Calcium homeostasis modulators (CALHMs) are voltage-gated, Ca2+-inhibited nonselective ion channels that act as major ATP release channels, and have important roles in gustatory signalling and neuronal toxicity1,2,3. Dysfunction of CALHMs has previously been linked to neurological disorders1. Here we present cryo-electron microscopy structures of the human CALHM2 channel in the Ca2+-free active or open state and in the ruthenium red (RUR)-bound inhibited state, at resolutions up to 2.7 Å. Our work shows that purified CALHM2 channels form both gap junctions and undecameric hemichannels. The protomer shows a mirrored arrangement of the transmembrane domains (helices S1–S4) relative to other channels with a similar topology, such as connexins, innexins and volume-regulated anion channels4,5,6,7,8. Upon binding to RUR, we observed a contracted pore with notable conformational changes of the pore-lining helix S1, which swings nearly 60° towards the pore axis from a vertical to a lifted position. We propose a two-section gating mechanism in which the S1 helix coarsely adjusts, and the N-terminal helix fine-tunes, the pore size. We identified a RUR-binding site near helix S1 that may stabilize this helix in the lifted conformation, giving rise to channel inhibition. Our work elaborates on the principles of CALHM2 channel architecture and symmetry, and the mechanism that underlies channel inhibition.
Subscribe to Journal
Get full journal access for 1 year
only $3.90 per issue
All prices are NET prices.
VAT will be added later in the checkout.
Rent or Buy article
Get time limited or full article access on ReadCube.
All prices are NET prices.
The cryo-EM density map and coordinates of EDTA–CALHM2hemi, EDTA–CALHM2gap, and RUR–CALHM2 have been deposited in the Electron Microscopy Data Bank (EMDB) under accession numbers EMDB-20788, EMDB-20790 and EMDB-20789, respectively, and in the Research Collaboratory for Structural Bioinformatics Protein Data Bank under accession codes 6UIV, 6UIX and 6UIW, respectively. The single subunit map(s) obtained from signal subtraction and associated mask have been deposited under the corresponding EMDB accession number.
Dreses-Werringloer, U. et al. A polymorphism in CALHM1 influences Ca2+ homeostasis, Aβ levels, and Alzheimer’s disease risk. Cell 133, 1149–1161 (2008).
Ma, Z. et al. Calcium homeostasis modulator 1 (CALHM1) is the pore-forming subunit of an ion channel that mediates extracellular Ca2+ regulation of neuronal excitability. Proc. Natl Acad. Sci. USA 109, E1963–E1971 (2012).
Taruno, A. et al. CALHM1 ion channel mediates purinergic neurotransmission of sweet, bitter and umami tastes. Nature 495, 223–226 (2013).
Maeda, S. et al. Structure of the connexin 26 gap junction channel at 3.5 Å resolution. Nature 458, 597–602 (2009).
Deneka, D., Sawicka, M., Lam, A. K. M., Paulino, C. & Dutzler, R. Structure of a volume-regulated anion channel of the LRRC8 family. Nature 558, 254–259 (2018).
Kefauver, J. M. et al. Structure of the human volume regulated anion channel. eLife 7, e38461 (2018).
Myers, J. B. et al. Structure of native lens connexin 46/50 intercellular channels by cryo-EM. Nature 564, 372–377 (2018).
Oshima, A., Tani, K. & Fujiyoshi, Y. Atomic structure of the innexin-6 gap junction channel determined by cryo-EM. Nat. Commun. 7, 13681 (2016).
Abbracchio, M. P., Burnstock, G., Verkhratsky, A. & Zimmermann, H. Purinergic signalling in the nervous system: an overview. Trends Neurosci. 32, 19–29 (2009).
Burnstock, G. Historical review: ATP as a neurotransmitter. Trends Pharmacol. Sci. 27, 166–176 (2006).
Ma, Z. et al. CALHM3 is essential for rapid ion channel-mediated purinergic neurotransmission of GPCR-mediated tastes. Neuron 98, 547–561 (2018).
Ma, Z., Tanis, J. E., Taruno, A. & Foskett, J. K. Calcium homeostasis modulator (CALHM) ion channels. Pflugers Arch. 468, 395–403 (2016).
