Functional characterization and overall architecture of Mrs2

Reduced growth on nonfermentable carbon sources represents a classical phenotype of eukaryotic cells lacking Mrs2 (ref. 38). We reproduced this observation through experiments using an in vivo assay based on wild-type (WT) and MRS2-knockout Saccharomyces cerevisiae strains (Extended Data Fig. 1a). To further dissect the function of the Mrs2 proteins, we selected a member from Chaetomium thermophilum, CtMrs2 (UniProt G0S186), which shares 30% and 43% sequence identity to hMrs2 and ScMrs2, respectively (Extended Data Fig. 2). First, we assessed the cellular localization in S. cerevisiae cells of a conservatively truncated CtMrs2 form fused to green fluorescent protein (GFP) to the N terminus. As observed from the bioimaging of live cells, the protein is targeted to internal compartments (Extended Data Fig. 1b), consistent with localization to mitochondria. Furthermore, the purified GFP-fused CtMrs2 protein sample displayed the expected homopentameric state following size-exclusion chromatography (SEC) and on blue native page (Fig. 1b). Thus, our data suggest that CtMrs2 serves as a Mg2+-conducting homopentameric channel resident in mitochondria like other Mrs2 members.

To further illuminate the structure and function of Mrs2 channels, the purified CtMrs2 sample was reconstituted into nanodiscs to facilitate single-particle cryo-electron microscopy (cryo-EM) (Methods). We first maintained Mg2+ in all procedures during sample preparation. Cryo-EM micrographs of the sample showed well-distributed particles, with the corresponding two-dimensional (2D) classifications demonstrating features of a soluble and a TM domain, respectively (Extended Data Fig. 3). The final reconstructed map reached an average resolution of 3.1 Å in the absence of imposed symmetry (Table 1). Upon map inspection, we also obtained a complementary map determined at 2.7-Å overall resolution when C5 symmetry was applied, in accordance with the established five-fold assembly of CorA/Mrs2 proteins31. The TM and soluble domains are overall well resolved in the C5 map, both consistent with five intermixed monomers, enabling de novo building of the polypeptide chains (Fig. 1c,d, Extended Data Figs. 3 and 4 and Supplementary Fig. 1a). The CtMrs2 architecture exhibits a similar cone-shaped fold to the prokaryotic CorA proteins (root-mean-square deviation (r.m.s.d.) of 3.4 Å to Protein Data Bank (PDB) 3JCF; alignments were performed using secondary-structure matching in Coot throughout the manuscript)24,25,30,31,32,33. Moreover, the structure shows an even higher resemblance to the structure of hMrs2 (r.m.s.d. of 3.8 Å to PDB 8IP3) (Extended Data Fig. 5)39,40,41. Accordingly, the membrane-spanning domain comprises two helices from each monomer. TM1 (residues 416–441) helices establish an inner ring ~30 Å across the mitochondrial inner membrane, surrounding the pore that is vertical to the membrane. TM1 helices also form a funnel extending ~100 Å into the matrix (sometimes referred to as the stalk helix, residues 371–440), while TM2 (residues 451–483) helices form an outer ring, wrapping around the TM1 helices. Because the TM1 helices are twisted (approximately 20° compared to TM2 helices), the TM2 helices interact with two TM1 helices, one within the monomer and the other in the adjacent polypeptide. Conversely, the NTD harbors six antiparallel β-sheets (β1–β6) and eight helices (α1–α8) (Fig. 1c,d and Extended Data Fig. 1c,d). Helices α4 and α5 (also known as willow helices) are parallel to the stalk helix, forming an acidic ring adjacent to the membrane interface through a range of negatively charged residues (Fig. 1e and Extended Data Fig. 1e). Adjacent to this acidic region, a basic ring is formed by positively charged residues of TM1 (K422) and the C terminus. The structure also exhibits an electronegative entry mouth from the intermembrane space (Fig. 1e), which may have a role in attracting hydrated Mg2+ from the surrounding environment. This region is established by the loop in between TM1 and TM2, which is possibly partly integrated in the mitochondrial inner membrane (Fig. 1c,d). Similarly, the inside of the funnel in the matrix is also highly negatively charged (Fig. 1e). These surface charge features are also preserved overall in hMrs2 (refs. 39,40,41).

