Figures and data
Structures of Cx43 and Cx26 predicted by AF3.
(A) α- and β-connexin sequence alignment with the known CO2-sensitive connexins at the top, Cx26, Cx30 and Cx32, followed by the α-connexin Cx43. The carbamylation motif is highlighted by the green box. (B) Cx26 and Cx43 AlphaFold3 structural alignments. (C) Cx26 and Cx43 hemichannel structural prediction.
Cx43 hemichannels can be opened by an increase in PCO2.
The CO2 sensitivity of Cx43 was assayed using three different methods. (A) Dye-loading with CBF from 20 (left) to 70 mmHg PCO2 (right). The inset represents the 0 Ca2+ control, white scale bar is 20 μm. Cumulative probability graph represents all the data points measured with more than 200 cells per condition, box plot shows the medians from each independent transfection. (B) Whole cell patch recordings performed on untransfected parental HeLa cells and HeLa cells expressing Cx43 at a holding potential of −50 mV with steps to −40 mV to measure whole cell conductance. The baseline was set to 35 mmHg PCO2 and three different concentrations were used (20, 55, 70 mmHg). The absolute conductance change evoked by a change in PCO2 was measured and plotted as a dose response curve (right side, median with interquartile range) with 20 mmHg being assigned to zero and the absolute value of all conductance changes plotted relative to this (see Methods for details). The points were fitted by the Hill equation H=6, EC50=43 mmHg (blue points, orange curve). The data for untransfected parental HeLa cells was plotted the same way (grey). (C) The genetically-encoded ATP sensor GRABATP was co-transfected alongside Cx43 into HeLa cells (n=22). Representative images at different PCO2 concentrations of the cells are shown, grey scale bar is 20 μm. Traces from the recording the images were selected from are shown; each line represent a different cell, the fluorescence was normalised by dividing all values by the baseline median pixel intensity before plotting. Box plot shows the µM ATP release as determined by normalisation to the 3 µM control solution applied. (D) Parental HeLa cells transfected only with GRABATP do not exhibit CO2 dependent ATP release (n=14), summary box plot shows the µM ATP release as determined by normalisation to 3 µM ATP calibration.
Cx43 hemichannels with a truncated C-terminus (Cx431-256) retain their sensitivity to CO2 and depolarisation.
(A) Changes in GRABATP fluorescence induced by changing PCO2 from 20 to 55 mmHg or KCl from 3 to 50 mM. (B) Summary graph showing the concentration of ATP release in response to 55 mmHg PCO2 and 50 mM KCl, n=41 cells from 3 independent transfections.
Effect of hypercapnic solutions on intracellular pH (pHi) of parental HeLa cells measured by BCECF fluorescence.
(A) Images showing BCECF fluorescence at 35, 55 and 70 mmHg, and then calibration to pH 7.0, 7.2 and 7.4 of the same cells after treatment with nigericin. (B) Quantification of the changes in fluorescence. C) Summary plot of the change in pHi induced by 55 and 70 mmHg PCO2. 3 independent seeds of cells, n=30.
Cx43 gap junction channels are insensitive to changes in PCO2.
(A) DIC image with mCherry fluorescence superimposed. Gap junctions are evident as red stripes between cells (yellow arrows). The fluorescence images show the patch pipette filled with NBDG its loading into the recorded cell and the neighbouring cells coupled via gap junctions. The numbers in each bottom left corner are the time of the recording from whole cell breakthrough in minutes. Scale bar 20 µm. (B) Summary graphs showing the time for the acceptor cell to reach 10% of the fluorescence of the donor cell at three different levels of PCO2 (n=6 for each condition).
The effect of single Lys mutations on the CO2 sensitivity of Cx43.
HeLa cells transfected with the mutant variant of Cx43 were subjected to the dye-loading protocol. Images of dye loading in response to CO2 and the 0 Ca2+ positive control (insets), together with the cumulative probability plots are shown for K105Q (A), K109Q (B), K144Q (C), K234Q (D). (E) AlphaFold3 prediction of Cx43 with hypothesised carbamylation bridge residues highlighted – K234 in orange and K109 in red. (F) Summary box plots for dye-loading showing the change in median pixel intensity from 20 mmHg PCO2 for each condition (70 mmHg PCO2 – green and 0 Ca2+ blue, n=5) for WT (recalculated data from Fig 2A for comparison) and each mutation. Scale bars are 20 µm.
