Structural titration of Slo2.2, a Na⁺-dependent K⁺ channel

Summary

The stable structural conformations that occur along the complete reaction coordinate for ion channel opening have never been observed. In this study we describe the equilibrium ensemble of structures of Slo2.2, a neuronal Na⁺-activated K⁺ channel, as a function of the Na⁺ concentration. We find that Slo2.2 exists in multiple closed conformations whose relative occupancies are independent of Na⁺ concentration. An open conformation emerges from an ensemble of closed conformations in a highly Na⁺-dependent manner, without evidence of Na⁺-dependent intermediates. In other words, channel opening is a highly concerted, switch-like process. The midpoint of the structural titration matches that of the functional titration. A maximum open conformation probability approaching 1.0 and maximum functional open probability approaching 0.7 implies that within the class of open channels there is a subclass that is not permeable to ions.

Graphical Abstract

Cryo-EM analyses of a protein ensemble under a range of conditions shows that a neuronal Na+-activated K+ channel uses an all or nothing approach to opening

Introduction

Ion channels are regulated by environmental stimuli that control their activity in cell membranes (Hille, 2001). The stimulus to activate (open) a ligand-dependent ion channel is sometimes a neurotransmitter, lipid, intracellular molecule, or an ion. Ion channels are thus similar to other proteins that undergo allosteric regulation, with certain advantages: the functional state of a channel can be monitored through electrical recording with high signal-to-noise ratio and time resolution in the millisecond to microsecond range (Hille, 2001). Through such measurements channel open probability as a function of stimulus and rates of transitions between functional states has been determined with high accuracy. These determinations support specific gating models that account for stimulus dependence of channel activity (Csanady et al., 2010; Horrigan and Aldrich, 2002; Lape et al., 2008; McManus and Magleby, 1988). These models invoke the existence of discrete conformational states of a channel in dynamic exchange (Csanady et al., 2010; Horrigan and Aldrich, 2002; Lape et al., 2008; McManus and Magleby, 1988).

In principle single particle cryo-electron microscopy (cryo-EM) offers the opportunity to inspect the ensemble of conformational states that underlie ligand-dependent channel activity. This would be impossible in crystallography because a useful crystal necessitates structural uniformity within the crystal lattice. By contrast, proteins on a cryo-EM grid are not necessarily uniform in conformation and therefore it should be possible, in principle, to characterize an equilibrium ensemble of conformational states and how a regulatory ligand influences the equilibrium. In other words, it should be possible to inspect the distribution of structures underlying a ligand titration and thus deduce the pathway of conformational changes by which an ion channel is activated.

Slo2.2 is a Na⁺-dependent K⁺ channel. In its physiological setting Slo2.2 is expressed in specific neurons where it is activated by high concentrations of intracellular Na⁺ that are believed to occur following repeated action potentials (Bader et al., 1985; Dryer et al., 1989; Haimann and Bader, 1989; Kameyama et al., 1984; Schwindt et al., 1989; Yan et al., 2012; Yuan et al., 2003). Its relevance to neuronal function is underscored by the occurrence of specific disease states in humans that are associated with intellectual disability and seizures correlated to genetic defects in the gene encoding Slo2.2 (Barcia et al., 2012; Epi et al., 2013; Heron et al., 2012; Ishii et al., 2013; Martin et al., 2014; McTague et al., 2013; Vanderver et al., 2014).

Recently we published a structure of chicken Slo2.2 in the absence of Na⁺ at an overall resolution of 4.5 Å using cryo-EM (Hite et al., 2015). In that structure the S6 “inner helices”, which form the gate in many K⁺ channels (Doyle et al., 1998; Jiang et al., 2002; Long et al., 2005; Whorton and MacKinnon, 2011, 2013), were constricted and thus the conformation was hypothesized to be closed, consistent with the absence of channel activity in membranes with 0 mM cytoplasmic Na⁺. Here we study the conformations of Slo2.2 as a function of Na⁺ concentration and compare these data with the functional titration curve.

Results

Detection of Na⁺-dependence in structural classification

Images of Slo2.2 channels vitrified in the presence of 20 mM, 40 mM, 80 mM and 160 mM Na⁺ were recorded. The particle images from all four conditions were merged together into a single data set (titration data set) and subjected to 3D refinement in Relion (Scheres, 2012). The particles were next classified by requesting 10 classes without refinement of angles or translations (Figure 1A). Then by referencing the Na⁺ concentration at which each particle had been vitrified we graphed the Na⁺ dependence of the relative population density of each class (Figure 1B). One class (class 3) is an outlier compared to the other 9: this class becomes increasingly populated as the Na⁺ concentration is increased. Upon detailed inspection we observed that class 3 is also structurally distinct from the others in that the inner helices are open and thus we refer to these as open Slo2.2 particles (Figure 1A). The other 9 classes are all similar to the previously described closed Slo2.2 structure determined in the absence of Na⁺ (Hite et al., 2015).

Figure 1. Heterogeneity of Slo2.2 particles in 20 – 160 mM NaCl.

