Real time dynamics of Gating-Related conformational changes in CorA

CorA, a divalent-selective channel in the metal ion transport superfamily, is the major Mg2+-influx pathway in prokaryotes. CorA structures in closed (Mg2+-bound), and open (Mg2+-free) states, together with functional data showed that Mg2+-influx inhibits further Mg2+-uptake completing a regulatory feedback loop. While the closed state structure is a symmetric pentamer, the open state displayed unexpected asymmetric architectures. Using high-speed atomic force microscopy (HS-AFM), we explored the Mg2+-dependent gating transition of single CorA channels: HS-AFM movies during Mg2+-depletion experiments revealed the channel’s transition from a stable Mg2+-bound state over a highly mobile and dynamic state with fluctuating subunits to asymmetric structures with varying degree of protrusion heights from the membrane. Our data shows that at Mg2+-concentration below Kd, CorA adopts a dynamic (putatively open) state of multiple conformations that imply structural rearrangements through hinge-bending in TM1. We discuss how these structural dynamics define the functional behavior of this ligand-dependent channel.


Lipid composition determines reconstitution density and morphology
In recent years, HS-AFM has demonstrated its power to observe molecular mechanisms and structural dynamics of single molecules under physiological conditions with extraordinary detail (Chiaruttini et al., 2015;Kodera et al., 2010;Marchesi et al., 2018;Preiner et al., 2014;Rangl et al., 2016;Ruan et al., 2018;Ruan et al., 2017). Here, we have used HS-AFM to track the conformational changes of CorA as a result of Mg 2+ -concentration changes in real-time. For this, wild-type (wt) CorA was reconstituted at low lipid-to-protein ratios (LPRs between 0.2 and 0.4) in the presence of Mg 2+ . The resulting proteo-liposomes were adsorbed onto freshly cleaved mica under saturating Mg 2+ (10 mM) condition and imaged at an acquisition rate of 1-2 frames s À1 with a resolution of 0.5 nm pixel À1 . CorA-containing vesicles were generated from protein reconstitutions in POPC/POPG (3:1, w:w). When imaged by HS-AFM, vesicles spread on the mica support resulting in large membrane patches with densely packed CorA (Video 1). These patches mostly exposed the periplasmic face. Smaller crowded areas of molecules expose the intracellular side ( Figure 1a, Video 1).
Cross-section analysis of the CorA membrane patches showed overall heights of~12 nm (Figure 1b). High-resolution movies of the periplasmic domains revealed protrusions of 0.9 ± 0.4 nm in height and 3.4 ± 0.4 nm in diameter, and resolved a central indentation where the channel pore is located (Figure 1c). Despite the fast diffusion of CorA molecules within the clusters exposing the intracellular side, some high-resolution snapshots allowed analysis of the surface structure. The intracellular face protruded~7 nm from the membrane (Figure 1-figure supplement 1a) with top-ring and outer diameters of~6 nm and~10 nm, respectively (Figure 1-figure supplement 1b,c; Video 2). These measurements are in good agreement with the molecular dimensions of CorA (Eshaghi et al., 2006;Lunin et al., 2006;Matthies et al., 2016).
However, it became clear that in order to study Mg 2+ -dependent conformational dynamics, more stably packed molecules were required (Müller et al., 1999;Ramadurai et al., 2010). We found that reconstituting the protein in DOPC/DOPE/DOPS (4:5:1, w:w:w) and adsorbed and imaged at lower pH 6.0 resulted in widespread surface coverage of CorA-crowded membranes with only slowly moving molecules. Moreover, under these conditions densely packed CorA patches were stacked and exposed the intracellular face of the channel (Figure 1d,e; Video 3), thus providing an excellent experimental platform for studying CorA Mg 2+ -dependent conformational changes. High-resolution HS-AFM topographs displayed a 'flower-shaped' surface structure corresponding to the CorA intracellular face. This view allowed to resolve individual subunits in the pentamer with top-ring diameter of 5.0 ± 0.9 nm and a center-to-center distance of the molecules, that is outer diameter, of 10.9 ± 2.1 nm ( Figure 1f).

