scrambling function was bolstered by the elegant structural and functional studies from the Dutzler and Accardi groups in which they were able to purify the TMEM16 fungal homologs, nhTMEM16 and afTMEM16, and demonstrated that they not only mediate Ca²⁺-dependent phospholipid scrambling [42, 43] but also conduct ions [43, 44]. Furthermore, the Hartzell and Galietta groups showed that the mammalian TMEM16F exhibits both ion channel and lipid scrambling activities in mammalian cells [45, 46]. Now it is generally accepted that TMEM16F is indeed a dual functional ion channel and lipid scramblase [10]. Although the biophysical properties of TMEM16C/D/G/J as lipid scramblases have not been fully characterized [47], TMEM16E and TMEM16K were recently shown to display both scrambling and ion channel activities [48–51].

6.3. Structure and Function of TMEM16 Proteins

6.3.1. Biophysical Properties of TMEM16 Ion Channels

Consistent with studies on endogenous CaCCs [52–54], channel gating of TMEM16A and TMEM16B involves the synergistic action of Ca²⁺ and membrane depolarization (Fig. 6.1c). Both channels display higher apparent Ca²⁺ sensitivity at more depolarizing voltages and higher concentrations of Ca²⁺ shift the conductance-voltage (G–V) relationship curves toward more negative voltages. TMEM16A is highly sensitive to Ca²⁺ with an estimated Ca²⁺ EC₅₀ in the low micromolar range from 0.4 to 1 μM at positive membrane potentials or from 0.7 to 6 μM at negative membrane potentials [23, 42, 55–62]. Despite sharing ~82% sequence identity with TMEM16A, TMEM16B displays a relatively lower Ca²⁺ sensitivity with an estimated Ca²⁺ EC₅₀ ranging from 1.2 to 3.3 μM at positive membrane potentials and from 1.8 to 4.9 μM at negative potentials [3, 27, 28, 63, 64], all of which are good agreement with studies on endogenous CaCC encoded by TMEM16B from olfactory sensory neurons [54, 65, 66].

Under low open probability (i.e., <1 μM Ca²⁺ for TMEM16A or 1–2 μM for TMEM16B), TMEM16A and TMEM16B currents show pronounced voltage-dependent outward rectification in addition to time-dependent activation and deactivation kinetics, with TMEM16B displaying faster activation and deactivation kinetics [21, 23, 27, 28]. The time dependence and voltage dependence quickly disappear under saturating Ca²⁺(i.e., >1 μM for TMEM16A or >10 μM for TMEM16B), and the currents follow Ohm’s law. Interestingly, similar to endogenous CaCCs [24], permeant anions also strongly impact the channel’s gating and kinetics. Larger and more permeable anions such as SCN^–, NO₃^–, or I^– pronouncedly enhance the channel’s open probability and accelerate channel activation kinetics [61, 67, 68]. This phenomenon could be explained by their higher occupancy within the pore which stabilizes the open state to allosterically modulate Ca²⁺-dependent gating [24].

In contrast to the CaCCs TMEM16A and TMEM16B, the small conductance and nonselective ion channel TMEM16F appears to be much less sensitive to Ca²⁺ with reported Ca²⁺ EC₅₀ values ranging from 3.4 up to 105 μM [2, 38, 41, 45, 69–72]. These large variations could be partly due to the pronounced desensitization of TMEM16F as well as the differences in experimental approaches including voltage-clamp protocols used and/or recording configurations. Nevertheless, similar to TMEM16A/B CaCCs, TMEM16F also exhibits voltage-dependent Ca²⁺ activation with depolarizing potentials enhancing its apparent Ca²⁺ sensitivity [2]. Most importantly, while TMEM16A/B CaCCs become constitutively open under saturating micromolar Ca²⁺, TMEM16F current remains strongly outward rectifying even at high Ca²⁺ and always requires membrane depolarization for activation (Fig. 6.1c).

Relatively small single-channel conductance values of mammalian TMEM16 channels hamper detailed single-channel analysis using patch clamp. TMEM16A’ single-channel conductance reportedly ranges from 2.6 pS [58], 3.5 pS [3] to 8.3 pS [23], whereas that of TMEM16B ranges from 1.2 to 3.5 pS [3, 27, 28]. TMEM16F has the smallest single-channel conductance of ~0.5 pS based on noise analysis [2]. Interestingly, the fungal afTMEM16 was reported to have ~300 pS of single-channel conductance based on reconstituted planar bilayer measurement [43]. Although nhTMEM16 was reported as a nonselective ion channel recently [44], despite the initial report suggesting that it functions as a sole CaPLSase without channel activity [42], its single-channel conductance remains unknown. Therefore, it is not clear whether the marked differences in single-channel conductance between the mammalian TMEM16 channels and the fungal afTMEM16 reflect the evolutionary divergence of different organisms or are solely due to the difference in the measurement methods.

6.3.2. Fluorescence Methods Enable Biophysical Characterization of TMEM16 CaPLSases

Due to the difficulties in monitoring phospholipid flip-flop at high resolution and high sensitivity, systematic functional characterization of TMEM16 CaPLSases is technically challenging. Flow cytometry is commonly used to monitor scramblase-mediated PS exposure from a large population of cells. In this method, Ca²⁺ ionophore-induced Ca²⁺ elevation activates CaPLSases, which translocate PS to the cell surface (Fig. 6.1d). The cell surface-exposed PS then recruits fluorescently tagged AnV, a PS-specific binding protein that serves as a readout of CaPLSase activity. Alternatively, fluorescently labeled phospholipids can be visualized via their internalization by CaPLSases. After bovine serum albumin (BSA) extraction of the uninternalized fluorescent phospholipids, the BSA-resistant fluorescence from the internalized phospholipids is measured by flow cytometry and used as a readout of CaPLSase activity. Nevertheless, the flow cytometry method lacks single-cell resolution and temporal resolution in monitoring PS exposure, which limits its application in studying the molecular mechanisms of TMEM16 CaPLSases.

The Hartzell group was first to demonstrate the feasibility of applying live-cell microscopy to monitor CaPLSase activities at single-cell resolution using microscopy [46]. In this method, the CaPLSase TMEM16F-mediated PS exposure was monitored by PS-binding probes LactoglobulinC2 fused to the Clover fluorescent protein (or LactC2) or AlexaFluor-conjugated Annexin V following activation by the Ca²⁺ ionophore A23187. This approach requires incubation of A23187 for 5 min in Ca²⁺-free solution followed by removal of A23187 and addition of 5 mM Ca²⁺ to mobilize Ca²⁺ entry via store-operated channels. Following activation, TMEM16F mediates PS externalization, and the binding of LactC2 or AnV to surface-exposed PS gives rise to the gradual increase in fluorescence signal at the cell membrane. Phospholipid scrambling of TMEM16F can be detected within several minutes following its activation. By combining this fluorescence-based assay with patch-clamp recordings in which Ca²⁺ was included in the pipette solution, the Hartzell group was also able to spontaneously measure CaPLSase and ion channel activities of TMEM16F for the first time. Also, similar to previous observations [38], even at high 200 μM Ca²⁺, TMEM16F’s current displayed a pronounced (nearly 10 min) delay before the development of both scrambling and channel activity.