Ma, Z., Saung, W. T. & Foskett, J. K. Action potentials and ion conductances in wild-type and CALHM1-knockout type II taste cells. J. Neurophysiol. 117, 1865–1876 (2017).
Siebert, A. P. et al. Structural and functional similarities of calcium homeostasis modulator 1 (CALHM1) ion channel with connexins, pannexins, and innexins. J. Biol. Chem. 288, 6140–6153 (2013).
Dreses-Werringloer, U. et al. CALHM1 controls the Ca2+-dependent MEK, ERK, RSK and MSK signaling cascade in neurons. J. Cell Sci. 126, 1199–1206 (2013).
Vingtdeux, V. et al. CALHM1 ion channel elicits amyloid-β clearance by insulin-degrading enzyme in cell lines and in vivo in the mouse brain. J. Cell Sci. 128, 2330–2338 (2015).
Cisneros-Mejorado, A. et al. Blockade and knock-out of CALHM1 channels attenuate ischemic brain damage. J. Cereb. Blood Flow Metab. 38, 1060–1069 (2018).
Martinez-Palomo, A., Braislovsky, C. & Bernhard, W. Ultrastructural modifications of the cell surface and intercellular contacts of some transformed cell strains. Cancer Res. 29, 925–937 (1969).
Caterina, M. J., Rosen, T. A., Tominaga, M., Brake, A. J. & Julius, D. A capsaicin-receptor homologue with a high threshold for noxious heat. Nature 398, 436–441 (1999).
Kirichok, Y., Navarro, B. & Clapham, D. E. Whole-cell patch-clamp measurements of spermatozoa reveal an alkaline-activated Ca2+ channel. Nature 439, 737–740 (2006).
Peier, A. M. et al. A heat-sensitive TRP channel expressed in keratinocytes. Science 296, 2046–2049 (2002).
Bai, X. C., Rajendra, E., Yang, G., Shi, Y. & Scheres, S. H. Sampling the conformational space of the catalytic subunit of human γ-secretase. eLife 4, e11182 (2015).
Foote, C. I., Zhou, L., Zhu, X. & Nicholson, B. J. The pattern of disulfide linkages in the extracellular loop regions of connexin 32 suggests a model for the docking interface of gap junctions. J. Cell Biol. 140, 1187–1197 (1998).
Purnick, P. E., Oh, S., Abrams, C. K., Verselis, V. K. & Bargiello, T. A. Reversal of the gating polarity of gap junctions by negative charge substitutions in the N-terminus of connexin 32. Biophys. J. 79, 2403–2415 (2000).
Verselis, V. K., Ginter, C. S. & Bargiello, T. A. Opposite voltage gating polarities of two closely related connexins. Nature 368, 348–351 (1994).
Oh, S., Rivkin, S., Tang, Q., Verselis, V. K. & Bargiello, T. A. Determinants of gating polarity of a connexin 32 hemichannel. Biophys. J. 87, 912–928 (2004).
Oh, S., Abrams, C. K., Verselis, V. K. & Bargiello, T. A. Stoichiometry of transjunctional voltage-gating polarity reversal by a negative charge substitution in the amino terminus of a connexin32 chimera. J. Gen. Physiol. 116, 13–31 (2000).
Michalski, K., Henze, E., Nguyen, P., Lynch, P. & Kawate, T. The weak voltage dependence of pannexin 1 channels can be tuned by N-terminal modifications. J. Gen. Physiol. 150, 1758–1768 (2018).
Moreno-Ortega, A. J., Ruiz-Nuño, A., García, A. G. & Cano-Abad, M. F. Mitochondria sense with different kinetics the calcium entering into HeLa cells through calcium channels CALHM1 and mutated P86L-CALHM1. Biochem. Biophys. Res. Commun. 391, 722–726 (2010).
Goehring, A. et al. Screening and large-scale expression of membrane proteins in mammalian cells for structural studies. Nat. Protocols 9, 2574–2585 (2014
Winkler, P. A., Huang, Y., Sun, W., Du, J. & Lü, W. Electron cryo-microscopy structure of a human TRPM4 channel. Nature 552, 200–204 (2017).
