1. Structural Biology and Molecular Biophysics
Download icon

Mechanism of pharmacochaperoning in a mammalian KATP channel revealed by cryo-EM

Research Article
  • Cited 0
  • Views 1,038
  • Annotations
Cite this article as: eLife 2019;8:e46417 doi: 10.7554/eLife.46417

Abstract

ATP-sensitive potassium (KATP) channels composed of a pore-forming Kir6.2 potassium channel and a regulatory ABC transporter sulfonylurea receptor 1 (SUR1) regulate insulin secretion in pancreatic β-cells to maintain glucose homeostasis. Mutations that impair channel folding or assembly prevent cell surface expression and cause congenital hyperinsulinism. Structurally diverse KATP inhibitors are known to act as pharmacochaperones to correct mutant channel expression, but the mechanism is unknown. Here, we compare cryoEM structures of a mammalian KATP channel bound to pharmacochaperones glibenclamide, repaglinide, and carbamazepine. We found all three drugs bind within a common pocket in SUR1. Further, we found the N-terminus of Kir6.2 inserted within the central cavity of the SUR1 ABC core, adjacent the drug binding pocket. The findings reveal a common mechanism by which diverse compounds stabilize the Kir6.2 N-terminus within SUR1’s ABC core, allowing it to act as a firm ‘handle’ for the assembly of metastable mutant SUR1-Kir6.2 complexes.

https://doi.org/10.7554/eLife.46417.001

Introduction

ATP-binding cassette transporters (ABC transporters) comprise a large protein superfamily responsible for transporting diverse molecules across cell membranes (Thomas and Tampé, 2018; Trowitzsch and Tampé, 2018; Wilkens, 2015). Uniquely, the sulfonylurea receptors SUR1 and SUR2 lack transport activity per se but instead are devoted to forming the ATP-sensitive potassium (KATP) channels, wherein four SUR.x subunits form a complex with four subunits of an inwardly rectifying potassium channel, Kir6.1 or Kir6.2 (Aittoniemi et al., 2009; Ashcroft and Ashcroft, 1990; Bryan et al., 2007; Nichols, 2006). KATP channels are gated by intracellular ATP and ADP, key features which enable them to regulate membrane excitability according to the energetic state of the cell (Nichols, 2006). Expressed in many cell types, they have critical roles in endocrine, cardiovascular, neurological and muscular functions (Aguilar-Bryan and Bryan, 1999; Foster and Coetzee, 2016; Nichols et al., 1996). In pancreatic β-cells, KATP channels composed of SUR1 and Kir6.2 couple glucose metabolism to insulin secretion (Aguilar-Bryan et al., 2001; Ashcroft, 2005). Gain-of-function mutations in the SUR1 gene ABCC8 or the Kir6.2 gene KCNJ11 are a major cause of neonatal diabetes, while loss-of-function mutations result in congenital hyperinsulinism (Ashcroft et al., 2017; Koster et al., 2005; Stanley, 2016). Among the latter, severe hyperinsulinism requiring pancreatectomy to prevent life-threatening hypoglycemia is often the result of KATP channel mutations that reduce KATP channel density on the β-cell surface by impairing channel biogenesis, assembly, and trafficking (referred to as trafficking mutations hereinafter) (Huopio et al., 2002; Vajravelu and De León, 2018).

Previously, we showed that KATP channel inhibitors commonly used to treat Type 2 diabetes are efficient pharmacochaperones for correcting surface expression defects of certain trafficking-impaired mutant KATP channels, including sulfonylureas (SUs) and glinides (Yan et al., 2004; Yan et al., 2007). Most trafficking mutations identified to date are in SUR1, perhaps due to its large size compared to Kir6.2 (Martin et al., 2013; Snider et al., 2013). The SUR1 protein has an N-terminal transmembrane domain referred to as TMD0 that interacts directly with the Kir6.2 subunit (Babenko and Bryan, 2003; Chan et al., 2003; Lee et al., 2017; Li et al., 2017; Martin et al., 2017b; Schwappach et al., 2000), followed by a cytoplasmic loop termed L0, and then the ABC core structure comprising two transmembrane domains TMD1 and TMD2 and two nucleotide binding domains NBD1 and NBD2 (Aguilar-Bryan et al., 1995; Martin et al., 2017b). Of particular interest, while trafficking mutations are found throughout SUR1, SUs and glinides have to date only corrected defects caused by mutations located in TMD0, the domain contacting Kir6.2. Recently, we found that carbamazepine (CBZ), an anticonvulsant structurally not related to SUs and glinides, also corrects KATP channel trafficking defects (Chen et al., 2013; Sampson et al., 2013) and surprisingly is effective only for mutations in TMD0 of SUR1 (Devaraneni et al., 2015; Martin et al., 2016). Furthermore, like SUs and glinides, CBZ inhibits KATP channel activity by reducing channel Po and abolishing channel responsiveness to MgADP (Zhou et al., 2014). The striking similarities of SUs, glinides and CBZ in their effects on the channel despite their chemical uniqueness suggest a shared pharmacochaperoning mechanism which remains elusive.

In this study, we took a comparative structural approach to understand how glibenclamide (GBC, a sulfonylurea), repaglinide (RPG, a glinide), and CBZ, interact with the channel to affect channel biogenesis and function. Using single-particle cryo-electron microscopy (cryoEM), we determined structures of the channel in the presence of GBC, RPG, and CBZ, and without pharmacochaperone. The structures show that like GBC (Martin et al., 2017a; Wu et al., 2018), RPG and CBZ occupy the same binding pocket in the transmembrane bundle above NBD1 of SUR1. Further, we undertook structural, biochemical and functional studies to determine the involvement of the distal N-terminal 30 amino acid stretch of Kir6.2, which is functionally critical to the actions of these drugs but whose structure has not yet been clearly defined. The combined results from these studies provide strong evidence that the distal Kir6.2 N-terminus is located in the cavity formed by the two halves of the SUR1 ABC core and is adjacent to the drug binding pocket. The study reveals how a chemically diverse set of KATP channel inhibitors allosterically control channel gating, and also promote the assembly and trafficking of nascent channels to the cell surface, by stabilizing a key labile regulatory interaction between the N-terminus of Kir6.2 and the central cavity of the ABC core of SUR1.

Results

Structure determination

For structure determination, a FLAG-tagged hamster SUR1 and a rat Kir6.2 (95% and 96% identical to human, respectively) were overexpressed in the insulinoma cell line INS-1, and the channel complex affinity purified via the FLAG-epitope tag as described for our recent cryoEM structure determination of KATP bound to GBC and ATP (Martin et al., 2017a; Martin et al., 2017b). These channels have been used extensively for structure-function and pharmacochaperone studies and are thus well characterized (Inagaki et al., 1995; Shyng et al., 1998; Yan et al., 2004). To allow direct comparison of channel structures containing bound RPG or CBZ with the GBC/ATP-bound structure solved previously (Martin et al., 2017a), data were similarly collected in the presence of ATP (1 mM without Mg2+) but varying the pharmacochaperone by alternatively including 30 μM RPG (referred to as the RPG/ATP state), 10 µM CBZ (referred to as the CBZ/ATP state) or the drug vehicle 0.1% DMSO (referred to as the ATP-only state) in the sample. Further, as an additional control we determined the structure of channels without any pharmacochaperone or ATP (referred to as the apo state).

Initial data processing in RELION with C4 symmetry imposed yielded one dominant 3D class for each dataset, with an overall reported resolution ranging from ~4 Å for the RPG/ATP and CBZ/ATP states to ~7 Å for the ATP-only and the apo states (Figure 1; Figure 1—figure supplements 14; Table 1). As with the GBC/ATP state structures we reported previously (Martin et al., 2017a; Martin et al., 2017b), increased disorder was observed at the periphery of the complex in all structures possibly due to minor deviations from C4 symmetry. To see whether resolution of SUR1 could be improved, focused refinement using symmetry expansion and signal subtraction with different masks was performed on all structures, including the previously collected GBC/ATP dataset (Martin et al., 2017a), as described in Materials and methods. This yielded maps with improved reported resolutions for SUR1 as follows: 3.74 Å for the GBC/ATP state, 3.65 Å for the RPG/ATP state, 4.34 Å for the CBZ/ATP state, 4.50 Å for the ATP-only state, and 4.55 Å for the apo state using the FSC cutoff of 0.143 (Table 1).

Figure 1 with 5 supplements see all
Structural determination and comparison.

(A) Unsharpened 3.9 Å C4 cryoEM reconstruction of the KATP channel bound to RPG and ATP. (B) Structural model of the channel in the RPG/ATP state. (C) Overlay of the RPG/ATP state structure, the ATP only state structure, and the apo state structure viewed from the top showing similarity of the dominant class of the ATP only and the apo state to the RPG/ATP state structure. (D) Left: Same model as (C) viewed from the side and focusing on the ABC transporter core module of SUR1 to illustrate the inward-facing conformation observed in all three structures. Right: Separation between Walker A and the signature motif in NBD1 and NBD2 (G716::S1483 and S831::G1382; Cα to Cα, indicated by the dashed line) in SUR1 bound to RPG and ATP (green) and ATP only (tan) viewed from the bottom.

https://doi.org/10.7554/eLife.46417.002
Table 1
Statistics of cryo-EM data collection, 3D reconstruction and model building.
https://doi.org/10.7554/eLife.46417.008
Data collectionRpg/ATPCbz/ATPGbc/ATPATP onlyApo
MicroscopeKriosKriosKriosKriosArctica
Voltage (kV)300300300300200
CameraFalcon IIIGatan K2Gatan K2Gatan K2Gatan K2
Camera modeCountingSuper-resolutionSuper-resolutionSuper-resolutionSuper-resolution
Defocus range (µm)−1.0 ~ −2.6−1.4 ~ −3.0−1.4 ~ −3.0−1.4 ~ −3.0−1.4 ~ −3.0
Movies57654413218023442047
Frames/movie24060606060
Exposure time (s)6015151515
Dose rate (e-/pixel/s)~0.78888
Magnified pixel size (Å)1.0451.72*1.72*1.71*1.826**
Total Dose (e-/Å^2)~40~40~40~40~40
Reconstruction
Whole channel
SoftwareRelion 3.0Relion 2Relion 2Relion 2Relion 2
SymmetryC4C4C4C4C4
Particles refined24,747138,00063,22780,30434,527
Resolution (masked)3.9 Å4.4 Å4.07 Å4.88 Å5.31 Å
SUR1 focused refinement
SoftwareRelion 3.0Relion 3.0Relion 3.0Relion 3.0Relion 3.0
SymmetryC1C1C1C1C1
Particles refined312,000499,095171,420123,75790,058
Resolution (masked)3.65 Å4.34 Å3.74 Å4.5 Å4.55 Å
Model Statistics(Includes KNt)(Includes KNt)
Map CC (masked)0.65590.77710.70650.81550.7677
Clash score3.377.523.222.603.11
Molprobity score1.581.951.601.461.5
Cβ deviations00000
Ramachandran
Outliers0%0%0%0%0%
Allowed7.03%9.66%7.98%6.29%5.93%
Favored92.97%90.34%92.02%93.71%94.07%
RMS deviations
Bond length0.0090.0080.0070.0050.009
Bond angles1.1661.1210.9921.0851.233
  1. *Super-resolution pixel size 0.86; **Super-resolution pixel size 0.913.

For model building, we used our previously published GBC/ATP structure (PDB: 6BAA) (Martin et al., 2017a) as a starting point and refined the models against the experimental data. For the GBC/ATP structure, the new map derived from focused refinement of the Kir6.2 tetramer plus an SUR1 subunit showed significantly improved cryoEM density in a number of regions of SUR1 previously not modeled in the GBC/ATP map (see deposited PDB and EMD files). This allowed us to build additional residues into the GBC/ATP structure. In particular, the ATP density in NBD1 became clearly resolved (Figure 1—figure supplement 5A; note ATP binds NBD1 in the absence of Mg2+[Ueda et al., 1997]). The linker between TMD2 and NBD2 (L1319-Q1342) also showed much improved density, allowing us to build a continuous polyalanine model (Figure 1—figure supplement 5B). Models for the other structural states were similarly built according to the highest resolution maps available for the focus-refined regions (for details see Materials and methods).

In all structures determined in the presence of pharmacochaperones and ATP, the Kir6.2 tetramer is in a closed conformation (Figure 1C), and the SUR1’s ABC core in an ‘inward-facing’ conformation wherein the NBD1 and NBD2 are separated (Figure 1D). This overall structure is very similar to that previously reported for the GBC/ATP state (Martin et al., 2017a; Martin et al., 2017b). Interestingly, the dominant class emerging from 3D classification for the ATP-only state as well as the apo state presented a similar conformation, with Kir6.2 tetramer closed and SUR1 inward-facing (Figure 1C and D). Of note, a dominant inward-facing conformation in the absence of ligand was also observed in another ABC transporter, the multidrug resistance protein MRP1 (Johnson and Chen, 2017). These observations suggest the closed channel conformation is the most stable for the apo state under our experimental conditions (i.e. no ATP and no exogenous PIP2, a lipid required for KATP channel opening [Nichols, 2006]). More data and extensive analyses will be needed to resolve whether other minor conformations are present in our samples and to understand the dynamics of these structures, especially in the ATP-only and the apo states. Here, we focus on analyzing the pharmacochaperone binding pocket and the mechanisms by which pharmacochaperones affect channel assembly and activity.

RPG and CBZ are located in the GBC binding pocket

We previously solved the KATP structure in the presence of GBC and ATP, revealing that GBC is lodged in the TM bundle above NBD1 (Martin et al., 2017a), a finding which has been independently confirmed (Wu et al., 2018). Here, in both the RPG/ATP and CBZ/ATP structures, we observe strong and distinctly shaped cryo-EM densities within the same GBC binding pocket (Figure 2B–D). Such density is absent in the ATP-only and the apo structures (Figure 2E and F), supporting assignment as the pharmacochaperone ligand.

Figure 2 with 1 supplement see all
Structural comparison of the pharmacochaperone binding pocket.