Nevertheless, endogenous TMEM16F is robustly expressed in HEK293 and various other commonly used cell lines [37]. Therefore, endogenous CaPLSase activity can significantly contaminate the measurements of heterologously expressed TMEM16 CaPLSases. This presents a serious concern for reliably assessing loss-of-function CaPLSase mutations or mutations claimed to convert TMEM16A CaCC to a CaPLSase. To circumvent this issue, we generated a TMEM16F-null HEK293T cell line and optimized a microscopy assay by utilizing the fluorescently conjugated AnV to study TMEM16F’s Ca²⁺-dependent scrambling at a single-cell resolution [70]. Furthermore, instead of relying on Ca²⁺ entry via store-operated channels as previously done by Yu et al., we optimized the concentration of the Ca²⁺ ionophore ionomycin that can trigger sufficient Ca²⁺ entry as well as Ca²⁺ release from internal stores for TMEM16F activation. This assay enabled us to reliably interrogate the structure–function relationship of both the mammalian TMEM16F and the Drosophila TMEM16 homolog Subdued without the confounding effects from endogenous TMEM16F [73].

The Jan lab developed a new approach in studying TMEM16F scrambling activity by taking advantage of the fact that cells lacking TMEM16F displayed a defect in microvesicle release [74–76]. By chemically inducing mouse embryonic fibroblasts (MEF) or HEK293 cells using a combination of paraformaldehyde (PFA), dithiothreitol (DTT), and Ca²⁺, robust generation of giant plasma membrane vesicles (GPMVs) can be observed [77]. Importantly, this generation of GPMVs begins after the initiation of PS exposure and requires both TMEM16F-dependent phospholipid scrambling and TMEM16F-mediated Ca²⁺ entry. By stably expressing TMEM16F, this approach allows the characterization of either gain-of-function or loss-of-function TMEM16F mutants. The time-dependent TMEM16F-mediated Ca²⁺ entry can also be monitored using a Ca²⁺ indicator, and the onset of Ca²⁺ influx usually takes place between 10 and 30 min. While this novel method allows measurements of both TMEM16F-dependent GPMV generation and Ca²⁺ entry, it is apparent that these cells may undergo apoptosis during the very long (up to 60 min) time of monitoring. As we showed previously, Ca²⁺ ionophore stimulation of overexpressed TMEM16F in HEK293 cells could lead cell death within 20–30 min after treatment [70]. Also, as this approach only monitors the generation of GPMVs, direct monitoring of PS exposure cannot be examined.

The Accardi and Dutzler groups first succeeded in purifying the fungal homologs afTMEM16 and nhTMEM16 and reconstituting them in liposomes composed of phospholipids conjugated with the fluorophore 7-nitro-2,1,3-benzoxadiazol (or NBD) [78]. Dithionite is added to quench the NBD-labeled phospholipids at the outer leaflet. In the absence of a scramblase, dithionite will reduce the fluorescence signal by half; when an active scramblase is present, additional quenching will be observed as inner leaflet NBD-labeled lipids are being transported to the outer leaflet for quenching by dithionite. This approach bypasses the challenges in overexpressing fungal TMEM16 scramblases in mammalian cells for scrambling assay. However, similar to TMEM16 ion channels, given the importance of membrane lipids such as PIP₂ on TMEM16 functions [57, 79–84], removal of these proteins from their native lipid environment could potentially lead to alterations of their functionality. Furthermore, it was often observed that while Ca²⁺ enhances lipid scrambling of reconstituted fungal scramblases, they can spontaneously mediate lipid scrambling in the absence of Ca²⁺ [85]. Whether this is due to an intrinsic property of these scramblases or the artificial environment remains to be established. Finally, while fungal TMEM16 scramblases have been extensively studied in reconstituted systems, the mammalian TMEM16 scramblases, namely TMEM16F, have yet to be successfully studied using this system. This could be due to the potential instability of the purified mammalian counterparts and/or their complex regulatory properties.

The Nagata group pioneered a single-molecule approach in studying TMEM16F-dependent phospholipid scrambling [86]. TMEM16F protein is purified from Ba/F3 cells stably expressing a high level of mouse TMEM16F. The purified TMEM16F is integrated into a microarray containing membrane bilayers with asymmetrically distributed fluorescently labeled phospholipids. Single-molecule scramblase assay is carried out in which TMEM16F-mediated translocation of phospholipids is activated by infusion of 100 μM Ca²⁺. This assay revealed for the first time the remarkable efficiency of TMEM16F scramblase, which transports lipids at an estimated rate of ~45,000 lipids/s. This experiment also confirmed the “channel-like” biophysical property of TMEM16F and that it non-specifically transports lipids down their concentration gradients. Despite some requirements such as purification of TMEM16F with high homogeneity and stability as well as experimental setup and optimization of the microarray of the lipid bilayer, this novel single-molecule approach provides valuable information that cannot be otherwise attained from live-cell imaging or liposome-reconstituted assays.

6.3.3. Overall Architecture of TMEM16 Proteins

The ground-breaking X-ray structure of the fungal scramblase nhTMEM16 by the Dutzler group provided unprecedented details into the structural understanding of TMEM16 [42] (Fig. 6.2a, b). Subsequent structural studies on the fungal afTMEM16, mouse TMEM16A, mouse TMEM16F, and human TMEM16K revealed that these proteins adopt a highly similar overall architecture [42, 69, 87–91]. Consistent with previous biochemical and biophysical characterizations [60, 62, 92–94], TMEM16 proteins are assembled as homodimers in which the transmembrane domain consists of 10 TM segments flanked by a large N-terminal cytosolic domain (NCD) and a short C-terminal extension of TM10 (Fig. 6.2a). In TMEM16A and TMEM16F structures, there is a large extracellular domain formed by the extracellular loops of TM1–2, 3–4, 5–6, 7–8, and 9–10 and is stabilized by four disulfide bonds whose disruptions led to dysfunctional channels [62]. Interestingly, this large extracellular domain is not observed in the fungal nhTMEM16 and afTMEM16 as well as the human endoplasmic reticulum (ER) TMEM16K scramblases [48], which may reflect the different phospholipid environments in fungi or within the ER. Also, except for nhTMEM16 whose “domain-swapped” organization of the N- and C-termini endows intensive dimer interactions, inter-subunit interaction in TMEM16A, F, and K and afTMEM16 is mediated mostly by the extracellular halves of TM10. As a result, two large hydrophobic cavities, or dimer cavities, are present at the central axis of all TMEM16 proteins. As will be discussed later, these dimer cavities could provide phospholipid-binding sites and play critical roles in the lipid-dependent modulation of TMEM16 proteins.