Mg 2+ -depletion induces large conformational changes of the intracellular face
We then monitored the structural changes of the channel upon Mg 2+ -depletion in real time at higher magnification, that is at scan sizes < 200 nm and pixel sampling of 0.5 nm pixel À1 . First, we studied the periplasmic face of CorA under saturating Mg 2+ (10 mM), in which all channels are expected to be in the closed conformation. Subsequently, membranes were imaged at reduced Mg 2+ -concentration near the reported apparent K d for Mg 2+ (~2 mM), and finally at 0 mM Mg 2+ to focus on the dynamically open Apo form.
Video 1. CorA reconstituted in POPC/POPG liposomes. In overview scans the proteoliposomes were opened by applying slightly increased loading forces, thereby revealing CorA membranes exposing both faces, the periplasmic face in crystalline packing and the intracellular side in crowded membrane areas.  Mg 2+ -depletion was achieved by live injection of EDTA, while continuously monitoring CorA. Using the closed state as reference, no structural changes of the periplasmic face of CorA were observed at 10 mM Mg 2+ , at 2 mM Mg 2+ (after~10 min) and in absence of Mg 2+ (after~12 min) ( Figure 2-figure supplement 1). At the present resolution (0.5 nm pixel À1 ) any putative changes in the conformation of the periplasmic face of the channel are beyond the current resolution limit of HS-AFM (although we cannot rule out that the structures of the small periplasmic loops might not change upon Mg 2+ -depletion). This is consistent with the~7 Å cryo-EM structures, which show little or no Mg 2+ -dependent changes on the periplasmic face (Matthies et al., 2016;Pfoh et al., 2012).
In stark contrast, previous EPR (Dalmas et al., 2010;Dalmas et al., 2014b) and cryo-EM (Matthies et al., 2016) analyses have revealed dramatic conformational rearrangements of the intracellular domains (resulting in a loss of symmetry) in the nominal absence of Mg 2+ (conditions favoring the functionally open state) ( Figure 2a). First, to measure the CorA Mg 2+ -affinity (Pilchova et al., 2017) in our experimental set-up, we used a microfluidic system connected to a constant pressure and flow pump , with which we slowly exchanged the complete 10 mM Mg 2+ measuring solution to a Mg 2+ -free buffer (containing additional 2 mM EDTA). Analysis of the distribution of symmetric and dynamically asymmetric CorA particles pointed to a K d of~2 mM Mg 2+ , which is in good agreement with the reported affinity ( Figure 2-figure supplement 2). Next, we monitored the structural and dynamical transition of CorA upon Mg 2+ -depletion by pipetting defined amounts of EDTA into the measurement fluid cell to achieve the following equilibrium concentrations: 10 mM Mg 2+ ,~2 mM Mg 2+ (~K d ), and 0 mM Mg 2+ (putative).
Consistent with the Mg 2+ -liganded cryo-EM structure ( Figure 2a, left), at saturating 10 mM Mg 2+ condition CorA channels revealed a stable, flower-like 5-fold symmetric structure (Figure 2b, left). However, once Mg 2+ -concentrations dropped to~2 mM and below, individual channels start to fluctuate between various structural states, occasionally assuming increased height and, after~20 min in absence of Mg 2+ , adopting an ill-defined conformation with significantly increased protrusion height (Figure 2b, right). The Mg 2+ -dependent loss of symmetry and increased protrusion height are also consistent with expected structural features of the Mg 2+ -free (open) CorA structures, where individual subunits move towards the former 5-fold axis and thus stand taller (Matthies et al., 2016) ( Figure 2a, bottom). The large,~1.5 nm, protrusion height difference (DHeight) of such a conformation change presents a useful signature to detect and follow the conformational dynamics of individual channels in the membrane. However, these two DHeight-states must not be mistaken with functional or even structural states. They are merely a way to topographically discriminate between the closed 5-fold symmetric state and any other state where single subunits stand up, that is move towards the 5-fold axis and thus appear higher.