(A) Structural model of the RPG-bound SUR1 ABC transporter core module viewed from the side showing the slice viewed from the top (indicated by the two black lines) and the pocket viewed from the side at higher magnification (indicated by the red box) in B-F. (B–F) The pharmacochaperone pocket viewed from the top and the side of the channel in the states indicated. To enable comparison, each map was sharpened and filtered to 4.6 Å (the resolution of the apo state reconstruction) with the Postprocessing procedure in RELION. Ligand density corresponding to RPG in (B) is shown in magenta, CBZ in (C) in red, and GBC in (D) in blue. The binding pocket is empty in both the ATP only state (E) and the apo state (F). Note the side chain of W1297 in TM17 and Y377 in TM7 are shown and labeled in panel (E) to serve as reference points. (G) Chemical structures of the three pharmacochaperones shown in B–D.

https://doi.org/10.7554/eLife.46417.009

Density for RPG appears compact and palm-shaped, and suggests that the molecule adopts a considerably folded shape upon binding to SUR1 (Figure 2B; Figure 2—figure supplement 1B). Interestingly, RPG possesses a carboxyl group adjacent a benzyl group, analogous to the sulfonyl group in GBC that is also adjacent a benzyl group (Figure 2G). Refinement of an RPG molecule into the observed binding pocket density orients this carboxylate towards N1245, R1246, and R1300 (Figure 3A), which coordinate the sulfonyl group in the GBC-bound structure (Figure 3B). However, unlike GBC, the RPG density is distant from S1238. This explains previous functional data that an SUR1 S1238Y mutation does not affect RPG’s ability to modulate channel function (Hansen et al., 2002; Yan et al., 2006). The helix on the opposite side of the binding pocket (i.e. TM8; Figure 3A) is lined with hydrophobic residues (W430, F433, and L434), which may support binding through a combination of van der Waals interactions and shape complementarity. Interestingly, although similar in overall structures, we noted subtle rearrangements within the binding pocket between GBC- and RPG-bound states (Figure 3A and B). The most obvious is in W1297, which in RPG is flipped down towards the ligand. Thus, there is sufficient flexibility of the binding pocket to accommodate diverse compounds with high affinity.

Models of the PC binding pocket.

(A) RPG binding site model, with residues mutated to alanine in C labeled in red. (B) GBC binding site model. (C) Western blots showing effects of alanine mutation of the selected residues on the ability of GBC, CBZ, and RPG to correct the processing defect caused by the F27S mutation in the TMD0 of SUR1. The arrow indicates the mature, complex-glycosylated SUR1. The lower band is the core-glycosylated immature SUR1. The thick vertical line in the bottom blot indicates samples from the same experiments run on two separate gels.

https://doi.org/10.7554/eLife.46417.011

CBZ is a smaller molecule with molecular weight about half of that of GBC and RPG (Figure 2G). However, in the CBZ-bound SUR1 structure the cryoEM density corresponding to the ligand has a size and shape that markedly resembles GBC (Figure 2C and D), with one end pointing towards S1238. While the close proximity to S1238 is in agreement with our previous finding that mutation of S1238 to Y diminishes the ability of CBZ to both inhibit and chaperone the channel (Devaraneni et al., 2015), the density is too large to be fitted by a single CBZ molecule (Figure 2C; Figure 2—figure supplement 1C). The structure of CBZ has been extensively studied and multiple polymorphic crystalline forms have been reported, including dimers (Florence et al., 2006; Grzesiak et al., 2003). Thus, an intriguing possibility is that CBZ may bind as a dimer to occupy the entire binding pocket (Figure 2—figure supplement 1C). Alternatively, the result could be explained by a CBZ molecule having multiple occupancies, and what is observed in the reconstruction is an average of the ensemble. Due to this uncertainly we did not model the CBZ molecule in the structure. To our knowledge this is the first protein structure determined in complex with CBZ. Thus, there are no structures for comparison. More studies are needed to distinguish these possibilities.

Structural evidence above indicates that the three pharmacochaperones occupy a common binding pocket to exert their effects. To seek functional evidence, we mutated five select SUR1 residues lining the binding pocket to Ala (Figure 3A) and monitored the effect of mutation on the ability of GBC, CBZ and RPG to correct the trafficking defects in SUR1-TMD0 mutants. The trafficking mutation F27S in SUR1 has been well-characterized in previous work, showing nearly undetectable mature complex-glycosylated form (upper band in Figure 3C) in the absence of pharmacochaperones but strong recoveries with both CBZ and GBC (Chen et al., 2013). In an F27S background, we introduced the five binding pocket mutations which have been shown previously to not significantly alter SUR1 maturation by themselves (Martin et al., 2017a), and examined the impact on the chaperone function of CBZ, GBC and RPG. In Western blots, there was a complete absence of the mature upper band for F27S-SUR1 in vehicle (DMSO) treated control. In contrast, there were strong upper bands when the mutant was expressed in the presence of CBZ, RPG or GBC (Figure 3C). This chaperone ability of the drugs was reduced to a variable extent for each of the binding site mutations tested in the F27S background (Figure 3C). Interestingly, we noted that while the sensitivity profile is very similar for GBC and CBZ, that for RPG is slightly different, in particular to the R1246A mutation. This suggests despite sharing a common binding pocket, RPG forms distinct chemical interactions with surrounding residues compared to GBC and CBZ. The combined structural and functional data presented above provide strong evidence that CBZ, RPG, and GBC exert their pharmacochaperoning effects by binding to a common binding pocket we have here identified.

Structure of the distal N-terminus of Kir6.2

The N-terminal ~30 amino acids of Kir6.2 (referred to as KNt hereinafter) has been known to be critical for channel assembly, gating, and interaction with sulfonylureas and glinides. However, underlying mechanisms remain poorly understood. Early studies showed that deletion of KNt significantly increases channel open probability (Babenko et al., 1999; Koster et al., 1999b; Reimann et al., 1999; Shyng et al., 1997), reduces ATP inhibition, and renders the channel less sensitive to sulfonylureas (Koster et al., 1999a; Reimann et al., 1999). KNt also appears to contribute to GBC binding (Kühner et al., 2012; Vila-Carriles et al., 2007), and is necessary for high affinity interaction with RPG (Hansen et al., 2005; Kühner et al., 2012). Deletion of amino acids 2–5 from the KNt shifts the binding affinity of RPG by more than 30-fold (Kühner et al., 2012). Moreover, KNt is critical for channel assembly and pharmacochaperoning (Devaraneni et al., 2015; Schwappach et al., 2000). Removal of KNt markedly reduces the ability of GBC and CBZ to rescue SUR1-TMD0 trafficking mutations (Devaraneni et al., 2015). Our recent studies showing that p-azidophenylalanine genetically incorporated at Kir6.2 amino acid position 12 or 18 was photocrosslinked to SUR1 and that the extent of crosslinking increased in the presence of GBC or CBZ further suggest physical interactions between KNt and SUR1 in a drug-sensitive manner (Devaraneni et al., 2015). Taken together, these studies led us to hypothesize that KNt is located near the pharmacochaperone binding pocket we have identified.

Close examination of the three pharmacochaperone-bound SUR1 structures reconstructed using focused refinement indeed revealed significant and continuous cryo-EM density immediately adjacent to the drug binding site, especially in the RPG-bound map (Figure 4A and B). This density appears as a roughly linear peptide inserting between the two TMDs of SUR1 from the intracellular side. By filtering the GBC/ATP and RPG/ATP maps to a lower resolution (6 Å), we were able to see contiguous density that extends out of the SUR1’s central cavity and connects to the density of the first structured residue R32 in Kir6.2 (Figure 4—figure supplement 1). Of note, a recent report by Wu et al. also pointed out a density in the SUR1’s central cavity that is nearly contiguous with the Kir6.2 cytoplasmic domain in our published GBC-bound structure and proposed it to be the N-terminus of Kir6.2 (Wu et al., 2018) (see Discussion).

Figure 4 with 3 supplements see all
Kir6.2 N-terminus cryoEM density in SUR1.

(A) RPG-bound SUR1 from focus-refined, unsharpened map viewed from the side. The major domains are labeled in different colors. (B) Vertical slice view of the map shown in (A) that reveals the bound RPG (in gold) and the cryoEM density of Kir6.2 N-term (in magenta). (C) Comparison of the Kir6.2 N-term cryoEM density in the different structures in horizontal slices viewed from the bottom. All maps are sharpened and filtered to 6 Å. Apo and ATP-only structures are displayed at 1.8σ; RPG/ATP, GBC/ATP, and CBZ/ATP structures are displayed at 2.2σ. Note lower threshold is needed for the Kir6.2 N-term density in the ATP-only structure to become visible.

https://doi.org/10.7554/eLife.46417.012

Interestingly, in the RPG-bound structure the piperidino moiety of RPG is in close proximity to the density that corresponds to the most N-terminus of the KNt (Figure 4—figure supplement 2A). This observation agrees with previous findings that the increased affinity to RPG caused by Kir6.2 is due to the peperidino group (Stephan et al., 2006), and that deletion of only a few amino acids from the Kir6.2 N-terminus decreases RPG binding affinity by more than 30-fold (Kühner et al., 2012). Of note, the KNt density is also seen in the ATP-only reconstruction though not as strong or well-defined as in the drug-bound reconstructions (i.e., the density disappears at lower σ values for the ATP-only map), but is largely absent in the apo state structure (Figure 4C). This indicates KNt enters into the central cavity of the inward-facing SUR1 even in the absence of drugs, and that drug binding stabilizes the location of KNt in the central cavity. We found that deletion of amino acids 2–5 or 2–10 increasingly reduced the ability of the pharmacochaperones to correct the processing defect of the F27S TMD0 mutant SUR1 (Figure 4—figure supplement 2B). Although the reduction in the F27S-SUR1 mature band intensity in these experiments could be due to a destabilization effect of KNt deletions on F27S mutant SUR1, it could also suggest that the KNt is needed for the pharmacochaperoning effect. Interestingly, the same Kir6.2 N-terminal deletion mutations also reduced the maturation of WT SUR1 and the ability of pharmacochaperones to enhance WT SUR1 maturation (Figure 4—figure supplement 2C). These structural and functional observations lead us to propose that KNt binds to the central cavity of the SUR1 ABC core and stabilizes SUR1-Kir6.2 association by simultaneous interaction with multiple SUR1 TMD helices to facilitate complex assembly. This mechanism not only is important for efficient biogenesis of WT channels but also likely underlies the ability of pharmacochaperones to enhance WT channel biogenesis efficiency and rescue mutant channel assembly (see Discussion).

Physical and functional interactions of the distal N-terminus of Kir6.2 with SUR1

While our cryoEM structures support the assignment of the KNt density, which is consistent with the many functional experiments described above (Babenko and Bryan, 2002; Babenko et al., 1999; Devaraneni et al., 2015; Koster et al., 1999b; Martin et al., 2016), direct physical evidence linking KNt to SUR1 residues lining the central cavity is still lacking. Identification of crosslinked SUR1 residue via p-azidophenylalanine engineered in KNt is technically challenging due to the complex chemistry of the azido-mediated crosslinks (Devaraneni et al., 2015). As an alternative approach, we performed crosslinking experiments using an amine-reactive homobifunctional crosslinker CBDPS (Cyanurbiotindimercaptopropionylsuccinimide) with purified KATP channels bound to GBC, followed by mass spectrometry to identify crosslinked peptides, as described in Materials and methods. One of the inter-SUR1-Kir6.2 crosslinks we identified connected Kir6.2 lysine five and SUR1 lysine 602 (Figure 4—figure supplement 3), which is near the drug binding site in our structural model. The distance between α carbons (Cα) of the crosslinked lysines in our model is ~18 Å, which is within the reported range for the linker (Brodie et al., 2019). This result provides the first evidence for the physical proximity of KNt to the SUR1 ABC core central cavity.

Because the crosslinking/mass spectrometry experiments were performed using purified channels in detergents, there is a possibility that the interaction we have identified does not occur in cell membranes under physiological condition. To further corroborate the crosslinking/mass spectrometry results and focus on the interaction, we engineered cysteine pairs guided by the structure to test whether KNt can be crosslinked to SUR1. Inspection of our structural model pointed to an endogenous cysteine C1142 in SUR1 with proximity to the distal end of KNt (Figure 4—figure supplement 2A). Accordingly, we mutated Kir6.2 amino acids 2–5 to cysteine and asked whether the engineered Kir6.2 cysteines formed disulfide bonds with SUR1-C1142. We hypothesized that crosslinking of KNt with SUR1-C1142 would trap KNt within the SUR1 ABC core and reduce channel open probability, based on previous studies that deletion of KNt increases channel open probability (Babenko and Bryan, 2002; Babenko et al., 1999; Koster et al., 1999a; Reimann et al., 1999; Shyng et al., 1997).

Using inside-out patch-clamp recording, channel activity was monitored in response to the oxidizing reagent I2 followed by exposure to the reducing reagent dithiothreitol (DTT) (Anishkin et al., 2008). We tested WT channels, channels formed by WT SUR1 (C1142) and Kir6.2-L2C, S3C, R4C or K5C, as well as channels formed by SUR1-C1142A and Kir6.2-L2C, S3C, R4C, or K5C. Current decay was observed to a variable extent in most patches even in bath solution without I2. This behavior, referred to as rundown is typical of KATP channels and is attributed to loss of PIP2 (Lin et al., 2003; Ribalet et al., 2000). Channel rundown tends to plateau and is not spontaneously reversible. Because of rundown, it is difficult to discern current decay caused by I2-induced crosslinking; however, the current decay component caused by crosslinking should be reversible following exposure to DTT. Strikingly, co-expression of Kir6.2-L2C with WT SUR1 resulted in channels that displayed current recovery upon exposure to DTT (10 mM) following the initial exposure to I2 (250 µM). Importantly, none of the other Kir6.2 cysteine mutations showed DTT-induced current recovery when co-expressed with SUR1-C1142 (R4C shown as an example in Figure 5B). WT channels or channels formed by co-expressing Kir6.2-L2C and a SUR1 in which C1142 was mutated to an alanine (C1142A) also failed to show current reversal in DTT. Note not all patches of channels formed by WT SUR1 and Kir6.2-L2C showed significant current recovery in DTT; these patches also had less current decay in I2 (Figure 5E). It is possible that these are deep patches where solution exchange was slower; alternatively we speculate that PIP2 loss might be slower in these patches such that KNt was for the most part not inserted in SUR1’s central cavity to be crosslinked. Regardless, our observations that more than half of the WT-SUR1/Kir6.2-L2C patches (8 out of 14) showed significant DTT-induced current recovery while none of the other channels tested did (>40 patches) support crosslinking between SUR1-C1142 and Kir6.2-L2C. Together our structural and functional data is consistent with the idea that insertion of KNt into the central cavity of SUR1 ABC core serves as a mechanism to reduce channel open probability, in addition to promoting channel assembly.

Patch-clamp recording of cysteine mutants to probe location of the Kir6.2 N-terminus.