(A) Ten 3D class averages of Slo2.2 vitrified in the presence of 20 mM, 40 mM, 80 mM and 160 mM Na⁺. The nine classes that closely resemble closed, Na⁺-free Slo2.2 are colored red and the open class is colored blue. (B) Fraction of particles in each class is shown for the four Na⁺ concentrations. The sum of fractions of all ten classes for each Na⁺ concentration is 1.0.

Open Slo2.2 particles represent only 7.4% of total particles in the titration data set (Figure 1). To determine a more accurate structure underlying the open particles we collected an additional data set at 300 mM Na⁺. This data set was collected on a different microscope and therefore could not be combined with the titration data set for simultaneous image processing. When particles were processed independently but in a similar manner, 3D classification yielded eight classes (representing 83% of particles) that resemble open Slo2.2 and two classes (17%) that resemble the closed channel structure (Figure 2A,B). Before considering the Na⁺ titration in further detail, we first describe the open Slo2.2 structure.

Figure 2. Cryo-EM structure of Slo2.2 in open and closed conformations.

(A, B) Cryo-EM density maps of open (A) and closed (B) Slo2.2 low-pass filtered to 3.8 Å resolution and 4.3 Å resolution, respectively, calculated from images of Slo2.2 vitrified in the presence of 300 mM Na⁺. (C, D) Ribbon diagrams of open (C) and closed (D) Slo2.2 colored by domain: S1–S4 domains green, pore domains yellow, RCK1 domains blue and RCK2 domains red. Grey lines indicate approximate borders of the membrane. See also Figures S1, S2, S3 and S7.

Structure of open Slo2

Figure 2 shows cryo-EM maps and models of the open and closed Slo2.2 channels derived from the 300 mM Na⁺ data set. The closed channel, determined at 4.3 Å resolution, is indistinguishable from the closed Slo2.2 structure described previously (Figures S1 and S2). Open Slo2.2, determined at 3.8 Å resolution, shows many structural differences from the closed structure (Figures 2, S1 and S2). It also reveals structural elements that were not resolved in the cryo-EM density of closed Slo2.2 due to its higher resolution and reduced conformational heterogeneity (Figures S1 and S3) (Hite et al., 2015). These elements include complete connectivity between transmembrane and cytoplasmic domains (S6-RCK1 linker) and between the pore and the S1–S4 domains (S4–S5 linker) (Figure S3A–C).

Each subunit of the four-fold symmetric Slo2.2 tetramer consists of an S1–S4 domain and pore domain (S5–S6) within the membrane and two RCK domains outside the membrane on the cytoplasmic side (Figure 2C,D) (Jiang et al., 2001). Pore domains form the ion pathway and are surrounded by the S1–S4 domains in typical 6-transmembrane K⁺ channel fashion. The short S4–S5 linker in Slo2.2 confirms that the S1–S4 domain is not domain-swapped to a neighboring subunit but instead is attached through tertiary interactions to a pore domain formed from the same subunit (Figure S3A,C). Thus, with respect to domain organization within the membrane, Slo2.2 is similar to Slo1 and Eag1 (Whicher and MacKinnon, 2016; Tao et al, 2016). Outside the membrane the RCK domains form a gating ring for sensing intracellular Na⁺ concentration and regulating the channel’s gate (Hite et al., 2015; Zhang et al., 2010).

The following features distinguish the open conformation from the closed. First, the gating ring is compressed more tightly against the transmembrane domain so the length along the pore-axis of the open channel is 5 Å shorter than that of the closed channel (Figure 2A,B). Second, a subdomain structure on the membrane-facing surface of the gating ring called the RCK1 N-lobe has adopted an expanded conformation relative to the closed channel; that is, it is tilted away from the 4-fold axis (Figure 3A). Third, expansion of the RCK1 N-lobes has displaced the S6 helices radially away from the pore axis (Figure 3A,B). This opening of the S6 helices produces an expansion of the narrowest segment of the ion pathway below the selectivity filter (Met 333) from about 6 Å diameter in the closed channel, which is less than the diameter of a hydrated K⁺ ion (~ 8 Å), to about 20 Å (Figure 3C). The mutant M333A exhibits residual current in the absence of Na⁺, supporting the notion that the S6 helices serve as a gate to complete pore closure with a functional constriction at M333 (Figure 3D, E).

Figure 3. Activation of Slo2.2 by intracellular Na⁺.

(A) Superposition of open and closed Slo2.2 channels aligned by the RCK2 domains of all four subunits. Front and rear subunits, S1–S4 domains and S5 are removed for clarity. The pore and RCK1 N-lobes are colored blue and red in the open and closed structures, respectively. Dashed lines indicate residues absent from the closed structure. (B) Superposition of open (blue) and closed (red) pore domains aligned by the pore helices and selectivity filters viewed from within the plane of the membrane (left) and from the cytoplasm (right). Spheres represent the Cα position of Met-333. (C) Plot of pore diameter (between van der Waals surfaces) for open (blue) and closed (red) Slo2.2. The narrowest part of the pore in closed Slo2.2 below the selectivity filter is located at Met-333 (6.2 Å). (D,E) Representative currents at 0 and 300 mM Na⁺, recorded from excised inside-out patches from cells expressing wild-type (D) or M333A (E) Slo2.2. Pipette solution contained 150 mM KCl, 5 mM EDTA and 10 mM Hepes (pH 7.4) and bath solution contained 150 mM KCl, 10 mM Hepes (pH 7.4), and either 0 mM (top row) or 300 mM (bottom row) NaCl. The membrane voltage was held at 0 mV and stepped to voltages from −80 mV to 0 mV in 10 mV increments.