Height section kymographs of individual molecules (    . We must highlight, however, that a total depletion of Mg 2+ probably never occurs under physiological conditions, thus the fully Mg 2+depleted asymmetric stable states observed by cryo-EM and likely adopted here at the end of the depletion experiments might not be visited in the cell. More likely, the 5-fold symmetric closed state interchanges with a highly fluctuating molecule where single subunits dissociate from the quaternary structure of the cytoplasmic ensemble, opening the channel.

CorA fluctuates between several conformations in the open state
Following the general observation that the stable closed and open conformations are interconnected by a regime in which the channels are highly dynamic, we pursued a detailed structural examination of CorA in this intermediate dynamic stage at low Mg 2+ -concentrations. High-resolution HS-AFM image sequences of individual CorA channels revealed that CorA undergoes conformational rearrangements of the entire cytosolic Mg 2+ -sensor domain at rates beyond the bandwidth of the current measurements (<550 ms, the imaging rate of our videos). We also acquired movies at 250 ms frame acquisition and found that individual sequential frames displayed different apparently unrelated conformations, an indication that the conformational fluctuations are even faster (Figure 4d). Movie snapshots were classified into four conformational classes (Figure 4, Video 6): First, the fully-liganded 5-fold symmetric closed conformation (Eshaghi et al., 2006;Lunin et al., 2006;Matthies et al., 2016;Payandeh and Pai, 2006) Figure 4d, 1.5 s, 1.75 s). It is clear that a key challenge in defining the conformational landscape of CorA will be the unbiased classification of discrete states in this highly flexible molecule, made evident by following a single CorA channel frame by frame (Figure 4c,d). Unfortunately, all our computational classification attempts failed likely due to the presence of protein-protein contacts with neighboring molecules and limited resolution due to the high mobility.
To work out a more detailed picture of the CorA-transitions we visually assigned these four different types of states to high-resolution HS-AFM movie sections recorded at different Mg 2+ - Figure 3 continued low-to-high ('up', red) events binned over a time window of 30 s. Turquois pentagons and red diamonds indicate average closing and opening event time-points of corresponding average dwell-time (right axis), respectively. These averages were calculated over a sliding window of 20 events along the time axis. Analysis included molecules from 2 experiments at 0 mM Mg 2+ (shown in (b) top and bottom) and 30,500 HS-AFM images thereof. Right: Histogram of average dwell-times in the 'high' (red) and 'low' (turquois) states (where the high state represents/comprises all conformational states with elevated subunits).  We suggest that molecules with increased height where one subunit moves up close to the channel axis represent Mg 2+ -free states.