(A) Three examples of traces of channels formed by WT(C1142) SUR1 and L2C Kir6.2 showing DTT-induced current recovery following I2 exposure. Channels were exposed to 250 µM I2 or 10 mM DTT as indicated by the bars above the traces. Baseline was obtained by exposing channels to 1 mM ATP. The arrow in the first trace indicates the blocking effect of the DTT that was readily reversed upon return to K-INT. Note as we have documented before (Lin et al., 2003; Pratt et al., 2012), DTT alone did not cause significant current run-up or run-down aside from the reversible blocking effect (not shown). (B, C, D) Representative traces of control channels formed by WT SUR1 and R4C Kir6.2, C1142A SUR1 and L2C Kir6.2, or WT SUR1 and Kir6.2. (E) Quantification of current changes at the end of 120 s exposure to I2, and at the end of subsequent 120 s exposure to DTT (expressed as % of currents at the start of I2 exposure). The asterisk indicates statistical significance comparing currents at the end of I2 exposure and the end of DTT exposure using paired student’s t-test (p<0.05).

https://doi.org/10.7554/eLife.46417.016

Discussion

Loss of membrane protein expression due to impaired folding and trafficking underlies numerous human diseases. In some cases, such defects can be overcome by small molecule ligands that bind to mutant proteins, referred to as pharmacochaperones (Bernier et al., 2004; Hanrahan et al., 2013; Powers et al., 2009; Ringe and Petsko, 2009). Although significant progress has been made in discovering pharmacochaperones, as exemplified for the CFTR mutation ΔF508 (Mijnders et al., 2017; Pedemonte et al., 2005; Veit et al., 2018), detailed structural mechanisms of how these compounds work are in most cases not well understood. This problem is even more complex in hetero-multimeric proteins such as the KATP channel where not only folding of the mutant protein itself but interactions with assembly partners may be involved. In this study, we set out to understand how structurally diverse compounds that inhibit KATP channels act as pharmacochaperones to correct channel trafficking defects. We present structural, biochemical, and functional evidence that GBC, RPG, and CBZ, despite diverse chemical structures, occupy a common pocket in SUR1, that this occupancy stabilizes a labile binding interaction between the N-terminus of Kir6.2 and the central cavity of the SUR1 ABC core, and that this interaction controls channel activity as well as the assembly and trafficking of nascent channels to the β-cell surface. Our findings shed light on the principles that govern the assembly of the KATP complex and offer insight into the mechanism by which channel inhibitors act allosterically to rescue channel trafficking defects caused by SUR1 mutations in TMD0 (see Figure 6).

Cartoon of pharmacochaperoning and channel inhibition mechanism.

(A) Top: During WT channel biogenesis, KNt insertion into the SUR1 central cavity facilitates SUR1-TMD0 interaction with Kir6.2 for successful channel assembly. For SUR1 containing TMD0 trafficking mutations, it is unable to assemble with Kir6.2 in the absence of a pharmacochaperone (PC), and is targeted for ER-associated degradation (ERAD). Upon binding of PC to the binding pocket in SUR1, the Kir6.2 N-term becomes stabilized in the central cavity of the SUR1 ABC core, allowing assembly of the mutant TMD0 with Kir6.2 and trafficking of the complex to the plasma membrane. (B) Insertion of KNt into SUR1’s central cavity prevents Kir6.2 from adopting an open conformation (top) to close the channel. However, this interaction is labile and reversible (indicated by the black arrow going both directions). Binding of PC in the SUR1 pocket stabilizes Kir6.2 N-term in the central cavity and pushes channel equilibrium towards a closed state (indicated by the thick red arrow pointing to the closed state). In addition, PC binding stabilizes SUR1 in an inward-facing conformation, unable to be stimulated by Mg-nucleotides. Crosslinking of SUR1’s endogenous C1142 with engineered Kir6.2-L2C also traps the Kir6.2 N-term in the central cavity, closing the channel (see Figure 5).

https://doi.org/10.7554/eLife.46417.017

Among ABC transporters and inward rectifier potassium channels, SUR1 and Kir6.2 are unique in having evolved to become functionally inter-dependent. The mechanisms by which these two structurally unrelated proteins form a heteromeric complex to regulate K+ transport and insulin secretion have been intensively investigated. Recently, several cryo-EM structures have been reported of KATP channels naturally assembled from separate SUR1 and Kir6.2 proteins (Li et al., 2017; Martin et al., 2017a; Martin et al., 2017b), or comprising engineered SUR1(C)-(N)Kir6.2 fusion proteins (Lee et al., 2017; Wu et al., 2018). These structures reveal the channel’s overall architecture as a Kir6.2 tetramer surrounded by four SUR1, wherein Kir6.2 makes direct contact with the TMD0 of SUR1. This structure explains why numerous trafficking mutations are found in TMD0 (Snider et al., 2013). However, we have shown in the structure of GBC-bound KATP complexes that, GBC resides well within the TM bundle of the SUR1 ABC core, above NBD1 and distant from TMD0 (Martin et al., 2017a). An allosteric mechanism by which binding of GBC rescues TMD0 trafficking mutations has remained obscure.

A critical missing piece of the puzzle was the KNt, which we have shown to be involved in pharmacochaperone rescue of TMD0 mutations (Devaraneni et al., 2015). KNt (Kir6.2 aa 1–31) was not well resolved in the initial published GBC-bound structures (Li et al., 2017; Martin et al., 2017a; Martin et al., 2017b). Recently, Wu et al. compared a GBC-bound cryoEM channel structure using a SUR1-Kir6.2 fusion construct with a 39aa linker to previously published GBC-bound structures of channels formed by co-expression of SUR1 and Kir6.2 as separate proteins (Wu et al., 2018). They noted a cryoEM density in SUR1’s central cavity that projects towards the Kir6.2 cytoplasmic domain in the SUR1 and Kir6.2 co-assembled structures but not the SUR1-Kir6.2 fusion structure. Despite this density not being contiguous with the structured Kir6.2 N-terminus, it was reasoned the density likely corresponds to KNt based on previous structure-function data. The authors argue that the lack of density in the fusion cryoEM map is likely due to the large size of the linker that cannot be accommodated in the central cavity. Here, through improved focused refinement algorithms and cross-correlation of multiple structures we were able to discern a low-resolution, contiguous cryoEM density emanating from the structured Kir6.2 R32 and ending near the top of the SUR1’s central cavity, allowing us to assign the KNt density with confidence. This assignment is further supported by the crosslinking/mass spectrometry data connecting Kir6.2-K5 to SUR1-K602. Interestingly, the crosslinking/mass spectrometry experiment also identified another crosslink between Kir6.2-K5 and SUR1-K205 (Figure 4—figure supplement 3). That Kir6.2-K5 is able to interact with both SUR1-K602 lining the central cavity and K205 in L0 is consistent with flexibility of KNt as reflected by the low cryoEM resolution of this stretch of amino acids.

The combined structural and functional data presented in this study provide a mechanism of pharmacochaperoning wherein pharmacochaperone binding acts in trans to overcome assembly defects caused by mutations within TMD0. Because the pharmacochaperone binding site is formed by the TM helices of the SUR1 ABC core, and TMD0 is a separate domain, it is unlikely that its occupancy by chemically diverse compounds overcome or prevent misfolding of TMD0 trafficking mutations at the site of the mutation. Supporting this notion, GBC, RPG, and CBZ fail to chaperone TMD0 trafficking mutants in the SUR1RKR-AAA background, a variant in which the RKR ER retention signal has been mutated to AAA to allow Kir6.2-independent trafficking to the cell surface (Devaraneni et al., 2015; Yan et al., 2006). Rather, occupancy of the pharmacochaperone binding site stabilizes the insertion of KNt into the central cavity of the SUR1 ABC core, an otherwise labile interaction that is nonetheless crucial for assembly and trafficking of the channel complex out of the ER (Figure 6A). Progressive deletion of KNt (2-5, 2-10) renders pharmacochaperones less and less effective in rescuing the trafficking mutant (Figure 4—figure supplement 2B), likely due to progressive weakening of pharmacochaperone-dependent KNt-SUR1 interactions. However, we cannot rule out the possibility that deletion of Kir6.2 KNt further destabilizes the SUR1 trafficking mutant such that the mutant SUR1 can no longer be rescued by pharmacochaperones. Importantly, progressive deletion of KNt (2-5, 2-10) also markedly reduces the biogenesis efficiency of WT channels and makes pharmacochaperones less and less effective in enhancing the biogenesis efficiency of the channel (Figure 4—figure supplement 2C). These results lead us to propose insertion of KNt into the central cavity of the SUR1-ABC core as a principal mechanism to stabilize SUR1-Kir6.2 interaction during channel assembly. We suggest that during normal channel assembly, insertion of KNt into the SUR1 central cavity facilitates interactions between SUR1-TMD0 and Kir6.2. Once assembled, the interactions between SUR1-TMD0 and Kir6.2 become the primary anchor holding the two subunits together and the KNt-SUR1 interaction, rather than serving a structural role, adopts the role of gating regulation (see below).

The location of the KNt density observed in our structures, including the ATP-only structure, also illuminates how Kir6.2 and SUR1 interact to modulate channel activity. First, by occupying the central cavity of the SUR1 ABC core, KNt prevents NBD dimerization and stabilizes SUR1 in an inward-facing conformation, thereby abolishing the ability of Mg-nucleotides to stimulate channel activity (Sikimic et al., 2019). Second, by trapping KNt in the central cavity, SUR1 prevents the Kir6.2 tetramer from undergoing conformational changes needed for channel opening. Our results that DTT exposure recovers currents of channels formed by SUR1-C1142 and Kir6.2-L2C previously exposed to I2 indicates crosslinking-induced decrease of channel activity and lends support to the second mechanism. It also explains early studies showing that deletion of KNt increased channel open probability and decreased the ability of sulfonylureas to inhibit the channel (Babenko and Bryan, 2002; Koster et al., 1999a; Reimann et al., 1999; Shyng et al., 1997). Interestingly, Wu et al. recently showed that inserting increasing number of amino acids in the linker between SUR1 and Kir6.2 in the SUR1-Kir6.2 fusion channels caused a gradual decrease in channel inhibition by GBC (Wu et al., 2018). They postulate that KNt acts as a chain which prevents Kir6.2 cytoplasmic domain from rotating to open the channel when trapped in the SUR1’s central cavity; increasing the linker length in the fusion protein reduces the tension exerted by the KNt-SUR1 interaction to reduce channel inhibition by GBC. Future determination of open channel structures will allow direct structural comparison of conformational differences in the Kir6.2 cytoplasmic domain. It is worth noting that the KNt density, albeit weaker, is found in the ATP-only structure. This suggests KNt spends significant time in SUR1’s ABC core cavity when Kir6.2 is in the ATP-bound closed state. By contrast, the KNt cryoEM density is absent in the apo structure, suggesting that when no ATP is bound to Kir6.2, KNt has little residence time in the ABC core central cavity. The fact that despite not seeing KNt, the Kir6.2 channel is still closed in the dominant class structure is likely due to loss of endogenous PIP2 during purification.

Our structures show that CBZ and RPG share the same binding pocket as GBC in KATP channels. Although both RPG and CBZ have been characterized extensively, with RPG being a commonly used oral hypoglycemic medication and CBZ an anticonvulsant, there are no known structures of these drugs bound to proteins. Thus, our structures provide the first examples of how these drugs interact with their target proteins. Despite occupying a common pocket, each of the three compounds appear to form distinct chemical interactions with residues lining the binding pocket. This is particularly clear when comparing RPG with GBC and CBZ. Curiously, we found that the density corresponding to CBZ cannot be fit with a single molecule. CBZ is a well-known Nav channel blocker (Lipkind and Fozzard, 2010). It will be important to determine whether CBZ adopts a similarly bound arrangement in other target proteins.

Despite functional divergence, all ABC transporters share some structural similarities, in particular the ABC core structure (Thomas and Tampé, 2018). An exciting finding here is how the Kir6.2 N-terminus can act as a plug or handle to regulate channel gating and assembly by inserting itself into the SUR1 ABC core central cavity formed by the two TM bundles. The exploitation of this structural space is akin to the mechanism by which viral peptide ICP47 enables immune-evasion of the pathogens (Blees et al., 2017; Oldham et al., 2016; Parcej and Tampé, 2010). In this case, ICP47 secreted by viruses such as Herpes simplex virus and Cytomegalovirus inserts into the inner vestibule formed by the ABC peptide transporters TAP-1 and TAP-2 (transporter associated with antigen processing) (Blees et al., 2017; Oldham et al., 2016), thus preventing the transport of cytosolic peptides into the ER for MHC complex loading and immune surveillance. Following this, molecules which can reside within this space may offer opportunities for functional modulation of ABC transporters such as KATP. Indeed, application of a synthetic Kir6.2 N-terminal peptide (a.a. 2–33) to the cytoplasmic face of KATP channels in isolated membrane patches has been shown to increase KATP channel open probability (Babenko and Bryan, 2002). This is presumably due to competition of the exogenous peptide with the N-terminus of Kir6.2 for binding to the SUR1 ABC core. With regard to the drug binding pocket, it is worth noting that both GBC and CBZ have been reported to bind or are substrates of other ABC transporters (Zhou et al., 2008). For example, Bessadok et al. (2011) recently showed that the multidrug resistance transporter P-glycoprotein (ABCB1) recognizes several SUR1 ligands including GBC, albeit with much lower affinity. Moreover, CBZ has been shown to correct the trafficking defect of ΔF508 CFTR (ABCC9), the most prevalent mutation underlying cystic fibrosis (Carlile et al., 2012). An intriguing possibility is that these other ABC transporters recognize SUR1 ligands through a similar binding pocket identified here via conserved residues within the SUR1 GBC binding pocket such as R1246 and W1297. It would be interesting to determine in the future whether and how CBZ binding affects ΔF508 CFTR structure to correct its processing defect (Sampson et al., 2013).

In summary, our study revealed a mechanism by which a diverse set of compounds modulate the gating and assembly/trafficking of the pancreatic KATP channel, a critical regulator of glucose-stimulated insulin secretion. Our findings may serve as a drug development platform in KATP channels and other ABC transporters for new pharmacological chaperones or modulators with improved efficacy and specificity.