Expansion of RCK1 N-lobes induced by Na⁺ is similar to the expansion documented in the Slo1 Ca²⁺- and voltage-activated K⁺ channel induced by Ca²⁺ (Wu et al, 2010; Yuan et al, 2012; Tao et al, 2016; Hite et al, 2016). The concomitant changes in the S1–S4 voltage-sensing domains observed in Slo1, however, are not observed in Slo2.2. While there is a slight rigid body rotation of the four S1–S4 domains around the pore when the S6 helices open, the domains do not undergo internal conformational change (Figure 4A). The four helices of the S1–S4 domain in Slo2.2 are tightly packed against one another through a network of hydrogen-bonded contacts (Figure 4B). These structural features, together with the absence of conformational change within the S1–S4 domain upon channel opening, are consistent with Slo2.2 being a purely ligand-gated K⁺ channel that has no net charge in its S4 helix (Yan et al., 2012; Yuan et al., 2003). The S1–S4 domain, although present in Slo2.2, does not transmit voltage changes to channel gating. Slo2.2 opens through ligand-mediated allosteric regulation only: cytoplasmic Na⁺ binding causes expansion of the gating ring, displaces the S6 helices, and opens the channel by exerting force via the S6-RCK1 linker (Video S1).

Figure 4. S1–S4 domain structure.

(A) Stereo view of superposed open (blue) and closed (red) S1–S4 domains. Main-chain R.M.S.D is 0.83 Å. (B) Stereo view of open Slo2.2 S1–S4 domain shown in ribbon. Residues in S1–S4 that form inter-helix hydrogen bonds are shown with sticks and colored by atom.

Na⁺ titration of the conformational states

The titration data set and 300 mM Na⁺ data set were combined to analyze structural class dependence on Na⁺ concentration. As described, the titration data set comprised images collected at four different Na⁺ concentrations processed as a single data set. Images from the 300 mM Na⁺ data set were processed separately but in a similar manner (Figure 5). Particles were selected using Relion’s autopicking algorithm followed by manual removal of false positives (selections that were obviously not single channels) (Scheres, 2015). To ensure that all particles were accounted for in an unbiased manner no further particles were excluded at any point in the processing. Particles were then subjected to 3D refinement of angles and translations followed by 3D classification into 10 classes without refinement of angles and translations. As shown in Figure 1, one of the 10 classes from the titration data set was unique in that it corresponded to the open structure described above. The other 9 classes were closed, but showed variation with respect to the angular relationship between the cytoplasmic and transmembrane domains, as was documented previously (Figure S4) (Hite et al., 2015).

Figure 5. Image analysis workflow.

Representative images of Slo2.2 channels recorded in the presence of 20 mM, 40 mM, 80 mM, 160 mM and 300 mM Na⁺. Slo2.2 particles were automatically selected from the images (green circles). The extracted particles from the 20 mM, 40 mM, 80 mM and 160 mM Na⁺ images were combined into a single titration data set for 3D refinement, while the 300 mM Na⁺ data set was refined independently. Using the angles and orientations obtained from the 3D refinement, the particles from the titration data set and the 300 mM Na⁺ data set were classified into 10 classes. The closed classes are colored red and the open classes are colored blue.

We next determined how reproducible the probability of the open class would be in independent runs of 3D classification (Figure 6A,B). Five independent runs each yielded a single class corresponding to the open conformation. The fraction of total particles contributing to the open class for each run is shown as a function of Na⁺ concentration. The small differences between each run indicate that the classification algorithm is precise. The precision was further assessed by comparing the particles classified as open during the five classifications. Of all the particles classified as open, greater than 70% were classified as such four or five times (Figure S5). Consistent with high precision, likelihood probability distributions show that 75% of the particles specified as open have a probability value of 0.99 or higher for being open (i.e. the distributions approach the delta function) (Figure S5).

Figure 6. Reproducibility of 3D classification.

(A, C, E) Cryo-EM density maps of open Slo2.2 (A) and the two closed Slo2.2 classes with the most extreme rotations (C - closed class 1, E - closed class 9). (B, D, F) Fraction of particles in the open class (B) and the two closed Slo2.2 classes with the most extreme rotations (D - closed class 1 and F - closed class 9) from the titration data set is plotted as a function of Na⁺ concentration for 5 independent 3D refinement and classification runs. See also Figures S4 and S5.

By combining the titration data set with the independently but similarly image processed 300 mM data set we generated a graph of the probability that a particle exists in the open conformation as a function of Na⁺ concentration (Figures 7A–E,H). The data follow a sigmoidal increasing function ranging from near 0 to 0.83. While we did not collect images in the presence of Na⁺ concentrations above 300 mM, the shape of the curve indicates that the function would approach unity at higher Na⁺ concentrations. The data fit reasonably well to a Hill function with a binding constant of 0.24 ± 0.04 M and Hill coefficient of 3.5 ± 1.3 (Figure 7H).