Fig. 6.2.

Structural basis of TMEM16 proteins. (a) Topology of TMEM16 proteins. (b) Structure of the Ca²⁺-bound fungal nhTMEM16 (PDB 4WIS). (c) Structure of the Ca²⁺-bound mouse TMEM16A showing its overall architecture (left), Ca²⁺ coordinating residues (middle), and conformational change of TM6 upon Ca²⁺ binding. (d, e) Residues that are implicated in controlling phospholipid scrambling in the fungal nhTMEM16 (PDB 4WIS) and mouse TMEM16F (PDB 6QP6). (f) The lipid permeation pathway of the human TMEM16K as viewed from the ER lumen in its closed state (PDB 6R6X) and open state (PDB 5OC9)

Each monomer functions as an independent ion- and/or lipid-conducting unit [58, 95] whose activation is controlled by two highly conserved Ca²⁺ binding sites located within TMs 6–8 (Fig. 6.2a, b). The asymmetric hourglass-shaped substrate permeation pathway is formed by numerous hydrophilic as well as nonpolar residues from TMs 3–7 in which a central constriction site is formed just above the Ca²⁺ binding sites. A large body of studies have supported this hydrophilic cavity as the non-selective permeation pathway for ions and lipids in both TMEM16 scramblases and ion channels [42, 48, 70, 85, 88, 90, 91, 96, 97]. The hydrophilic grooves of both fungal afTMEM16 and nhTMEM16 and the human TMEM16K have been observed in an “open” conformation, which could represent a lipid-conductive state. Nevertheless, all current Ca²⁺-bound structures of TMEM16A and TMEM16F adopt a closed permeation pathway in which the pores are too narrow to allow ion passage. These diverse functional states will be discussed in detail in the later section.

6.3.4. Ca²⁺-Dependent Activation Mechanism

Prior to structural determination, mutagenesis studies have already provided valuable clues into the Ca²⁺-dependent activation of TMEM16A. Yu et al. identified E698 and E701 (TM7) and Tien et al. identified three additional residues, E650 (TM6) and E730 and D734 (TM8), as the potential Ca²⁺ binding residues in TMEM16A [60, 62] (Fig. 6.2c, numbering based on TMEM16A(a) isoform lacking the EAVK segment in the first intracellular loop). Subsequent structural studies not only validated these electrophysiological findings but also revealed three additional asparagine residues (N646 and N647 of TM6 and N726 of TM8) as important for Ca²⁺ binding [42, 90, 96]. Within each TMEM16 monomer, these acidic and asparagine residues together form two highly conserved Ca²⁺ sites, S1 and S2. While S1 is located toward the intracellular side and is coordinated by the carboxylate groups of E650, E698, and D734, S2 is coordinated by E701, D730, and the asparagine N647 (also partially by N646 and N726 [90]) (Fig. 6.2c). In an excellent agreement with previous predictions [24], the Ca²⁺ binding sites of TMEM16 are located within the membrane field. Another important feature of TMEM16 proteins is that their Ca²⁺ binding sites are located within the immediate vicinity permeation pathway, which not only confers their efficient Ca²⁺-dependent activation but also explains the synergistic coupling between permeant anions and Ca²⁺ ligands.

The ion conduction pathway of TMEM16A is partially enclosed by the interaction at the extracellular portions of TM4 and TM6 (Fig. 6.2c). The peripheral TM6 has been implicated in Ca²⁺-dependent channel activation of TMEM16A [90, 96]. In the absence of Ca²⁺, TM6 is entirely helical and adopts a kinked confirmation at the highly conserved G640, which causes the C-terminal segment of TM6 to swing away from TM7 and TM8. The kinked conformation of TM6 exposes both Ca²⁺ sites accessible to the cytosolic environment for Ca²⁺ binding from the intracellular side. The unliganded Ca²⁺ binding sites also make the intracellular vestibule of TMEM16A highly electronegative, which acts to impede Cl^– entry from the intracellular side [91, 98]. It was suggested that binding of Ca²⁺ ions to 4 highly acidic residues from TM7 and TM8 takes place first, and this then allows TM6 to form stabilizing interactions with the bound Ca²⁺ ions via N647 and E650 residues. The interactions between N647 and E650 with S2 Ca²⁺ and S1 Ca²⁺, respectively, cause a slight rotation of TM6 and result in the formation of a π-helix at the G640 position that is stabilized by the interaction between the carbonyl of Q642 with S2 Ca²⁺ ion (Fig. 6.2c). Ca²⁺-dependent rearrangement of TM6 triggers partial widening of the central constriction site, although the captured structure is still in a nonconductive state. Consistent with the importance of TM6 as a gating element, disrupting its G640 hinge via alanine and especially proline substitutions not only enhances the channel’s apparent Ca²⁺ sensitivity but also confers basal channel activity in the absence of Ca²⁺ [68, 90]. Also, Q645A and I637A mutations on TM6 strongly enhance the channel’s Ca²⁺ sensitivity and allow channel activation via membrane depolarization in the complete absence of intracellular Ca²⁺ [68, 96, 98]; by contrast, the P654A mutation, located toward the cytosolic end of TM6, markedly reduces the channel’s Ca²⁺ sensitivity.

In the dual-functional TMEM16F, a similar yet different conformational transition of the gating TM6 was also observed. In the absence of Ca²⁺, the cytosolic end of TM6 appeared mobile and moved away from TM4 in a direction opposite to that seen in TMEM16A’s TM6. The binding of two Ca²⁺ ions neutralizes the negatively charged and polar Ca²⁺ binding acidic residues, including N620, N621, and E624 of TM6, E667 and E670 of TM7, and E699 and D703 of TM8, all of which strikingly resemble those in TMEM16A. This binding allows TM6 to approach TM7 and TM8 via a rigid body swinging movement around the highly conserved G615, which is equivalent to TMEM16A’a G640. However, because TMEM16F’s TM6 lacks the insertion of a residue near its G615 hinge, Ca²⁺ binding did not result in the partial unwinding and π-helix formation of TM6. A similar transition from bent to straight conformations of TM6 was also observed in the structures of TMEM16F with zero or 1 Ca²⁺ bound, respectively [69]. While the fungal afTMEM16 and nhTMEM16 scramblases lack the conserved glycine hinge in mammalian TMEM16, their TM6 also undergoes a similar swinging movement around the equivalent region upon Ca²⁺ binding [88, 89]. Together, these studies underscore the functional importance of TM6 in both TMEM16 ion channel and lipid scramblase gating.