Table 1 Cryo-EM data collection, refinement and validation statisticsA closed configuration with multiple bound Mg2+ ions

While the ion permeation pore overlays with that of the CorA proteins, it is substantially longer in CtMrs2, stretching ~70 Å across the mitochondrial inner membrane and remaining narrow a further ~5 Å into the matrix, limited by the five TM1 or stalk helices throughout (Fig. 2a,b and Extended Data Fig. 6). The channel starts at the negatively charged entrance in the intermembrane space and ends where the funnel commences to widen, stretching from N443 to R406, as also visualized using the MOLE online server42 and HOLE software analysis43 (Fig. 2a,b). This pore harbors four strong nonproteinaceous features in the cryo-EM density assigned as Mg2+ ions, as supported by similar observations for certain of the sites in hMrs2, also in nonsymmetric C1 maps39,40 (Fig. 2c,d and Extended Data Fig. 7).

Fig. 2: The closed permeation pathway and Mg2+-binding sites.

One selected monomer is shown in orange throughout. a, MOLE software analysis of the conducting pore, shown as electrostatic surface, with residues lining the pathway shown as sticks. b, HOLE software calculation of pore diameter along the pore. c, Location of putative Mg2+ ions (shown in green) in the closed homopentamer. d, Close-up views with supporting cryo-EM density in the symmetry-applied C5 map of the putative Mg2+-binding sites close to or in the conducting pathway. Site U is positioned next to the loop in between TM1 and TM2 and site S is positioned next to the GMN motif of the selectivity filter. Sites P1 and P2 are located in the TM domain. e, Close-up views of sites M1–M4 in the NTD of CtMrs2. f, Close-up views of the corresponding Mg2+-binding sites (shown in e) in hMrs2 (PDB 8IP3), equivalent to sites M1 (top) and M4 (bottom). g, Growth phenotypes on YPD and YPG (the latter requiring mitochondrial respiration) media of WT Mrs2 from S. cerevisiae (ScMrs2), the equivalent MRS2-knockout strain (−MRS2), and cells based on −MRS2 with an empty vector or a vector containing different mutant forms of ScMrs2 (CtMrs2 numbering).

The first Mg2+ is bound by the loops connecting TM1 and TM2, interacting with the conserved N443 (of the GMN motif), E449 and perhaps E450. This is relevant as it has been shown that this region is important for the Mg2+ conductance of CorA and Mrs2 proteins5,44. Indeed, N443 and E449 are omnipresent within the CorA/Mrs2 family and E450 is also conserved among Mrs2 (Extended Data Figs. 2 and 5a). We propose that E449 and E450 are responsible for the initial uptake of fully hydrated Mg2+ (with a diameter of 9.5 Å)1, as permitted by the 10.6-Å and 17.8-Å diameters of the pores at these residues (pore calculations were computed using the software HOLE throughout the manuscript), respectively, a hypothesis also supported by previous efforts on hMrs2 (ref. 5). This arrangement may prime hydrated Mg2+ for the pore, thereby augmenting the local concentration before uptake, and contribute to establishing specificity. We coin this site U, as it likely is important for the uptake of partially dehydrated Mg2+, considering the 4.6-Å distance to N443. Directly associated, a second Mg2+ ion is coordinated to the main and side chains of GMN motif residues G441 and N443, respectively (Fig. 2c,d and Extended Data Fig. 7). The narrow width of the pore (~4.5-Å diameter at both G441 and N443) suggests that the ion is further dehydrated, as previously reported for this site of TmCorA31,32,33 and hMrs2 (ref. 39); thus, Mg2+ is partially water-stripped during the delivery from site U. Further supporting a critical role of the second site, it is generally found in CorA and Mrs2 structures, including hMrs2 (we name this site S for specificity, as discussed later)31,32,33,39.

Finally, two additional features of the cryo-EM density in the pore are assigned as Mg2+ ions, the first located at T434 (glycine in hMrs2) and the second positioned at the conserved T427 (Fig. 2c,d and Extended Data Fig. 7). We designate these pore sites P1 and P2; because of the narrow pore diameters of 5.3 and 7.4 Å, respectively, it is likely that the ions at these sites are only partially hydrated.

