Hypothesised Cx32 mediated CO2 signalling cascade in peripheral myelin.

Three fingers of a myelin paranode have been used for illustrative purposes. Restoration of transmembrane ionic gradients following action potential propagation via the actions of Na+/K+ ATPases incurs a metabolic cost and increases production of ATP and CO2. AQP1, permeable to CO2, provides a pathway for CO2 to leave the node and enter the paranode and bind to Cx32 on the intracellular loop. This triggers opening of Cx32 and release ATP. Carbonic anhydrase(CA) catalyzes the combination of CO2 and H2O and ultimately the production of HCO3- and H+ ions and ebectively competes with Cx32 for CO2.

AQP1 localizes to both the Schwann cell paranode and also the axonal node.

A, B) Representative confocal SIM images from a single optical plane showing the localisation of Caspr, Cx32 and AQP1 in isolated mouse sciatic nerve. Arrowheads indicate the node. Scale bars, 10 μm. C) Boxplots showing degree of colocalization between Cx32 and AQP1 at the node/paranode. Control measurements (to right of dashed line) used these same images with one channel flipped 90° and the same thresholds as when measuring colocalization. Kruskal Wallis ANOVA p < 0.0001.

CAII localizes to myelinating Schwann cells, in particular to the axonal node and the Schwann cell paranode.

(A and B) Representative confocal LSM images in single optical plane showing the localisation of CAII and Cx32 in isolated mouse sciatic nerve. Arrow heads indicate the node. Intense CAII staining, denoted by a white asterisk (*) is present in non-myelinated fibres. Scale bar applies to A and B: 10μm.

A membrane impermeable dye, FITC, loads into Schwann cell paranodes in a CO2 dependent manner through a hemichannel.

A) Representative images showing the FITC loading into mouse sciatic nerve bundles. Arrow heads indicate the node. Little FITC loading occurs in response to control (35mmHg, 10 mins) aCSF. FITC loading was greatly increased by 70 mmHg aCSF (10 mins) and application of 100 μM FCCP. FITC loading in 70 mmHg aCSF was blocked by carbenoxolone (CBX) but not the TRPA1 antagonist HC030031. B) Boxplot showing intensity of FITC fluorescence under the diberent conditions. Each point represents a separate ROI from 5 diberent nerves for each condition. Scale bar – 10μm.

Activity dependent loading of FITC into Schwann cell paranodes.

A) Representative images showing the FITC loading into mouse sciatic nerve bundles in response to diberent lengths of stimulation (30 Hz). Arrows indicate the paranode. B) An isolated mouse sciatic nerve fibre loaded with the membrane impermeable dye FITC (30 Hz, 5 min) and counterstained with Caspr, an axonal membrane protein which is expressed only in the paranodal region. White arrows indicate the paranode of interest. Note lack of loading into the axon. Scale bars – 15 μm.

Activity dependent loading of FITC is sensitive to manipulation of carbonic anhydrase activity.

A) Representative images showing the FITC loading into mouse sciatic nerve bundles in response to 1 min of electrical stimulation in the absence (left) or presence (right) of the carbonic anhydrase inhibitor, acetazolamide. Arrows indicate the position of paranodes. Brightfield inset shows the presence of the nerve fibre and the arrowhead in the fluorescence image indicates its position. B) Summary plot showing how the pixel intensity of Schwann cell paranodes, and therefore FITC loading, vary in response to stimulus duration and the presence of acetazolamide (each point mean ± SD). C) Representative images showing the FITC loading into mouse sciatic nerve bundles in response to five minutes of stimulation in the absence (left) or presence (right) of the carbonic anhydrase enhancer, L-Phenylalanine. Brightfield inset shows the presence of the nerve fibre and the arrowhead in the fluorescence image indicates its position. D) Boxplot showing the ebect L-Phenylalanine had on FITC loading into mouse Schwann cell paranodes. Scale bars – 15 µm. Each point represents a separate ROI from 5 diberent nerves. L-Phe vs control MW test: p<0.0001.