6.3.5. Voltage-Dependent Activation of TMEM16 Ion Channels

Compared to Ca²⁺ activation, the mechanism of voltage-dependent activation of TMEM16 channels is still under debate. This is mainly because TMEM16 proteins do not possess canonical voltage sensors as seen in the 6-TM cation channels, and it is challenging to separate the Ca²⁺- and voltage-dependent activation. In contrast to the Ca²⁺- and voltage-activated BK K⁺ channel, voltage alone cannot activate TMEM16 channels in the absence of Ca²⁺. As early studies predicted and recent structural and functional studies demonstrated [42, 52, 53, 90, 99], multiple Ca²⁺ binding sites are physically located in the membrane field electrical field. It is thus plausible that voltage-dependent Ca²⁺ binding is likely responsible for the voltage dependence of TMEM16 channels sites [52, 53, 99]. Nevertheless, a number of gain-of-function (GOF) mutations in the ion permeation pore such as Q645A or G640P challenged this mechanism as they can be activated solely by membrane depolarization without Ca²⁺ binding [68, 90]. To further understand the paradoxical voltage dependence of TMEM16A, Lam and Dutzler proposed that the apparent voltage dependence is derived from the highly negatively charged nature of the Ca²⁺ binding residues in TMs 6–8 which serve as an electrostatic gate in the ligand-free state [98]. The authors provided compelling electrophysiological and Poisson-Boltzmann calculations to demonstrate that binding of Ca²⁺ affects anion conduction by altering the electrostatics at the intracellular opening of the narrow neck via long-range Coulombic interactions. This mechanism nicely explains the strongly outward rectifying currents observed in Q645A and G640P GOF mutant CaCCs, in which impediment of Cl^– entry to the intracellular vestibule may be caused by the highly negatively charged residues of the Ca²⁺ binding sites whereas membrane depolarization favors Cl^– entry from the extracellular side. However, strong voltage dependence is also a hallmark of a number of TMEM16 dual-functional scramblase channels, some of which are more permeable to cations [2, 72, 73]. It is therefore unclear whether the same voltage-dependent mechanism proposed by Lam and Dutzler can also apply to the other TMEM16 channels that are not Cl^- selective.

By carefully characterizing two GOF mutations Q645A and I637A, which confer voltage-dependent activation without the requirement of Ca²⁺ binding, the Jan group proposed that TMEM16A could adopt multiple different open states depending on the Ca²⁺ occupancy at its binding sites [68]. The authors suggested that binding of a single Ca²⁺ triggers a partial opening of the steric gate to allow Cl⁻ to enter the channel pore, which then can undergo a fast voltage-dependent conformational change in TM6 to eventually allow Cl⁻ conduction to give rise to the outward rectifying current activation. More permeable ions such as I⁻ or SCN^- have a longer dwell time within the pore and thus could augment this voltage-dependent conformational change to enhance channel activation. The binding of the second Ca²⁺ allows TM6 to adopt a fully activated conformation and result in a linear Cl⁻ conductance. Nevertheless, the underlying voltage-sensing residue(s) remain unknown, although the authors suggested that K641 or even the acidic Ca²⁺ binding residues may play a role.

Interestingly, a mutagenesis study by Xiao et al. showed that neutralization of the 4 consecutive glutamates (444-EEEE-447) in the cytosolic TM2–3 loop, while not affecting Ca²⁺ dependent gating, significantly reduced the voltage-dependent activation of TMEM16A by right shifting the G–V curve [61]. On the other hand, deletion of the adjacent “c” splicing segment (448-EAVK-451) reduced both the Ca²⁺ sensitivity and the voltage-dependent channel activation in addition to enhancing the dissociation of Ca²⁺. These results are consistent with a previous study from the Galietta group who showed that skipping the EAVK segment reduced the voltage sensitivity of human TMEM16A [56]. Interestingly, E367 residue and equivalent acidic stretch 386-EEEEE-390 in the first intracellular loop of TMEM16B were also found to be important for voltage-dependent activation [63]. These findings highlight the importance of the TM2–3 loop in regulating TMEM16A’s voltage-dependent activation. However, how an intracellular loop affects voltage dependence is still unclear.

Taken together, while these structural and functional studies have provided important insights into the voltage-dependent activation mechanism of TMEM16A CaCC, future mechanistic studies are required for a comprehensive understanding of the voltage-dependent gating mechanism of both TMEM16 CaCCs and TMEM16 dual-functional channels and CaPLSases.

6.3.6. Ion Selectivity of the TMEM16A/B CaCCs

The anion selectivity of TMEM16A/B CaCCs partly arises from the presence of numerous basic residues located at both extracellular and intracellular vestibules of the channel that create an attractive environment for anions [91]. The central constricted region is formed by both polar and nonpolar residues. During permeation, these polar residues could compensate for the loss of anion-coordinating water molecules, whereas hydrophobic residues increase the energetic penalty for smaller anions and thus favor larger anions [90]. This mechanism explains the lyotropic permeability sequence of SCN⁻ > NO₃⁻ > I⁻ > Br⁻ > Cl⁻ > F⁻ seen in both TMEM16A and TMEM16B [3, 22, 23, 27, 28, 59, 61, 64, 100], which strikingly resembles endogenous CaCCs [24, 67, 101, 102]. Also, large anions can shed their hydration shells faster than Cl⁻, allowing them to enter the pore quickly [24]. However, these large ions display poor conductance, which reflects how fast the ions dissociate from the pore and traverse the channel. In other words, large anions enter quickly but get lodged within the pore. It should be noted that anionic conductance in CaCCs displays a bell-shaped relationship with their hydration energies in which anions with the lowest or highest hydration energies have the lowest conductance; this is simply because anions with the lowest permeabilities cannot easily enter the pore due to their large hydration energies, and anions with the highest permeabilities enter the pore quickly but interact strongly with the pore and are thus poorly conducted. Finally, hydrophobic anions such as SCN^- or C(CN)₃^-, which are highly permeable to CaCCs, exert pore blockage effects on CaCCs, consistent with the presence of hydrophobic residues within the channels’ permeation pathway [90, 96, 102].