Figure 7. Na⁺ titration of Slo2.2.

(A–E) Representative cryo-EM images of Slo2.2 vitrified in the presence of 20 mM (A), 40 mM (B), 80 mM (C), 160 mM (D) and 300 mM Na⁺ (E). Particles marked with a red circle were classified as closed and those with a blue circle were classified as open. (F) Representative macroscopic current traces of intracellular Na⁺ activation. Currents were recorded using using the inside-out patch-recording configuration; pipette solution contained 150 mM KCl, 5 mM EDTA, and 10 mM Hepes (pH 7.4). Bath solution contained 150 mM KCl, 10 mM Hepes (pH 7.4), and 0 mM to 600 mM NaCl controlled by bath perfusion. Membrane voltage was held at 0 mV and stepped to −60 mV. (G) Variance (σ²)-mean current (<I>) relationship for traces shown in (F) and fit to a parabola (equation 1, methods). A fit to this equation yielded estimates of unitary conductance (i) and channel number (N) through application of equations 1 and 2 (methods). (H) Open probability deduced from data as in (F) is graphed as a function of intracellular Na⁺ (black, error bars represent SEM for 5 experiments). The structural open probability based on cryo-EM is plotted on the same graph (blue, error bars represent standard deviation for 5 independent classifications). A Hill function fit to the electrophysiology data yields a binding constant for Na⁺ of 0.20 ± 0.02 M, a Hill coefficient of 3.0 ± 0.5 and maximum value 0.68 ± 0.03. A Hill function fit to the structural data with a fixed maximum of 1.0 yields a binding constant of 0.24 ± 0.04 M and a Hill coefficient of 3.5 ± 1.3. See also Figure S6.

Figure 7 correlates the dependence of the open conformation on Na⁺ concentration with the dependence of channel activity on Na⁺ concentration. Slo2.2 channels were expressed in HEK cells and their activity recorded using the inside-out patch-recording configuration (Figure 7F). Current traces at different Na⁺ concentrations were subjected to variance-mean analysis to estimate the open probability as a function of Na⁺ (Figures 7G and S6) (Sigworth, 1980). The open probability graph shows the mean and SEM for 5 independent experiments. The Na⁺ dependence of channel activity follows a similar sigmoidal function but approaches a maximum value less than unity. A Hill function fit to the data yields a binding constant for Na⁺ of 0.20 ± 0.02 M, Hill coefficient of 3.0 ± 0.5 and maximum value 0.68 ± 0.03 (Figure 7H). The lower maximum value of the channel activity plot compared to the structure titration implies that not all channels that adopt the open conformation conduct ions all the time. It is well known that many ion channel types never reach a functional open probability of 1.0 even in the setting of a maximum stimulus.

We next addressed whether the probability of different closed classes depends on Na⁺ concentration. Because the closed classes are similar to one another and appear to represent a continuous distribution of conformations, which differ mainly in the degree of gating ring rotation with respect to the transmembrane domain over a 7° range of angles, it was difficult to identify with certainty a specific corresponding class from one 3D classification to the next. We therefore focused our analysis of closed conformations on two classes on the opposite ends of the rotational distribution (i.e. the two most different closed conformations, which we have named closed class 1 and closed class 9). The fraction of particles for these two classes as a function of Na⁺ concentration for each of the five 3D classifications is shown (Figure 6C–F). We observe no systematic change in the fraction of particles as a function of Na⁺ concentration. We note that the variation between runs of 3D classification is greater for the closed conformations than for the open conformation (Figure 6B,D,F). This finding can be explained on the basis of relatively small differences between the closed structures, which result in broader likelihood probability distributions (i.e. reduced precision) (Figure S5A, B).

Discussion

The analysis of class probability relies on the 3D classification algorithm in Relion (Scheres, 2012). By carrying out repeated trials the reproducibility indicates that the probability determination for the open class is quite precise. Through indirect reasoning we also think that the probability determination for the open conformation is reasonably accurate as well: the map of the open structure is of very high quality (Figures 2, S1 and S3). If closed channels were contributing significantly to the map we should expect significant blurring of density, especially in the regions known to undergo large conformational change, which includes a 5 Å displacement of the transmembrane domain relative to the gating ring. But this is not the case (Figure S1). Even the relatively small number of open particles (7%) in the titration data set yielded a map to a resolution of 4.1 Å with well-defined density throughout the molecule. It is evident that the precision and likely the accuracy are lower for determining closed conformation probabilities. However, we think that this analysis can reliably distinguish among the most different closed state conformations, classes 1 and 9.

The appearance of the open conformation depends on Na⁺. The probability of being in any of the closed conformations of course also depends on Na⁺ (because closed equals not open), however, we do not observe a systematic change in the probability of being in a particular closed conformation relative to other closed conformations - at least based on our analysis of the two most easily distinguishable extremes (Figures 6 and S4). That Na⁺ dependence is expressed only in the appearance of a unique open conformation is somewhat unexpected. One might have imagined multiple Na⁺-dependent conformations on a reaction pathway leading from closed to open conformations, but the data are not consistent with such a picture. Instead, there appears to be an ensemble of Na⁺-independent closed conformations from which the open conformation emerges in a highly Na⁺-dependent fashion. Even when we processed images without imposing symmetry, specifically looking for asymmetric conformations, we did not observe any. Thus, at the level of structural detail currently accessible to us we conclude that stable intermediate conformations between the closed ensemble and the open conformation do not exist. In other words, channel opening is highly concerted, giving rise to a steep switch-like activation function.