Huang, Y., Roth, B., Lü, W. & Du, J. Ligand recognition and gating mechanism through three ligand-binding sites of human TRPM2 channel. eLife 8, e50175 (2019).
Fan, C., Choi, W., Sun, W., Du, J. & Lü, W. Structure of the human lipid-gated cation channel TRPC3. eLife 7, e36852 (2018).
Huang, Y., Winkler, P. A., Sun, W., Lü, W. & Du, J. Architecture of the TRPM2 channel and its activation mechanism by ADP-ribose and calcium. Nature 562, 145–149 (2018).
Haley, E. et al. Expression and purification of the human lipid-sensitive cation channel TRPC3 for structural determination by single-particle cryo-electron microscopy. J. Vis. Exp. 143, e58754 (2018).
Mastronarde, D. N. Automated electron microscope tomography using robust prediction of specimen movements. J. Struct. Biol. 152, 36–51 (2005).
Zheng, S. Q. et al. MotionCor2: anisotropic correction of beam-induced motion for improved cryo-electron microscopy. Nat. Methods 14, 331–332 (2017).
Zhang, K. Gctf: real-time CTF determination and correction. J. Struct. Biol. 193, 1–12 (2016).
Scheres, S. H. RELION: implementation of a Bayesian approach to cryo-EM structure determination. J. Struct. Biol. 180, 519–530 (2012).
Punjani, A., Rubinstein, J. L., Fleet, D. J. & Brubaker, M. A. cryoSPARC: algorithms for rapid unsupervised cryo-EM structure determination. Nat. Methods 14, 290–296 (2017).
Emsley, P., Lohkamp, B., Scott, W. G. & Cowtan, K. Features and development of Coot. Acta Crystallogr. D 66, 486–501 (2010).
Afonine, P. V. et al. Towards automated crystallographic structure refinement with phenix.refine. Acta Crystallogr. D 68, 352–367 (2012).
Chen, V. B. et al. MolProbity: all-atom structure validation for macromolecular crystallography. Acta Crystallogr. D 66, 12–21 (2010).
Delano, W. The PyMOL Molecular Graphics System. https://pymol.org/ (2002).
Pettersen, E. F. et al. UCSF Chimera—a visualization system for exploratory research and analysis. J. Comput. Chem. 25, 1605–1612 (2004).
Heymann, J. B. & Belnap, D. M. Bsoft: image processing and molecular modeling for electron microscopy. J. Struct. Biol. 157, 3–18 (2007).
Drozdetskiy, A., Cole, C., Procter, J. & Barton, G. J. JPred4: a protein secondary structure prediction server. Nucleic Acids Res. 43, W389–W394 (2015).
We thank G. Zhao and X. Meng for the support with data collection at the David Van Andel Advanced Cryo-Electron Microscopy Suite; the HPC team of VARI for computational support; D. Nadziejka for technical editing and E. Haley for proofreading. J.D. is supported by a McKnight Scholar Award, a Klingenstein-Simon Scholar Award and the National Institutes of Health (NIH) (grant R01NS111031).
The authors declare no competing interests.
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Peer review information Nature thanks Kenton Swartz and the other, anonymous, reviewer(s) for their contribution to the peer review of this work.