The polar and non-polar pore-lining residues that form the central constriction neck in TMEM16A have been shown to be important for ion selectivity in TMEM16A [96]. Alanine mutations including N542A, D550A, N587A, V595A, Q705A, and F712A further enhanced the channel’s permeability toward large anions such as I⁻ and SCN⁻ by increasing P_I/P_Cl and P_SCN/P_Cl ratios, whereas S635A reduced both. Notably, in addition to their roles in discriminating anions [68, 96], mutations of these pore-lining residues were also found to modulate the channel’s Ca²⁺-dependent activation, perhaps via their allosteric effects on the channel activation gate [68, 70, 96]. Dang et al. showed that while N542A, I546A, Y589A, I592A, and F708A significantly enhanced, V595A and L639A reduced the channel’s apparent Ca²⁺ sensitivity [96]. Remarkably, I546A and I637A mutations, located in TM4 and TM6, respectively, not only enhance the Ca²⁺ sensitivity but also allow the channel to conduct basal Cl⁻ currents in the complete absence of Ca²⁺ [68, 90]. We showed that the L543K mutation in TM4 of TMEM16A, which provides prominent steric hindrance to the central constriction gate, not only endows voltage-dependent activation but remarkably also turns the TMEM16A mutant channel into a constitutively active lipid scramblase [70]. Together with previous observations from endogenous CaCCs [24], these studies suggest that in TMEM16A/B CaCCs, anion occupancy within the channel’s amphipathic central constricted region strongly influences Ca²⁺-dependent channel gating.

6.3.7. Ion Selectivity of the Dual-Functional TMEM16F

The nature of ion selectivity of TMEM16 scramblases, most notably TMEM16F, remains highly controversial. We previously showed that TMEM16F is actually more permeable toward cations such as Ca²⁺ and Na⁺ than anions with a P_Na/P_Cl of ~6.8 [2]. However, other studies reported that TMEM16F is rather a poorly selective channel with a P_Na/P_Cl of ~1.3 [71] or ~1.4 [46] or even more selective towards Cl⁻ with a P_Na/P_Cl of 0.5 [45] or 0.3 [41]. The discrepancies in ion selectivity values of TMEM16F among different labs may reflect the different methodologies as well as the assumption of using “impermeant” ions such as Cs⁺, aspartate, or NMDG⁺. In fact, TMEM16F was shown to display significant permeability toward NMDG⁺ [2, 46] and aspartate with a P_Asp/P_Cl = 0.5 [41]. The fact that TMEM16F conducts Cs⁺, NMDG⁺, or aspartate implies a possibility that the presumably “impermeant” ions such as Cs⁺, NMDG⁺, and MES⁻ used in these studies could contribute significantly to the measured currents and therefore confound the permeability measurements. This promiscuous selectivity of TMEM16F may be explained by the “lipidic pore” model proposed by the Hartzell group who suggested that the observed current in TMEM16F could be a result of a leaky product during lipid permeation through the large hydrophilic cavity [103]. Adding to this complexity is the recent finding that TMEM16F’s ion permeation undergoes a dynamic change in its ion selectivity [72]. Ye et al. showed that this change is not due to different channel activation states but is likely caused by an alteration in the electrostatic field of the permeation pathway. Given these observations, it is tempting to speculate that TMEM16F pore is a highly dynamic structure that can undergo significant widening or opening to accommodate large molecules such as phospholipids and different ions. Future studies are needed to further understand the dynamic nature of ion permeation through the TMEM16 dual-functional channels and CaPLSases.

6.3.8. Phospholipid Permeation Through TMEM16 CaPLSases

How phospholipids and ions permeate through TMEM16 CaPLSases has been extensively studied using various structural, functional, and computational methods. Different models have been proposed to describe the dual permeation of phospholipids and ions. In this section, we summarize the major findings that support these models.

6.3.8.1. Classical “Credit Card” Model for Phospholipid Permeation

The X-ray structure of the fungal scramblase nhTMEM16 by the Dutzler lab [42] beautifully demonstrated that TMEM16 CaPLSases may also utilize the “credit card” mechanism (Fig. 6.2b), a prevailing model for transporter-mediated phospholipid flip-flop [104], to conduct phospholipid permeation. According to this “credit card” mechanism, the phospholipid headgroup slides through a hydrophilic protein groove facing the hydrophobic membrane core with the phospholipid acyl tails remaining within the membrane core (Fig. 6.1a). Indeed, each nhTMEM16 monomer has a hydrophilic furrow or groove that consists of both polar and charged residues and faces the lipid environment with TM4 and TM6 at the protein–lipid interface [42]. Coarse-grained molecular dynamics simulations of nhTMEM16 in phosphatidylcholine from the Samson group first showed that lipid headgroups could indeed occupy this hydrophilic groove [105].

To understand how lipid permeation may take place in nhTMEM16 and how lipids interact with the hydrophilic groove, Bethel and Grabe performed atomistic molecular dynamics simulations and continuum membrane-bending calculations using the nhTMEM16 structure (PDB 4WIS) as a template [106]. They discovered two lipid-interacting sites flanking the hydrophilic groove that could serve as “stepping stones” for lipid permeation: an extracellular S_E site of E313 (on TM3) and R432 (on TM6) and a cytosolic S_C site of E352 and K353 (both at the intracellular end of TM4) (Fig. 6.2d). While residues constituting the S_E site appear to be switched in positions in mammalian TMEM16, they are highly conserved with the exceptions of TMEM16H and TMEM16K. Consistent with their roles in lipid scrambling, Feng et al. showed that the equivalent S_E site R478A mutation delayed the onset of GPMV generation, whereas Gyobu et al. showed that both R478A and E604C mutations significantly impaired PS exposure in TMEM16F [69, 107]. In TMEM16A CaCC, S_E sites residues R511 and E619 are equivalent to E313 and K432, respectively, and R511 was previously shown to be important for TMEM16A’s ion selectivity [100] while E619 plays an important role in proton sensing [55]. The S_C site, which resides in the scrambling domain previously proposed by the Hartzell group [46], serves to nucleate headgroup-dipole stacking interactions and to promote lipid penetration into the groove (Fig. 6.2e). Using a chimeric approach, the Hartzell group narrowed down to a stretch of 35 amino acids from N525 to Q559 of TMEM16F, which is equivalent to TMEM16A D550 to K584, that function as a “scrambling domain” (SCRD) in TMEM16F [46] (Fig. 6.2e). Remarkably, the introduction of this segment to TMEM16A endows its ability to scramble lipids while retaining its Cl⁻ selective ion conduction. Gyobu et al. further showed that substitution of the equivalent SCRD of the ER TMEM16E scramblase also confers scrambling activity to TMEM16A [50]. However, we later found that a single mutation TMEM16A L543K, which is outside of this SCRD, is sufficient to convert this CaCC into a constitutively active scramblase [70]. This raises the question of whether the scrambling activity in TMEM16 scramblases indeed requires this 35 amino acid stretch and how the residues within the long stretch control lipid scrambling.