Three more constriction regions of the pore follow P2, marked by D420 (asparagine in hMrs2), the conserved M417–R413 pair that is positioned at the membrane interface to the matrix and the R406 (phenylalanine in hMrs2)–D410 (invariant) pair. Yet, we do not observe cryo-EM density for ions at these three constrictions. However, on the basis of the hMrs2 structures, the presence of Cl− and Mg2+ was suggested at the equivalent of R413 and D410, respectively39,40. Nonetheless, the combination of invariant positively charged residues at R413 and a hydrophobic or positively charged lock for cations at R406, along with the small diameters at these constrictions (4.0 and 2.0 Å, respectively, representing the narrowest parts of the permeation pathway), indicates that the CtMrs2 structure is closed (Fig. 2b).

Intermonomer Mg2+ binding to the soluble domain

The cryo-EM density reveals the presence of four more putative Mg2+-binding sites per monomer, present as a ‘glue’ in between separate polypeptides in the soluble fraction of CtMrs2. One pair of these sites (M1 and M2; M for the matrix of the mitochondria) are positioned at the tip of the protein (Fig. 2c,e and Extended Data Fig. 7). M1 is tightly coordinated by E374 and E378 of the stalk helix of one monomer and linked to D349 and D350 next to α5 of the adjacent monomer (Fig. 2c,e and Extended Data Fig. 7). Notably, this site is essentially conserved in hMrs2 as E293, E297 and E261 (Fig. 2f). Similarly, M2 interacts with H370 and the conserved E374 and loosely binds to the conserved D349 of the neighboring protein chain (Fig. 2c,e and Extended Data Fig. 7). M2 may also exist in hMrs2 as E262 and E293 or E261 and E263 but this is not supported by the hMrs2 structures39,40,41. We conclude that the approximate position of M1 likely is conserved, although the binding amino acids vary somewhat among Mrs2 members. Divergent to the Mg2+ sites at the funnel tip in CorA proteins, M1 and M2 are located inside the negatively charged funnel, which may imply a somewhat different role of these sites (or rather M1) in Mrs2 proteins than the corresponding in CorA proteins (Extended Data Fig. 6c).

The second pair of putative Mg2+-binding sites (here denoted M3 and M4) are located in between the stalk helix of one monomer and helix α5 of the adjacent monomer (Fig. 2c,e and Extended Data Fig. 7). M3 and M4 are positioned deeper into the funnel than M1 and M2, closer to the pore end. M3 is coordinated by S396 and S397 of the stalk helix and K325, S328 and T329 (through the side chains for all except the lysine) in the α5 helix of the neighboring monomer. Conversely, the directly associated M4 interacts with N393 and S396 of the stalk helix and S328, T329 and Q332 of helix α5 of the adjacent protein chain (Fig. 2c,e and Extended Data Fig. 7). Interestingly, T329 and N393 are conserved as glutamates in many Mrs2 members including hMrs2. Moreover, despite being positioned somewhat more peripherally, M4 is maintained in hMrs2, bound by the conserved E312 of the stalk helix and E138, E243 and D247 of helix α5 (Fig. 2f)39. Similarly, sites M3 and M4 are also conserved in certain but not all CorA members, such as MjCorA (PDB 4EV6)25 (Extended Data Fig. 5d). Thus, in addition to site M1, it is likely that the approximate location of site M4 is maintained across Mrs2 proteins (subtle differences observed even among hMrs2 structures, sometimes involving T246)40 and that M4 is also present in many CorA members. It is possible that the highly electronegative environment of the funnel serves to ensure that these cation sites are occupied when the Mg2+ levels in the matrix are elevated (Fig. 1e).

On the functional role of the Mg2+-binding residues

To further investigate Mg2+ binding, conductance and regulation, we exploited the previously mentioned S. cerevisiae-based assay, assessing the WT and mutant forms of ScMrs2 with maintained expression profiles (Extended Data Figs. 5a and 8 and Supplementary Fig. 1b). Supporting a role in ion uptake and permeation for site U, growth on a nonfermentable carbon source was impaired with the E449K;E450K double substitution (Fig. 2g). Similarly, the G441A;M442A;N443A mutant exhibited reduced cell proliferation, which is consistent with previous studies demonstrating that interruptions of the GMN motif impair the protein function23,44,45,46. Surprisingly, substitutions targeting sites P1 and P2 (T434A and T427A, respectively) left the cells essentially unaffected. Our interpretation of this observation is that certain parts of the ion conductance pathway are somewhat insensitive to amino acid changes, as also supported by the fact that T434 is replaced by a glycine in hMrs2 (Extended Data Figs. 2 and 5a). Conversely, the cell growth was abolished when selected, putative, gating residues of the pore were substituted (M417A, N412A;R413A and D410A), indicating a crucial role of these amino acids for protein function (Fig. 2g). However, substitution of two of the residues contributing to the M1 and M2 sites (E374A;E378A) left the cell proliferation unaffected. This can be interpreted as the latter form having little consequence on protein function, as discussed later.