Activity dependent loading of the FITC is reduced by inhibition of the Krebs cycle.

A) Representative images showing the FITC loading into mouse sciatic nerve bundles in response to 5 minutes of 30 Hz stimulation in the absence (left) or presence (right) of the 50 µM H2O2 which blocks aconitase and the Krebs cycle. Brightfield inset shows the presence of the nerve fibre and the arrows in the fluorescence images indicates position of paranodes. Scale bar – 15 μm. B) Boxplot showing the ebect 50 µM H2O2 had on FITC loading into mouse Schwann cell paranodes. Each point represents a separate ROI from 5 diberent nerves. H2O2 vs control MW test: p<0.0001.

Activity dependent loading of the FITC is reduced by inhibition of AQP1.

A) Representative images showing the FITC loading into mouse sciatic nerve bundles in response to 5 minutes of 30 Hz stimulation in the absence (left) or presence (right) of the AQP1 blocker TC AQP1-1. Brightfield inset shows the presence of the nerve fibre and the arrows in the fluorescence images indicates position of paranodes. Scale bar – 15 μm. B) Boxplot showing the ebect TC AQP1-1 had on FITC loading into mouse Schwann cell paranodes. Each point represents a separate ROI from 5 diberent nerves. TC AQP1-1 vs control MW U test: p<0.0001.

Activity dependent dye loading into Schwann cells depends on CO2 binding to Cx32.

A) Images of single axons dissected from a sciatic nerve transduced with AAV-Mpz-Cx32DN-IRES-mCherry. Axons that do not express Cx32DN do not exhibit mCherry fluorescence and show robust dye loading to 15 Hz stimulation. By contrast axons that express mCherry are Cx32DN+ve and do not show dye loading during stimulation. Scale bar 15 µm. B) Summary graph showing the pixel intensity of axons that are Cx32DN+ve versus the control Cx32DN-ve. MW p<0.0001, each circle an individual paranode from n=5 nerves.

Simple model of CO2 signalling at the paranode reproduces experimentally observed patterns of activity dependent FITC loading.

A) Adaptation of the Matsuda et al model to incorporate CO2 production and metabolism via CAII. Action potentials provide the input to the model as activity variable X, which determines variables Y and Z, which determine ATP production and consumption. CO2 is proportional to ATP production, via the rate constant α. Code for the model is provided in the Matlab files CO2Cx32Model.m and CO2Cx32Plot.m. B) Incorporation of the modified Matsuda et al model into a single cell that possesses a K+ leak channel and Cx32 and represents the paranode, albeit including the nodal mitochondria. (C) Outputs from the model to show how the change in [CO2] and the consequent FITC loading evoked by two diberent durations of stimulation (1 and 5 min) varies with the Vmax of CAII. Reduction from the control value (15 mM/s) to 1 mM/s simulates the ebect of acetazolamide, whereas an increase of Vmax to 30 mM/s simulates application of L-Phe. D) A summary graph showing how FITC loading varies with stimulus duration with the 3 diberent values for the VMax of CAII.

CO2 dependent Ca2+ influxes into Schwann cells are hemichannel dependent.

A) Representative trace showing change in normalised Fluo4 fluorescence in response to 70mmHg aCSF (red bar), 35mmHg aCSF with the non-specific hemichannel blocker carbenoxolone (100 µM, orange bar) and 70 mmHg aCSF plus 100 µM carbenoxolone (blue bar).B) Representative images showing changes Fluo4 fluorescence in response to hypercapnic aCSF. The circles in the first panel show the measurement ROIs drawn around the paranodes. Scale bar = 10 µm. C) Boxplot showing the change in normalised fluorescence (ΔF/F0) in Fluo4 loaded Schwann cell paranodes evoked by 70 mmHg aCSF in the presence and absence of 100 µM carbenoxolone (CBX). Each datapoint consists of a paranode, with all the data collected from 4 sciatic nerves. Control vs CBX, MW test, p = 0.0087.