6.3.8.2. Membrane Bending/Distortion Is a Common Feature in TMEM16 Scramblases

Molecular dynamics and structural studies all revealed an interesting phenomenon in which TMEM16 scramblases pronouncedly deform the membrane, albeit at different regions. Using fast and continuum membrane-bending calculations, Bethel and Grabe first showed that nhTMEM16 induces large-scale deformation of the membrane, which results in thinning of the bilayer across the hydrophilic groove of ~36%, thus significantly shortening the travel distance for lipids. Similar membrane distortions were also observed in several simulations studies by other groups [85, 97]. Remarkably, this phenomenon was subsequently confirmed by structural studies of nhTMEM16 and afTMEM16 either in a lipid-like nanodisc environment or detergent [88, 89]. Importantly, this bending of the membrane by TMEM16 scramblases does not depend on Ca²⁺ binding as it was observed in all nanodisc-reconstituted TMEM16 scramblases. Falzone et al., on the other hand, observed pronounced membrane bending at the dimer cavity, which is likely caused by the fact that at this dimer interface, TM3 and TM5 of one monomer are longer than TM1 and TM2 of the other monomer. Interestingly, in the open active afTMEM16 scramblase, the electron density of the nanodisc at the subunit cavity appears weaker, which is consistent with a thinner and/or disordered membrane that facilitates lipid permeation. In fact, in the ceramide-inhibited afTMEM16 structure, although the Ca²⁺-induced conformation is similar to that of the active scramblase, the density of nanodisc at the subunit cavity is stronger, presumably explaining its inhibitory effect by modulating the lipid environment and not Ca²⁺-dependent conformational transition. In the structures of the mouse TMEM16F, Alvadia et al. and Feng et al. did not observe any obvious membrane distortion in the nanodisc-reconstituted mTMEM16F. However, Feng et al. did observe that PIP₂ supplementation allows TM6 to adopt a kink conformation that in turn causes membrane thinning at TM3 and TM4 and that this membrane distortion is essential to lipid scrambling in TMEM16F [69]. Taken together, these atomistic and structural analyses suggest a possibility that TMEM16 CaPLSases have evolved a special arrangement at their protein–lipid interface to facilitate phospholipid permeation by membrane thinning and deformation.

6.3.8.3. “Lipidic Pore” Dual Permeation Model Derived from the “Credit-Card” Model

Experimental studies and MD simulations from the Hartzell and Tajkhorshid groups suggested that at least in the nhTMEM16 scramblase, both ions and lipids traverse the subunit cavity or hydrophilic groove [97] (Fig. 6.3a). The authors found three different sites for lipid interaction: S_int of Q374 and N378 of TM5, R505 of TM7; S_cen (about one third into the membrane from the intracellular side) of N378, T381, S382 (TM5), and T340 (TM4); and S_ext (near the extracellular entrance) of E313, N317 (TM3), K325, Q326, T333 (TM3), R432, N435, and Y439 (TM6). R432 of TM6 is part of the S_E site identified in Bethel and Grabe. Notably, two full events of PS lipid permeation from the inner leaflet to the outer leaflet were observed during which the PS headgroup interacts mostly with the oxygen or nitrogen atoms of R505 of the S_int site. Furthermore, T333 and Y439 of the S_ext appear to form a steric gate to control the lipid permeation pathway. In the absence of Ca²⁺, TM4 transitions closer toward TM6 to establish an interaction between T333 and Y439, thereby forming a constriction site to close the lipid pathway. This observation was supported by the greatly reduced scrambling activity by the T333V mutation, likely due to the increased hydrophobicity. However, in a reconstituted system, T333W showed no obvious reduction in scrambling [85]. Interestingly, the single TMEM16A mutations V543S, V543T, and K588N, when expressed in HEK293 cells that robustly express endogenous TMEM16F [37, 73], convert the ion channel into a scramblase [97]. These observations are interesting but are not readily reconcilable as V543 is not part of the previously proposed scrambling domain.

Fig. 6.3.

Mechanistic models of phospholipid and ion permeation by dual-functional TMEM16 CaPLSases/channels. The proposed models for lipidic pore (a), out-of-the-groove (b), and alternating pore-cavity (c) mechanisms are shown

By performing Trp mutagenesis on residues lining the hydrophilic groove of nhTMEM16, Lee et al. found three distinct regions that are important for lipid permeation: (1) the lower constriction site of L302, T430, T381, and S382 near the midpoint of the membrane, (2) the extracellular entrance comprising E313 and R432, and (3) A395, Q436, Y439, and F440 residues between the lower constriction site and the extracellular entrance [85]. Their MD simulations studies further showed that a triad of charged residues, E313, E318, and R432, are critical for lipid permeation gating in nhTMEM16. These residues not only interact with the permeating lipids but also undergo dynamic rearrangement during the gating process. In fact, the interaction of phospholipids with R432 of TM6 promotes sequential disengagement from E313 and E318 of TM3 during permeation. Due to this disengagement, TM6 rotates away from TM3 and allows Y439 of TM6 to move away from T333 of TM4, thereby widening the groove to enable lipid permeation. Furthermore, Y439 is not only important for gating but is also critical for lipid coordination. Thus, Lee et al. proposed that lipid translocation requires a widening of the hydrophilic groove, a process that likely cannot be achieved by channel-only TMEM16 proteins such as TMEM16A and TMEM16B. Interestingly, Trp mutations that result in reduced scrambling also affect the ion conduction of nhTMEM16, which is consistent with the hypothesis that ions and lipids share a common permeation pathway formed by the hydrophilic groove.

6.3.8.4. “Ions-in-the-Pore and Lipids-Out-of-the-Groove” Dual Permeation Model

The Accardi group reconstituted purified afTMEM16 and nhTMEM16 scramblases into liposomes and investigated whether they can transport phospholipids whose headgroups are derivatized with polyethylene glycol (PEG) moieties of various sizes [43]. Remarkably, both fungal scramblases can scramble lipids conjugated with PEG with sizes ranging from 2000 to 5000 Da, which are equivalent to roughly 8–40 Å in diameter and are thus much larger than the hydrophilic groove. Notably, these large lipids are transported at rates that are comparable to those of normal-sized lipids. This led to the proposal that these large lipids surf on the surface of TMEM16 scramblases without having to permeate through the hydrophilic cavity, or via an “out-of-the-groove” mechanism (Fig. 6.3b), resembling the proposed phospholipid permeation mechanism for the Ca²⁺-independent scramblase opsin [108]. Using the lipid pathway R432W mutant, Malvezzi et al. showed this mutation more strongly impaired scrambling of the smaller lipids than PEG-conjugated lipids, which is consistent with PEG-conjugated lipids permeating outside the groove whereas permeation of small lipids occurs within the groove. As TMEM16 scramblases also mediate ion transport, believed to occur through the hydrophilic groove [69, 87, 90, 91, 96], translocation of large lipids via this out-of-the-groove model should therefore impose minimal effects on their ion permeation. Indeed, afTMEM16-mediated scrambling of PEG-conjugated lipids does not affect ion permeation or ion selectivity to NMDG⁺ of afTMEM16, further supporting this out-of-the-groove model. This also suggests that ion permeation in TMEM16 proteins likely takes place via a proteinaceous pore (ions-in-the-pore) and that permeation of large lipids does not induce dilation of the pathway.

