Nonspecific Membrane Interactions Can Modulate BK Channel Activation

Large-conductance potassium (BK) channels are transmembrane (TM) proteins that can be synergistically and independently activated by membrane voltage and intracellular Ca2+. The only covalent connection between the cytosolic Ca2+ sensing domain and the TM pore and voltage sensing domains is a 15-residue “C-linker”. To determine the linker’s role in BK activation, we designed a series of linker sequence scrambling mutants to suppress potential complex interplay of specific interactions with the rest of the protein. The results revealed a surprising sensitivity of BK activation to the linker sequence. Combing atomistic simulations and further mutagenesis experiments, we demonstrated that nonspecific interactions of the linker with membrane alone could directly modulate BK activation. The C-linker thus plays more direct roles in mediating allosteric coupling between BK domains than previously assumed. Our results also suggest that covalent linkers could directly modulate TM protein function and should be considered an integral component of the sensing apparatus. One-sentence summary The covalent linker between BK intracellular and transmembrane domains is likely a part of the sensing apparatus that regulates the channel activation.


Introduction
Widely distributed in nerve and muscle cells, large-conductance potassium (BK) channels are characterized by a large single-channel conductance (~ 100 -300 pS) (1)(2)(3)(4)(5) and dual activation by both intracellular Ca 2+ and membrane voltage (6)(7)(8), thus an ideal model system for understanding the gating and sensor-pore coupling in ion channels. BK channels are involved in numerous vital physiological processes including intracellular ion homeostasis and membrane excitation, and are associated with pathogenesis of many diseases such as epilepsy, stroke, autism and hypertension (9). Functional BK channels are homo-tetramers, each containing three distinct domains (Figure 1a). The voltage sensor domain (VSD) detects membrane potential, the pore-gate domain (PGD) controls the K + selectivity and permeation, and the cytosolic tail domain (CTD) senses various intracellular ligands including Ca 2+ . The VSD and the CTD also form a Mg 2+ binding site for Mg 2+ dependent activation (Yang 2008 NSMB). The tetrameric assembly of CTD domains is also referred to as the "gating ring". VSD and PGD together form the trans-membrane domain (TMD) of BK channels. Previous studies on mouse BK channels (10) and recent atomistic structures of full-length Aplysia californica BK channel (aSlo1) (11,12) reveal that CTD of each subunit reside beneath the TMD of the neighboring subunit in a surprising domain-swapped arrangement ( Figure S1a).
The only covalent connection between CTD and TMD of BK channels is a 15-residue peptide referred to as the "C-linker" (R329 to K343 in the human BK channel, hSlo1) (green in Figure 1a). This linker directly connects the pore lining S6 helices in the PGD (yellow in Figure S1b) to the Nterminus of CTD (known as RCK1 N-lobe; blue in Figure S1b), and is believed to play an important role in mediating the gating ring-pore coupling (1,8,9). For example, Niu et al. (13) observed that lengthening the C-linker through inserting poly-AAG (Table S1) was accompanied with right shifted G-V curves, while shortening the C-linker led to left-shifted G-V curves, clearly demonstrating the importance of C-linker in BK gating. Intriguingly, the voltage required for half activation, V0.5, displayed a highly linear relationship with the number of residues inserted or deleted in the absence of Ca 2+ . This led to the proposal that the linker-gating ring behaves as a "passive spring" in activation of BK channels (13), suggesting that the C-linker largely provides an inert linkage.
Evidence has also accumulated to suggest that the C-linker may play more direct roles in mediating allosteric coupling of BK channels. Both Ca 2+ -free and bound Cryo-EM structures of full-length aSlo1 (11,12) were found to contain wide-open pores and thus did not provide clue to how the channel might be gated. Atomistic simulations later revealed that, due to the movement of pore-lining S6 helices (Figure 1b), the BK pore cavity become narrower, more elongated, and crucially, more hydrophobic in the metal-free state (i.e., without bound Ca 2+ and Mg 2+ ; presumably the closed state) (14). As such, the pore can readily undergo hydrophobic dewetting transition to give rise to a vapor barrier that prevents ion permeation (14). Recognition of hydrophobic gating in BK channels provides a mechanistic basis for further understanding how the C-linker may mediate the sensor-pore coupling. Specifically, key movements of S6 helices involve bending at the glycine hinge (G310, G311) toward the membrane upon Ca 2+ binding (Figure 1b), leading to ~6 Å expansion of the pore entrance near I323. Considering that C-linkers are directly connected to S6 helices ( Figure S1b), their interactions with the rest of the channel as well as membrane and water will likely have an effect on the S6 orientation and consequently BK channel activation.
It has also been proposed that BK channels may be gated at the selectivity filter (15)(16)(17), the conformation of which could be modulated by the S6 helix orientation and thus C-linker interactions. The notion that interactions of C-linkers can modulate BK activation has actually been demonstrated in a recent study, where the R329KK331 residues in the C-linker was proposed to form alternating interactions with E321 and E224 from the neighboring chains and membrane lipids during each gating cycle (18).
A key challenge of using mutagenesis to delineate the roles of a specific residue or interaction in protein function is that multiple competing effects may be perturbed simultaneously. The C-linker, in particular, is involved in extensive interactions with the gating ring and VSD ( Figure S2), making it very difficult to derive unambiguous interpretation of its role in BK activation (18). In this work, we examine a set of BK channels with scrambled C-linker sequences to determine if the C-linker is largely inert in mediating the sensor-pore coupling of BK channels. Combining atomistic simulation, mutagenesis and electrophysiology, we discover that the C-linker is a major pathway of gating ring-pore communication and can play a much more direct and specific role in mediating BK activation than previously thought. In particular, we show that nonspecific interactions of the C-linker with the membrane/solvent environment, namely, membrane anchoring effects, can directly modulate voltage gating of BK channels. This observation has been verified by additional experiments performed on BK constructs either lacking CTD or with the membrane anchoring residue mutated. Orientation of the pore lining S6 helices and C-linkers in the metal-free (green) and metal-bound (red) states; The dash lines indicate the approximate positions of membrane interfaces. The location of Tyr 332 and Gly 310 are indicated by yellow and blue spheres in the metal-bound and metal-free states, respectively. The black arrow shows S6 movement upon metal binding. c. Macroscopic currents of WT, K0, K2 and K7 hSlo1 channels. The currents were elicited in 0 [Ca 2+ ]i by voltage pulses from -30 to 250 mV with 20 mV increments for WT and K2 and voltage pulses from -80 to 200 mV with 20 mV increments for K0 and K7. The voltages before and after the pulses were -50 and -80 mV, respectively. d. Conductance-voltage (G-V) curves for WT, K0, K2 and K7 hSlo1 channels in 0 [Ca 2+ ]i showing significant shifts in the activation voltage (V0.5); All solid lines were fit to the Boltzmann relation (see Methods), with V0.5 of 183.4 ± 3.2 mV for WT; 89.6 ± 3.5 mV for K0; 195.5 ± 3.5 mV for K2 ; and 48.7 ± 4.7 mV for K7.