The open state

While not evident from the previously available structures of CorA/Mrs2, Mrs2 must open for Mg2+ influx when the levels of the cation are low. Consequently, to further illuminate the Mg2+ conductance mechanism, we prepared a new CtMrs2 sample using a mild isolation strategy with the inclusion of EDTA from cell lysis and until final usage. Two cryo-EM maps, with and without applied C5 symmetry, were determined at an overall resolution of 3.2 and 3.5 Å, respectively (Extended Data Fig. 9 and Methods). However, the two maps superpose well, yielding a single structure (Fig. 3a, Extended Data Fig. 10 and Table 1).

Fig. 3: The open state of Mrs2.

The cryo-EM density panels refer to the C5 map throughout. a, Left, the 3.2-Å overall resolution cryo-EM density of the CtMrs2 homopentamer open state. Right, cartoon representation of the corresponding structure. The inset shows the feature assigned as cardiolipin, with the equivalent cryo-EM density in gray. b, Surface electrostatics of the open structure shown from the membrane plane, from the intermembrane space and from the matrix. c–f, Structural comparisons of the open and closed CtMrs2 structures along the ion conductance pore at E449-E450 (c; view from the intermembrane space), D420 (d; view from the matrix), M417 (e; view from the intermembrane space) and R413 and R406 (f; view from the intermembrane space). g,h, MOLE and HOLE software analyses of the pore of the open structure, shown as the electrostatic surface, with residues lining the pathway shown as sticks (g) and with the diameter of the pore along the pathway (h). i, Close-up views of the RDLR motif and the R314 wedge in the open structure. j, Putative Mg2+-binding sites at the GMN motif selectivity filter (S site) and at the T427 (P1) and T434 (P2) rings observed in the cryo-EM maps calculated with five-fold symmetry (left) and without symmetry (right). k, Mg2+-binding stoichiometry in various CtMrs2 forms under different conditions, as determined by ICP-MS (stoichiometries refer to Mg2+ per CtMrs2 pentamer). Data points represent the means of five independent measurements, each from one purified sample, and error bars indicate the s.d.

Source data

The EDTA-induced structure preserves the homopentamer form (Fig. 3a and Extended Data Fig. 10). Indeed, the TM domain including the pore is well maintained overall and the intermembrane space remains negatively charged, presumably facilitating Mg2+ uptake (Fig. 3b–d). However, as partly achieved by rotation of these side chains away from the center (discussed below), the constriction rings of the pore marked by D420, M417–R413 and R406 in the closed structure display a striking enlargement of about 1, 5 and 13 Å (Fig. 3d–f). As a consequence, site S at the GMN motif is the narrowest point of the pore (4.0–4.7 Å in diameter) in the open structure. Lastly, the funnel on the matrix side of the membrane is also widened. Thus, we conclude that this structure represents a conducting conformation, allowing the influx of partially hydrated Mg2+ to the matrix (Fig. 3c,g,h).

Few intermonomer contacts remain in the funnel of the open structure. However, while the conserved R314 has a peripheral orientation in the closed configuration, it serves as a wedge in between two adjacent stalk helices in the open conformation, through interaction with the invariant D410 and N414 of the neighboring monomer, thus directly influencing the matrix-facing end of the conducting pathway (Fig. 3a,i). Furthermore, toward the funnel tip, residues R200 and R203 of the conserved RDLR motif of the α1–α3 subdomain interact with E346, D348, E374 and E378 of the stalk helix of the adjacent monomer (side chain interactions except for E346), residues that assist in establishing the M1 and M2 sites in the closed structure (Fig. 3a,i). It is, thus, tempting to speculate that sites M1 and M4 (and sites M2 and M3 in CtMrs2) stabilize the closed state, while the RDLR site serves the same purpose in the open conformation. However, the RDLR motif is only present in the eukaryotic Mrs2 proteins, implying that this interaction is specific for the eukaryotic members.

