Distinct classes from classification of Cx26 solubilised in LMNG.

a) Overall density associated with LMNG-NConst viewed from the cytoplasmic face. b) Density associated with LMNG-NFlex. c) As (a) focussed on the KVRIEG motif and the link between the N-terminus and TM1. Left: the cartoon has been coloured through the colours of the rainbow with blue at the N-terminus to red at the C-terminus, except for the KVRIEG motif, which is shown in magenta. Right: stick representation with the same colouring showing the interactions between the residues on the link between the N-terminus and TM1 (blue), residues on TM2 (green) and the KVRIEG motif (magenta). d) Cartoon representation of the cytoplasmic region of the LMNG-NConst structure. The two neighbouring subunits to the central subunit in the figure have been made semi-transparent. The dotted lines show the proximity of K125 of one subunit to R104 of the neighbouring subunit. Trp24 on TM1 is in the region of TM1 that adopts an altered conformation with respect to the previously solved structures of Cx26.

Cryo-EM data collection and processing statistics

Cryo-EM refinement and validation statistics

Comparison of LMNG-NConst and LMNG-NFlex structures

a) Overall superposition showing the movement of TM2 and the link between the N-terminus and TM1. LMNG-NConst in cyan and LMNG-NFlex in yellow (alternate subunits have been coloured in lighter shades). The KVRIEG motif has been coloured magenta with a sphere indicating the position of K125. The residues between the N-terminus and TM1 for the LMNG-NConst structure have been coloured blue. b) As (a) but focussed on TM1. The conformation of TM1 differs between the two structures as seen by the change in position of Trp24. TM2’ is from the neighbouring subunit. HC denotes the hydrocarbon chain from a lipid. The positions of T86’ and L89’ of TM2 in the NFlex conformation are not compatible with F31 and I30 TM1 in the NConst conformation.

Focussed classification of a single subunit results in density for the cytoplasmic loop consistent with models from AlphaFold.

a) Models generated by AlphaFold for a single subunit (left; coloured according to confidence level) and for the hexamer (right; in wheat with the position of K125 depicted by red sticks and the position of R104 in blue). b) Focussed classification of a single subunit (highlighted by an oval and coloured as in Figure 1d with the cytoplasmic loop in magenta) resulted in clear density for part of the cytoplasmic loop in a conformation consistent with the models from AlphaFold. This does not extend to the neighbouring subunits (coloured grey). c) Superposition of the single subunit built into the density (cyan) on the AlphaFold model (wheat). Showing the change in position of the helix in the cytoplasmic loop (highlighted by an arrow in the relevant colour). d) Reconstituting a hexamer by replicating the conformation of the subunit seen in (b) to all 6 subunits of the hexamer results in an apparently more closed conformation of the hemichannel, though there are also residue clashes, especially at the N-terminus. Lys 125 and Arg 104 are depicted with red and blue sticks respectively.

Cx26K125E gap junctions are constitutively closed at a PCO2 of 35 mmHg.

a-c) Montages each showing bright field DIC image of HeLa cells with mCherry fluorescence corresponding to the Cx26K125E-mCherry fusion superimposed (leftmost image) and the permeation of NBDG from the recorded cell to coupled cells. Yellow arrow indicates the presence of a gap junction between the cells; scale bars, 20 µm. The numbers are the time in minutes following the establishment of the whole cell recording. In Cx26WT expressing cells (a), dye rapidly permeates into the coupled cell. For Cx26K125E expressing cells, no dye permeates into the neighbouring cell even after 10 minutes of recording at either 35 mmHg (b) or 55 mmHg (c) PCO2 despite the presence of morphological gap junctions. d) Quantification of fluorescence intensity in the recorded cell (donor) and the potentially coupled cell (recipient) for both Cx26WT and Cx26K125E (7 pairs of cells recorded for WT and K125E at 35 mmHg and 6 pairs of cells for K125E at 55 mmHg, data presented as mean ± SEM). While dye permeation to the recipient cell follows the entry of dye into the donor for Cx26WT, no dye permeates to the acceptor cell for Cx26K125E. Note that the fluorescence intensity in the donor cell for Cx26K125E at both levels of PCO2 is higher than for Cx26WT at 35 mmHg, presumably because the dye remains trapped in the donor cell rather than diffusing to the recipient cell.

Density associated with the K125E mutant.

a) Superposition of density for K125E90 D6 averaged map (blue) on the density for the K125R90 D6 averaged maps (orange). The ovals show the position of TM2 from each subunit and the arrows show the direction of the difference between TM2 in the two structures. b) As (a) but focussed on TM2 in a view approximately perpendicular to the membrane. c) Superposition of density for WT90 Cx26 (PDB ID 7QEQ; pink) D6 averaged maps on the density for R125E90 (orange). d) Density associated with one subunit of the K125E90 structure (unsharpened map). The structure has been coloured as in Figure 1d. e) Superposition of K125E90 structure (light blue) on the structure of LMNG-NConst (cyan) showing the similarity between the two structures.