Sequence scrambling of the C-linker dramatically modulates BK voltage activation
To further resolve whether the C-linker is largely an inert linkage (as part of the linker-gating ring "passive spring") or it has more direct roles (mediated through various specific and nonspecific interactions of C-linker residues), we investigated a series of BK mutants where the C-linker sequence is scrambled (Table 1). If the linker-gating ring largely acts like a passive spring with an inert C-linker, the expectation is that these scrambling mutant BK channels would have similar gating properties. Table 1 shows experimental results on the scramble mutant channels. Among these mutants K3, K5 and K6 did not show functional expression, while other mutant channels showed robust currents (Figure 1c). We measured voltage dependent activation of these channels and the conductance-voltage (G-V) relationships were fitted using the Boltzmann function to derive V0.5, the voltage where G/GMax reaches 0.5 (Figure 1d). Left shift of G-V (V0.5 Table 1: C-linker scrambling mutations and measured V0.5 in the full-length and Core-MT BK channels at [Ca 2+ ] = 0 and [Mg 2+ ] = 0. The Core-MT constructs are based on the TMD, C-linker of mSlo1, and an 11-residue tail from KV 1.4 of the mouse Shaker family (19,20). The location of the nearest Tyr to the S6 C-terminal is highlighted in green. K0 (Y330G) was designed to remove the Tyr sidechain in the K0 background. Importantly, as shown in Figure 1d and  Figure S2). Atomistic simulations also suggest that the homology models and Cryo-EM structures lead to similar structural and dynamical properties (e.g., see Figure S3). The linker forms extensive contacts with the RCK1 N-lobe (H344 to N427) of the gating ring, mainly mediated by the Y332GGSYSA338 segment in the C-linker, and S0' of VSD ( Figure S3). In particular, the two conserved tyrosine residues (Y332 and Y336) are fully embedded in a hydrophobic RCK1 N-lobe pocket, apparently maintaining a tight packing between the C-linker and RCK1 N-lobe ( Figure S3). Several positively charged residues (R329, K330, K331, R342 and K343) flank the above segment and are exposed to solvent, likely mediating the solubility of the C-linker. Importantly, these two short tracks of charged residues appear to be tightly anchored by the C-linker-RCK1 contacts.