Materials and methods

Key resources table
Reagent type
(species)
or resource
DesignationSource or referenceIdentifiersAdditional
information
Gene (Cricetus cricetus)ABCC8 (SUR1)UniProt databaseQ09427
Gene (Rattus norvegicus)KCNJ11 (Kir6.2)UniProt databaseP70673
Recombinant adenovirus (Cricetus cricetus)FLAG-tagged hamster SUR1PMID: 28092267FLAG-epitope inserted at the N-terminus of SUR1 and cloned into AdEasy vector for production of adenovirus
Recombinant adenovirus (Rattus norvegicus)Rat Kir6.2PMID: 28092267N/AConstructed using the AdEasy vector for production of adenovirus
Recombinant adenovirustTAPMID: 28092267N/AAdenovirus for over-expression of Tetracycline-controlled transactivator (tTA) under CMV promoter used for Tet-Off system
Recombinant DNA reagent (Cricetus cricetus)FLAG-tagged ham SUR1 in pECEPMID: 11226335N/AFLAG-epitope inserted at the N-terminus of SUR1
Recombinant DNA reagent (Rattus norvegicus)Rat Kir6.2 in pcDNA3PMID: 14707124N/A
Cell line (Rattus norvegicus)INS-1 clone 832/13PMID: 10868964RRID:CVCL_7226
Cell line (Chlorocebus aethiops)COSm6PMID: 11226335RRID:CVCL_8561
Chemical compound, drugDigitoninCalbiochemCAS 11024-24-1
Chemical compound, drugATPSigma-AldrichA7699
Chemical compound, drugGlibenclamideSigma-AldrichG0639
Chemical compound, drugRepaglinideSigma-AldrichR9028
Chemical compound, drugCarbamazepineSigma-AldrichC4024
Chemical compound, drugCBDPS-H8/D8Creative Molecules, IncCat. Number: 014S
Chemical compound, drugFuGENE6PromegaE2691
PeptideFLAG-peptideSigma-AldrichF3290
AntibodyAnti-FLAG M2 affinity gelSigma-AldrichA2220
AntibodyAnti-SUR1 (rabbit polyclonalPMID: 17575084N/A(1:100)
AntibodyHorseradish Peroxidase conjugated goat anti-rabbit secondaryJackson ImmunoResearchCode: 111-035-144(1/1000)
Software, algorithmSerial EMPMID: 16182563http://bio3d.colorado.edu/SerialEM
Software, algorithmMOTIONCOR2PMID: 28250466http://msg.ucsf.edu/em/software/motioncor2
Software, algorithmCTFFIND4PMID: 26278980http://grigoriefflab.janelia.org/ctffind4
Software, algorithmDoGPickerPMID: 19374019https://sbgrid.org/software/titles/dogpicker
Software, algorithmRelion-3PMID: 30412051https://www2.mrc-lmb.cam.ac.uk/relion
Software, algorithmCOOTPMID: 20383002http://www2.mrc-lmb.cam.ac.uk/personal/pemsley/coot
Software, algorithmUCSF ChimeraPMID: 15046863http://www.cgl.ucsf.edu/chimera
Software, algorithmPymolSchrödingerhttps://pymol.org/2

Cell lines used for protein expression

Request a detailed protocol

INS-1 cells clone 832/13 and COSm6 cells were used for protein expression (see below). The identity of these cell lines has been authenticated (see Key Resources Table above). These cell lines are not on the list of commonly misidentified cell lines maintained by the International Cell Line Authentication Committee. The mycoplasma contamination testing was performed routinely in the lab and shown to be negative for the work described here.

Protein expression and purification

Request a detailed protocol

KATP channels were expressed and purified as described previously (Martin et al., 2017a; Martin et al., 2017b). Briefly, the genes encoding pancreatic KATP channel subunits, which comprise a hamster SUR1 and a rat Kir6.2 (94.5% and 96.2% sequence identity to human, respectively), were packaged into recombinant adenoviruses (Lin et al., 2005; Pratt et al., 2009); both are WT sequences, except for a FLAG tag (DYKDDDDK) engineered into the N-terminus of SUR1 for affinity purification. INS-1 clone 832/13 (Hohmeier et al., 2000), a rat insulinoma cell line, was infected with the adenoviral constructs in 15 cm tissue culture plates. Protein was expressed in the presence of 1 mM Na butyrate as well as 5 µM RPG (for the RPG/ATP structure) or 10 µM CBZ (for the CBZ/ATP structure) to enhance expression and formation of the channel complex (Chen et al., 2013; Yan et al., 2006). At 40–48 hr post-infection, cells were harvested and cell pellets flash frozen in liquid nitrogen and stored at −80°C until purification.

For purification, cells were resuspended in hypotonic buffer (15 mM KCl, 10 mM HEPES, 0.25 mM DTT, pH 7.5) and lysed by Dounce homogenization. The total membrane fraction was resuspended in buffer A (0.2M NaCl, 0.1M KCl, 0.05M HEPES, 0.25 mM DTT, 4% sucrose, pH 7.5) and solubilized with 0.5% Digitonin. Note for the ATP-only dataset, 1 mM ATP was included throughout the purification; for the CBZ/ATP dataset, 10 µM CBZ and 1 mM ATP were present throughout; for the RPG/ATP dataset, 1 mM ATP and 30 µM RPG were included throughout. The soluble fraction was incubated with anti-FLAG M2 affinity agarose for 4 hr. The agarose beads were washed three times with 4X beads volume of buffer A (without sucrose) and bound proteins eluted with buffer A (without sucrose) containing 0.25 mg/mL FLAG peptide. Purified complexes were concentrated to ~1–1.5 mg/mL and used immediately for cryo grid preparation. For the RPG/ATP sample, the final concentration of RPG was 30 µM, and ATP 1 mM; for the CBZ/ATP sample, the final concentration of CBZ was 10 µM, and ATP 1 mM; for the ATP-only sample, the final ATP concentration was 1 mM.

Sample preparation and data acquisition for cryo-EM analysis

Request a detailed protocol

3 µL of purified KATP channel complex was loaded onto UltrAufoil gold grids which had been glow-discharged for 60 s at 15 mA with a Pelco EasyGlow. The sample was blotted for 2 s (blot force −4; 100% humidity) and cryo-plunged into liquid ethane cooled by liquid nitrogen using a Vitrobot Mark III (FEI).

Single-particle cryo-EM data was collected on a Titan Krios 300 kV using a Falcon III detector (Thermo Scientific) for the RPG/ATP dataset and a Gatan K2 Summit detector for the CBZ/ATP and the ATP only datasets. The apo dataset was collected on a Talos Arctica 200 kV microscope with a Gatan K2 Summit detector using SerialEM. Data collected using the Gatan K2 Summit direct electron detector were performed in the super-resolution mode, post-GIF (20 eV window), at a physical pixel size of 1.72 Å (Krios) or 1.826 Å (Arctica). For the RPG/ATP dataset, the calibrated pixel-size of the Falcon III was at 1.045 Å. Defocus was varied between −1.0 and −3.0 µm across the datasets. Detailed imaging parameters are provided in Table 1.

Image processing

Request a detailed protocol

The raw frame stacks were gain-normalized and then aligned and dose-compensated using Motioncor2 (Zheng et al., 2017) with patch-based alignment (5 × 5). CTF parameters were estimated from the aligned frame sums using CTFFIND4 (Rohou and Grigorieff, 2015). Particles were picked automatically using DoGPicker (Voss et al., 2009) with a broad threshold range to reduce bias. Subsequently, each image was analyzed manually to recover particles missed by automatic picking and remove bad micrographs. 2D classifications were done using RELION-2 (Kimanius et al., 2016). Classes displaying fully assembled complexes and high signal/noise were selected and particles re-extracted at 1.72 Å/pix (or 1 Å/pix for the RPG/ATP dataset) and then used as input for 3D classification in RELION-2 (see Table 1 and Figure 1—figure supplements 14).

Extensive 3D classification was performed to sample heterogeneity within the data. Symmetry was not imposed at this step in order to select only true four-fold classes. Up to four consecutive rounds of classification were performed, specifying 4 or five classes per round. Individual classes and combinations of classes were refined for each dataset to achieve the best reconstructions (Figure 1—figure supplements 14). A soft mask encompassing the entire complex was used during refinement in RELION, with C4 symmetry imposed to yield an overall channel reconstruction using the gold-standard FSC cutoff. The map was B-factor corrected and filtered using RELION-2’s Postprocessing procedure, with the same mask used for refinement.

Focused refinement

Request a detailed protocol

Focused refinement of SUR1 was carried out in RELION-3 using symmetry expansion, partial signal subtraction that removes signals outside the masked region, followed by further 3D refinement of signal subtracted particles (Scheres, 2016). Different masking strategies were applied to the different datasets for focused refinement to obtain the best reconstruction (Figure 1—figure supplements 14). For the RPG/ATP dataset, refinement against the SUR1 ABC core module following symmetry expansion and signal subtraction led to significantly improved resolution (Figure 1—figure supplement 1). For the CBZ/ATP dataset, refinement against the entire SUR1 was performed, which also yielded an improved density map for SUR1 (Figure 1—figure supplement 2).

Similar focused refinement strategies were initially applied for the previously published GBC/ATP dataset and the new ATP-only and the apo datasets. However, this resulted in deterioration of map quality and resolution compared to the starting reconstructions. Therefore, alternative signal subtraction strategies were tested for focused refinement, which led to the use of a mask that includes the Kir6.2 tetramer and one SUR1 for the three datasets. Thus, for the GBC/ATP dataset, the particles included in the final C4 reconstruction were subjected to C4 symmetry expansion and signal subtraction using a mask that includes the Kir6.2 tetramer and one SUR1 (Figure 1—figure supplement 3). Although the resulting map showed little improvement in overall resolution compared to our previously published C4 map (Martin et al., 2017a), the local resolution of the SUR1 in several regions was significantly improved, especially NBD1 and the linker between TMD2 and NBD2 (Figure 1—figure supplement 5). A similar scheme with some modifications was employed for the ATP-only and the apo datasets (Figure 1—figure supplement 4).

Modeling

Request a detailed protocol

Modeling was performed for the RPG- and GBC-bound SUR1 maps, which following focused refinement have the highest resolutions among all the structures included in this study.

For RPG-bound SUR1, the final map was B-factor corrected and filtered using RELION-3’s Postprocessing tool to optimize observable side chain features. For model building, we used SUR1 from our previously published structure (PDB:6BAA) as the initial model. The ABC core structure including the two transmembrane domains (TMD1 and TMD2) and two cytosolic nucleotide binding domains (NBD1 and NBD2) was docked into the density map by rigid-body fitting using Chimera's ‘Fit in’ tool. The model was further optimized by rigid-body refinement using ‘Phenix.real_space_refinemet’ with default parameters (Afonine et al., 2018). Of note, additional densities not observed in the previously published KATP cryoEM density maps are now clearly resolved, particularly the ATP density in NBD1 and the density for the linker regions between TMD2 and NBD2 (Figure 1—figure supplement 5). To build additional residues in these regions, the final map was sharpened by B-factor using RELION 3.0 (Zivanov et al., 2018) to optimize observable map features. As there are no homology models available for these disordered regions, models were built manually de novo in COOT (Emsley et al., 2010) followed by refinement in PHENIX, as detailed below. Initially, poly-alanines were built into the maps with some clear bulky side chains in COOT; refinement was performed by ‘Phenix.real_space_refine’ with the following parameters: global minimization, morphing, and SUR1 initial model as a restraining reference model. Bulky side chains in the new loop regions were added, and the region between D1060-C1079 that was previously modeled as a non-helical structure was corrected to a complete helix by adjustment of residues in accordance with the density map. The model was refined by iterative manual inspection and side chains adjustment to fit the densities in COOT and real space refinement using ‘Phenix.real_space_refine’ with tight secondary structure and torsion angle restrains. Previously published NBD domains (PDB: 6BAA) was inserted as a reference model to provide additional restrain and to minimize overfitting. The Kir6.2 N-terminus (amino acids 1–19) was built as a poly-alanine model with the side chain of L2 modeled to show its relation to SUR1 C1142. For modeling RPG in the cryoEM density map, a molecular topology profile for RPG was created using eLBOW in PHENIX, and refined into the cryoEM density in COOT. The RPG molecule with new coordination was then added to the ABC core structure model for further refinement.

Model building for the GBC-SUR1 map was similarly performed. Because the GBC-bound SUR1 map was refined using the Kir6.2 tetramer with one full SUR1 subunit as a module, we created an initial model for SUR1 by merging TMD0 from the previous model (PDB:6BAA) with the final ABC core structure built using the RPG-SUR1 map described above, which was then docked into the GBC-SUR1 map using Chimera and refined using COOT and PHENIX. GBC was docked into the ligand density in the GBC-bound SUR1 map derived from focused refinement. In addition to the regions noted above for the RPG-SUR1 map, densities corresponding to K329-G353 are of sufficient quality to build poly-alanines with some bulky side chains de novo in COOT. Extra density at Asn10 likely corresponds to glycosylated Asn10, although the sugar moiety is insufficiently resolved for modeling.

SUR1 model building in the CBZ-bound SUR1 map was also performed following the same procedure outlined for GBC-SUR1, except that the initial model was the final ABC core structure from the RPG-SUR1 model. Note for the CBZ map, no attempt was made to incorporate the CBZ molecule due to uncertainty in whether the CBZ binds as a monomer or dimer, as discussed in the main text. Also, due to insufficient resolution ATP molecule bound at NBD1 was not modeled.

Functional studies

Request a detailed protocol

Point mutations were introduced into hamster SUR1 cDNA in pECE using the QuikChange site-directed mutagenesis kit (Stratagene). Mutations were confirmed by DNA sequencing. Mutant SUR1 cDNA in pECE and rat Kir6.2 in pcDNA3 were co-transfected into COS cells using FuGENE6, as described previously (Devaraneni et al., 2015), and used for Western blotting and electrophysiology as described below.

For Western blotting, cells were lysed in a buffer containing 20 mM HEPES (pH 7.0), 5 mM EDTA, 150 mM NaCl, 1% Nonidet P-40, and cOmplete protease inhibitors (Roche) 48–72 hr post-transfection. To rescue trafficking-impaired SUR1-F27S mutant, 0.1% DMSO (vehicle control), 1 µM GBC, 1 µM RPG or 10 µM CBZ were added to cells 24 hr before cell harvest. Proteins in cell lysates were separated by SDS/PAGE (8%), transferred to nitrocellulose membrane, probed with rabbit anti-SUR1 antibodies against a C-terminal peptide of SUR1 (KDSVFASFVRADK), followed by HRP-conjugated anti-rabbit secondary antibodies (Amersham Pharmacia), and visualized by chemiluminescence (Super Signal West Femto; Pierce) with FluorChem E (ProteinSimple).

For electrophysiology experiments testing the effects of crosslinking, cells co-transfected with SUR1 and Kir6.2 cDNAs along with the cDNA for the green fluorescent protein GFP (to facilitate identification of transfected cells) were plated onto glass coverslips 24 hr after transfection and recordings made in the following two days. All experiments were performed at room temperature as previously described Devaraneni et al. (2015). Micropipettes were pulled from non-heparinized Kimble glass (Fisher Scientific) on a horizontal puller (Sutter Instrument, Co., Novato, CA, USA). Electrode resistance was typically 1–2 MΩ when filled with K-INT solution containing 140 mM KCl, 10 mM K-HEPES, 1 mM K-EGTA, pH 7.3. ATP was added as the potassium salt. Inside-out patches of cells bathed in K-INT were voltage-clamped with an Axopatch 1D amplifier (Axon Inc, Foster City, CA). ATP (as the potassium salt), I2 (diluted from 250 mM stock in ethanol), or dithiothreitol (DTT) were added to K-INT as specified in the figure legend. All currents were measured at a membrane potential of −50 mV (pipette voltage = +50 mV). Data were analyzed using pCLAMP10 software (Axon Instrument). Off-line analysis was performed using Microsoft Excel programs. Data were presented as mean ± standard error of the mean (s.e.m.).