The curve describing open conformation probability appears to approach 1.0 at high Na⁺ concentrations while the curve for functional open probability reaches a maximum value of only about 0.7 (Figure 7H). This discrepancy is possibly due to differences in the experimental conditions of the function and structure experiments. Channel activity was measured in membranes at room temperature while structural measurements were carried out in detergent, with the sample being cooled during vitrification. Further complicating the comparison, the vitrification procedure creates an uncertainty about the final buffer composition of a cryo-EM grid because of evaporation prior to freezing. Given these differences and uncertainties we were actually surprised to see how well the Na⁺-dependence curves correlated, particularly with respect to their midpoint values and general shape.

It is entirely possible that the maximum value discrepancy is not related to experimental differences, but instead to the existence of a functional state of the channel we have yet to detect in the structure experiments. Many ion channels never reach a functional open probability of 1.0 even in the setting of strong stimulus (e.g. a high ligand concentration or a strongly depolarized membrane voltage) (Islas and Sigworth, 1999; Jahr, 1992; Mak and Foskett, 1994; Wang et al., 2016). The explanation is that the stimulus causes the channel to conduct by energetically favoring an open, conductive conformation relative to a closed, non-conductive conformation; however, once the open conformation is reached additional non-conductive states can occur. This is the general concept behind ‘inactivation’ in voltage-gated channels and ‘desensitization’ in ligand-gated channels. Therefore, the discrepancy in maximum values lends itself to the following possible interpretation. Na⁺ drives the equilibrium such that all of the channels undergo a large global conformational change that we recognize as an open conformation. However, approximately 30% of those open channels are not conductive at any given moment.

The non-conductive channels are somehow distinct, perhaps they are inhibited by a blocking ion, or they may contain a small conformational change elsewhere along the ion conduction pathway that we do not see in our density maps at the current resolution. With better resolution we can hope to someday distinguish such a second class of open conformation channels to explain the difference in maximum values of the activity and structure titrations.

Methods

CONTACT FOR REAGENT AND RESOURCE SHARING

Requests for reagents may be directed to Lead Contact Roderick MacKinnon (mackinn@mail.rockefeller.edu).

EXPERIMENTAL MODEL AND SUBJECT DETAILS

Cell lines

Sf9 (Spodoptera frugiperda) cells were used for protein expression and maintained in Grace’s Media supplemented with 10% Fetal Bovine Serum at 27° C. Electrophysiological recordings were performed using HEK293T cells maintained in high glucose DMEM supplemented with 10 % FBS and 1% L-Glutamine at 37° C in 5% CO₂.

METHOD DETAILS

Constructs

A synthetic gene fragment encoding residues 1 to 1201 of Chicken Slo2.2 was purchased from Bio Basic Inc. The fragment was cloned into a modified pFastbac vector (Invitrogen) containing green fluorescent protein and a 1D4 antibody recognition sequence (TETSQVAPA) (Wong et al., 2009) on the C terminus (Hite et al, 2015). The plasmid was transformed into DH10Bac E. coli cells to generate bacmid DNA. Recombinant baculovirus was produced by three rounds of viral amplification in Sf9 cells.

For electrophysiological recordings, the Slo2.2 gene fragment was cloned into a modified pCEH vector with a GFP and His₈-tag on the C terminus. Site directed mutagenesis was performed to mutate Met-333 to Ala.

Expression and purification

For large-scale expression, Sf9 (Spodoptera frugiperda) cells infected with baculovirus were cultured at 27°C for 72 hours in supplemented Grace’s insect cell medium (Invitrogen). Cells were washed with ice-cold phosphate-buffered saline and extracted for 3 hours at 4°C with buffer containing 50 mM Hepes pH 7.4, 300 mM KCl, 300mM NaCl and 40 mM dodecyl-β-D-maltopyranoside (DDM) in the presence of a protease inhibitor cocktail (2 μg/ml leupeptin, 2 μg/ml aprotinin, 2 μg/ml pepstatin A, 1 mM benzamidine, 100 μg/ml 4-(2-aminoethyl) benzenesulfonyl fluoride hydrochloride, and 100 μM phenylmethane sulphonylfluoride). The insoluble fraction was removed by centrifugation at 35,000 x g for 45 minutes at 4°C and the remaining soluble fraction was incubated with 1D4-affinity resin pre-equilibrated with 20 mM Hepes pH 7.4, 300 mM KCl, 300mM NaCl and 4 mM DDM. The suspension was mixed for 4 hours at 4°C. Beads were collected on a column by gravity and then washed with 10 column volumes of wash buffer (20 mM Hepes pH 7.4, 300 mM KCl, 4 mM DDM and 0.1 mg/ml 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphoethanolamine (POPE): 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphoglycerol (POPG) (3:1 w/w)). The protein was digested with PreScission protease (20:1 w/w ratio) on the column overnight at 4°C to remove the affinity tag and then eluted with 2 column volumes of wash buffer. Concentrated protein was further purified by size-exclusion chromatography on a Superose 6 (GE Life Sciences) column in 20 mM Hepes pH 7.4, 300 mM KCl, 300mM NaCl, 1.5 mM DDM and 0.05 mg/ml POPE:POPG (3:1 w/w). Peak fractions were pooled and concentrated to 6 mg/ml for cryo-EM analysis. For the 20mM, 40 mM, 80mM or 160mM Na⁺ data sets, NaCl was excluded from the purification buffers and added to a final concentration of 20mM, 40 mM, 80mM or 160mM immediately prior to EM sample preparation. Total ionic strength of the buffer was equilibrated by the addition of KCl.

