Connexin26 mediates CO2-dependent regulation of breathing via glial cells of the medulla oblongata

Abstract

Breathing is highly sensitive to the PCO₂ of arterial blood. Although CO₂ is detected via the proxy of pH, CO₂ acting directly via Cx26 may also contribute to the regulation of breathing. Here we exploit our knowledge of the structural motif of CO₂-binding to Cx26 to devise a dominant negative subunit (Cx26^DN) that removes the CO₂-sensitivity from endogenously expressed wild type Cx26. Expression of Cx26^DN in glial cells of a circumscribed region of the mouse medulla - the caudal parapyramidal area - reduced the adaptive change in tidal volume and minute ventilation by approximately 30% at 6% inspired CO₂. As central chemosensors mediate about 70% of the total response to hypercapnia, CO₂-sensing via Cx26 in the caudal parapyramidal area contributed about 45% of the centrally-mediated ventilatory response to CO₂. Our data unequivocally link the direct sensing of CO₂ to the chemosensory control of breathing and demonstrates that CO₂-binding to Cx26 is a key transduction step in this fundamental process.

Fig. 1. Rationale for the design of a dominant-negative Cx26 subunit (Cx26^DN) to inhibit CO₂-mediated hemichannel opening.

a Ribbon diagram, showing a single connexin26 subunit, highlighting the two amino acid residues vital for CO₂ sensing—R104 (white arrowhead) and K125 (red arrowhead, Cx26^WT, left). In the Cx26^DN mutant, these two residues are mutated—R104A and K125R (Cx26^DN, right). This will prevent carbamylation of the subunit and formation of a salt “carbamate” bridge with adjacent connexin subunit in the hexamer. b CO₂-dependent whole-cell clamp conductance changes from HeLa cells stably expressing Cx26^WT (data from ref. ). The grey line is drawn with a Hill coefficient of 6, suggesting that hemichannel opening to CO₂ is highly cooperative. c Hypothesised coassembly of Cx26^WT and Cx26^DN into heteromeric hemichannels that will be insensitive to CO₂ because insufficient carbamate bridge formation will occur to induced channel opening.

Fig. 2. FRET signal between connexin variants.

Example images of acceptor depletion Förster resonance energy transfer (ad-FRET) experiments. HeLa cells were co-transfected with equal amounts of DNA transcripts (to express connexin–fluorophore constructs) and PFA fixed after 48 h. Two channels were recorded: 495–545 nm (Clover, green) and 650–700 nm (mRuby2, magenta), and images were acquired sequentially with 458- and 561-nm argon lasers, respectively. Photobleaching was performed using the 561-nm laser for 80 frames at 100% power, targeting ROIs. Three combinations of connexin–Clover (donor, green) and connexin–mRuby2 (acceptor, red) are shown before and after bleaching. Colocalisation is shown in white in the overlay images, along with bleached areas (white ovals) and background references (yellow ovals). Following acceptor bleaching, ROIs in Cx26^WT-Clover + Cx26^WT-mRuby2 and Cx26^WT-Clover + Cx26^DN-mRuby2 samples show enhanced fluorescence intensity of donor (Clover, green) and reduced fluorescence intensity of acceptor (mRuby2, magenta). ROIs in Cx43^WT-Clover + Cx26^WT-mRuby2 samples show almost no change in fluorescence intensity of donor (Clover, green) and reduced fluorescence intensity of acceptor (mRuby2, magenta). Colocalisation measurements decrease mainly due to photobleaching of mRuby2 and subsequent elimination of its fluorescence. All co-transfection combinations showed statistically significant colocalisation at the highest resolution for the images. Colocalisation threshold calculations were carried out in ImageJ using the Costes method; 100 iterations, omitting zero–zero pixels in threshold calculation. Statistical significance was calculated for entire images and individual ROIs, with no difference between either calculation (p value = 1 for all images, with a significance threshold of p > 0.95). Scale bar, 10 μm.

Fig. 3. FRET efficiency of coexpressed connexin variants.

a Box and whisker plots showing the difference in FRET efficiency (%E) across different connexin co-expression connexin samples. FRET efficiency was calculated from background-adjusted ROIs as: %E = 100 × (clover_post − clover_pre)/clover_post. One-way ANOVA was carried out in SPSS. Post hoc testing revealed all individual comparisons to be significant at ***p < 0.001, with the exception of the comparison between Cx26^WT + Cx26^WT and Cx26^WT + Cx26^DN data sets, which were not significantly different. Each dot represents a different ROI. b Box and whisker plot showing mRuby2 bleaching efficiency during the acceptor depletion step. While a small amount of bleaching occurred in untargeted regions (presumably due to light scattering and/or reflection), targeted ROIs received drastically greater bleaching. Importantly, all targeted regions showed highly similar bleaching efficiencies. Bleaching efficiency was calculated from background-adjusted ROIs as: Bleaching = (1 − (mRuby2_post/mRuby2_pre)). Boxes show the interquartile range, the median is indicated by the horizontal line within the box, and the mean is indicated by the cross within the box. Range bars show minimum and maximum values.