Crosslinking and mass spectrometry

Request a detailed protocol

Purified KATP channels were crosslinked using a 50:50 mixture of light (H8) and heavy (D8), amine-reactive, homobifunctional crosslinker CyanurBiotinDimercaptoPropionylSuccinimide (CBDPS) [14 Å span] (Creative Molecules Inc) (Petrotchenko et al., 2011). CBDPS was dissolved in anhydrous DMSO at 100 mM concentration, then immediately added to the purified sample. The final reaction buffer contained 150 mM NaCl, 50 mM KCl, 50 mM HEPES, pH 7.5, 0.05% digitonin, 1 mM CBDPS, 1% DMSO, 1 μM GBC, and 0.30 mg/ml purified protein in a volume of 180 µl. The crosslinking reaction was allowed to proceed for 20 min on ice, then quenched with 100 mM Tris, pH 8.0. Protein was then methanol/chloroform precipitated by: addition of 400 µl of methanol, vortexing, addition of 100 µl of chloroform, vortexing, addition of 300 µl of water, vortexing, and the mixture centrifuged at 16,000 x g at 4°C for 10 min. The upper aqueous layer was then removed, being careful not to remove the precipitated protein at the interface, 500 µl of methanol was added, the sample vortexed, then centrifuged as above. The supernatant was then removed and the pellet was washed twice by addition of 500 µl of methanol, gentle vortexing, and centrifugation as above. The pellet was dissolved by shaking for 30 min at 37°C in 15 µl of 8M urea and 20 µl of 50 mM ammonium bicarbonate containing 0.2% ProteaseMAX detergent (Promega). The sample was then reduced by addition of 40.3 µl of 50 mM ammonium bicarbonate, 1 µl of 0.5M dithiothreitol, and heating at 57°C for 30 min. The sample was then alkylated by addition of 2.7 µl of 0.55M iodoacetamide and incubation in the dark for 15 min. Digestion of the sample was then performed overnight at 37°C with shaking after addition of 1 µl of 1% ProteaseMAX detergent and 20 µl of MS-Grade trypsin (Thermo Scientific) dissolved at 0.1 µg/µl concentration in 1 mM HCl. Following digestion, trifluoroacetic acid was added to a final 0.5%, the sample incubated at room temperature for 5 min, centrifuged at 16,000 x g for 15 min, and the supernatant removed. Crosslinked peptides were then affinity purified using reagents and hardware provided in a Cleavable ICAT Reagent Kit for Protein Labeling (monoplex version) using the manufacturer’s recommended protocol (Sciex), except the digest was applied directly to the avidin affinity cartridge without the prior cation exchange purification, the digest was diluted in 1.0 ml the manufacturer’s avidin cartridge loading buffer, and the pH of the solution was adjusted to seven by addition of 1.0 M Tris, pH 8.0 buffer prior to injection onto the avidin cartridge. Purified peptides were then dried by vacuum centrifugation, dissolved in 20 µl of 5% formic acid and analyzed by liquid chromatography/mass spectrometry. The digest was injected onto an Acclaim PepMap 100 μm x 2 cm NanoViper C18, 5 μm trap (Thermo Scientific), at 5 µl/min for 5 min in mobile phase A containing water, 0.1% formic acid, then switched on-line to a PepMap RSLC C18, 2 μm, 75 μm x 25 cm EasySpray column (Thermo Scientific). Peptides were then eluted using a 7.5–30% mobile phase B (acetonitrile, 0.1% formic acid) gradient over 90 min at a 300 nl/min flow rate. Data-dependent tandem mass spectrometry analysis was performed using an Orbitrap Fusion instrument fitted with an EasySpray source (Thermo Fisher Scientific). Survey scans (m/z = 400–1500) and data-dependent MS2 scans were performed in the Orbitrap mass analyzer at a resolution = 120,000, and 30,000, respectively, following higher energy collision dissociation (HCD) using a collision energy of 35 following quadrupole isolation at a 1.6 m/z isolation width. Peptides of charge states 3–7 were selected with signal intensities over 5 × 104 and having a targeted inclusion mass difference of 8.05 to select peptides containing the mass shifted CBDPS cross-linkers. The instrument was also configured to collect MS/MS scans for only the heavy labeled peptide pair. The method also used dynamic exclusion with 30 s duration and mass tolerance of 10 ppm. Cross-linked peptides were identified using version 2.0.0.5 of MeroX software (Iacobucci et al., 2018) using cross-linker masses of 509.0974 and 517.1476 for the (H8) and (D8) forms of the CBDPS cross-linker respectively, and mass precision tolerances of 5 and 10 ppm for precursors and fragment ions, respectively. The instrument mzML file, detailed instrument settings, MeroX result file, FASTA file containing Kir6.2 and SUR1 sequences, and MeroX settings file can be downloaded at ProteomeXchange Consortium via the PRIDE (Perez-Riverol et al., 2019) partner repository with the dataset identifier PXD014498.

Note added in proof

Request a detailed protocol

While this paper was in review, a paper has been published by Ding et al. showing a structure of a KATP channel formed by a SUR1-Kir6.2 fusion protein bound to repaglinide (Ding et al., 2019).

References

  1. 1
  2. 2
  3. 3
  4. 4
  5. 5
  6. 6
  7. 7
  8. 8
  9. 9
  10. 10
  11. 11
  12. 12
  13. 13
  14. 14
  15. 15
  16. 16
  17. 17
  18. 18
  19. 19
  20. 20
  21. 21
  22. 22
  23. 23
    Features and development of coot
    1. P Emsley
    2. B Lohkamp
    3. WG Scott
    4. K Cowtan
    (2010)
    Acta Crystallographica. Section D, Biological Crystallography 66:486–501.
    https://doi.org/10.1107/S0907444910007493
  24. 24
  25. 25
  26. 26
  27. 27
  28. 28
  29. 29
  30. 30
  31. 31
  32. 32
  33. 33
  34. 34
  35. 35
  36. 36
  37. 37
  38. 38
  39. 39
  40. 40
  41. 41
  42. 42
  43. 43
  44. 44
  45. 45
  46. 46
  47. 47
  48. 48
  49. 49
  50. 50
  51. 51
  52. 52
  53. 53
  54. 54
  55. 55
  56. 56
  57. 57
  58. 58
  59. 59
  60. 60
  61. 61
  62. 62
  63. 63
  64. 64
  65. 65
  66. 66
  67. 67
  68. 68
  69. 69
  70. 70
  71. 71
  72. 72
  73. 73
  74. 74
  75. 75
  76. 76
  77. 77
  78. 78
  79. 79
  80. 80
  81. 81
  82. 82
  83. 83
  84. 84
  85. 85
  86. 86
  87. 87
  88. 88

Decision letter

  1. Richard Aldrich
    Senior Editor; The University of Texas at Austin, United States
  2. Gary Yellen
    Reviewing Editor; Harvard Medical School, United States
  3. Vera Y Moiseenkova-Bell
    Reviewer; Perelman School of Medicine, University of Pennsylvania, United States
  4. Colin G Nichols
    Reviewer; Washington University School of Medicine, United States
  5. Jeffrey Agar
    Reviewer; Northeastern University, United States

In the interests of transparency, eLife includes the editorial decision letter and accompanying author responses. A lightly edited version of the letter sent to the authors after peer review is shown, indicating the most substantive concerns; minor comments are not usually included.

Thank you for submitting your article "Mechanism of pharmacochaperoning in KATP channels revealed by cryo-EM" for consideration by eLife. Your article has been reviewed by four peer reviewers, and the evaluation has been overseen by a Reviewing Editor and Richard Aldrich as the Senior Editor. The following individuals involved in review of your submission have agreed to reveal their identity: Vera Moiseenkova-Bell (Reviewer #1); Colin G. Nichols (Reviewer #3); Jeffrey Agar (Reviewer #4).

The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.

Summary:

The reviewers appreciate the importance of the new cryogenic-EM structural data providing key evidence of the commonality of binding sites for different channel inhibitors and pharmacochaperones. This is also the first observation in cryo-EM of ATP binding to NBD1 of SUR1 in the absence of Mg2+, which provides an excellent confirmation of earlier photoaffinity labeling studies.

However, as the paper is focused on the hypothesis that the pharmacochaperone effect occurs through drug stabilization of the interaction between the Kir6.2 N-terminus (KNt) and the SUR ABC core domain, revisions are needed to address weaknesses in the evidence for this hypothesis.

Essential revisions:

1) Please make proper acknowledgment of the prior structural evidence, in the 2018 Wu et al. paper, for binding of the KNt to the SUR core domain. This paper is mentioned, en passant, as revealing a weak and disconnected cryoEM density that 'fueled speculation' that the KNt can be located within the central cavity of the SUR1 core. But this is a bit dismissive. The Wu et al. paper makes a rather strong case for this argument, revealing near contiguous density coming off the Kir6.2 subunit towards the cavity and with only a short gap to the density in the cavity, whereas the current manuscript shows what seems to be a more disembodied density in the cavity, and appears to be at no better resolution than that of Wu et al. Moreover, Wu et al. showed additional correlation in so far as there was no density in the SUR1-Kir6.2 fusion structure (in which the N-terminus is not free to act as an extended peptide).

Please also comment on whether in the current structures there is any continuity between Kir6.2 and the density attributed to the N-terminus of Kir6.2 within the core of SUR1 (as seen by Wu et al.), and what in the new structures makes it convincing that it is the N-terminus. The present functional data are not sufficient to make this conclusion.

2) Address the weaknesses in the new experimental evidence for the binding of the KNt and its role in the pharmacochaperone effect:

A) Potentially the strongest new evidence is the MS identification of cross-linked peptides including the KNt and the ABC core. However, there are major concerns that the XL-MS workflow used could produce false positives, particularly because of the absence of an alkylation step to reduce the chance of disulfide scrambling. These experiments should be repeated using an alkylation step and with tighter mass tolerances, using a well-established or validated XL-MS workflow. (Please see the detailed comments of reviewer #4 on this point, included below.) Complete details of the experimental methods and the XL-MS raw data files should be included.

It should also be acknowledged that the cross-linking is done on purified protein and not on channels in membranes, so it is possible the protein adopts conformations it might not normally do under physiological conditions.

On this question of false positive interactions, note that some of the SUR1-SUR1 crosslinking residues seems very far apart in both propeller and quatrefoil structures. For instance, K915 to K1412 looks to be ~70Å yet the linker is only 15A.

B) The electrophysiological tests of cross-linking of the Kir6.2-L2C mutant aim to provide functional evidence for KNt contacting the ABC core domain in an intact membrane under physiological conditions, however, in the opinion of the reviewers and the Reviewing Editor, these current results do not provide convincing support for the hypothesis. The text states that "co-expression of Kir6.2-L2C with WT SUR1 resulted in channels that displayed an accelerated current decrease in I2 (250µM) that was partially reversible by subsequent exposure to DTT[…]" In fact, both wild type channels and Kir6.2-L2C/SUR1-C1142A channels show a similar extent of inhibition during iodine application. If the difference between Kir6.2_L2C and controls is in the rate of current decrease, there is no quantification to support this claim. It is true that there was no DTT reversal apparent in the controls, but this reversal is only apparent in a subset (Figure 5E) of the Kir6.2-L2C records. Were the experiments that showed more profound reversal only the ones with small currents (like the middle record in Figure 5A)? Was the effect of DTT alone tested on wild-type or mutant channels? As for the drug block experiments, how are the effects of iodine differentiated from the effects of rundown (clearly apparent in the right-hand record in Figure 5A)?

It is unclear how best the authors can address these concerns and shortcomings in the physiological cross-linking experiments. Possibly more experiments along these same lines will make the effects more clear. Alternatively, given the apparent distance between the putatively cross-linked sites in the structure, it may be helpful to investigate the effects of a rapid, non-zero-length crosslinker such as a bis-MTS reagent. Although it would be preferable to include an improved version of these experiments, if this is not possible then these experiments should probably just be omitted from the revised manuscript.

C) Although deletions at the N-terminus of Kir6.2 are said to abolish the pharmacochaperone effect (by interfering with the putative interaction of the KNt with the drug bound within the SUR core), there is a very clear pharmacochaperone effect on the ΔN5 mutation in Figure 4—figure supplement 1C (and data are not shown for the ΔN10 mutation). If indeed the KNt deletion prevents the interaction with the SUR-bound drug, it suggests that the pharmacochaperone effect may be independent of the KNt interaction. The weaker (but still not absent) pharmacochaperone effect with the addition of the F27S mutation (panel B) may simply indicate that the KNt deletion destabilizes the protein and the F27S destabilizes it further, with both effects independent of the drug binding. At a minimum, the results with ΔN10 (with no F27S) should be shown, and this alternative interpretation should be discussed.

3) In Figure 3—figure supplement 1B,C, it is not possible to deduce if the mutations in SUR1 had any effect on drug block because it is very hard to distinguish current rundown from block in the records provided. The difficulty is even greater when the currents are very small as for the mutant channel. How was the% current measured in panel D? There really seems to be no steady state block by the drug (see B,C) and there is no clear, rapid onset of block. The problem with these drugs is that they are largely irreversible and they block slowly at low concentrations. It would be better to test a single drug concentration (preferably 100nM) on each patch. Alternatively, the S1238Y mutation of SUR1 could be exploited for comparisons as sulfonylureas block these mutant channels reversibly and at lower affinity.

4) The value of the Rb+ efflux experiments in Figure 3 is unclear and the conclusions drawn from these experiments may be overstated. The drugs were applied for up to 40 min. Is this long enough to affect surface trafficking of SUR1? Can it be concluded if the data primarily represent the effect of the drug on channel inhibition? The authors argue that differences between mutational effects on the ability of repaglinide and carbamazepine to inhibit flux confirm that the mutations are directly in the drug binding site and do not have an indirect-allosteric effect. At first glance, the effects of mutating these residues in the binding pocket on the ability of the two drugs appear to be similar (Figure 3C,D). Is flux a sensitive enough measurement to make such distinctions? Why were mutations tested using Rb+ efflux, rather than patch clamp? Efflux is both determined by the number of channels in the membrane and by the ability of the drug to block the channel. So they can't state 'mutations reduced the sensitivity of the channel to inhibition' – unless they have shown that changes in surface expression do not occur on the time scale of their experiments.

Reviewer #1:

In this manuscript, Dr. Shyng and colleagues presented cryo-EM structures of KATP channels bound to pharmacochaperones glibenclamide, repaglinide, and carbamazepine. All three compounds are clearly resolved in these structures and showed that chemically diverse KATP channel inhibitors interact with the channel through the same binding pocket. In addition, this study provided structural information on how N-terminus of the channel interacts with the transporter and stay in the closed conformation.

Manuscript is very well written, a lot of structural and functional data presented to confirm conclusions drawn from the structural work.

Overall, I do not have cryo-EM specific experimental concerns and consider this manuscript be ready for publication in eLife.

Reviewer #2:

The authors have provided five cryo-EM structures of the KATP channel complex in association with ATP and either glibenclamide (GBC, a refinement of their previous structure), repaglinide (RPG), carbamazepine (CBZ), or no drug, and in the absence of both drug and ATP. The data show that all 3 drugs bind in the same pocket on SUR1. They also identify some density near the drug binding site in the ABC core of SUR1 that they attribute to the unresolved N-terminus of Kir6.2 (as previously suggested by Wu et al. Protein Sci, 2018). The structural data are complemented by data showing the effect of mutations in the drug binding pocket on the ability of drugs to chaperone SUR1 to the plasma membrane (as indicated by complex glycosylation of SUR1).