Electron microscopy sample preparation and imaging

3.5 μl of purified channel at a concentration of ~6 mg/ml was pipetted onto glow-discharged copper Quantifoil R 1.2/1.3 holey carbon grids (Quantifoil). Grids were blotted for 4 s with a blotting force of 1 and a humidity of ~88% and flash frozen in liquid-nitrogen-cooled liquid ethane using an FEI Vitrobot Mark IV (FEI). Grids were then transferred to an FEI Titan Krios electron microscope operating at an acceleration voltage of 300 keV. Images were recorded in an automated fashion on a Gatan K2 Summit (Gatan) detector set to super-resolution counting mode with a super-resolution pixel size of 0.65 Å using SerialEM (Mastronarde, 2005). Images of the 20 –160 mM Na⁺ samples were recorded for 15 s with a subframe exposure time of 300 ms at a defocus range of −1.0 to −3.0 μm and a dose of approximately 10 electrons per pixel per second for a total accumulated dose of approximately 89 electrons per Ų on the specimen over 50 subframes (approximately 1.8 electrons per Ų per subframe). Images of the 300 mM Na⁺ samples were recorded for 10 s with a subframe exposure time of 200 ms at a defocus range of −1.0 to −3.0 μm and a dose of approximately 10 electrons per pixel per second for a total accumulated dose of approximately 59 electrons per Ų on the specimen over 50 subframes (approximately 1.2 electrons per Ų per subframe).

Image processing and map calculation

For the 20 –160 mM Na⁺ images, the dose-fractionated super-resolution images were 2 × 2 down sampled using Fourier cropping (resulting in a pixel size of 1.30 Å) followed by whole-frame motion correction and dose-filtration using Unblur (Grant and Grigorieff, 2015b). Contrast transfer function parameters were estimated using CTFFIND4 (Rohou and Grigorieff, 2015). ~5,000 particles were interactively selected using RELION to generate templates representing different views for automated particle selection (Scheres, 2012, 2015). The autopicked particles were manually inspected to remove false positives, resulting in 62K, 102K, 157K, and 140K particle images for the 20 mM, 40 mM, 80 mM and 160 mM Na⁺ data sets, respectively. Following auto-picking and manual inspection, all 461k particles were subjected to 3D auto-refine in RELION (Scheres, 2012; Kimanius et al, 2016). The angular and translational parameters determined using 3D auto-refine were fixed and maintained throughout 25 cycles of 3D classification specifying 10 classes in RELION with the map generated by 3D auto-refine low-pass filtered to 60 Å serving as the initial model. In order to characterize the reproducibility of the 3D classification algorithm, the 461k particles were subjected to 5 separate cycles of 3D auto-refine and 3D classification. Manual inspection identified one open class as well as two closed classes with the most extreme rotations (closed class 1 and closed class 9) for each calculation. The contributions of the 4 data sets (20 mM, 40 mM, 80 mM and 160 mM Na⁺) to each of the 10 3D classes were determined using identifiers uniquely associated with each particle during particle extraction.

For the 300 mM Na⁺ images, the dose-fractionated super-resolution images were 2 × 2 down sampled using Fourier cropping (resulting in a pixel size of 1.30 Å) and corrected for anisotropic scaling with mag_distortion_correct followed by whole-frame motion correction with Unblur (Grant and Grigorieff, 2015a, b). The parameters of the contrast transfer function were estimated by ctffind4 (Rohou and Grigorieff, 2015). Following whole-frame motion correction, 160k particles were interactively selected using RELION. The particles were then corrected for per-frame motion correction and dose filtered using alignparts_lmbfgs (Rubinstein and Brubaker, 2015). The ~160k particle images were subjected to 3D auto-refine in RELION, followed by 3D classification into 10 classes with fixed angular and orientation parameters. Eight of the classes adopted the open conformation, totaling ~130k particles, and were combined for refinement in FREALIGN (Lyumkis et al., 2013). The first iteration of FREALIGN, the global search, was performed with the reference map limited to a resolution of 10 Å. Successive rounds of refinement were performed with higher resolution reference maps, up to a maximum resolution of 6 Å for the final iterations. The resolution of the final map was estimated to be 3.8 Å as assessed by Fourier shell correlation using the 0.143 cut-off criterion (Figure S1) (Rosenthal and Henderson, 2003). During the later stages of refinement a soft mask was placed around the protein density. Outside of the mask, the reference was limited to a resolution of 30 Å for alignment. The two closed classes, totaling 32k particles, were combined and refined using a similar procedure in FREALIGN. The final map was calculated using a reference that was resolution-limited to 6 Å and achieved a resolution of 4.3 Å as assessed by Fourier shell correlation using the 0.143 cut-off criterion (Figure S1). The final maps were sharpened using an isotropic b-factor of −150 Ų. Local resolution estimations were calculated using RESMAP (Kucukelbir et al., 2014).

