In vivo real-time dynamics of ATP and ROS production in axonal mitochondria show decoupling in mouse models of peripheral neuropathies
Mitochondria are critical for the function and maintenance of myelinated axons notably through Adenosine triphosphate (ATP) production. A direct by-product of this ATP production is reactive oxygen species (ROS), which are highly deleterious for neurons. While ATP shortage and ROS levels increase are involved in several neurodegenerative diseases, it is still unclear whether the real-time dynamics of both ATP and ROS production in axonal mitochondria are altered by axonal or demyelinating neuropathies. To answer this question, we imaged and quantified mitochondrial ATP and hydrogen peroxide (H2O2) in resting or stimulated peripheral nerve myelinated axons in vivo, using genetically-encoded fluorescent probes, two-photon time-lapse and CARS imaging. We found that ATP and H2O2 productions are intrinsically higher in nodes of Ranvier even in resting conditions. Axonal firing increased both ATP and H2O2 productions but with different dynamics: ROS production peaked shortly and transiently after the stimulation while ATP production increased gradually for a longer period of time. In neuropathic MFN2R94Q mice, mimicking Charcot-Marie-Tooth 2A disease, defective mitochondria failed to upregulate ATP production following axonal activity. However, elevated H2O2 production was largely sustained. Finally, inducing demyelination with lysophosphatidylcholine resulted in a reduced level of ATP while H2O2 level soared. Taken together, our results suggest that ATP and ROS productions are decoupled under neuropathic conditions, which may compromise axonal function and integrity.
KeywordsAxonal activity Demyelination MFN2 Mitochondria ROS
Coherent Anti-stokes Raman Scattering
Cyan fluorescent protein
Electron transport chain
Fluorescence resonance energy transfer
Mitochondria Associated Membranes
Oxidant Receptor Peroxidase 1
Phosphate buffered saline
Peripheral nervous system
Redox sensitive GFPs
Reactive oxygen species
Yellow fluorescent protein
While the nervous system, and the brain in particular, represents around 2% of the body mass, it consumes up to 20% of the glucose we mobilize every day . This high energy expenditure in the nervous system is firstly due to the synaptic activity that requires high amounts of ATP . A second energy demanding process is the propagation of action potentials (APs) along axons. Indeed, conduction of APs involves ion exchanges through the plasma membrane, first through voltage-gated ion channels to depolarize, then through the sodium-potassium ATPase to repolarize [4, 5] Therefore, the production of ATP in axons is crucial for repeated regeneration of APs. Both in the CNS and in the peripheral nervous system (PNS), myelination significantly reduces the energy cost of AP propagation through the sequestration of the AP firing machinery at the node of Ranvier and axon initial segment .
Mitochondria appear to be the main source of cellular ATP and these organelles are abundant in CNS [24, 48] and PNS  axons. Some reports indicated that mitochondria were more abundant in the node of Ranvier , but this remains controversial . In addition, axonal AP activity and axo-glial junctions regulates the recruitment of mitochondria in the nodal area . Therefore, a well-accepted idea is that axonal AP activity could locally stimulate ATP production in axonal mitochondria .
As an intrinsic by-product of ATP production, mitochondria also produce reactive oxygen species (ROS) [31, 50]. The mitochondrial electron transport chain (ETC) and NAPDH oxidases are the major sources of superoxide (O2•−), a highly reactive type of ROS, within cells . O2•− is rapidly converted to H2O2 by superoxide dismutase (SOD) and H2O2 is then reduced to H2O and O2 by mitochondrial glutathione peroxidase (GPx) and cytosolic enzymes . H2O2 can also react with iron or O2•− to form hydroxyl radicals (OH•), which is then reduced by molecular hydrogen to H2O . If not reduced by antioxidants, all types of ROS are highly toxic for the cell and in particular for mitochondria, where they are produced. Indeed ROS can directly oxidize and damage DNA, carbohydrates, proteins or lipids .
In several neurodegenerative diseases and axonopathies, such as Parkinson’s disease , multiple sclerosis (MS)  and Charcot-Marie-Tooth diseases (CMT) , mitochondrial dysfunctions result in increased levels of ROS. The importance of mitochondrial homeostasis for the normal function of myelinated axons [55, 65] is highlighted by mutations in the mitofusin 2 (MFN2) gene. Indeed, MFN2 has several functions in mitochondria physiology, among which mitochondrial fusion  and transport along axons , and mutations in its gene result in axonal form of peripheral neuropathy CMT2A in humans and in mice [13, 83]. In particular, MFN2 loss-of-function in axonal mitochondria decreases the expression of oxidative phosphorylation subunits , suggesting that MFN2 mutations also affect ATP and ROS production. Axonal mitochondria dysfunction is as well involved in MS, a chronic inflammatory disease affecting myelinated tracts in the brain . MS is characterized by demyelinated lesions resulting from the loss of the myelin sheath and the progressive form of MS shows successive events of myelin degeneration and regeneration that damage axons . An increase of mitochondrial ROS contributes to progressive MS  and a common hypothesis is that demyelinated axons are more demanding in ATP in order to maintain their functions and their integrity [45, 64].
So, several important questions remain unanswered: How different is the production of ATP and ROS in nodal mitochondria versus internodal mitochondria? How does action potential propagation regulate the dynamics of mitochondrial ATP and ROS production? In what way do axonal or demyelinating neuropathies affect ATP and ROS production in mitochondria of peripheral nerve axons?