In addition, the authors provide data showing the effect of drug binding site mutations on Rb+ efflux (which reflects KATP channel activity in the whole cell) and excised patch recordings of the effects of GBC and RPG on KATP currents in WT and W1297 mutant channels. The location of the N terminus within the SUR1 cavity is explored in two ways: by chemical crosslinking followed by mass spectrometry, and by examining the ability to crosslink residues in SUR1 and Kir6.2 by engineered cysteine pairs.

The cryo-EM structures in complex with RPG and CBZ are novel, and should be published. This is also the first observation in cryo-EM of ATP binding to NBD1 of SUR1 in the absence of Mg2+, which provides an excellent confirmation of earlier photoaffinity labeling studies. The surface expression data is also interesting and supports the idea that the drugs chaperone SUR1 to the plasma membrane by binding within the same drug binding pocket as that which is linked to channel inhibition. Surprisingly, since the lab has an excellent reputation for KATP channel electrophysiology, the functional data presented in this paper is weak and it is difficult to draw strong conclusions as to the mutational effects on drug binding or on the cross-linking of the N-terminus of Kir6.2 to SUR1 from the data as presented.

Figure 3—figure supplement 1B,C. It is not possible to deduce if the mutations in SUR1 had any effect on drug block because it is very hard to distinguish current rundown from block in the records provided. The difficulty is even greater when the currents are very small as for the mutant channel. How did they measure the% current in panel D – there really seems to be no steady state block by the drug (see B,C) and there is no clear, rapid onset of block. The problem with these drugs is that they are largely irreversible and they block slowly at low concentrations. It would have been better to have tested a single drug concentration (preferably 100nM) on each patch. Alternatively, the S1238Y mutation of SUR1 could be exploited for comparisons as sulfonylureas block these mutant channels reversibly and at lower affinity.

The authors provide persuasive structural evidence as to which residues line the drug-binding pocket. As such, the value of the Rb+ efflux experiments in Figure 3 is unclear and the conclusions drawn from these experiments may be overstated. The drugs were applied for up to 40 min. Is this long enough to affect surface trafficking of SUR1? Can it be concluded if the data primarily represent the effect of the drug on channel inhibition? The authors argue that differences between mutational effects on the ability of repaglinide and carbamazepine to inhibit flux confirm that the mutations are directly in the drug binding site and do not have an indirect-allosteric effect. At first glance, the effects of mutating these residues in the binding pocket on the ability of the two drugs appear to be similar (Figure 3C,D). Is flux a sensitive enough measurement to make such distinctions? Why were mutations tested using Rb+ efflux, rather than patch clamp? Efflux is both determined by the number of channels in the membrane and by the ability of the drug to block the channel. So they can't state 'mutations reduced the sensitivity of the channel to inhibition' – unless they have shown that changes in surface expression do not occur on the time scale of their experiments.

Figure 5. Most of their functional evidence for the 6.2 N-terminus contacting the ABC core domain comes from this figure. The authors state that "co-expression of Kir6.2-L2C with WT SUR1 resulted in channels that displayed an accelerated current decrease in I2 (250µM) that was partially reversible by subsequent exposure to DTT[…]" In fact, both wild type channels and Kir6.2-L2C/SUR1-C1142A channels show a similar extent of inhibition during I2 application. If the difference between Kir6.2_L2C and controls is in the rate of current decrease, there is no quantification to support this claim. It is true that the authors observe no DTT reversal in their controls, but this reversal is only apparent in a subset (Figure 5E) of their Kir6.2-L2C records. Were the experiments that showed more profound reversal only the ones with small currents (like the middle record in Figure 5A)? Did the authors test the effect of DTT alone on wild-type or mutant channels? As for the drug block experiments, how are the authors differentiating the effects of iodine from the effects of rundown (clearly apparent in the right-hand record in Figure 5A)?

Reviewer #3:

This is an interesting paper, in a series with the earlier reports of Martin et al. on KATP channel structure, that provides key evidence of the commonality of binding sites for different channel blockers, and intriguing mechanistic connection between the SUR1 subunit and the Kir6.2 subunit. There are a few points that should be addressed:

1) The paper is generally very well written, but should be carefully checked for minor typos and use of commas. One point of contention: 'evidence' is for or against an argument, 'evidence' does not 'establish' anything scientifically. In a couple of places, 'evidence' should be replaced by 'data' or perhaps 'results'.

2) The authors mention the 2018 Wu et al. paper, en passant, as revealing a weak and disconnected cryoEM density that 'fueled speculation' that the KNt can be located within the central cavity of the SUR1 core. But this is a bit dismissive. The Wu et al. paper makes a rather strong case for this argument, revealing near contiguous density coming off the Kir6.2 subunit towards the cavity and with only a short gap to the density in the cavity, whereas the current manuscript shows what seems to be a more disembodied density in the cavity, and appears to be at no better resolution than that of Wu et al. Moreover, Wu et al. showed additional correlation in so far as there was no density in the SUR1-Kitr6.2 fusion structure (in which the N-terminus is not free to act as an extended peptide). This is not to detract from the authors' conclusions here, but more credit should be given to the previous paper.

3) The major potentially definitive evidence in the current paper for the disposition of KNt is the mass spec data. In this regard, it is common to designate a fragmentation pattern as observed in the peptide fragmentation spectrum from tandem MS on the corresponding peptide sequence. Addition of the fragmentation information would make it easier to interpret the data.

4) There are a couple of assertions of likelihood in the Discussion for which the evidence seems a bit thin. In particular, why is it the case that by trapping KNt in the central cavity, SUR1 is 'likely' to prevent a PIP2-bound conformation in Kir6.2? The PIP2 binding site is not in the distal N terminus.

Reviewer #4:

This review concerns only the cross-linking MS (XLMS) results. A summary of my review is that the XLMS workflow is not a "typical" workflow, has at least one fundamental flaw and lacks sensitivity, and the cross-linking results could well be false positive.

XLMS is the most technically demanding and specialized type of biological MS experiment. As a result, expert users tend to rely upon well-established XLMS workflows that minimize false-positive IDs. It appears, however, that the authors have created a novel XLMS workflow (i.e. a workflow predicated upon ETD-dissociation of the cross-linked precursor, followed by MS3 of the resulting fragments, followed by atypical XL-MS data analysis), and combined this with novel methods for sample preparation (i.e. no alkylation to prevent S-S scrambling). While the individual components of their workflow (other than omitting alkylation) are reasonable, have they been proven to work well (in combination) during XL-MS?

This reviewer's recommended "fix" is to employ a start-to-finish XL-MS workflow (e.g. sample preparation, cross-linkers, MS, and DA methods) from a well-established group. The leaders in this field are Albert Heck (https://experiments.springernature.com/articles/10.1038/nmeth.3603) and Lan Huang (https://pubs.acs.org/doi/abs/10.1021/acs.analchem.7b04431). Note that all expert groups employ specialized, fit-for-purpose data analysis and validation methods (e.g. XLinkX or equivalent). If the authors (and editor) choose the risky approach of using the author's current workflow, I include suggestions below for some de-risking.

Five major concerns of the current XL-MS methods include: 1) The authors do not appear to employ a well-established XL-MS workflow (or do not reference one). The potential for false positives (via S-S scrambling) increases, the sensitivity of their analysis suffers, and alternative results (i.e. cross-links that don't fit the proposed model) will not be detected. 2) The cross-linking reaction is not described in sufficient detail to repeat or to critique and a reference to the cross-linking method does not appear to have been included. 3) The omission of the reduction step (i.e. for typical reduction and alkylation) is necessary with this cross-linker and was appropriate. However, the omission of the alkylation step will result in "free" cysteine containing peptides, which will certainly exchange (via thiolate-disulfide interchange) with the DSP S-S bond. This study must be performed with a denaturing alkylation step immediately following cross-linking and quench (prior to digestion). 4) No proper cross-link validation step was included. 5) The engineered disulfides should be detected directly by XL-MS analysis (if ETD cleaves all disulfides, the same workflow used by the authors for other cross-links should work here, provided S-S scrambling is prevented by alkylation).

[Editors’ note: the revised article was subsequently rejected after discussions between the reviewers, but the authors were invited to resubmit after an appeal against the decision. The decision letter after the re-review is shown below.]

Thank you for resubmitting your work entitled "Mechanism of pharmacochaperoning in a mammalian KATP channel revealed by cryo-EM" for consideration by eLife. Your article has been re-reviewed by four peer reviewers, and the evaluation has been overseen by a Reviewing Editor and a Senior Editor. The following individuals involved in review of your submission have agreed to reveal their identity: Vera Y Moiseenkova-Bell (Reviewer #1); Colin G. Nichols (Reviewer #3); Jeffrey Agar (Reviewer #4); Michael Puljung (Reviewer #5).

Our decision has been reached after consultation between the reviewers. Based on these discussions and the individual reviews below, we regret to inform you that your work will not be considered further for publication in eLife.

We appreciate the valuable new data in the paper, but our expert mass spectrometry reviewer still has serious concerns that the MS crosslinking results are too prone to false positives, without alkylation to prevent disulfide scrambling and a higher m/z resolution analysis of the crosslinking products. Given this concern combined with the weaknesses already pointed out in the functional disulfide crosslinking work, it seems that the central contention of the paper about the interaction between the KNt and the SUR1 core is not adequately supported by the experimental results.

We understand that you may disagree with these concerns, but we consider them too important to overlook. At this point, we are rejecting the submitted manuscript, but we would welcome a new submission of a manuscript that addressed these concerns, most likely with new experimental results. Alternatively, if you think that we have just got it wrong, our rejection of the paper frees you to submit the paper to other journals.

Reviewers #1 and #3 were satisfied with the revisions.

Reviewer #4:

The authors did formulate a general response to XL-MS concerns, and this stated (more or less) that the authors did not share my concerns (although they did provide their raw data, which is great) and therefore did not perform additional experiments. I would not myself consider publishing biologically relevant findings using a new method of analysis that hasn't undergone proper validation experiments, or an MS/MS accuracy cutoff of 1 Da, or not validating any putative cross-links with higher accuracy follow-up scans. However, I will admit to being on the more conservative side in such matters.

My serious concern about false positives arises from free thiols that are part of the protein (which could be "lone thiols" or arise from small amounts of native disulfides that never formed). There is a vast literature that not only predicts that thiolates (i.e. free cys on digested peptides) will attack the disulfide bond of DSP, displacing one of the two peptides, but that can even predict the rate constants of this reaction (which will be near their fastest possible rates at the pH used by these authors, exhibiting half-lives on the millisecond timescale). In my own experience, if you have free thiols (on peptides) in the presence of disulfide bonds (on peptides), there will be complete scrambling of disulfide bonds. Here's the most worrisome part- following attack on the DSP disulfide bond by a Cys residue (from another peptide), the leaving group has Lys-conjugated DSP fragment that terminates in a thiolate. This can then attack another DSP cross-linked peptide and form a completely spurious crosslink (which releases another Lys-conjugated DSS fragment, and the mess continues until all crosslinks are artifacts).

This is a fatal design flaw and is why I suggested they alkylate prior to digestion (which they didn't do). This problem of scrambling is more important than any concern about "overalkylation".

Here is a link to a review describing the exchange rates: https://gmwgroup.harvard.edu/files/gmwgroup/files/369.pdf

And here are a few published descriptions of disulfide scrambling in practice:

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3082428/

https://www.ncbi.nlm.nih.gov/pubmed/17609855

https://www.ncbi.nlm.nih.gov/pubmed/12441146

https://www.ncbi.nlm.nih.gov/pubmed/10972999

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2143640/

If we assumed for the sake of argument that there were no problems with scrambling, the 1Da MS/MS cutoff is another fatal flaw (irrespective of other aspects of the XLMS experiment).

The authors also use DSP in a way that it has never been used (I asked them to include a reference to their workflow and they didn't). DSP is usually chemically reduced before MS experiments (and compared to a non-reduced experiment). However, these authors extended Huang's concept of the "MS-cleavable" crosslinker from Collisional activation (CID) acting upon a sulfoxide bond (where it has been validated with crosslinkers), to ETD acting upon disulfides (where it has never been validated with crosslinkers, only intact proteins in still-controversial studies). Prior to using such a workflow to reach biologically significant decisions, there are a number of hoops authors jump through, one being a methods development study where they show a histogram of the length of the ID'd crosslinks (where the 70Å cross-link they ID'd would have been a problem). Here I invited them to prove their workflow on their engineered disulfide bond peptides, which like all other of my requests, they declined.

And the authors don't really detect half-DSP conjugated lys residues- they detect peptides that happen to have modifications that sum to 88 +/- 1 Da (and there are a lot of ways to get to that mass).

They use the insoluble crosslinker, rather than water soluble analogue.

Unlike the studies of established labs, they don't bother using isotope labeled DSP to validate their results. These are commercially available.

Most of these concerns are in my previous review.

Reviewer #5:

I find the new version of the Martin et al. manuscript to be greatly improved over the original. The inclusion of new data (particularly in Figure 3) was helpful as was their expanded discussion, particularly in reference to the Wu et al. paper. I am still not completely persuaded by their electrophysiology data in Figure 5 (I2 cross linking), but appreciate the changes they made in the main text regarding this figure, which is a much better description of the data. My remaining concerns are minor and can be addressed as changes to the Discussion.

https://doi.org/10.7554/eLife.46417.040

Author response

Essential revisions:

1) Please make proper acknowledgment of the prior structural evidence, in the 2018 Wu et al. paper, for binding of the KNt to the SUR core domain. This paper is mentioned, en passant, as revealing a weak and disconnected cryoEM density that 'fueled speculation' that the KNt can be located within the central cavity of the SUR1 core. But this is a bit dismissive. The Wu et al. paper makes a rather strong case for this argument, revealing near contiguous density coming off the Kir6.2 subunit towards the cavity and with only a short gap to the density in the cavity, whereas the current manuscript shows what seems to be a more disembodied density in the cavity, and appears to be at no better resolution than that of Wu et al. Moreover, Wu et al. showed additional correlation in so far as there was no density in the SUR1-Kir6.2 fusion structure (in which the N-terminus is not free to act as an extended peptide).

Please also comment on whether in the current structures there is any continuity between Kir6.2 and the density attributed to the N-terminus of Kir6.2 within the core of SUR1 (as seen by Wu et al.), and what in the new structures makes it convincing that it is the N-terminus. The present functional data are not sufficient to make this conclusion.

We regret to have given the reviewers the impression that we were being dismissive of the study by Wu et al. In response, we have revised the manuscript and provided a detailed account of the evidence described in their paper for the KNt cryoEM density (Results and Discussion sections) to properly acknowledge their findings.