Model building and refinement

The cryo-EM structure of the Na⁺-free structure (PDB: 5A6E) was docked into the open map using UCSF Chimera and manually rebuilt in coot to fit the density (Emsley et al., 2010; Pettersen et al., 2004). The register of the S1–S4 domain was assigned using large side chains. The cryo-EM density map of one of the half-maps corresponding to a new smaller unit cell that extended 5 Å from the model in all directions was extracted. Cycles of real space refinement using phenix.real_space_refine and reciprocal space refinement using REFMAC were alternated with manual rebuilding in Coot (Adams et al., 2011; Brown et al., 2015). Geometric and secondary structure restraints were tightly maintained throughout refinement to minimize over-fitting. To monitor the effects of over-fitting, the Fourier shell correlation of the refined model was determined for the half-map used during refinement (FSC work) and the half-map that was not used at any point during refinement (FSC free) (Figure S7). The final refined structure of the contained residues 69–118, 139–618, 720–1019, 1098–1135 and 1138–1171.

Following refinement, the open structure was docked into the closed map using UCSF Chimera and manually rebuilt in coot to fit the density. Residues for which the density was missing or poor were deleted from the model. The model was then refined using phenix.real_space_refine and REFMAC. Local resolution estimates were calculated using ResMap. The final refined structure of the contained residues 77–117, 140–235, 243–334, 341–618, 720–858, 864–1021 and 1097–1169. All structure calculations were performed using software compiled by SBGrid (Morin et al., 2013). Structure figures were prepared with UCSF Chimera and Pymol (Pymol version 1.7.2 Schrodinger LLC).

Electrophysiological recordings from HEK 293T cells

Chicken Slo2.2 with green fluorescent protein was cloned into pCEH vector with GFP at the C-terminus for mammalian cell expression. HEK 293T cells were maintained in DMEM supplemented with 10% FBS. Cells were transfected using FuGene HD according the manufacturers protocol (Promega). 48 hours following transfection, the cell culture dish was transferred and media was replaced with bath solution. All recordings were performed at room temperature on an Axopatch 200B amplifier (Molecular Devices) using the inside-out patch-recording configuration. Recordings were filtered at 1 kHz using a low-pass Bessel filter and digitized at 10 kHz using a Digidata 1322A analogue-to-digital converter (Molecular Devices). Pipettes of 2–3 MΩ resistance were pulled from borosilicate and fire polished. Currents were recorded during voltage steps from −80 to 0 mV from a holding potential of 0 mV. NaCl concentration was controlled by local perfusion at the patch. The pClamp software suite (Molecular Devices) was used to control membrane voltage and record current. Pipette solution was 150 mM KCl, 5 mM EDTA, and 10 mM Hepes (pH 7.4). Bath solution was 150 mM KCl, 3 MgCl₂, 1 mM CaCl₂, and 10 mM Hepes (pH 7.4). Perfusion solutions were 0–600 mM NaCl, 150 mM KCl, and 10 mM Hepes (pH 7.4).

The unitary conductance and number of channels in the patch were calculated from the non stationary noise generated by channel opening and closing at −60 mV in the presence of varying Na⁺ concentrations (Sigworth, 1980). Data between the time interval 220 to 800 ms were used for calculating the mean ionic current (<I>) and variance (σ²). The number channels (N) and the unitary conductance (i) were determined by fitting the plot of the variance (σ²) of the current trace against the mean ionic current (<I>) to the parabolic function:

$${\mathrm{\sigma }}^2={\mathrm{i}}^{\ast }<\mathrm{I}>-<\mathrm{I}{>}^2/\mathrm{N}$$ (1)

The P_O at each Na⁺ concentration can be calculated as the ratio of mean ionic current (<I>) and theoretical maximal current, which is the product of the unitary conductance (i) and the number of channels (N):

$${\mathrm{P}}_{\mathrm{O}}=<\mathrm{I}>/({\mathrm{i}}^{\ast }\mathrm{N})$$ (2)

P_O is plotted as function of the Na⁺ concentration and fit with the following Hill equation:

$${\mathrm{P}}_{\mathrm{O}}={\mathrm{P}}_{\mathrm{O}}\_max/(1+{(\mathrm{K}/[{\text{Na}}^{+}])}^{\mathrm{n}})$$ (3)

where K is an estimate of the apparent affinity for Na⁺ and n is the Hill coefficient. Data were fit using OriginPro (OriginLab Corp.).

QUANTIFICATION AND STATISTICAL ANALYSIS

CryoEM

All reported resolutions are based upon the 0.143 Fourier Shell Correlation criterion (Rosenthal and Henderson, 2003). Error bars for the fraction of particles in each class (Figures 7H) represent standard deviation of the fraction particles classified into the open class during 5 independent classifications of the images.