Cx43K144R hemichannels are insensitive to changes in PCO2.
(A) Cumulative probability plot from 5 independent transfections showing dye loading in response to 20 nd 70 mmHg and the zero Ca2+ stimulus. (B) Summary graph showing the change in median fluorescence intensity caused by 70 mmHg and zero Ca2+ in Cx43WT (data recalculated from Fig 2) and Cx43K144R.
The effect of single Lys mutations on CO2-dependent ATP release from Cx43.
The genetically-encoded ATP sensor GRABATP was co-transfected alongside Cx43 mutant variations, K105Q (A) n=23, K109Q (B) n=20, K144Q (C) n=12, K234Q (D) n=15. Representative trace measurements were selected; each line represents a different cell; the fluorescence was normalised by dividing all values by the baseline median pixel intensity before plotting. Box plot on the right (E) and (F) represent the µM ATP released as determined through normalisation to the 3 µM ATP application.
The effect of single Lys mutations on the CO2 dose-response properties of Cx43.
Comparison of Cx43WT (data replotted from Fig 2C), Cx43K105Q, Cx43K109Q (data replotted from Fig 4E) and Cx43K144Q (data replotted from Fig 4F). Data is plotted as medians with lower and upper quartiles. For the WT, K105Q and K144Q the fitted curve is drawn according to a modified Hill equation: [ATP]=MaxATP.[(PCO2/K)H/(1+(PCO2/K)H)].[1−(PCO2/Ki)Hi/(1+(PCO2/Ki)Hi)] Where K and H are the affinity and Hill coefficient of the channel for opening by CO2, Ki and Hi are the affinity and Hill coefficient for inhibition of the channel by CO2, and MaxATP is the asymptotically maximum release of ATP to CO2. For K109Q which did not exhibit CO2-dependent inhibition at high levels of PCO2, the Hill equation was used. The parameters for the curves are:
Paired Lys mutations are required to abolish the CO2 sensitivity of Cx43.
Representative images for all combinations tried are shown – K109Q K144Q (A), K105Q K144Q (B), K105Q K109Q (C), K105Q K234Q (D), K144Q K234Q (E). Insets represent the 0 Ca2+ positive control. For each construct, the pixel intensity in expressing cells was measured from 5 individual transfections, with at least 40 cells per condition. Cumulative probability plots display all measured data points. (F) The box plots shows the change in median pixel intensity from 20 mmHg PCO2 for each transfection for 70 mmHg PCO2 (green) and 0 Ca2+ (blue).
Paired Lys mutations abolish CO2 dependent ATP release via Cx43.
HeLa cells co-expressing Cx43 and GRABATP were subjected to changing PCO2-levels and fluorescence was recorded. Representative traces (A-E) for each of the Cx43 mutations – K105Q K144Q (A) n=23, K109Q K144Q (B) n=19, K105Q K234Q (C) n=17, K109Q K234Q (D) n=16, K144Q K234Q (E) n=13. The traces indicate normalised fluorescence changes (ΔF/F0). 3 µM ATP was applied at the end of each experiment to confirm sensor functionality. Furthermore, as a positive control, 50 mM KCl was applied to depolarise the cells and confirm channel function. (F-G) Box plots summarising the total ATP release in µM for each double mutant in 55 mmHg PCO2 and 50 mM KCl. Data points represent individual measured cells.
Introduction of negative charge into the carbamylation motif via Lys to Glu mutations has mixed effects on CO2 sensitivity of Cx43 hemichannels.
Expressing cells pixel intensity for each construct was measured from 5 individual transfections, with at least 40 cells per condition. (A-C) Representative cell images for 20 (left), 70 (right) with insets displaying the 0 Ca2+ control for each mutant K105E (A), K144E (B), K234E (C). Scale bar represents 20 µm. Cumulative probability graphs of pixel intensities are shown on the right for each mutant with three conditions 20 mmHg PCO2 (blue line), 70 mmHg PCO2 (orange) and 0 Ca2+ (grey line). The box plot shows change in median pixel intensity from 20 mmHg PCO2 for each transfection for 70 mmHg PCO2 (green boxes) and 0 Ca2+ (blue boxes).