The recent structural studies suggested that the “ions-in-the-pore and lipids-out-of-the-groove” model may also apply to mammalian TMEM16 CaPLSases. Alvadia et al. reported that TM4 is in contact with TM6 to enclose the permeation pathway in both unliganded and Ca²⁺-bound TMEM16F structures [87]. The authors thus speculated that in mammalian TMEM16 scramblases, at least in the case of TMEM16F, lipid permeation could take place via the “out-of-the-groove” mechanism. The Jan and Cheng groups also made similar observations for TMEM16F structures in either digitonin or nanodisc conditions [69]. Feng et al. observed that while PIP₂ supplementation promoted widening of the intracellular portion of the pore, the putative lipid permeation pathway still adopted an enclosed conformation analogous to that observed in TMEM16A [90, 91, 96] and TMEM16F [87]. Owing to the lack of an open hydrophilic groove, Feng et al. proposed a different “out-of-the-groove” model that is dependent upon PIP₂ binding, which causes a kink in TM6 as well as membrane distortion and thinning. The authors suggested that TMEM16F harbors distinct permeation pathways for ions and lipids, and that membrane distortion plays a key role in lipid scrambling and that lipid scrambling may take place even in the absence of an open hydrophilic groove.

6.3.8.5. “Alternating Pore-Cavity” Dual Permeation Model

To reconcile the difference between the “lipidic pore” model and the “out-of-the-groove” model, an alternative model of the “alternating pore-cavity” model was proposed [87] (Fig. 6.3c). In this model, ions and lipids are proposed to traverse by distinct conformational states of TMEM16F. Lipid permeation requires a widened hydrophilic groove as that seen in the structures of nhTMEM16 [42, 89], afTMEM16 [88], or human TMEM16K [48], whereas ion conduction is mediated by an enclosed pathway, which is likely captured in the Ca²⁺-bound TMEM16F [69, 87] and resembles that of TMEM16A structures [90, 96]. Although the “alternating pore-cavity mechanism” helps reconcile the discrepancy between the “lipidic core” model and the “ions-in-the-pore and lipids-out-of-the-groove” model, it is unclear how CaPLSases can precisely control the fast switching between the ion permeation state and the phospholipid permeation state with high efficiency and high fidelity. Additional functional and computational studies are needed to validate this model.

6.3.9. TMEM16 CaPLSase Gating

Scramblases and ion channels are passive membrane transporters that permeate their substrates at high speed. This requires that their permeation must be tightly controlled or gated between closed and open conformations. Numerous recent structural and functional studies have provided valuable insights into TMEM16 lipid permeation gating.

Owing to the lack of a mammalian TMEM16 scramblase structure at the time, we utilized the structures of nhTMEM16 (4WIS) and TMEM16A (5OYB and 5OYG) as templates to generate homology models of TMEM16F scramblase in an open, intermediate, or closed state [70]. We then performed atomistic MD simulations of these TMEM16F models to observe lipid permeation as well as the potential conformational transitions between the open and the closed states. We noticed that three hydrophobic residues, F518 (TM4), Y563 (TM5), and I612 (TM6), together form a hydrophobic constriction gate that impedes lipid permeation. Interestingly, whereas the side chains of these residues are in proximity in the closed state, they separate to widen the hydrophilic groove in the Ca²⁺-bound open state as if they serve as an inner gate at the cytosolic mouth of the permeation pathway. Indeed, by applying our optimized microscopy-based lipid scrambling assay, we found that removing steric hindrance of the inner gate via alanine substitutions of F518, Y563, and I612 strongly promote lipid scrambling in TMEM16F following Ca²⁺ ionophore treatment. Consistent with this observation, introducing large hydrophobic side chains (i.e. F518L or Y563W mutations) strongly reduces their scrambling rates. By contrast, introducing polar or charged mutations to these three hydrophobic residues make TMEM16F constitutively open. In the case of F518K or Y563K, the gain-of-function effects are highly potent such that they result in constitutive scrambling activities even when the Ca²⁺ binding sites are disrupted. These results led us to propose a “clamp shell” gating model for TMEM16 CaPLSases. We suggested 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. F518 in TM4 and I612 in TM6 likely serve as gate-keepers for the opening of this interface, whereas Y563 in TM5 likely serves as a cap that stabilizes the inner gate and obstructs phospholipid permeation in the closed state. Ca²⁺ binding-induced conformational changes lead to the separation of the TM4–TM6 interface as well as the inner gate residues, subsequently exposing the interior of the hydrophilic groove to the surrounding phospholipids such that their phospholipid headgroups can enter and translocate via the credit-card mechanism. On the one hand, introducing smaller, polar, or charged residues to these critical locations removes the steric hindrance, resulting in enhanced phospholipid permeation. On the other hand, the inner activation gates of F518K and Y563K mutants are severely disrupted that phospholipids can freely go through the constitutively open gate in the absence of Ca²⁺ binding.

Interestingly, we found that channel opening by two TMEM16A mutations L543K and I637K, which are equivalent to TMEM16F F518K and I612K, respectively, can be elicited solely via membrane depolarization without Ca²⁺. Strikingly, L543K also conferred constitutively active lipid scrambling activity to the CaCC in the absence of Ca²⁺. These experiments support the notion that TMEM16 CaCCs and CaPLSases may utilize a similar set of hydrophobic residues at their inner activation gate to control ion and phospholipid permeation. Instead of opening like a “clam shell” in TMEM16 CaPLSases to separate TM4 and TM6, the helices maintain in contact in TMEM16 CaCCs to exclude phospholipid headgroups to penetrate and permeate. Ca²⁺ binding only dilates the proteinaceous pore to allow Cl⁻ to go through. When a positive charge is introduced to L543 in TM4, the interaction between TM4 and TM6 is likely weakened such that the mutant TMEM16A can open like a clam shell to allow spontaneous permeation of phospholipids. Our findings thus suggest that TMEM16 CaPLSases and CaCCs may share similar Ca²⁺-dependent gating mechanisms 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.