Extended data figures and tables
a, b, Representative current traces recorded in whole-cell mode at −60 mV for cells expressing wild-type CALHM2, tsA control cells, tsA cells transfected with empty pEGC vector (a) or tsA cells expressing CALHM2(E37R) (b). Cells were switched from bath buffer that contained 5 mM Ca2+ to one that contained 0.5 mM EGTA (0 mM Ca2+) to induce the current. The current was inhibited using a buffer that contained 100 μM RUR and 0.5 mM EGTA. n = 13, 15, 10 or 6 biologically independent experiments were performed for wild-type CALHM2, CALHM2(E37R), tsA control or pEGC control, respectively. c, Quantification of current amplitude in 0.5 mM EGTA, 100 μM RUR and 0.5 mM EGTA, and 5 mM Ca2+ conditions for CALHM2-expressing and control cells from a, b. Two-tailed paired and unpaired t-tests were applied to calculate P values (using GraphPad Prism 7), from within and outside of each cell type, respectively. Each dot indicates the value of one single independent experiment. RUR inhibition was nearly abolished in the CALHM2(E37R) mutant. d, Representative current–voltage relationships obtained by applying 500-ms voltage pulses ranging from 140 mV to −140 mV from a holding potential of 0 mV (20-mV steps) to cells expressing wild-type CALHM2 (top), tsA control cells (middle) and tsA cells transfected with empty pEGC vector (bottom). Currents were recording in the presence of 5 mM Ca2+, 0.5 mM EGTA (0 mM Ca2+) or 100 μM RUR and 0.5 mM EGTA. n = 7, 6 or 4 biologically independent experiments were performed for wild-type CALHM2, tsA control or pEGC control, respectively. e, The data of wild-type CALHM2 in d were normalized to the amplitude of the current recorded in the presence of EGTA at 140 mV and calculated as mean ± s.e.m. from 7 cells. f, Averaged current amplitude of cells expressing wild-type CALHM2 (n = 7), cells transfected with empty pEGC vector (n = 4) and tsA control cells (n = 6) in the presence of 0.5 mM EGTA, plotted as mean ± s.e.m. g, FSEC profiles showing that human CALHM2 N-terminally tagged with GFP (CALHM2N) and human CALHM2 C-terminally tagged with GFP (CALMH2C) have the best biochemical properties among human CALHM1, CALHM2 and CALHM3 when solubilized in digitonin. This experiment was repeated three times yielding similar results. h, Size-exclusion chromatography profiles of CALHM2C in digitonin before (blue) and after (red) GFP cleavage. After cleavage, the main peak shifted towards a smaller molecular weight. Purification of CALHM2C was repeated multiple times (>10), yielding similar results in each case. i, SDS gel of purified CALHM2C (indicated by red arrowheads) before (left band) and after (right band) GFP cleavage. j, Stability test of purified CALHM2C in the presence of RUR at two concentrations using FSEC, showing that a high concentration of RUR decreases CALHM2 stability. This experiment was repeated five times, each yielding similar results.
a, A representative micrograph of purified CALHM2N by negative-stain electron microscopy; selected two-dimensional classes are shown below. Fuzzy areas (indicated by red arrowheads) are caused by flexible GFP tags; this is also implied in c and d. Only hemichannels, and not gap junctions, appear in the micrograph. b, A representative micrograph of purified CALHM2C (from which the GFP tag has been cleaved) by negative-stain electron microscopy. The fuzzy areas in a, c, d disappeared in the two-dimensional classes, which confirms that they are indeed due to the GFP tag. One of the two-dimensional class averages represents a gap junction (red circle). c, A representative micrograph of purified CALHM2N by cryo-EM in the presence of 1 mM EDTA, collected using the Titan Krios; selected two-dimensional classes are shown below. Only hemichannels, and not gap junctions, are observed. d, A representative micrograph of purified CALHM2C by cryo-EM using the Talos Arctica; selected two-dimensional classes are shown below. Only hemichannels, and not gap junctions, are seen. e, FSEC profile of wild-type CALHM2 and the CALHM2(N168A) mutant, expressed in tsA 201 and N2a cells. This experiment was repeated three times, each yielding similar results. f, The wild-type CALHM2 ran at the same height in SDS gel before and after treatment using 250 U of Endo H and 250 U PNGase at optimum pH. Moreover, the CALHM2(N168A) mutant ran at the same height as the wild-type CALHM2. These data suggest CALHM2 is not glycosylated. This experiment was repeated multiple times (Endo H, n = 6; PNGase, n = 3), all yielding non-glycosylation phenotypes.
A total of 4,809 movies was collected using a Titan Krios equipped with K2. Particles were autopicked using Gautomatch, and visually examined in RELION to eradicate false-positive selections. After manual clean-up, particles were subjected to two rounds of two-dimensional classification in RELION. Ab initio reconstruction with two classes was performed in CryoSPARC to separate hemichannel and gap-junction particles, and to generate initial models of the hemichannel and gap junction. Hemichannel and gap-junction particles were further cleaned up using two-dimensional and three-dimensional classification in RELION. Particles from three-dimensional classes that showed high-resolution features and obvious C11 symmetry were combined and refined in RELION. To assess the structural heterogeneity of the helix S1, an approach that combined symmetry expansion and signal subtraction was carried out, in which all of the subunits were subtracted and classified without image alignment in RELION.