Lys to Glu mutations abolish CO2-dependent ATP Release.
GRABATP fluorescence traces of ATP release from cells expressing single mutant connexins: K105E (A) n=18, K144E (B) n=17, K234E (C) n=13, in response to 55 mmHg PCO2 (red bars), 50 mM KCl depolarisation control (orange bars) and a 3 µM ATP control application at the end of all experiments to confirm sensor functionality. Traces show changes in normalised fluorescence over time (ΔF/F), indicating ATP release. On the right, the box plots display the quantified ATP release in µM for each mutant under 55 mmHg PCO2 and 50 mM KCl. Each data point represents a measurement from an individual cell.
The Cx43 double mutant, K105E K109E, is constitutively open.
(A) Representative fluorescence images of cells expressing the Cx43 K105E K109E mutant under low (20 mmHg PCO2, left) and high CO2 (70 mmHg PCO2, right), inset shows the 0 Ca2+ control. Scale bar is 20 µm. Cumulative probability plot of pixel intensity for each condition is shown on the right, overall indicating higher baseline fluorescence and perpetually open hemichannels. (B) Dye-loading with 20 mmHg PCO2 (left) and 20mmHg PCO2 with 100 µm LaCl3. Cumulative probability plot for pixel intensity under these conditions is shown on the right. (C) Fluorescence traces of ATP release from cells co-expressing GRABATP and Cx43 K105E K109E (n=16) under 20 mmHg PCO2 and 20 mmHg PCO2 with 100 µM LaCl3 (black bar). Fluorescence is normalised to baseline. (D) Fluorescence traces of ATP release from cells co-expressing GRABATP and Cx43 WT under 20 mmHg PCO2 and 20 mmHg PCO2 with 100 µM LaCl3 (black bar). (E) Box plots summarizing median pixel intensity under the different conditions, showing a significant reduction in intensity in the presence of LaCl3 (p = 0.03). (F) Box plot shows normalised fluorescence changes values for the difference between – 20 mmHg PCO2 (representing baseline normalisation) and the application of LaCl3 for both the K105E K109E and WT Cx43 constructs.
PCO2-dependent modulation of amplitude of fEPSPs in hippocampus is mediated via Cx43.
(A-C) Time-course plots showing the average amplitude of the normalised fEPSP (± SEM) amplitude in response to different conditions. Inserts display representative fEPSPs. (A) Control condition showing an increase in fEPSP amplitude in response to a modest change to PCO2 (20 to 35 mmHg), red bar represent the application of 35 mmHg PCO2, baseline conditions – 20 mmHg PCO2. (B) EPSP responses in the presence of Gap26 (black bar) and subsequent wash. (C) EPSP responses with scrambled Gap26 peptide applied (black bar). (D) Box plots representing the normalised fEPSP size difference (baseline was subtracted) with colour coded conditions: orange – control 35 mmHg PCO2, green – Gap26 with pre (35G), after (35S Wash).
Pathological mutations of Cx43 remove its sensitivity to CO2.
A,B) L90V prevents CO2-dependent dye loading. The dye loading assay shows no change in fluorescence between 20 and 70 mmHg, yet functional channels are expressed as shown by the zero Ca2+ positive control (inset). Box plot shows the change in median pixel intensity from 20 mmHg PCO2 to 70 mmHg PCO2 (green boxes) and 0 Ca2+ (blue boxes) for each transfection. C-E) GRABATP recordings shown that L90V and A44V also abolish CO2-dependent ATP release (baseline 20 mmHg PCO2, red bar 70 mmHg PCO2).
Occurrence of the carbamylation motif in the alpha connexin clade.
Names in red indicate the presence of the motif. The sequences were aligned in TCoffee to check for the presence of the motif. The molecular phylogenetic tree was constructed from the aligned traces by the average distance using Blosum62 in JalView (Waterhouse et al., 2009). Accession numbers for the protein sequences in the tree are given in the accompanying table supplement.