Electrophysiology

Error bars in Figure 7H represent standard error of the mean for five independent experiments.

DATA AND SOFTWARE AVAILIBILITY

Data Resources

Maps of open and closed Slo2.2 have been deposited with Electron Microscopy Data Bank with accession codes EMD-8515 and EMD-8517, respectively. Atomic coordinates for open and closed Slo2.2 have deposed with the Protein Data Bank with accession codes 5U70 and 5U76, respectively.

Supplementary Material

1. Figure S1. Cryo-EM reconstruction of chicken Slo2.2, related to Figure 2.

(A) Representative micrograph and 2D class averages of Slo2.2 vitrified in the presence of 300 mM Na⁺. Scale bar represents 500 Å. (B) Fourier shell correlation curve for open (blue) and closed (red) chicken Slo2.2. Overall resolution estimated to be 3.78 Å for the open conformation and 4.27 Å for the closed conformation on the basis of the FSC = 0.143 (dashed line) cut-off criterion. (C,D) Angular distribution plots for open (C) and closed (D) Slo2.2 reconstructions. (E,F) cryo-EM density maps of open (E) and closed (F) Slo2.2 colored by local resolution using Resmap (Å).

2. Figure S2. Comparison of open and closed Slo2.2 with Na⁺-free Slo2.2, related to Figure 2.

(A) Superposition of open Slo2.2 (blue) with Na⁺-free Slo2.2 (black) aligned using all main-chain atoms. Dashed lines represent the width of the section at right. (B) Superposition of closed Slo2.2 (red) with Na⁺-free Slo2.2 (black) aligned using all main-chain atoms. Dashed lines represent the width of the section at right.

3. Figure S3. Cryo-EM density map segments of open Slo2.2, related to Figure 2.

(A) Ribbon digram of open Slo2.2 colored by domain. S1–S4 domain is colored green, pore domain is colored yellow, RCK1 domain is colored blue and RCK2 domain is colored red. Dashed boxes correspond to regions shown in panels B and C. (B) Ribbon diagram of S6-RCK1 linker with S6 is colored in yellow and RCK1 colored in blue. Cryo-EM density is displayed as mesh. (C) Ribbon diagram of S4–S5 linker with S4 colored in green and S5 is colored in yellow. Cryo-EM density is displayed as mesh. (D) Segments of the open Slo2.2 cryo-EM density map corresponding to S0 and S1, S2, S3, S4, S5, S6, αA, βD, αD, βL, αQ and βR shown as grey wire mesh. Channel is shown as sticks and colored by atom.

4. Figure S4. Heterogeneity of Slo2.2 particles imaged in 20 – 160 mM Na⁺, related to Figure 6.

(A) Cryo-EM density maps of closed class 1 (red) and open class (blue) aligned by their gating rings. Dashed lines represent the width of the density section at right. (B) Cryo-EM density maps of closed class 9 (cyan) and open class (blue) aligned by their gating rings. Dashed lines represent the width of the density section at right. (C) Cryo-EM density maps of closed class 1 (red) and closed class 9 (cyan) aligned by their gating rings. Dashed lines represent the width of the density section at right. Black lines extend from the central axis through the S0 density.

5. Figure S5. Precision of 3D classification of open and closed Slo2.2, related to Figure 6.

(A,B) Histogram of MaxValueProbabilityDistribution for open (A) and closed (B) Slo2.2 for all particles of the titration data set following 3D classification. Number of particles for each distribution to normalized to total of 1. (C) Plot of the fraction of particles classified n times into the open class during five independent 3D refinement and classification runs.

6. Figure S6. Na⁺-dependent activation of Slo2.2 in cellular membranes, related to Figure 7.

(A,C,E) Macroscopic current traces of intracellular Na⁺ activation. Currents were recorded using the inside-out patch-configuration; pipette solution contained 150 mM KCl, 5 mM EDTA, 10 mM Hepes (pH 7.4) and bath solution contained 150 mM KCl, 10 mM Hepes (pH 7.4), and 0 mM to 600 mM NaCl controlled by bath perfusion. Membrane voltage was held at 0 mV and stepped to −60 mV. (B,C,D) Variance (σ²)-mean current (<I>) relationship fit to equation 1 (methods). The fit yielded estimates of unitary conductance (i) and channel number (N) (equations 1 and 2, methods).

7. Figure S7. Validation of the refined model, related to Figure 2.

(A) Coordinate refinement statistics for Slo2.2 open and closed models. (B,C) Fourier shell correlation curves of refined open (B) and closed (C) models versus unmasked maps for cross-validation. The black curve is the refined model compared to the full data set (FSC sum), the blue curve is the refined model compared to half-map 1 (FSC work, used during refinement) and the red curve is the refined model compared to half-map 2 (FSC free, not used during refinement).

8. Video S1. Activation of Slo2.2, related to Figure 2.

This video shows the conformational changes that occur during activation of Slo2.2.

Highlights.

• Cryo-EM and functional titrations of Slo2.2

• Structures of Slo2.2 in open and closed conformation

• Switch-like activation mechanism between open and closed conformations

• Evidence of non-conductive channels in an open conformation