Mutation
The stability of the C-linker conformation as a structured loop is further confirmed by atomistic molecular dynamic (MD) simulations, which showed that the C-linker maintained stable conformations and contacts with VSD and RCK1 N-lobe throughout the 800 ns simulation time Dynamic network analysis further reveals that the C-linker is a key pathway of dynamic coupling between the gating ring and PGD. Such analysis utilizes correlation of residue motions during MD simulations to uncover probable pathways of allosteric coupling in biomolecules (22)(23)(24). The optimal and suboptimal paths of coupling are then identified as the shortest paths with the highest pairwise correlations, which should possess the highest probabilities of information transfer (22).
We analyzed the optimal and suboptimal pathways of coupling between I323 in the S6 helices, where substantial conformational change occurs during the gating event (14), and critical metal binding residues, including D895 and R514 in the RCK2 and RCK1 Ca 2+ binding site, respectively, and E374 in the Mg 2+ binding site residing in the CTD. The results reveal that the main pathways of communication from Ca 2+ and Mg 2+ binding sites to the PGD, go through the C-linker for all four chains (Figure 2c-d and S9). This is not necessarily surprising since C-linker is the only covalent connection between the domains. Interestingly though, in the metal-free state the main path from the RCK2 Ca 2+ binding site to PGD goes from the neighboring monomer in every other chain (Figure 2c-d). This can be attributed to much tighter S6 helix packing in the metal-free state (14), and, combined with the domain swapped arrangement of BK tetramers, may help enforce cooperative gating response upon metal binding.

Scrambling mutations minimally perturb C-linker structure and dynamics
Atomistic simulations suggest that the changes in voltage dependent gating are unlikely to derive from a change of the structural features or functional roles of the permutated linkers. The structure of the channel appears minimally perturbed by the mutated linkers, with the overall RMSD below ~5 Å and the TM domain (core) RMSD around 2-3 Å from the initial cryo-EM-derived structures for both WT and mutant channels ( Figure S8a-b). All mutant channels can readily undergo hydrophobic dewetting transitions as observed for the WT channel ( Figure S8c). The linker region also maintains similar backbone conformations in WT and all mutants ( Figure S7), even though it becomes slightly more dynamic in the K7 mutant as reflected in the RMSF profile ( Figure S7b).
Furthermore, scrambling mutations do not appear to perturb long-range coupling properties either; the C-linker remains to provide the key pathway of dynamic coupling between CTD and PGD.
Sequence properties, particularly distributions of charged residues, can modulate intrinsic conformational properties of disordered peptides including chain extension (25), which could explain the changes in channel activation in terms of the linker-gating ring passive spring model.
To investigate this hypothesis, we performed atomistic simulations of the isolated C-linker segments in the ABSINTH implicit solvent (26), which has been shown to reliably predict the inherent conformational extension of disordered peptides (27). The results suggest that V0.5 of linker scrambling mutants is not correlated with the inherent linker extension ( Figure S10) and is inconsistent with the prediction based on the previous linker-gating ring passive spring model (13) (dashed line in Figure S10). This observation further supports the notion that the C-linker is unlikely an inert component of the linker-gating ring passive spring.
Finally, since the C-linker is a structured loop with extensive interactions with the CTD and VSD ( Figure S2, S5-6), we considered whether some specific subsets of mutations could perturb the coupling among VSD, PGD and CTD to modulate voltage dependent activation (V0.5), even though sequence scrambling is designed to suppress such effects. To test this, we analyzed Clinker residue contact probability maps from MD trajectories to identify potential group of contacts that could be correlated to V0.5 ( Figure S5-6). However, none of the observed changes in specific protein-protein contacts could explain the measured V0.5 variations. As discussed above, key conformational changes involved in activation of BK channels include the re-orientation of the pore-lining S6 helices, which bends outwardly and toward the membrane at the glycine hinge (G310 and G311) (Figure 1b). Directly connected to the S6 helix, C-linker residues would be moved closer to the membrane interface during activation (e.g., comparing red vs green cartoons in Figure 1b). As such, it can be anticipated that nonspecific interactions of the C-linker with the membrane interface can affect channel activation by stabilizing the bended conformation of S6 helices, thus modulating the activation voltage. Among various amino acids, aromatic ones such as Tyr and Trp are known to be "membrane anchoring", with a strong preference towards localizing at the membrane-water interface (28,29). While the sequence scrambling was designed to suppress the potential effects of specific interactions of the C-linker, the two Tyr residues (Y332 and Y336 in WT) were placed at different separations from the membrane interface (Table 1, highlighted in green fonts). Indeed, the measured V0.5 shows a strong correlation with the sequence separation between E324 (the approximate location of membrane interface) and the nearest C-linker Tyr residue ( Figure 3)  To more directly examine if membrane anchoring plays a role in modulating BK activation, we further analyzed the atomistic trajectories to understand the details of Tyr interaction with the membrane interface. As illustrated in Figure 4, Tyr sidechains contain both aromatic rings and polar groups that allow them to embed their aromatic rings in the lipid tail region and at the same time direct the polar -OH groups toward the lipid headgroup region to form hydrogen bonding interactions with water and lipid headgroups (29). In addition, the aromatic ring could also form pcation interactions with positively charged cholines in lipid headgroups (30). Importantly, results from simulation analysis confirm that positioning of the C-linker in the metal-bound state allows more extensive interactions between Try sidechains and the membrane interface (e.g., compare  has been shown to be extremely slow at the multi-µs level or longer (31).