Activity dependent increase of intracellular Ca2+ in Schwann cell paranodes.

A) Superimposed brightfield (BF) and fluorescence image of GCaMP8 transduced nerve. To show expression at the paranode. B) Representative GCaMP8 traces showing change in normalised fluorescence in response to 15 Hz electrical stimulation (red bar) in the presence of acetazolamide (AZ, green bar) or TC-AQP1-1 (purple bar). C) The fluorescence images show before (Baseline), during stimulation of the nerve (15 Hz), during stimulation in presence of AZ (15 + Az) and stimulation in the presence of TC AQP1-1 (15 + TC AQP1-1). Scale bar = 5 µm; black dashed line indicates the shape of an individual fibre, white arrow indicates hotspot of GCaMP8 fluorescence at the paranode. D) boxplot showing the change in normalised GCaMP8. fluorescence (ΔF/F0) evoked Schwann cell paranodes in response to 15 Hz electrical stimulation in the control, with AZ and with TC AQP1-1. Kruskal Wallis ANOVA, p < 0.0001. Pairwise MW comparisons: control vs AZ, p = 0.0142; control vs TC AQP1-1, p<0.0001. Each datapoint consists of a paranode, with all the data collected from 4 sciaSc nerves.

CO2 dependent slowing of conduction velocity following high frequency stimulation.

The CAP was evoked at 1 Hz and then 10 mins of 30 Hz stimulation was given prior to remeasuring the CAP at 1 Hz stimulation frequency. Representative CAPs from mouse sciatic nerve prior to high frequency stimulation (Black trace), and after (Blue trace) for WT nerves: A) in the absence of any compound; B) with 100 µM acetazolamide; C) with 1 mM L-Phenylalanine;D) with 80 µM TC AQP1-1. E) Boxplot showing the change in latency (time to peak of CAP) before and after high frequency stimulation for Control, Acetazolamide (AZ), L-Phe and TC AQP1-1. Kruskal Wallis ANOVA p=0.0016. Pairwise MW tests: control vs AZ, p=0.0317; control vs L-Phe, p=0.0159; control vs TC AQP1-1, p=0.0079.

Markers for the node, paranode and juxtaparanode in the isolated mouse sciatic nerve.

A) Location of the axonal node (KCNQ2) and Schwann cell paranode (Caspr) within an isolated mouse sciatic nerve fibre. KCNA2 is expressed in the outer myelin layer. Arrowheads indicate the node. Scale bar = 15 µm. Confocal LSM image, single optical plane. B) Schematic indicating the same locations within an isolated nerve fibre.

Cx32 colocalises with mitochondria in the Schwann cell paranode, alongside SFXN1.

A, B) Representative SIM images in single optical plane showing the localisation of Cx32, CytC and SFXN1 in isolated mouse sciatic nerve. Arrowheads depict the node. Scale bars – 10μm. C) Boxplots showing degree of colocalization between Cx32 and CytC and SFXN1 and CytC. M1 is the proportion of Cx32 (or SFXN1) that colocalises with CytC and M2 is the reverse proportion of CytC that colocalises with Cx32 (or SFXN1). Control measurements (to right of dotted line for each pair) used these same images with one channel flipped 90° and the same thresholds as when measuring colocalization. Kruskal Wallis ANOVA: Cx32 vs CytC, p = 0.0001; SFXN1 vs CytC, p < 0.0001.

The localization of AQP1 in relation to mitochondria, CytC, in isolated mouse sciatic nerve.

Representative SIM images in single optical plane showing the localisation of AQP1 and CytC in isolated mouse sciatic nerve. Arrows indicate the node. Scale bar – 10μm.

Cx31.3 does not open in response to hypercapnia.