Interestingly, while we expected a Mg2+-depleted structure, we observed cryo-EM density at the S, P1 and P2 sites, which we tentatively assigned as Mg2+ despite the EDTA treatment of the sample (the P2 site is not as distinct as the other sites in the map calculated without imposed symmetry) (Fig. 3j). To confirm whether Mg2+ was still present in the presence of chelator, we determined Mg2+ binding stoichiometries to CtMrs2 using inductively coupled plasma mass spectrometry (ICP-MS). The results show that Mg2+ was indeed present in the sample subjected to EDTA (Fig. 3k). Surprisingly, we detected a similar Mg2+-to-protein ratio for all samples (Mg2+-treated sample, EDTA-treated sample and two separate double mutants targeting sites M3 and M4, respectively), equivalent to ~2 Mg2+ ions per CtMrs2 pentamer as observed in the open structure (Fig. 3k). This indicates that CtMrs2 adopts an open conformation in the absence of Mg2+, as we had to remove Mg2+ in the last step of purification for the ICP-MS analyses. Considering that site S is formed by the hallmark GMN motif of the CorA/Mrs2 proteins, it is conceivable that its presence in the open configuration relates to an important role in establishing Mg2+ specificity and perhaps also in preventing the (back)flow of Mg2+, other ions or even water molecules from the matrix.

Conformational changes between the closed and open states

In comparison to the TM domain, dramatic differences occur in the soluble domains between the closed and open structures. Overall, as seen from the mitochondrial matrix side, the soluble domains rotate counterclockwise and widen from the closed to the open structure, with accentuating changes from the start (linked to the pore) to the end (at the tip) of the funnel (Fig. 4a,b).

Fig. 4: Conformational changes between the open and the closed states.

a,b, Alignments of the TM domains of the open and closed CtMrs2 structures shown from the membrane plane, from the intermembrane space and from the matrix (a) and of a single monomer (b). Residues lining the pore are shown as sticks in b. The intermembrane space loop and the TM domain (formed by TM1 and TM2) are highly similar between the two states, whereas the NTD is rotated, providing a wider funnel in the open structure. Nonetheless, the five-fold symmetry is maintained in both configurations. c,d, Conformational changes at the M1 and M2 (c) and M3 and M4 (d) Mg2+ ion-binding sites between the closed (left) and the open (right) states, with residues involved in ion binding shown as sticks.

Remarkably, among the most displaced structural features are the stalk and α5 helices of adjacent monomers, which are well separated in the open Mrs2 setting, thus eliminating the M1–M4 sites. Specifically, the residues forming the M1 and M2 sites are separated by more than 20 Å in the open structure (Fig. 4c). Similarly, the amino acids that bind to sites M3 and M4 are displaced by approximately 10 Å (Fig. 4d). The consequence is that the stalk helix is straightened and shifted outward in the open structure, thereby being mainly responsible for the widening of the D420, M417–R413 and R406 constrictions of the pore. Moreover, the stalk helix preceding the soluble domain (the α1–α3 subdomain) within the monomer is brought along from a funnel-lining to a more peripheral position. Notably, this greatly reduces contacts between monomers in the funnel in the open configuration, with the RDLR arginines essentially replacing the M1 site in the closed structure.

The M1–M4 Mg2+ site regions control the shape and conductance of Mrs2

Considering the dramatic rearrangements of the M1–M4 ion-binding regions between the two configurations, we aimed to further refine the roles of these sites. First, we used an Escherichia coli-based Ni2+ sensitivity assay, which has been exploited, for example, for studies of hMrs2 (ref. 40), assessing Mrs2-faciliated increased uptake of the metal. As expected, E. coli cells expressing CtMrs2 showed increased sensitivity toward Ni2+ compared to control cells and the toxicity could be prevented through the N443A substitution that interferes with the GMN selectivity filter (Fig. 5a). Conversely, disruption of the M1 and M2 (E374A;E378A) or the M3 and M4 (S328A;T329A and S396A;S397A) sites increases cell toxicity (the same mutants were applied in all below-mentioned assays). This suggests that such forms correlate with open CtMrs2. Instead, an intention to stabilize the M1 and M2 sites (E374R, mimicking bound Mg2+) resulted in decreased Ni2+ sensitivity, congruent with closed Mrs2. Interestingly, supplementation of Mg2+ partially rescued all protein forms, indicating that CtMrs2 has higher affinity for Mg2+ than for Ni2+ and further illustrating that the M1–M4 Mg2+-binding sites have an important mechanistic role (Fig. 5a,b).