Fig. 4. Cx26^DN removes CO₂ sensitivity in HeLa cells stably expressing Cx26^WT.

Dye loading under 35 mmHg PCO₂ (control) or 55 mmHg PCO₂ (hypercapnic) conditions revealed how the CO₂ sensitivity of HeLa cells stably expressing Cx26 (Cx26-HeLa cells) changes over time after transfection with Cx26^DN. a, b Representative dye-loading images of Cx26-HeLa cells at day 4 and day 6 days after cells were either untreated a or transfected with Cx26^DN b. In b, the inset represents a Zero Ca²⁺ control to demonstrate the presence of functional hemichannels even when the HeLa cells showed no CO₂-dependent dye loading. c, d Cumulative probability distributions comparing mean pixel intensity for each condition at day 4 and day 6 (untreated Cx26-HeLa cells, c; Cx26-HeLa cells transfected with Cx26^DN, d; n > 40 cells per treatment repeat, with at least 5 independent repeats for each treatment). The cumulative distributions show every data point (cell fluorescence intensity measurement). e Median change in pixel intensity caused by 55 mmHg PCO₂ and Zero Ca²⁺ from baseline (35 mmHg) over days 4, 5, and 6 post-transfection. At day 6, median pixel intensities (from 7 independent repeats) were compared using the Kruskal–Wallis ANOVA (χ² = 9.85, df = 2, p = 0.007**) and post hoc with Mann–Whitney U test (Cx26^WT vs Cx26^WT+Cx26^DN, W = 51, p = 0.003***; Cx26^WT+Cx26^DN Zero Ca²⁺ vs Cx26^WT+Cx26^DN, W = 46, p = 0.002***). Each circle represents one independent replication (independent transfections and cell cultures). Boxes show the interquartile range, the median is indicated by the horizontal line within the box, and the whisker is 1.5 times the interquartile range. Scale bars, 40 μm.

Fig. 5. Connexin26-mediated hypercapnic breathing response in conscious mice.

a Mice aged 11–14 weeks were bilaterally injected with lentivirus (LV) at the ventral medullary surface (VMS) to introduce either the Cx26^DN or Cx26^WT gene under the control of a GFAP promoter into genomic DNA. Whole-body plethysmographic measurements of frequency (f_R), tidal volume (V_T), and minute ventilation (V_E) were recorded for each mouse at 0, 3, 6, and 9% CO₂, before (week 0) and after (weeks 2 and 3) LV transduction. Two weeks after transduction, there was a difference (two-way mixed-effects ANOVA followed by post hoc t test: V_T F = 4.245, df = 3, p = 0.008; V_E F = 3.738, df = 3, p = 0.015; p values on figure given for post hoc comparisons) in adaptive changes in tidal volume to 6% CO₂ when comparing mice expressing Cx26^WT (empty circles, n = 12 mice) and mice expressing Cx26^DN (black circles, n = 12 mice). The median is indicated as a horizontal line within the box, and the mean is represented by a cross within the box. Range bars show minimum and maximum values. b Location of GFAP:Cx26 LV construct expression (green) in the sagittal plane—scale bars, 1 mm (left), 200 µm (right). R rostral chemosensitive site, C caudal chemosensitive site, NA nucleus ambiguous, VII facial nucleus, preBot preBötzinger complex, Bot Bötzinger complex, rVRG rostral ventral respiratory group, RTN retrotrapezoid nucleus, cVRG caudal ventral respiratory group, LRt lateral reticular nucleus, D dorsal, V ventral, R rostral, C caudal.

Fig. 6. The putative chemosensory cells extend processes dorsal, medial, and rostral.

Top three images: Coronal sections. Bottom three images: parasagittal sections showing the location of the transduced cells (left) and examples of the cells at higher magnification (middle and right). Cells expressing GFAP:Cx26:Clover (green) at the ventral medullary surface have a morphology unlike that of astrocytes, with a cell body at the very margin of the ventral surface and long processes that extend deep into the brain in the direction of respiratory nuclei. The ventrolateral respiratory column lies caudal to the VII nucleus and ventral to the nucleus ambiguous (dashed ovals). Choline acetyltransferase staining (magenta). py pyramids. Scale bars: bottom-left, 1 mm; top-left and top-middle, 200 µm; top-right and bottom-middle and bottom-right, 50 µm. M medial, L lateral, D dorsal, V ventral, R rostral, C caudal.

Fig. 7. Schematic of the location of glial cells which, when transduced with Cx26^DN at the medullary surface, reduced the chemosensitivity of breathing.

Location of cells is shown in the coronal (a) and parasagittal (b) planes. The cells are in an area ventral to the lateral reticular nucleus (LRt) and inferior olive (IO). The area reaches laterally to the same parasagittal plane as the nucleus ambiguus (NA) extends medially to the pyramids (py). NTS nucleus tractus solitarius, 12n hypoglossal nucleus, ROb raphé obscurus, RPa raphé pallidus, SP5I spinal trigeminal nucleus interpolar part, VII facial nucleus, C1 lateral paragigantocellular nucleus C1.