Tyr membrane anchoring affects BK channel activation similarly without the gating ring
If the observed effects on the voltage gating of full-length BK channels are indeed mainly due to the nonspecific interactions between the C-linker and membrane interface, they should persist in the Core-MT truncated channel (19,20), in which the whole gating-ring is removed and there is no CTD coupling with either VSD or PGD via the C-linker (32) (Fig 5a). To test this prediction, Core-MT channels with three C-linker scrambling mutations, K0, K2 and K7, were expressed ( Fig   5b) and their voltage dependent activation measured. Mutations of the C-linker shifted the G-V relation of the Core-MT constructs in the same directions as observed for the full-length channels (Table 1), with K0 and K7 stabilizing channel activation (Fig 5c, e) while K2 making activation at higher voltages (Fig 5d). However, the reduction in the activation voltage by K0 mutation is only ~42 mV in Core-MT, compared to ~94 mV in the full-length channel (Figure 6c), and by K7 mutation is ~67 mV in Core-MT, compared to ~135 mV in the full-length channel (Figure 6e). This is not considered surprising as the C-linker is more flexible in the absence of the gating ring, thus weakening the effects of membrane anchoring. The observation that linker sequence scrambling mutations can modulate BK activation even in the Core-MT background is remarkable, providing a direct evidence that the C-linker is more than an inert, passive covalent linker for coupling the gating ring and PGD. Instead, the linker plays a more direct and more specific role in modulating the opening of BK pore, such as through its interactions with the membrane environment. As such, the linker could be considered an integral component of the sensing apparatus.

Removal of membrane anchoring Tyr in a C-linker BK mutant recovers WT-like gating
We note that Tyr is not the only type of residues being shuffled in the sequence scrambling (Table   1) and that interactions of other residues, particularly charged ones (18), with membrane and water could also affect the open/close equilibrium of the channel. This may explain why K1 and K2 mutant channels have different V0.5, even though the nearest Tyr is at position 331 in both mutants (Table 1). To further examine if Tyr residues indeed provide the dominant contributions, we replaced Y330 with Gly in the K0 mutant to completely remove the aromatic side chain. The mutant expressed robust currents (Figure 7a). Strikingly, K0 Y330G mutation abolished the effects of K0 on G-V relation and largely shifted the G-V back to that of the WT, with a V0.5 of 169.8 ± 5.0 mV as compared to 183.4 ± 3.2 mV for the WT (Figure 7b). This, together with the correlation shown in Figure 3, provides a direct support that membrane anchoring effects of Tyr are mainly responsible for modulating the activation voltage in the C-linker sequence scrambling mutants.