A) Representative images of GRABATP fluorescence, encoded by a 16 colour LUT in Cx31.3 transfected HeLa cells in response to control (35 mmHg), hypercapnic (70 mmHg) and depolarising (50 mM KCl) aCSF. Scale bar 20 μm. B) Top, traces showing the normalised GRABATP fluorescence for cells transfected only with GRABATP. Bottom, traces from the cells transfected with both Cx31.3 and GRABATP as shown in (A). C) Summary statistic box plots showing the [ATP] release, calculated as a ratio from the normalised fluorescence change evoked by a certain solution or stimuli and the fluorescence change evoked by 3 µM ATP. Cells expressing only GRABATP do not exhibit changes in fluorescence to 70 mmHg or 50 mM KCl aCSF. Each data point represents a cell, with all cells coming from at least three transfections.

Acetazolamide and L-Phe do not alter the CAP.

A) Current - voltage input-output curves for the CAP, comprising data from all nerves subjected to either 100 µM acetazolamide or 1 mM L-Phenylalanine. Curves were produced before application of the respective compound (green circles), at the end of the pre-incubation of the compound (red circles) and at the end of dye loading (black circles) to show that the compounds had no ebect on the amplitude of the CAP. N=5 nerves for each condition. Insets show the averaged CAP during incubation with the compound. B) Boxplots showing the half maximum CAP amplitude for each respective condition and compounds laid out in (A). Kruskal Wallis ANOVA: Acetazolamide, p = 0.4441; L-Phenylalanine, p = 0.9917.

50 µM H2O2 does not alter the CAP.

A) Current -voltage input-output curves for the CAP comprising data from all nerves subjected to 50 µM H2O2. Curves were produced before application of H2O2 (green circles), at the end of the pre-incubation (red circles) and at the end of dye loading (black circles) to show that H2O2 had no ebect on the amplitude of the CAP. N=5 nerves for each condition. Inset shows the averaged CAP during incubation with the compound. B) Boxplots showing the half maximum CAP amplitude for each respective condition in (A). Kruskal Wallis ANOVA: p = 0.6808.

TC AQP1-1 does not alter the CAP.

A) Current -voltage input-output curves for the CAP comprising data from all nerves subjected to 80 µM TC AQP1-1. Curves were produced before application of the TC AQP1-1 (green circles), at the end of the pre-incubation with TC AQP1-1 (red circles) and at the end of dye loading (black circles) to show that TC AQP 1-1 had no ebect on the amplitude of the CAP. N=5 nerves for each condition. Inset shows averaged CAP during incubation with TC AQP1-1. B) Boxplots showing the half maximum CAP amplitude for each respective condition in (A). Kruskal Wallis ANOVA: p = 0.2516.

GDPβS blocks ATP-induced increases in paranodal Ca2+ but does not alter activity dependent FITC loading into the paranode.

A) Recording of ATP-induced increase in intracellular Ca2+ at the paranode (blue bar). When ATP was applied after 100 µM GDPβS (red bar), no increase in Fluo4 fluorescence was seen. B) Images of Fluo4 fluorescence showing ebect of ATP with and without GDPβS (same recording as A). Scale bar = 5 µm. C) Boxplots summarizing that GDPβS blocks the change in fluorescence induced by ATP. Each point is a single paranode from a single nerve. D) Images of FITC loading into sciatic nerve induced by 5 min electrical stimulation with and without 100 µM GDPβS. Arrows point to paranode, insets show corresponding brightfield images. Scale bar 10 µm. E) Boxplots showing that GDPβS has no ebect on activity dependent (30 Hz stimulation) FITC loading. Each point represents a separate ROI from 5 diberent nerves.

NH4Cl does not alter FITC loading into the paranode.

A) Images of FITC fluorescence in isolated nerves stimulated for 5 minutes at 30 Hz, in the control (35 mmHg) and after treatment with 100 µM NH4Cl. Arrows indicate paranode. Scale bar 15 µm. B) Summary graph of FITC fluorescence in nerves in response to 5 minutes stimulation in each condition. Each point represents a separate ROI from 5 diberent nerves.