Discussion
On the basis of real-time HS-AFM imaging, we find that CorA Mg 2+ -dependent gating can be best described in three phases: Above the apparent Mg 2+ -affinity of the intracellular Mg 2+ -sensor sites (~2 mM Mg 2+ ), the channel adopts a stable 5-fold symmetric state reminiscent of the high-resolution structures (phase 1). After several minutes of exposure to a Mg 2+ -free solution, CorA transitions into a set of asymmetric architectures, which are characterized by elevated cytoplasmic   Color scale was adapted for each condition separately with a gradient from green (lowest occurrence of transition) over yellow and orange to red (highest occurrence of transition). Numbers in the center of boxes of 4 state transitions represent the sum of transitions between states with elevated subunits (blue dashed square) and between transitions of strongly elongated structures (red square). (c) CorA conformational transition model based on the HS-AFM observations. Within~10 min of Mg 2+ -depletion, the 5-fold symmetric, fully Mg 2+ -liganded CorA transit into dynamically fluctuating molecules with flexible subunits until their conformation stabilizes in a Mg 2+ -free highly asymmetric structure with increased membrane protrusion height. Figure 5 -Information Supplement 1: Estimation of thermally activated TM1 helix motions We estimated the theoretical range of helical motion by considering that TM1 behaves like a flexible rod undergoing thermally excited motions. The helix (rod) is characterized by a specific persistence length L P that is related to the bending stiffness K S through L P ¼ KS kBT . The basic description for the change in curvature between two points on the rod is given by qt s ð Þ qs , with s being the arc length andt a unit tangent vector at position (s). In an ideal system, the total elastic energy E ela of a particular conformation is given by the integral of the bending energies accumulated along a rod with contour length L: Ks 2 qt s ð Þ qs Þ 2 ds Assuming only circular curvatures along the rod, qt s ð Þ qs ¼ 1 r , where r is the radius of curvature. Using this basic description of an elastic polymer rod and considering the persistence length of protein a-helices L p = 100 nm (as described in Choe and Sun, 2005) and a contour length L = 11 nm (length of TM1, see domains (phase 3). These structures are adopted through major conformational changes implicating hinge-bending of TM1 and reorientation of the cytoplasmic domains. This picture is wholly consistent with data derived from functional (electrophysiology), biochemical (crosslinking), biophysical (EPR, Fluorescence) and structural (cryo-EM) approaches. However, these results contrast with recent crystallographic data of CorA Mg 2+ binding site mutants that reported limited or no structural changes compared to the closed wt channel, thereby challenging the hypothesis that Mg 2+ -depletion leads to large conformational changes and channel opening (Kowatz and Maguire, 2019). We argue, that the constrains of the crystal lattice on CorA limits the range of conformational changes that can be observed in X-ray structures. In contrast, HS-AFM imaging under physiological conditions shown here is more consistent with the single particle cryo-EM data reported in Cleverley et al. (2015); Matthies et al. (2016) and the EPR spectroscopic data (Dalmas et al., 2010;Dalmas et al., 2014b), that is when the Mg 2+ is investigated under lattice-free conditions. These facts must be considered a critical factor in explaining these differences. Among multiple conformational states within phase 3, we identified CorA structures that resembled the low-resolution cryo-EM open-I and open-II structures. These phases 1 and 3 are characterized by structural stability or at least limitations by the structural freedom of CorA. Here, more work is required to identify and clearly distinguish between different sub-structures in the Mg 2+ -free elevated state of CorA. In contrast, at intermediate Mg 2+concentrations, and/or after short incubation times at 0 mM Mg 2+ , the channel reveals a highly mobile and fluctuating state (phase 2). Active subunits change and the entire molecule displays high structural variability, likely adopting many more conformations than the three states so far identified by cryo-EM. Indeed, while we can assign the fully 5-fold symmetric state with certainty, our assignment of molecules to state open-I and open-II must be considered as putative. Perhaps more importantly, at intermediate Mg 2+ -concentrations~50% of the molecules could not be assigned and were pooled in a class of undefined quaternary structures (open-+). In this context, the details of the cryo-EM study become relevant: open-I and open-II only pooled 26,271 and 27,416 particles, respectively, of the total 173,653 particles, and thus a significant portion of the particles remained outside of these classes (Matthies et al., 2016). We note, however, that this initial assignment is ultimately constrained by the resolution limitations of both cryo-EM structural assignments and the present application of HS-AFM. We expect that given a larger number of particles to process (in cryo-EM) or improvements in the spatial and/or temporal resolution of HS-AFM, additional states are likely to emerge. We propose that the various CorA structural states and the associated subunit movements observed by HS-AFM may reflect variations of number and location of Mg 2+ -ions bound to the cytosolic domain (Figure 5c).