Conclusion
We have combined atomistic simulations and experiment to determine the role of the 15-residue C-linker in the sensor-pore coupling of BK channels. As the only covalent connection between PGD and CTD of BK channels, the C-linker has been widely assumed to play an important role in mediating the information transfer and domain coupling. Our analysis show that C-linker is a structured loop with extensive contacts with CTD and VSD, and remain highly stable in both metalbound (activated) and metal-free (deactivated) states. Dynamic coupling analysis confirms that C-linker is the key pathway of the senor-pore coupling. However, the linker is not just an inert Our conclusion that the C-linker is unlikely an inert component of the linker-gating ring "passive spring" is also consistent with other structural and functional studies of BK channels. For example, the Cryo-EM structures of full-length BK channels (11,12) reveal that the previously published poly-AAG insertion site (13), right after residue S337 (Table S1), actually locates in a short loop following the C-linker segment -Y332GGSY336-that forms stable contacts with RCK1 N-lobe ( Figure   S2). The inserted residues would project away from the channel ( Figure S13), and are very unlikely to affect the effective C-linker length (or the gating ring-pore distance) as originally designed. Instead, the observed effects in BK gating V0.5 upon insertion/deletion of C-linker residues could likely be attributed to certain nontrivial structural and/or dynamical impacts, such as weakening of the VSD/CTD interactions. Another important evidence that is inconsistent with the passive spring model comes from the study of Core-MT BK channels. Since the whole gating ring is removed, the Core-MT construct should correspond to a state where the passive spring is fully relaxed and thus V0.5 maximizes. Yet, the V0.5 of WT Core-MT is only ~52 mV larger than the full-length BK channel (Table 1). This is far below what may be expected based on poly-AAG insertion mutants, which can increase V0.5 by ~142 mV with (AAG)3 inserted (13) (also see Table  S1). Interestingly, the recently published Cryo-EM structures confirms that b subunits make extensive contacts with the C-linker for influencing the channel gating and manipulating the channel function (21).
TM ion channels and receptors frequently contain separate TM domains, which directly mediate function such as ion permeation, and intracellular and extracellular sensing domains, which often control the function of TM domains in response to various chemical signals (33). The general role of the covalent linkers connecting the sensing and TM domains is of great general interest (34,35). A central question is whether the linker mainly provides an inert and passive connection between sensing and TM domains or it should be considered an element of signal sensing itself.
Dissecting the potential roles of covalent linkers is challenging because multiple sources of interactions and conformational transitions could impose competing strains on the linker to modulate functional regulation. It is difficult to design experiments that can unambiguously test and validate whether a particular type of strain imposed on the covalent linker could lead to predictable functional outputs. Linker sequence permutation could provide an effective strategy to suppress the potential consequence of specific (but unknown) linker interactions, allowing one to test the functional role of a single type of strain imposed on the linker (such as membrane anchoring). Our findings show that non-specific interactions of the C-linker can regulate BK voltage gating. Therefore, covalent linkers of membrane proteins could serve as sensors of signals that perturb their interaction with the environment, which in term can modulate the functional center in the TM domain, may it be the gate of ion channels or intramolecular signaling pathway in receptors.

Acknowledgements:
We thank Rohit Pappu for the original design of C-linker mutants. This work was supported by National Institutes of Health Grants R01 HL70393 and GM114300 (to Chen).
We also would like to thank Roderick MacKinnon and Xiao Tao for sharing the hSlo1 Cryo-EM structures prior to their release in PDB. Computing for this project was performed on the Pikes cluster housed in the Massachusetts Green High-Performance Computing Center (MGHPCC).