Bushell et al. applied X-ray crystallography and cryo-EM to capture the human TMEM16K in multiple different states, including closed, intermediate, and open states [48]. Most interestingly, their X-ray structure of the Ca²⁺-bound TMEM16K revealed an open hydrophilic groove, which is thus far the only open groove confirmation of a mammalian TMEM16 scramblase (Fig. 6.2f). The transition from an open to a closed groove conformation is associated with several prominent structural rearrangements, including a ~10-degree rotation of the N-terminal cytosolic domain (NCD), separation of the interdimeric interaction between the C-terminal segments of TM10 (TM10’ or αTM10) that results in a ~30% reduction in the dimer interface, and finally the pivoting movements of TM3 and TM4 around TM5, all of which culminate in the closure of the hydrophilic groove at the ER luminal side. In this closed groove conformation, an intensive network of residues from TM4–7, including Y366, A367, L416, S415, T435, L436, and Y507, form a hydrophobic gate and likely occlude lipid permeation.

By determining the cryo-EM structures of the afTMEM16 scramblase, Falzone et al. observed global conformational changes upon Ca²⁺ binding [88]. In the absence of Ca²⁺, the bending of both TM4 and TM6 results in a complete enclosure of the subunit cavity that is accompanied by the upward movement of TM3 and the dilation of the Ca²⁺ binding sites. Notably, Ca²⁺ binding results in the straightening of both TM4 and TM6, allowing TM6 to disengage from TM4 to open the lipid translocation pathway. The pronounced bending of TM4 is likely mediated by P324 and P333 residues (equivalent to nhTMEM16 P332 and P341), which was also observed in the nhTMEM16 scramblase [89]. Falzone et al. speculated that afTMEM16 could also adopt an “intermediate” state in which TM4 is bent while TM6 is straight, and that this intermediate state is only ion conductive.

To understand the structural transition during Ca²⁺-dependent gating of the nhTMEM16 scramblase, the Paulino and Dutzler groups determined the Ca²⁺-bound and Ca²⁺-free cryo-EM structures of nhTMEM16 in detergent DDM and lipid-like nanodisc (2N2) [89]. These structures revealed the remarkable conformational changes that nhTMEM16 may undergo upon Ca²⁺ binding. In the absence of Ca²⁺, TM4 undergoes a pronounced movement toward TM6 to enclose the subunit cavity from the membrane’s hydrophobic environment, resembling the apo structures of TMEM16A. In the presence of Ca²⁺, nhTMEM16 can adopt three major conformations at its subunit cavity, tentatively assigned as “Ca²⁺-bound closed,” “intermediate,” and “Ca²⁺-bound open.” First, in the “Ca²⁺-bound closed” state, nhTMEM16 structure highly resembles that of the Ca²⁺-free closed state, and the subunit cavity is shielded from the membrane via tight interactions between TM4 and TM6. Second, in its “intermediate” conformation, while the subunit cavity is still enclosed from the membrane, the cavity has widened relative to that of the closed state. Third, in the “Ca²⁺-bound open” state, the subunit cavity of nhTMEM16 is exposed to the lipid environment reminiscent of the Ca²⁺-bound open structure in detergent. This widening of the subunit cavity is likely mediated by the pronounced orientation of both TM3 and TM4 toward TM10 the adjacent subunit. In transitioning from the closed to the open state, TM4 undergoes the largest helical rearrangements around the potential pivot points consisting of P332 and P431 as well as G339. In fact, mutations of P431 and G439 both significantly reduced nhTMEM16’s lipid scrambling both in the absence and in the presence of Ca²⁺.

This open state is believed to represent the lipid scrambling-competent state of nhTMEM16. On the other hand, observation of the intermediate state, which likely does not promote scrambling, led Kalienkova et al. to propose an equilibrium model for the Ca²⁺-bound and Ca²⁺-free states in which the partially enclosed “intermediate” state represents the ion-conductive state whereas the full open cavity represents the lipid-conductive state. In agreement with these observations, Khelashvilli et al. showed that nhTMEM16 can transition from an open to an intermediate state which only allows ion conduction based on atomistic simulations and structural determination [109]. The middle region of TM4 moves toward TM6 to partially enclose the hydrophilic groove, thus impeding lipid permeation. They also revealed that the hydrophobic interaction between TM3 and TM4, mediated by L302 of TM3 and I343 and L347 of TM4, helps stabilize the open membrane-exposed lipid pathway. Consistent with this observation, disrupting this hydrophobic interaction via the L302A mutation which, despite strongly impairing lipid permeation, retains its ion channel activity. The cryo-EM structure of L302A mutant nhTMEM16 further confirmed that the scramblase indeed adopts an intermediate state that strikingly resembles their MD simulated structure and is similar to the intermediate state of the WT nhTMEM16 reported by Kalienkova et al. Interestingly, in the nhTMEM16 L302A structure, the interaction between E313 and R432 (the S_E site) is disrupted, resulting in an open extracellular gate as well as a continuous pore that is wide enough to accommodate water, Na⁺ or K⁺ ions.

6.4. Future Prospective

Since the identification of TMEM16A as a CaCC in 2008, the collective efforts of the TMEM16 field have greatly advanced our understanding of these intriguing membrane transporters and their contributions to human health and diseases. Given the wealth of structural information available and various established functional assays, we are at the exciting stage to comprehensively understand the molecular mechanisms of different TMEM16 proteins, including (1) the precise biophysical properties of the intracellular TMEM16 proteins; (2) the molecular basis of TMEM16 voltage dependence and its cooperativity with Ca²⁺ in activating TMEM16s; (3) the allosteric gating mechanisms that couple different parts of the proteins; (4) the relationship between ion and phospholipid permeation through the dual-functional TMEM16 proteins and their underlying structural basis; (5) the regulatory mechanisms of TMEM16 proteins by various factors and cell signaling pathways; (6) the identification of TMEM16-specific pharmacological reagents and their working mechanisms to influence TMEM16 functions. At the cellular level, we need to gain a better understanding of (1) the cellular functions of the intracellular TMEM16 CaPLSases; (2) how TMEM16 proteins are activated and regulated under physiological and pathological conditions; and (3) the cellular functions mediated by TMEM16 channels and CaPLSases in various cell types. There is also an urgent need to decipher the physiological roles of TMEM16 proteins using animal models as well as the pathophysiology of human TMEM16 mutations. We anticipate that answers to these questions will significantly promote our understanding of this important family of membrane transport proteins and facilitate the future design of TMEM16-specific therapies to treat numerous diseases such as heart attack, stroke, asthma, epilepsy, ataxia, muscular dystrophy, viral infection, pregnancy complications, and cancer.