Among the main structural features of CorA is its long (~11 nm) first transmembrane helix (TM1), ranging all the way from the periplasmic face to the surface of the cytoplasmic Mg 2+ -sensor. TM1 is actually the only connection between the TM region and the cytoplasmic domain ( Figure 5-figure  supplement 1). Thus, it seems plausible that the large fluctuations of the cytoplasmic domains might be directly translated by TM1 into fluctuations within the channel pore. We suggest this mechanism as the basis for ion conductance through a particularly long (in comparison to other channels)~55 Å pore, in which moving TM1s allow the asynchronous progression of ions through the pore. Considering the typical persistence length of an a-helix of~100 nm (Choe and Sun, 2005), allowed us to estimate how much an 11 nm long thermally activated helix fluctuates ( Figure 5 -Information Supplement 1). We find that at 1k B T TM1 could bend at its cytoplasmic end~2 nm out of axis, in good agreement with observed bending of this helix in the open-I and open-II cryo-EM structures and the HS-AFM movies of subunit displacement. Thus, we propose that in the open state the pore is by far less narrow than assumed based on static structures. Indeed, constantly fluctuating TM1s Video 8. CorA molecules monitored just after complete Mg 2+ -depletion. The majority of the molecules are in a highly dynamic state constantly switching between different structural states. Video settings: Full frame size: 100 nm, 200 pixels, scan rate: 1.8 frames s À1 , full color range: 5 nm.
https://elifesciences.org/articles/47322#video8 would provide a considerable channel diameter towards the cytoplasmic side. This model would predict that amino acids close to the periplasmic side of the pore play the key role for gating. In agreement with this hypothesis, we find that the Mg 2+ -channel signature motif (GMN) is located right at the periplasmic end of the channel, whereas amino acids towards the cytoplasmic face of TM1 are less conserved ( Figure 5-figure supplement 1).
The idea that there are multiple conductive conformations of CorA, or, as a paradigm shift, the conductive conformation of CorA is a fluctuating molecule, is favored by both, the subunit movement analysis and state transition analysis. In the long-term absence of Mg 2+ , CorA molecules adopt a rather stabilized asymmetric state resembling the putatively open state cryo-EM structures. However, a living cell under physiological conditions is unlikely to ever reach Mg 2+ -concentrations below~1 mM. Thus we hypothesize that in cellula the physiologically relevant open state is a collection of fluctuating conformational states. Thus, the conformational energy landscape of CorA gating would consist of two deep energy minima representing the stable closed (symmetric) and putatively open (asymmetric) conformations connected by a wide plateau in which a variety of open conformations are adopted through permanent molecular fluctuations. The compelling utility of HS-AFM for the study of macromolecular dynamics at high spatio-temporal resolution and in native-like conditions is demonstrated, elucidating a process so far inaccessible to other structural or biophysical techniques. The present data set and our proposed mechanistic interpretation of the conformational dynamics of Mg 2+ -dependent CorA gating sets the stage for an unprecedented understanding of CorA as a 'reverse' polarity ligand gated ion channel with an unique gating mechanism.

Protein purification
CorA from T. maritima was expressed and purified as previously described (Dalmas et al., 2010). Briefly, the CorA-Pet15b vector was used to transform, then express CorA in E. coli BL21 DE3. After cell harvesting and disruption, membranes were collected by ultracentrifugation and gently solubilized. The sample was then cleared by ultracentrifugation and purified using cobalt high affinity chromatography column (Clontech Laboratories). The concentrated protein sample (AMICON 100 kDa cutoff membrane filters, EMD Millipore) was homogenized by gel filtration (Superdex 200 10/300 GL column, GE Healthcare Bio Sciences) and equilibrated in 50 mM HEPES, pH 7.3, 200 mM NaCl, 20 mM MgCl 2 , and 1 mM DDM).

Sample preparation for HS-AFM
A 1.5 mm diameter muscovite mica sheet was glued on a HS-AFM glass rod sample support and mounted on a HS-AFM scanner. Reconstituted CorA membranes were adsorbed on freshly cleaved mica for~5 min. Subsequently, the sample was rinsed with imaging buffer (50 mM MES, pH6.0, 200 mM NaCl) containing 10 mM Mg 2+ .