The paper by Wu et al. compared their GBC-bound cryoEM structures using the SUR1-Kir6.2 fusion construct with a 39aa linker, their previous GBC bound structure of channels formed by co-expression of SUR1 and Kir6.2 as separate proteins, and the GBC/ATP bound structures also of channels formed by individual SUR1 and Kir6.2 proteins we published. They noted cryoEM density in SUR1’s central cavity in structures of channels formed by separate SUR1 and Kir6.2 proteins but not by the SUR1-Kir6.2 fusion. More specifically, they used our published GBC/ATP structure to show near contiguous density coming off the Kir6.2 subunit towards the SUR1 cavity. They proposed the density likely corresponds to KNt based on previous structure-function data and argued the lack of density in the fusion cryoEM map is due to the large size of the linker that cannot be accommodated in the central cavity.

We have now included additional cryoEM map figures showing that when our GBC/ATP and RPG/ATP-bound cryoEM maps are filtered to a lower resolution (6Å), contiguous density emanating from the first structured residue in Kir6.2 (R32), following along SUR1-L0, and reaching the ABC core central cavity is seen (revised Figure 4—figure supplement 1). While we agree Wu et al. offered a strong case of argument for the KNt density assignment and our own refined maps showing contiguous density further supports the assignment, there could be alternative structural interpretation given the low resolution of this density. For example, the cryoEM density of TMDs-NBDs linkers as well as the NBD1-TMD2 linker is largely missing in all the published structures and there is a possibility that the proposed Kir6.2 N-terminus density actually corresponds to these missing parts. For this reason, we decided to take the time to design biochemical and functional crosslinking experiments before rushing to publication.

Although the reviewers felt that our functional data was not sufficient (however, see our response below), we believe this data further bolsters the structural data and offer insight into the role of the KNt-SUR1 central cavity interaction in modulating channel activity (see revised Figure 6B). Thus, while the cryoEM structure, crosslinking-mass spectrometry, and functional crosslinking results individually do not provide definitive proof for the KNt density assignment, taken together they offer, in our view, compelling experimental evidence for the structural model proposed by Wu et al. and us.

2) Address the weaknesses in the new experimental evidence for the binding of the KNt and its role in the pharmacochaperone effect:

A) Potentially the strongest new evidence is the MS identification of cross-linked peptides including the KNt and the ABC core. However, there are major concerns that the XL-MS workflow used could produce false positives, particularly because of the absence of an alkylation step to reduce the chance of disulfide scrambling. These experiments should be repeated using an alkylation step and with tighter mass tolerances, using a well-established or validated XL-MS workflow. (Please see the detailed comments of reviewer #4 on this point, included below.) Complete details of the experimental methods and the XL-MS raw data files should be included.

It should also be acknowledged that the cross-linking is done on purified protein and not on channels in membranes, so it is possible the protein adopts conformations it might not normally do under physiological conditions.

We thank reviewer 4 for his comments and offer the following explanations for our experiments. Although the method the reviewer recommended from the Heck lab is a good one, it requires a completely different informatics workup (XlinkX) and is better suited for the analysis of cross-links in more complex digests. The second paper from Lan Huang is a review of cross-linking, but also highlights the sulfoxide containing cleavable crosslinkers developed in their laboratory that was also used by the Heck lab. While a sulfoxide crosslinker could be used to repeat these experiments, and XlinkX software used to interpret results, we don’t feel that it’s necessary for analyzing crosslinks in a purified protein preparation such as ours. We respectfully disagree with the reviewer’s assessment that because we didn’t follow a “mainstream” method that our results are invalid. We and others have used ETD to break disulfide bonds in native proteins in many previous studies. As the disulfide bond in DSP is essentially identical to the disulfide bonds in peptides (-CH2-S-S-CH2-), there is no reason to believe that ETD wouldn’t also preferentially fragment the disulfide bond in DSP, and that is precisely what we have observed. The fact that we could perform the peptide identifications using a common search engine (Sequest) using MS3 data is a strength, not a liability. In fact, using MS3 based peptide identification in conjunction with a cleavable crosslinker is a common workflow when using a cleavable crosslinker, and this is outlined in Figure 2A of the Huang review.

We also appreciate the reviewer’s concern about disulfide bond scrambling that may occur due to our omission of an alkylation step. This is certainly a concern when analyzing disulfides between cysteine residues in proteins. However, it does not pose a problem when using our crosslinking method with DSP, because it would not lead to misidentification of crosslinks. If a free cysteine carried out a disulfide exchange with a DSP crosslink, the resulting mass of the ETD liberated peptide would not lead to an identification, because it wouldn’t have a DSP remnant of 88 mass units. The worst case scenario would just be that we would have a loss of sensitivity if one of the DSP crosslinks underwent a disulfide exchange reaction with a free cysteine. We chose not to include the iodoacetamide alkylation to avoid over-alkylating the proteins, because the usual second addition of DTT that we normally do to scavenge the excess IAA after alkylation is complete couldn’t be done.

Another one of the reviewer’s concerns was that we didn’t describe the cross-linking reaction in detail. We have now provided more experimental details on the reactions in Materials and methods.

The reviewer is correct that the analysis of crosslinks using mass spectrometry is a demanding experiment. However, we feel that sufficient precautions were made to prevent false identification of the reported crosslinks. This is because: 1) the SEQUEST search of the two liberated peptides coming from a single MS2 event had to simultaneously identify unique peptides coming from either SUR1 or Kir6.2 when using a rat Swiss-Prot protein database containing over 8,000 entries, 2) both identified peptides had to have the +88 mass unit increase localized to a lysine residue due to the expected fragmentation of the disulfide bond within the DSP crosslinker, 3) the lysine identified at the crosslinking site had to exhibit a missed trypsin cleavage, 4) the identified peptides in the SEQUEST search had to have less than a 10 ppm mass error, and 5) the two identified peptides in the crosslink had to have a combined mass that matched the mass of the still crosslinked peptide detected in the MS survey scan. We apologize for the error in the mass tolerance of the precursor ions during the SEQUEST search previously given in the Materials and methods section. This has now been corrected.

A suggestion was made to share the raw files so other groups have the option of validation. The raw files used for the crosslinking experiment has been deposited to the ProteomeXchange Consortium via the PRIDE repository (dataset identifier PXD013873, see Materials and methods under the subheading “DSP crosslinking and mass spectrometry”).

Finally, we have acknowledged in the Results section that the crosslinking is done on purified proteins and not on channels in membranes, so it is possible the protein adopts conformations it might not normally do under physiological conditions.

On this question of false positive interactions, note that some of the SUR1-SUR1 crosslinking residues seems very far apart in both propeller and quatrefoil structures. For instance, K915 to K1412 looks to be ~70Å yet the linker is only 15A.

We thank the reviewer for pointing this out. We have now added in the Materials and methods and figure legend that because we use detergent solubilized, purified channel proteins for crosslinking, we cannot rule out the possibility of inter-SUR1 crosslinks. When we interpret the cross-linking/mass spectrometry results we do keep this in mind. The SUR1 K915-K1412 crosslink is very likely a result of inter-SUR1 occurrence within the same channel. As noted by the reviewers, K915 and K1412 are very far apart in the single SUR1 protein in our structure. However, the residues from adjacent SUR1 subunits in the same channel could reach crosslinking distance if SUR1 subunits are flexible. In this regard, it is worth noting that we indeed observe flexibility of SUR1 subunits that deviate from C4 symmetry (Results section). While we cannot rule out inter-channel SUR1 crosslinks, the probability of detecting such random crosslinks would be expected to be quite low. An inter-channel crosslink is even less likely the case for the Kir6.2-K5 and SUR1-K602 pair considering that the Kir6.2 tetramer is surrounded by SUR1 subunits. Nevertheless, we cannot completely rule out the possibility that this crosslink arises from dissociated subunits.

For the caveats discussed above we only used the CL/MS data as supporting evidence in our manuscript (see response to point 1 above). We note however, in our published study examining the effects of GBC and CBZ on photo-crosslinking of azidophenylalanine engineered at amino acid 12 or 18 position of Kir6.2 to SUR1, crosslinking was performed in intact cells. While we were unable to identify SUR1 residues to which Kir6.2-12AzF or 18AzF form crosslinks, our observation that crosslinking is enhanced by GBC and CBZ again provide supporting evidence of pharmacochaperone-dependent KNt interactions with SUR1.

B) The electrophysiological tests of cross-linking of the Kir6.2-L2C mutant aim to provide functional evidence for KNt contacting the ABC core domain in an intact membrane under physiological conditions, however, in the opinion of the reviewers and the Reviewing Editor, these current results do not provide convincing support for the hypothesis. The text states that "co-expression of Kir6.2-L2C with WT SUR1 resulted in channels that displayed an accelerated current decrease in I2 (250µM) that was partially reversible by subsequent exposure to DTT […]" In fact, both wild type channels and Kir6.2-L2C/SUR1-C1142A channels show a similar extent of inhibition during iodine application. If the difference between Kir6.2_L2C and controls is in the rate of current decrease, there is no quantification to support this claim. It is true that there was no DTT reversal apparent in the controls, but this reversal is only apparent in a subset (Figure 5E) of the Kir6.2-L2C records. Were the experiments that showed more profound reversal only the ones with small currents (like the middle record in Figure 5A)? Was the effect of DTT alone tested on wild-type or mutant channels? As for the drug block experiments, how are the effects of iodine differentiated from the effects of rundown (clearly apparent in the right-hand record in Figure 5A)?

It is unclear how best the authors can address these concerns and shortcomings in the physiological cross-linking experiments. Possibly more experiments along these same lines will make the effects more clear. Alternatively, given the apparent distance between the putatively cross-linked sites in the structure, it may be helpful to investigate the effects of a rapid, non-zero-length crosslinker such as a bis-MTS reagent. Although it would be preferable to include an improved version of these experiments, if this is not possible then these experiments should probably just be omitted from the revised manuscript.

We appreciate the reviewers’ comments. These experiments turned out to be more challenging compared to our previous crosslinking pairs in the cytoplasmic domain of the channel (including Kir6.2-E229C/Kir6.2-R314C and Kir6.2-Q52C/SUR1-E203C). Cytoplasmic cysteine pairs are easily crosslinked by H2O2 or copper phenanthroline or even form disulfide bond spontaneously upon patch excision (Lin et al., 2003; Pratt et al., 2012).

For the Kir6.2-L2C and SUR1-C1142 crosslinking, we had to test many different conditions. Initially we tried H2O2 and copper phenanthroline at different concentrations as we did in prior studies and tested more than 20 SUR1-Kir6.2 cysteine pair combinations, however, no clear reversal of currents by DTT after exposure to oxidizing agents was ever observed. After reviewing membrane protein crosslinking literature, we learned that I2 works better for residues in the transmembrane regions that are less solvent exposed. Indeed, systematic testing of multiple SUR1/Kir6.2 pairs using a paradigm of I2 exposure followed by DTT exposure, we were finally able to see a clear DTT reversal effect on the currents of the SUR1-C1142/Kir6.2-L2C pair. As noted by the reviewers, not every patch showed clear reversal and those that did not tended to show less current reduction with I2 exposure. While we cannot offer definitive explanations, a likely reason is that I2 did not induce significant crosslinking in these patches. This could happen if the patches were deeply recessed and slow to respond to reagents present in the bath solution (this is not uncommon in inside-out recordings based on our >20 years of experience). The other possibility is that channels in those patches had higher open probability such that the probability of KNt reaching the SUR1 central cavity was low. Consistent with the latter explanation, when channels were exposed to KINT solution containing 1mM EDTA which significantly reduces rundown caused by Mg2+-dependent loss of PIP2 (Lin et al., 2013), we did not see rapid current decay in I2 nor subsequent reversal by DTT. This is the reason why we performed the crosslinking experiments in KINT without EDTA. Under this condition, WT channels display a variable degree of “rundown” as well appreciated by researchers in the field. To our knowledge what accounts for this variability is not well understood, which makes it difficult to differentiate rundown from I2-induced decrease of currents. In light of the reviewers’ comments we have removed the phrase “accelerated current decrease” and the revised statement now reads: "co-expression of Kir6.2-L2C with WT SUR1 resulted in channels that displayed current decrease in I2 (250μM) that was partially reversible by subsequent exposure to DTT […]".

We agree with the reviewers that performing more experiments will not resolve the intrinsic difficulties of these crosslinking experiments. However, we respectfully disagree that the data presented should be removed as we believe that the DTT-induced current reversal (despite not in every patch) provides strong functional evidence for the structural interpretation and for a role of this interaction in channel activity. Taken as a whole, our structural, biochemical, and functional results converge and corroborate to support the mechanism we propose on how pharmacochaperones inhibit channel activity and correct channel trafficking defects. Nonetheless, we appreciate the weaknesses pointed out by the reviewers and have tried our best to revise the manuscript to explain the crosslinking experiments more clearly (Figure 5 legend).

Finally, regarding testing the effect of DTT alone on WT or mutants, we have documented such control experiments in our previous publications (Lin et al., 2003; Pratt et al., 2012). As such, we do not feel the need to show these recordings again. We have now included in the Figure 5 legend to specifically state that DTT alone does not cause significant channel run-up or run-down aside from an acute reversible blocking effect as indicated by the

arrow in the first recording (Figure 5A) and cited our published papers where this was documented.

C) Although deletions at the N-terminus of Kir6.2 are said to abolish the pharmacochaperone effect (by interfering with the putative interaction of the KNt with the drug bound within the SUR core), there is a very clear pharmacochaperone effect on the ΔN5 mutation in Figure 4—figure supplement 1C (and data are not shown for the ΔN10 mutation). If indeed the KNt deletion prevents the interaction with the SUR-bound drug, it suggests that the pharmacochaperone effect may be independent of the KNt interaction. The weaker (but still not absent) pharmacochaperone effect with the addition of the F27S mutation (panel B) may simply indicate that the KNt deletion destabilizes the protein and the F27S destabilizes it further, with both effects independent of the drug binding. At a minimum, the results with ΔN10 (with no F27S) should be shown, and this alternative interpretation should be discussed.

As requested we have now included a blot of WT-SUR1 co-expressed with ΔN10-Kir6.2 and treated with vehicle, glibenclamide, repaglinide, or carbamazepine (see revised Figure 4—figure supplement 2C) showing a further attenuation of drug effects on the maturation of WT-SUR1, similar to that seen for F27S-SUR1.