Ebects of acetazolamide and NH4Cl on the intracellular pH of Schwann cell paranodes.

A) representative trace showing the normalised BCECF fluorescence ΔF/F0 in response to changing pH to 6.5 and 7.15, 100μM acetazolamide (AZ) and 100 μM NH4Cl. B) representative images for the trace for the BCECF fluorescence in each condition. Scale bar = 5 µm. C) Summary graph of ebect of AZ and NH4Cl on intracellular pH. Each point represents a nerve.

Cx32DN coassembles with Cx32WT.

A) Images of Clover and mRuby2 fluorescence before and after bleaching of the mRuby2 acceptor. The Clover fluorescence becomes brighter following bleaching for the Cx32DN and Cx32WT or Cx32WT and Cx32WT pairs but not for the Cx32WT and Cx43WT pair. B) Values for the FRET ebiciency and a comparison of bleaching ebiciency for the diberent pairings.

Cx32DN coassembles with Cx32WT.

The dependence of the FRET ebiciency (E) on the Acceptor level (A) and Donor to Acceptor (D-A) Ratio. A negative correlation between E and D-A ratio indicates coassembly into hexamers as opposed to random association in the membrane.

Cx32DN forms functional gap junctions.

A) Images of cells expressing Cx32 under brightfield (BF), the mCherry tag (red) and NBDG, a fluorescent glucose analogue (green). NBDG is present in the patch pipette and readily dibuses between the cells. Scale bar 20 µm. B) Changes in NBDG fluorescence over time in the Donor and Recipient cells showing the ready transfer of NBDG via Cx32DN gap junction channels (n=8).

Inhibition of carbonic anyhydrase increases loading of FITC into paranodes and outer myelin in the absence of electrical stimulation.

The images show the FITC fluorescence in the control and presence of 100 µM acetazolamide (AZ). The boxplot shows quantification of the fluorescence and that the increase in background fluorescence is significant (control vs AZ, MW test, p<0.0001). Each point represents a separate ROI from 5 diberent nerves.

Activity dependent Ca2+ influxes into Schwann cell paranodes are dependent on CO2.

A) Representative trace showing change in normalised Fluo4 fluorescence in response to 15 Hz electrical stimulation (red bar) in the presence of 100 µM acetazolamide (blue bar), in control aCSF (green bar), or in the presence of 80 µM TC AQP1-1 (purple bar). B) Boxplot showing the change in normalised fluorescence (ΔF/F0) evoked in Fluo4 loaded Schwann cell paranodes in response to high frequency electrical stimulation in each condition. Each datapoint consists of a paranode, with all the data collected from 4 sciatic nerves. Kruskal Wallis ANOVA, p < 0.0001. Pairwise MW comparisons: control vs AZ, p < 0.001; control vs TC AQP1-1, p<0.001. C) Brightfield and fluorescence images showing the activity dependent Ca2+ increase in a paranode in the presence of acetazolamide (AZ) and the lack of an activity dependent increase in Ca2+ in the presence of TC AQP1-1.

Model of the CAP to show how changing conduction velocity alters the rise time, time to peak and peak amplitude of the CAP.

A) Shows the profile of an individual action potential (AP, top), the distribution of conduction delays (based on the distribution of diameters of fibres in the sciatic nerve (middle), and the CAP arising from summation of 2000 individual APs occurring with diberent conduction delays (bottom, light blue line) and three diberent slowing factors. B) Shows how the rise time, peak time and amplitude vary with the slowing factor. To mimic slowing, every conduction delay is multiplied by the slowing factor, thus when this is 1.0 no slowing occurs and when for example it is 1.2 each conduction delay is slowed by 20%. Matlab code for the model is provided in compoundAP.m.