HS-AFM
All experiments were performed using HS-AFM (Ando et al., 2001) (SS-NEX, Research Institute of Biomolecule Metrology Co.) operated in amplitude modulation mode, using ultra-short cantilevers (8 mm) with a nominal spring constant of~0.15 N/m and a resonance frequency of~600 kHz in liquid (USC, NanoWorld). Videos of CorA membranes were recorded with imaging rates of~1-2 frames s À1 and at a resolution of 0.5 nm pixel À1 . The energy input by the AFM tip (estimated to~1.5 k BT, considering a 90% imaging amplitude of a 1 nm free amplitude)  was minimized by continuously adapting the drive and setpoint amplitude and optimizing the feedback parameters.

Structural titration experiments
Monitoring the transition of CorA from Mg 2+ -saturated to low/no Mg 2+ -conditions was achieved by depleting Mg 2+ by either adding EDTA into the measuring solution or alternatively by buffer solution exchange to Mg 2+ -free buffer using an integrated constant-pressure and constant-flow pump system . Experiments were performed with CorA membranes prepared from three different purifications and~10 different reconstitutions. In total, about 50 transition experiments on different membranes patches were performed on different days (over~30 days) using two different, but similar HS-AFM systems, and~10 USC cantilevers. All recordings showed a clear structural transition from stable 5-fold symmetrical proteins to very dynamic molecules with increased height. About 20% of the recorded CorA transition movies were considered for further analysis.

Data analysis
HS-AFM images were first-order flattened and contrast adjusted using laboratory-made routines in Igor Pro software (WaveMetrics). Videos were then drift corrected with respect to the membrane patch, or aligned on individual CorA molecules using an in-house developed analysis software plugin for ImageJ (Fechner et al., 2009;Husain et al., 2012). Dimensions of the CorA were calculated by height histogram analysis (n = 20,500 height values) and cross section analysis (n = 25) for each condition. Estimation of dwell-times was based on CorA protrusion height: DHeight/time traces were generated by subtracting the minimum pixel value from the maximum pixel value in the 5 Â 5 nm center area of the molecule. For the molecular height transition detection the DHeight traces were analyzed by a Step Transition and State Identification (STaSI) method in a MatLAB (MathWorks) routine (Heath and Scheuring, 2018;Shuang et al., 2014). StaSI indicated a minimum description length (MDL) for fitting the data with two DHeight states (where the increased height states comprises all conformational states with elevated topography). For in depth analysis, 25 molecules of two individual experiments were tracked over 900 and 1600 frames at different Mg 2+ conditions, resulting in~30,500 analyzed frames. Molecules in different height states and the corresponding dwell-times were binned either over 20 events along the time axis. Note, the transition analysis only discerned between states where subunits fluctuate between different height levels and did not classify between sub-states that exposed equivalent height levels to the putative closed (symmetric) state or the putatively open, activated (asymmetric and elevated) state. Such, the height/time traces did also not discriminate which of the open states was assumed. Notably however, our experiments showed that structures exposing increased protrusion heights are likely representatives of Mg 2+depleted molecules and thus detection of elevated height is a valuable fingerprint for the state transition. For conformational transition analysis membrane patches of~20 CorA channels were imaged at Mg 2+ concentrations of 10 mM, 3 mM, 0 mM, and after re-addition of 25 mM. Extracted CorA molecules from high-resolution movie sections (1,000-2000 molecular representations extracted from about 20 molecules for each Mg 2+ condition) were manually assigned to different structural states of (C) symmetric flower-shaped, (O-1) asymmetric elevated, (O-2) dome-shaped, elevated and (O-+) all others. The significance of the occurrence of each conformation under the four tested Mg 2+ concentrations was tested using a two-tailed students test.