Materials and Methods
Homology modeling and atomistic simulations. As described previously (14), homology models of the WT metal-bound and metal-free hSol1 channels were built using Modeller v9.14 (36) based on sequence alignment of Tao et al (11). Sequence alignment shows 55.95% identity for the full-length channels. The sequence identities in PGD is higher at 61.96%. The high level of sequence identity suggests that the homology models are likely reliable. This has been confirmed by direct comparison with the recently published Cryo-EM structures of hSlo1 (21) ( Figures S2). The backbone RMSD between the model and new structure is only 2.18 Å at the whole channel level and as low as 0.87 Å in the PGD. Structures for C-linker scrambling mutants were build based on the WT hSlo1 models using CHARMM (37).
Using CHARMM-GUI server (38), the hSlo1 structures were inserted in POPC lipid bilayers followed by solvation using the TIP3P water model (39). 450 mM KCl was then added, same as used in Cryo-EM structure determination (11,12). Each system was first energy minimized, Structural and dynamic analysis: All analysis were performed using a combination of in-house scripts, MDAnalysis (46) and Gromacs2016 (47,48) software. Only snapshots from the last 150 ns of all production MD trajectories were used for the calculation of Tyr-membrane interactions (SASA of burial, hydrogen bonding, p-cation and carbon-carbon contacts) as well as RMSF and the C-linker contact map (SI). A (hydrophobic) carbon-carbon contact was considered formed if the distance is no greater than 4.5 Å. The p-cation interaction was identified when the distance between the center of mass of the Tyr aromatic ring and Nitrogen atom of POPC choline group is no greater than 5.0 Å. Similarly, the cutoff was set at 5.0 Å for calculation of the C-linker contact map (SI) while only the heavy atoms of each residue were considered. Hydrogen bonds were analyzed using the MDAnalysis "HyrogenBondAnalysis" class with default criteria. The number of pore water molecules was calculated using the same criteria as described previously (14).
Dynamic network analysis was performed using the Networkview (22) plugin of VMD (49). For this, snapshots were extract every 1 ns from the 800 ns molecular dynamic trajectories. To build the network, each amino acid was represented as a single node at their Cα position and a contact (edge) was defined between two nodes if the minimal heavy-atom distance between them was within a cutoff distance (4.5 Å) during at least 75% of the trajectory. The resulting contact matrix were then weighted based on the correlation coefficients of dynamic fluctuation (Cij), calculated using the Carma software (50), as wij = -log (|Cij|), where Cij = <Dri(t).Drj(t)> / (<Dri(t) 2 > <Drj(t) 2 >) 1/2 and Dri(t) = ri(t) -<ri(t)>, ri(t) is the position of the atom corresponding of the i th node and <> denotes ensemble average (over the MD trajectory). The path length between the desired nodes were then calculated as the sum of the edge weights. The shortest (optimal) path, calculated using Floyd-Warshall algorithm (51), is believed to represent the dominant mode of communication.
Slightly longer (suboptimal) paths were also calculated. VMD was used for preparing all molecular illustrations. Electrophysiology Data analysis. Relative conductance (G) was obtained by measuring macroscopic tail current at -80 mV or -120mV. The conductance-voltage (G-V) relationships was plotted to fit with the Boltzmann function:

Mutations and
Where G/GMax means the ratio of conductance to maximal conductance, z means the number of          . Optimal (red) and suboptimal (green) dynamic coupling pathways between the critical residues in the Ca 2+ and Mg 2+ binding sites (Ca colored as purple sphere) and I323 (Ca colored as yellow sphere) in PGD. Whole channel is shown as transparent ribbon with each chain colored differently. S6 and C-linker of chain C are colored in gray and shown with a cartoon representation. a. Path between R514 (located in RCK1 Ca 2+ binding site in CTD; purple sphere) in chain C to I323 (located in S6 helix of PGD; yellow sphere) in chain C. b. Path from E374 (located in Mg 2+ binding site in CTD; purple sphere) in chain C to I323 (located in S6 helix of PGD; yellow sphere) in chain C. c. Path from D99 (located in Mg 2+ binding site in VSD; purple sphere) in chain B to I323 (located in S6 helix of PGD; yellow sphere) in chain C. See Methods for additional details. Figure S10. Correlation between the average intrinsic end-to-end distance of free C-linkers and measured V0.5 of the WT hSlo1 and mutants. The average distances were calculated from ABSINTH simulations of isolated C-linker peptides. The red dashed line shows the linear relationship derived from the C-linker insertion/deletion study (Niu et al 2004; also see Table S1). . Figure S12. A representative snapshot of metal-bound K0 channel showing membrane distortion around the protein. a. Top view, b. side view. The POPC phosphorous atoms have been colored according to their distances to the membrane center. S6 helices are colored in magenta and Tyr 330 side chains are shown as pink spheres. Figure S13. A representative structural model of hSlo1 with 12 residues (3x AAG units; colored red) inserted after S337 in the C-linker region (colored green). a. side view; b. top view. For clarity, only the C-linker and S6 helix (yellow) of chain A (colored as blue ribbon) are highlighted.