Regarding the alternative interpretation that KNt deletions make SUR1 less stable and F27S makes the protein even less stable, our response is as follows. By looking at the SUR1 upper band intensity, this may seem like a logical alternative interpretation. However, an important reminder here is that the upper band only appears when the channel complex is fully formed and has passed the ER quality control, which requires successful assembly of SUR1 with Kir6.2. Thus the reduced upper band indicates less stable SUR1/Kir6.2 complex and not SUR1 alone. We think it is unlikely that KNt deletion significantly impacts SUR1 stability as this would be expected to lead to increased ER associated degradation and decreased lower band intensity, which is not what we observed (see revised Figure 4—figure supplement 2B, C)

How do we explain the residual drug effects seen in SUR1 co-expressed with ΔN5- or even ΔN10-Kir6.2? We believe the most logical interpretation is that the distal ~20 amino acids of KNt is critical for SUR1 and Kir6.2 interactions during the initial assembly process even in the absence of drugs. Drug binding boosts the interaction in part by participation of KNt in drug binding but also by stabilizing SUR1 in an inward-facing conformation that allows KNt to interact with the central cavity for channel assembly. In addition, we have previously shown using metabolic pulse-chase experiments that drug binding slows ER degradation of mutant SUR1 (A116P; Yan et al., 2004), which would allow more time for Kir6.2 to interact before the mutant SUR1 is targeted for ERAD. Progressive deletion of KNt leads to a gradual loss of PCs’ ability to chaperone channel assembly by making KNt an increasingly weaker anchor. We have now revised the Results and Discussion sections and the cartoon figure (Figure 6) to explain our data more clearly.

Reviewer 2 also commented that in our study deletion of Kir6.2 N-terminus markedly reduced maturation of WT SUR1; however, previous studies co-expressing WT SUR1 and Kir6.2Δ5 or Kir6.2Δ10 showed descent currents. We note that using oocyte or mammalian cell transient expression, it is always possible to obtain membrane patches containing sufficient channels from individual cells that do not reflect the overall expression level. Also, because channels lacking the Kir6.2 N-terminus have higher Po, it can give the impression that their expression is not affected.

3) In Figure 3—figure supplement 1B,C, it is not possible to deduce if the mutations in SUR1 had any effect on drug block because it is very hard to distinguish current rundown from block in the records provided. The difficulty is even greater when the currents are very small as for the mutant channel. How was the% current measured in panel D? There really seems to be no steady state block by the drug (see B,C) and there is no clear, rapid onset of block. The problem with these drugs is that they are largely irreversible and they block slowly at low concentrations. It would be better to test a single drug concentration (preferably 100nM) on each patch. Alternatively, the S1238Y mutation of SUR1 could be exploited for comparisons as sulfonylureas block these mutant channels reversibly and at lower affinity.

The W1297 functional testing experiments were meant to provide further support to our drug binding site structural model. However, in revising the manuscript, we believe this is not necessary and may cause more confusion (based on reviewers’ feedback). As such, we have removed this supplementary figure.

For reviewers’ information, we have used a similar experimental scheme (10 and 100nM GBC) to study other binding pocket mutations in our recent publication where a number of mutations but not all render GBC inhibition highly reversible (see Figure 7 in Martin et al., eLife, 2017b).

4) The value of the Rb+ efflux experiments in Figure 3 is unclear and the conclusions drawn from these experiments may be overstated. The drugs were applied for up to 40 min. Is this long enough to affect surface trafficking of SUR1? Can it can be concluded if the data primarily represent the effect of the drug on channel inhibition? The authors argue that differences between mutational effects on the ability of repaglinide and carbamazepine to inhibit flux confirm that the mutations are directly in the drug binding site and do not have an indirect-allosteric effect. At first glance, the effects of mutating these residues in the binding pocket on the ability of the two drugs appear to be similar (Figure 3C,D). Is flux a sensitive enough measurement to make such distinctions? Why were mutations tested using Rb+ efflux, rather than patch clamp? Efflux is both determined by the number of channels in the membrane and by the ability of the drug to block the channel. So they can't state 'mutations reduced the sensitivity of the channel to inhibition' – unless they have shown that changes in surface expression do not occur on the time scale of their experiments.

Although the experiments were conducted over a 40 min period, the difference in efflux rate was already apparent during the first few minutes when we look at the efflux profile over time (see Figure 7A in Martin et al., eLife, 2017b). We don’t have any experimental evidence that the number of channels does not change over the course of the efflux experiment in the present study for every mutant. However, our previous surface biotinylation studies have not found significant changes in surface expression of WT channels in COS cells over 30min (e.g. Bruederle et al. Traffic, 2011).

In our previous publication where we tested the effect of GBC binding site mutations on the ability of GBC to inhibit channel activity using Rb efflux assays, we did include an Rb efflux profile example in addition to summary data. In the current study we repeated these experiments using the same set of mutations but testing response to RPG and CBZ with the intent to further validate our structural models for the RPG and CBZ binding pockets. We therefore only presented summary data and referred to our previous publication for the initial GBC experimental dataset. Rb efflux is an efficient way to assess drug response. In our previous study we used both Rb efflux and electrophysiology data to demonstrate effects of binding site mutations (see Figure 7 and Figure 7—figure supplement 1 in Martin et al., eLife, 2017b). As we believed the main point of this experiment is to show reduced inhibition by the mutants to confirm the binding pocket model, we did not think it is necessary to perform electrophysiology experiments especially considering the focus of this paper is on the pharmacochaperoning effects of the drugs. In place of electrophysiology, we performed pharmacochaperone rescue experiments to show that binding pocket mutations also compromised PC rescue effects.

In light of the reviewers’ comments, we have removed the Rb efflux results in Figure 3 and only show the impact of the binding site mutations on chaperoning effects of the drugs. Note, we also replaced the original Western blot showing the effect of RPG with a new one showing the effect of GBC and RPG side-by-side to facilitate direct comparison (now Figure 3C).

[Editors’ note: the author responses to the re-review follow.]

We appreciate the valuable new data in the paper, but our expert mass spectrometry reviewer still has serious concerns that the MS crosslinking results are too prone to false positives, without alkylation to prevent disulfide scrambling and a higher m/z resolution analysis of the crosslinking products. Given this concern combined with the weaknesses already pointed out in the functional disulfide crosslinking work, it seems that the central contention of the paper about the interaction between the KNt and the SUR1 core is not adequately supported by the experimental results.

We understand that you may disagree with these concerns, but we consider them too important to overlook. At this point, we are rejecting the submitted manuscript, but we would welcome a new submission of a manuscript that addressed these concerns, most likely with new experimental results. Alternatively, if you think that we have just got it wrong, our rejection of the paper frees you to submit the paper to other journals.

Reviewers #1 and #3 were satisfied with the revisions.

Reviewer #4:

[…] The authors also use DSP in a way that it has never been used (I asked them to include a reference to their workflow and they didn't). DSP is usually chemically reduced before MS experiments (and compared to a non-reduced experiment). However, these authors extended Huang's concept of the "MS-cleavable" crosslinker from Collisional activation (CID) acting upon a sulfoxide bond (where it has been validated with crosslinkers), to ETD acting upon disulfides (where it has never been validated with crosslinkers, only intact proteins in still-controversial studies). Prior to using such a workflow to reach biologically significant decisions, there are a number of hoops authors jump through, one being a methods development study where they show a histogram of the length of the ID'd crosslinks (where the 70 Å cross-link they ID'd would have been a problem). Here I invited them to prove their workflow on their engineered disulfide bond peptides, which like all other of my requests, they declined.

And the authors don't really detect half-DSP conjugated lys residues- they detect peptides that happen to have modifications that sum to 88 +/- 1 Da (and there are a lot of ways to get to that mass).

They use the insoluble crosslinker, rather than water soluble analogue.

Unlike the studies of established labs, they don't bother using isotope labeled DSP to validate their results. These are commercially available.

Most of these concerns are in my previous review.

We thank Dr. Agar for his constructive criticisms and valuable suggestions to improve our XL-MS experiment. Two fatal design flaws were pointed out, one concerns potential false positives due to disulfide scrambling, and the other is the 1Da MS/MS cut-off. On the second point, we apologize for not making this clear in our R1 version of the manuscript. The 1.0 Da value is for the MS3 scan in the Orbitrap Fusion’s ion trap and is within the setting commonly used (this low mass accuracy reflects the ion trap's inability to resolve isotopes for sparse or highly charged peptides; please see https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3226415/; https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4966894/), while the precursor tolerance is set at 10 ppm.

Nonetheless, given the reviewer’s serious concerns about our results and that many validation steps were requested to support our data which will take significant amount of time, we decided to repeat the XL-MS experiment using a different crosslinker that has been well documented in a number of published studies.

Specifically, we used an isotopically coded, CID-cleavable, biotinylated amine-reactive homobifunctional crosslinker, CBDPS (14Å space arm) (https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3033670/) for our experiments. The purified proteins were reduced with DTT and alkylated immediately after crosslinking and before trypsin digestion. The digested peptides containing the crosslinker were affinity purified via the biotin tag and MS analysis was conducted with mass precision tolerances of 5 and 10 ppm for precursors and fragment ions, respectively (details in the revised Materials and methods section).

Below are some examples of published studies which have used CBDPS with similar workflow to derive protein structural information:

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3814369/

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3642325/

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5501500/

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6561701/

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6453469/

The new experiment identified 2 inter-protein peptides between Kir6.2 and SUR1 and 4 intra-protein peptides in SUR1 as listed below (also shown in the revised Figure 4—figure supplement 3), with the distance between α carbons of the two cross-linked lysine residues in our model indicated where possible.

Kir6.2-SUR1:

Kir6.2 K5-SUR1 K205: 40.7Å? (see discussion below)

Kir6.2 K5-SUR1 K602 (identified in the previous DSP XL-MS experiments): 17.6Å

SUR1-SUR1:

K276-K394: 21Å

K276-K644: K644 is not modeled in the structure but is predicted to be located below K276

K941-K948: in NBD1-TMD2 linker not modeled in the structure but should be close given residue numbers

K1344-K1412: 17.6Å

The crosslinks identified using CBDPS are not identical to those identified in our previous experiments using DSP (12Å space arm, without DTT reduction and alkylation to prevent potential disulfide scrambling). Importantly, however, the same crosslinked peptide linking Kir6.2-K5 and SUR1-K602 was identified in both experiments (please see the revised Figure 4—figure supplement 3). The new XLMS results validate the physical proximity between Kir6.2-K5 and SUR1-K602 in our purified channels under the stated experimental conditions and strongly support our assignment of the Kir6.2 distal N-terminus density in SUR1’s ABC core central cavity. Note, in addition to the crosslink between Kir6.2-K5 and SUR1-K602, we also identified another crosslink between Kir6.2-K5 and SUR1-K205. According to our dominant class of GBC-bound cryoEM structure the Kir6.2 N-terminus is modelled within the SUR1 ABC core central cavity such that the distance between Kir6.2-K5 and SUR1-K205, which is in the proximal L0 region close to the Kir6.2 tetramer would be quite far (40.7Å). However, we should point out that in our cryoEM map, Kir6.2 N-terminus density is relatively poorly resolved compared to the rest of the structure, an indication of its flexibility even in the GBC-bound state. Thus, the identification of this alternative crosslink is consistent with our overall cryoEM structure data and suggests Kir6.2 N-terminus could dissociate from the SUR1 central cavity such that Kir6.2-K5 is within reach of SUR1-K205 via the CBDPS crosslinker.

Because the potential problems raised by the reviewer on our previous DSP crosslink experiments, we have deleted the original figure and methods and replaced them with the new CBDPS XLMS result figure and methods. The raw data have been deposited into the proteomic database PRIDE.

https://doi.org/10.7554/eLife.46417.041

Article and author information

Author details

  1. Gregory M Martin

    Department of Biochemistry and Molecular Biology, Oregon Health & Science University, Portland, United States
    Contribution
    Conceptualization, Data curation, Formal analysis, Funding acquisition, Investigation, Visualization, Methodology, Writing—original draft, Writing—review and editing
    Contributed equally with
    Min Woo Sung
    Competing interests
    No competing interests declared
  2. Min Woo Sung

    Department of Biochemistry and Molecular Biology, Oregon Health & Science University, Portland, United States
    Contribution
    Data curation, Formal analysis, Investigation, Visualization, Writing—review and editing
    Contributed equally with
    Gregory M Martin
    Competing interests
    No competing interests declared
  3. Zhongying Yang

    Department of Biochemistry and Molecular Biology, Oregon Health & Science University, Portland, United States
    Contribution
    Data curation, Investigation, Visualization
    Competing interests
    No competing interests declared
  4. Laura M Innes

    Department of Biochemistry and Molecular Biology, Oregon Health & Science University, Portland, United States
    Contribution
    Formal analysis, Investigation
    Competing interests
    No competing interests declared
  5. Balamurugan Kandasamy

    Department of Biochemistry and Molecular Biology, Oregon Health & Science University, Portland, United States
    Contribution
    Formal analysis, Investigation
    Competing interests
    No competing interests declared
  6. Larry L David

    Department of Biochemistry and Molecular Biology, Oregon Health & Science University, Portland, United States
    Contribution
    Formal analysis, Investigation, Visualization, Writing—original draft
    Competing interests
    No competing interests declared
  7. Craig Yoshioka

    Department of Biomedical Engineering, Oregon Health & Science University, Portland, United States
    Contribution
    Resources, Data curation, Software, Formal analysis, Supervision, Validation, Investigation, Methodology, Writing—original draft
    For correspondence
    yoshiokc@ohsu.edu
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-0251-7316
  8. Show-Ling Shyng

    Department of Biochemistry and Molecular Biology, Oregon Health & Science University, Portland, United States
    Contribution
    Conceptualization, Resources, Data curation, Formal analysis, Supervision, Funding acquisition, Investigation, Visualization, Methodology, Writing—original draft, Project administration, Writing—review and editing
    For correspondence
    shyngs@ohsu.edu
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-8230-8820

Funding

National Institutes of Health (DK57699)

  • Show-Ling Shyng

National Institutes of Health (DK066485)

  • Show-Ling Shyng

National Institutes of Health (F31 DK105800)

  • Gregory M Martin

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

We thank Dr. Christopher B Newgard for the INS-1E cell line clone 832/13 and the staff of the Multi-Scale Microscopy Core at the Oregon Health and Science University for help with imaging and data collection. We also thank Dr. Bruce Patton for comments on the manuscript. This work was supported by US National Institutes of Health grants DK57699 (S-LS), DK066485 (S-LS) and F31 DK105800 (GMM).

Senior Editor

  1. Richard Aldrich, The University of Texas at Austin, United States

Reviewing Editor

  1. Gary Yellen, Harvard Medical School, United States

Reviewers

  1. Vera Y Moiseenkova-Bell, Perelman School of Medicine, University of Pennsylvania, United States
  2. Colin G Nichols, Washington University School of Medicine, United States
  3. Jeffrey Agar, Northeastern University, United States

Publication history

  1. Received: February 26, 2019
  2. Accepted: July 22, 2019
  3. Accepted Manuscript published: July 25, 2019 (version 1)
  4. Version of Record published: August 19, 2019 (version 2)

Copyright

© 2019, Martin et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 1,038
    Page views
  • 253
    Downloads
  • 0
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Download citations (links to download the citations from this article in formats compatible with various reference manager tools)

Open citations (links to open the citations from this article in various online reference manager services)

Further reading

    1. Structural Biology and Molecular Biophysics
    Jie Li et al.
    Research Article
    1. Epidemiology and Global Health
    2. Structural Biology and Molecular Biophysics
    Hemanth KN Reddy et al.
    Research Article Updated