A total of 5,025 movies was collected using a Titan Krios equipped with K2. Particles were autopicked using Gautomatch, and visually examined in RELION to eradicate false-positive selections. After manual clean-up, particles were subjected to two rounds of two-dimensional classification in RELION. Three-dimensional classification using the map of EDTA–CALHM2hemi as an initial model yielded three classes with high-resolution features. Particles from the three classes were refined without applying symmetry, and particles from the class that showed obvious C11 symmetry were further refined using C11 symmetry. The two non-symmetric classes are highlighted by the red ellipses. The densities of the S1 helix and NTH that extend to the pore centre in one of the non-symmetric class are labelled. To assess the structural heterogeneity of helix S1, an approach that combined symmetry expansion and signal subtraction was carried out, in which all of the subunits were subtracted and classified without image alignment in RELION.
a, e, Representative electron micrograph of RUR–CALHM2 (a, out of 5,025 micrographs) and EDTA–CALHM2 (e, out of 4,809 micrographs). b, f, i, Selected two-dimensional class averages of the electron micrographs of RUR–CALHM2 (b), EDTA–CALHM2hemi (f) and EDTA–CALHM2gap (i). c, g, j, The gold-standard Fourier shell correlation (FSC) curves for the electron microscopy maps of RUR–CALHM2 (c), EDTA–CALHM2hemi (g) and EDTA–CALHM2gap (j) are shown in black, and the Fourier shell correlation curves between the atomic model and the final electron microscopy map are shown in red. d, h, k, The angular distribution of particles used for the refinement of RUR–CALHM2 (d), EDTA–CALHM2hemi (h) and EDTA–CALHM2gap (k).
Extended Data Fig. 6 Representative densities of the reconstructions of RUR–CALHM2, EDTA–CALHM2hemi and EDTA–CALHM2gap.
a, c, e, Local-resolution estimation of the structure of RUR–CALHM2 (a), EDTA–CALHM2hemi (c) and EDTA–CALHM2gap (e), calculated using Bsoft46. b, d, f, Representative densities of RUR–CALHM2 (b), EDTA–CALHM2hemi (d) and EDTA–CALHM2gap (f). The putative RUR-binding site density is shown in the panel on the right in b.
Extended Data Fig. 7 Comparison of EDTA–CALHM2hemi and EDTA–CALHM2gap with the connexin-46 gap junction, innexin-6 gap junction and a VRAC.
a, b, Overall structure comparison viewed parallel to the membrane (a, cartoon representation) and viewed from the intracellular side (b, surface representation), showing notable differences in symmetry, size and shape. The VRAC in b is viewed from the extracellular side. The size of the VRAC in b represents the largest diameter of the transmembrane domain. c, d, Single-subunit comparison in two different views viewed parallel to the membrane. The intracellular domains are highlighted using grey ellipses, and the grey rectangles represent cell membranes. The buried surface area in each pair of protomers in the CALHM2 gap junction is 378 Å2 (4,161 Å2 for an undecamer), which is substantially smaller than the equivalent buried surface area in connexin (94 Å2; or 5,654 Å2 for a hexamer) and innexin (1,550 Å2; or 12,397 Å2 for an octamer).
Extended Data Fig. 8 Secondary structure arrangement and domain organization of human CALHM2, and sequence alignment of the human CALHM family.
a, The secondary structure prediction of human CALHM2, and sequence alignment of CALHM family members CALHM1, CALHM2, CALHM3, CALHM4, CALHM5 and CALHM6. Secondary structure prediction was performed using the JPred online server47. Sequences were aligned using the Clustal Omega program and coloured using BLOSUM62 by conservation. Residues involved in RUR binding are marked with black filled circles. The extracellular helix (EH)0 in the S3–S4 linker is formed upon the docking of two hemichannels. P86 in human CALHM1 is boxed in red. b, Domain organization of the human CALHM2 protomer.