Fig. 5: The roles of the M1–M4 Mg2+-binding sites.

a,b, Comparison of the growth of E. coli with different CtMrs2 forms at different ion concentrations. The empty vector represents a control without Mrs2. c, Overlay of representative Mg2+ currents recorded in oocytes expressing WT and mutants. The red bar indicates the application period of Mg2+-containing recording solution. The control denotes water-injected oocytes. d, Summary of the recorded currents with peak current amplitudes following Mg2+ perfusion (Ipeak; left) and spontaneous current decay during Mg2+ perfusion (ΔIamp (%); right). Data shown as the mean ± s.e.m. (n = 5–19). Statistical analysis was conducted using a one-way ANOVA followed by Dunnett’s multiple-comparisons test in comparison to WT. NS, nonsignificant (P > 0.05; 0.16 for S328A-T329A and 0.79 for S396A-S397A); **P = 0.0036 and ****P < 0.0001. Further data and details are provided in Supplementary Fig. 2 and the Methods. ND, not determined (because of no detectable currents). e, Limited proteolysis assay (performed twice independently) using purified CtMrs2 forms using two separate proteases. The bands above 30 kDa represent CtMrs2, while the bands between 25 and 30 kDa are GFP, as confirmed by mass spectrometry. f, Binding isotherms for Mg2+ binding to WT CtMrs2 (at two stock concentrations) and mutant forms showing the heat of injection as a function of the molar ratio between Mg2+ and CtMrs2 monomer. For all mutants, 50 mM Mg2+ was used. For the mutants, the presented values are the means of two independent ITC titrations. Error bars show one s.d. and are estimated from the baseline uncertainties provided by NITPIC as previously described54 and from the uncertainty between injection from the two datasets, where i refers to the injection number (1–19). The inset graph shows the raw thermogram, before integration by NITPIC, for the titration of 50 mM Mg2+ into WT. The complete ITC data are presented in Supplementary Fig. 4 (Methods).

Source data

To further dissect the conductance, we performed electrophysiological recordings of CtMrs2 and mutants, which were expressed in Xenopus oocytes and perfused extracellularly with Mg2+ solution, guided by previous studies41,46 (Fig. 5c,d and Supplementary Fig. 2). The results show that WT displayed clear inward currents that peaked within seconds of the supplementation of Mg2+ and decayed slowly, reaching ~40% of the peak current by the end of the 2-min period that Mg2+ was supplied. In contrast, no current was observed for the N443A form or in water-injected oocytes (control). Moreover, the inward currents were rapidly abolished upon return to a Mg2+-free extracellular solution and were significantly reduced in the presence of the established inhibitor cobalt hexammine, all supporting that the CtMrs2 protein mediates the Mg2+ influx. In agreement with the Ni2+ sensitivity assay, significantly larger inward currents were detected for the ion-binding interfering forms (M1–M2 and M3–M4), which were even more pronounced for the M3–M4 mutants, in accordance with these amino acid changes favoring the open state. In comparison, no current was detected for the E374R mutant, supporting that this form prefers the closed conformation. Conversely, the decay of the flux during the 2-min Mg2+ exposure was similar for WT and for the M3–M4 alanine substitutions but the conductance was better maintained for the M1–M2 alanine mutant.

Next, we set up a limited proteolysis assay using purified protein. All the purified CtMrs2 forms maintained the pentameric assembly as assessed by SEC and all samples were susceptible to proteases (trypsin or chymotrypsin) at low Mg2+ concentrations (Fig. 5e and Supplementary Fig. 3). However, at elevated Mg2+ levels, WT and the two M3–M4 site mutants preserved protease resistance (less so for the S328A;T329A form). Conversely, the two M1–M2 substitutions (E374A;E378A and E374R) displayed almost complete susceptibility to protease degradation in the presence of Mg2+ (some uncleaved CtMrs2 left with trypsin for E374R). This indicates that the open state is more vulnerable for cleavage and that the two M3–M4 mutant forms can shift from the open to the closed configuration when Mg2+ is supplemented, whereas the M1–M2 alanine substitution may not.

Lastly, we conducted isothermal titration calorimetry (ITC) measurements to detect Mg2+ binding to purified CtMrs2 (Fig. 5f and Supplementary Fig. 4). The binding isotherms were recorded by titration of Mg2+ into the protein, starting with conditions under which an open protein conformation is expected. The binding isotherms of WT were approximately biphasic, reflecting multiple bindin