Comparison of the two structures derived from the LMNG classification with other structures of connexins.

a) Superposition of a single subunit from the LMNG-NConst (cyan) and LMNG-NFlex (yellow) structures on: Cx26 crystal structure (chartreuse, PDB ID 2ZW3); Cx32 (wheat, 7zxm) Cx50 (white, 7JJP); Cx43 in up (red, 7XQF), intermediate (chocolate, 7XQI) and down (salmon, 7F94) conformations; Cx36 in down (pink, 7XNH) and flexible (raspberry, 7XKT) conformations. The structures were superposed based on all chains of the hexamer. For clarity, only TM1, TM2 and the N-terminal helices are shown for each structure. b) As (a) for the beta connexins Cx26 and Cx32 (left), alpha connexins Cx43 and Cx50 (middle) and the gamma connexin, Cx36 (right) structures separately. Trp24 in each of the Cx26 and Cx43 structures has been depicted with a sphere representation. The isoleucine in the corresponding position is shown for Cx36. The sequence identities for common residues to Cx26 are 63% for Cx32, 49% for Cx50, 43% for Cx43 and 35% for Cx36.

Schematic representation of conformational changes

Schematic view of the cytoplasmic region of two opposing subunits within one hemichannel of the gap junction. The open structure on the left and the constricted structure on the right are in a dynamic equilibrium. The introduction of a negative charge on Lys125 of the KVRIEG motif (magenta) pushes the equilibrium to the right. In going from one conformation to the other: ① the cytoplasmic region of TM2 (green) flexes around Phe83 and the cytoplasmic loop adopts a more defined conformation; ② the cytoplasmic region of TM1 (pink) rotates, illustrated here by the movement of Trp24; ③ the N-terminal helix (blue), which will be affected by both ① and ②, adopts a position within the pore that constricts the entrance to the channel.

Workflow for initial processing of cryo-EM data for WT sample, solubilised in LMNG in CO2/HCO3- buffer.

The star denotes the classifications with the appearance of the NConst conformation that refine to a resolution greater than 4Å. The maps in the lower panel are coloured according to resolution as estimated in Relion 4.

Density for the transmembrane helices and the N-terminal helix associated with each the LMNG-NConst and LMNG-NFlex structures.

The residues with white carbon atoms are not included in the final structure.

Comparison of structures derived from LMNG solubilised protein with the structure derived from the DDM solubilised protein.

a) Density for lipid-like molecule. Left: Lipid-like density (red surface) in the pore of the protein seen in WT Cx26 solubilised in DDM (PDB 7QEQ, EMD-13937). Right: Maps associated with the LMNG solubilised protein superposed on that of the DDM solubilised protein (LMNG-NConst (cyan) and LMNG-NFlex (yellow), EMD-13937 (white) showing that the density remains irrespective of which of the two detergents is used to solubilise the protein. b) Superposition of the LMNG-NConst structure on the similar conformation of the protein derived from DDM-solubilised protein (7QEW). The conformation of TM1 differs between the two structures, as shown by the position of W24. We attribute the difference to incomplete particle separation of the DDM-derived protein during the classification procedure.

Workflow for single subunit classification for LMNG solubilised sample.

The maps in the lower panel are coloured according to resolution as estimated in Relion 4.

Density for the transmembrane helices and the N-terminal helix associated with each the LMNG-NConst-mon and LMNG-NFlex-mon structures.

The residues with white carbon atoms are not included in the final structure.

Workflow for processing of cryo-EM data for K125E sample in CO2/HCO3- buffer.

The star denotes the classifications with the appearance of the PCN conformation that refine to a resolution greater than 4Å. The maps in the lower panel are coloured according to resolution as estimated in Relion 4.

Density for the transmembrane and N-terminal helix associated with the K125E structure.

Workflow for processing of cryo-EM data for K125R sample in CO2/HCO3- buffer.

The star denotes the classifications with the appearance of the PCN conformation that refine to a resolution greater than 4Å.

Workflows for processing of cryo-EM data for samples in HEPES buffer.

Comparison of density maps from wild type and K125E Cx26 purified in HEPES buffer at pH 7.4:

a) WT Cx26 at 4.9Å resolution sharpened with a B-factor of –100. b) K125E sharpened with B-factor of –273 and low pass filtered to 5Å. c) Superposition of the two maps.