a, Representative current traces recorded in whole-cell mode at −60 mV in cells expressing CALHM2(R10A) or CALHM2(ΔN20). Human CALHM2 has three positively charged and two negatively charged residues in the NTH, resulting in one net positive charge. Out of these charged residues, R10 is the only residue that is conserved across CALHM1, CALHM2 and CALHM3. Cells were switched from bath buffer that contained 5 mM Ca2+ to one that contained 0.5 mM EGTA (0 mM Ca2+) to induce current. Current was inhibited using a buffer that contained 100 μM RUR and 0.5 mM EGTA. n = 14 or 10 biologically independent experiments were performed for CALHM2(R10A) or CALHM2(ΔN20), respectively. b, Quantification of current amplitude in 0.5 mM EGTA, 100 μM RUR and 0.5 mM EGTA, and 5 mM Ca2+ conditions for cells expressing CALHM2(R10A) or CALHM2(ΔN20), from a. Two-tailed paired t-tests were applied to calculate P values for comparisons using GraphPad Prism 7. Data are mean ± s.e.m. Each dot indicates the value of one single independent experiment. c, Current–voltage relationships were obtained by applying 500-ms voltage pulses that ranged from 140 to −140 mV from a holding potential of 0 mV (20-mV steps) to cells that express CALHM2(R10A) or CALHM2(ΔN20). Currents were recorded in the presence of 5 mM Ca2+, 0.5 mM EGTA and 100 μM RUR. Data were normalized to the amplitude of the current recorded in the presence of EGTA at 140 mV, and calculated as mean ± s.e.m. n = 7 biologically independent experiments were performed each for CALHM2(R10A) and CALHM2(ΔN20).
a, Surface representation of a gap junction viewed parallel to the membrane. Two paired subunits are highlighted. b, Surface representation of a hemichannel in the gap junction, viewed from extracellular side. c, Interface remodelling when docking two hemichannels (shown in grey; top) into a gap junction (shown in colour; bottom). d, Cartoon representation of the interface between two hemichannels. The two disulfide bonds are shown. The grey segment of S3–S4 linker represents deleted residues in a mutant (CALHM2(Δ143–146)). Only parts involved in the docking of hemichannels and disulfide bonds are highlighted in colour. e, Selected two-dimensional class averages of CALHM2(Δ143–146). This mutant yielded only hemichannels, and not gap junctions. f, Superimposition of single subunits of EDTA–CALHM2hemi (grey) and EDTA–CALHM2gap (pink) using the CTD. Only parts with conformational changes are highlighted in colour. To understand the conformational changes upon docking, we compared single subunits of a hemichannel and a gap junction. In the hemichannel, the loop connecting segment S3c and EH1 in the S3–S4 linker is flat and lacks extensive contact with the rest of the protein, giving rise to a flexible area that is probably required for the initiation of docking. Indeed, the S3–S4 linker is defined better in RUR–CALHM2 than in EDTA–CALHM2hemi, by forming interactions with the adjacent subunit. We suggest that the restricted S3–S4 linker in the RUR–CALHM2 hinders the docking of the hemichannels. g, Enlargement of the box in f, showing the remodelling of the S3–S4 linker from EDTA–CALHM2hemi (grey) to EDTA–CALHM2gap (pink). Upon docking, the S3c–EH1 loop remodels into two short loops and a short α-helix (EH0); the EH0–EH1 loop forms the primary interface and EH0 forms the minor interface in d. This motion accompanies an elevation of the S3–S4 linker and segment S3c that leads to an outward flexing of S3a, which breaks the loose interface between S3a and helix S1 in f. As a consequence, the S1 helix is detached from S3 and moves into a lifted conformation. The conformational changes of TMD upon docking in EDTA–CALHM2 are notably consistent with those induced by RUR. Moreover, the two docked hemichannels in the gap junction have similar conformations of their S1 helices.
About this article
Cite this article
Choi, W., Clemente, N., Sun, W. et al. The structures and gating mechanism of human calcium homeostasis modulator 2. Nature 576, 163–167 (2019) doi:10.1038/s41586-019-1781-3