To answer these questions, we set up an in vivo approach using a H2O2-sensitive GFP (roGFP-Orp1)  and a fluorescent ATP sensor (ATeam)  targeted to mitochondria to image and measure the dynamics of ATP and H2O2 production in mitochondria of myelinated axons of the PNS. Using this approach, we show that both ATP and H2O2 levels are increased in mitochondria residing in nodes of Ranvier and that nerve stimulation sharply increases ATP and H2O2 production by mitochondria in few minutes. While MFN2 mutations had no effect on basal levels of ATP and H2O2 in axons, it prevented the increase of ATP production after nerve stimulation. At the opposite, an increased H2O2 production was observed. Moreover, pathological demyelination reduced ATP production while enhancing mitochondrial ROS, showing that both neuropathic conditions decoupled the production of ATP from the production of ROS.
Materials and methods
pLPCX mito-roGFP-Orp1 (Addgene; #64992) was digested with XhoI/ClaI, blunted and cloned into pAAV-MCS (Cell Biolabs, Inc.) under the control of a CMV promoter. Clones were validated by sequencing. Likewise, pcDNA-mito-ATeam (from H. Imamura, Tokyo, Japan) was digested with XhoI/HindIII, blunted and cloned into the CMV promoter controlled pAAV-MCS. Mitochondria-targeting tags were CoxVIII for roGFP-Orp1 and two tandem copies of CoxVIII for mito-ATeam. AAV viral particles were produced at the viral vector production centre at the Centre de Biotecnologia Animal I Teràpia Gènica in Barcelona, Spain or the University of Nantes, France.
Either Swiss mice (Janvier, France) or MFN2R94Q mice  were used in the performed experiments. Mice were kept in the animal facility of the Institute for Neurosciences of Montpellier in clear plastic boxes and subjected to standard light cycles. All animal experiments were conducted in accordance with the French Institutional and National Regulation CEEA-LR-11032.
In vivo virus injection in spinal cord
A thin borosilicate glass capillary (Harvard Apparatus, Ref. 30-0016) was pulled with a Vertical Micropipette Puller (Sutter Instruments, P30–682) to form a glass needle. The glass needle was filled with viral solution (1 μl) and a 1 day old mouse is restrained in a position that exposes the lower back. The needle is injected through the skin using a micromanipulator and introduced into the spinal cord. The viral solution was injected over 2 min with short pressure pulses using a Picopump (World Precision Instrument) coupled to a pulse generator. After injection, the injection site is cleaned with betadine solution (Vetoquinol, cat. no. 3042413) for disinfection and the pup is placed back with its mother and littermates.
One month after the viral injection, the sciatic nerve was dissected and fixed in Zamboni’s fixative  for 10 min at room temperature. After fixation, the dissected sciatic nerves are washed in PBS and incubated in successive glycerol baths (15, 45, 60, 66% in PBS) for 18 to 24 h each before freezing at − 20 °C. The nerves were cut in small pieces in 66% glycerol and the epineurium sheaths removed. Small bundles of fibers were teased in double-distilled water on Superfrost slides and dried for 3 h at room temperature. Some nerves were frozen in O.C.T. Compound (Tissue-Tek, Ref. 4583) and longitudinal sections were cut using a cryostat (Leica Biosystems, CM3050). For immunostaining, the teased fibers or longitudinal sections were incubated for 1 h at room temperature in blocking solution (10% goat serum and 0.3% TritonX100 in PBS). Then, the samples were incubated with TOM20 (FL-145) primary rabbit antibody (1/500, Santa Cruz, Ref. sc-11,415), mouse anti-Myelin Basic Protein (1/500, Merck, Ref. NE1019) or rabbit NF-200 (1/500, Sigma, Ref. N4142) in blocking solution overnight at 4 °C. The next day, the samples were washed in PBS and incubated for 1 h at room temperature with secondary donkey antibodies coupled to Alexa568 (1/1000, Invitrogen, Ref. A10042) or Alexa488 (1/1000, Invitrogen, Ref. A21202). Finally the samples were washed in PBS and mounted in Immu-mount (Thermo Scientific). Images were acquired at room temperature using a 20× or 40× objective, a Zeiss confocal microscope LSM710, and its associated software.
Imaging of sciatic nerve in living mice
One month after the viral injection, the mice were anesthetized with a constant flow (1.5 l/min) of oxygen + 5% of isoflurane in an anesthesia box (World Precision Instruments, Ref. EZ-B800) for 5 min. Thereafter the anesthesia was maintained with a mask delivering 2% isoflurane at 0.8 l/min. The eyes were protected by eye protection gel (Ocry-gel, TVM, cat. no. 48026 T613/3). Intraperitoneal injection of 0.1 mg/kg buprenorphine was used for pre-surgery analgesia. The mouse was placed in a silicone mold, lying on its belly, shaved on his hind paw and the paws immobilized using small pins. The incision area was disinfected with betadine solution (Vetoquinol, cat. no. 3042413). The skin was cut using scissors, fat tissue removed, the gluteus superficialis and biceps femoris muscles were separated to reveal a cavity crossed by the sciatic nerve, and the sciatic nerve was lifted up using a small spatula. A long plastic strip was placed underneath the sciatic nerve and this strip fixed using magnets. The sciatic nerve was kept in an aqueous environment of either sterile PBS buffer or artificial cerebrospinal fluid (148 mM NaCl, 3 mM KCl, 1.4 mM CaCl2•2H2O, 0.8 mM MgCl2•6H2O, 0.2 mM NaPO4•H2O in sterile H2O) to prevent drying. At that point the mouse was placed under the two-photon microscope objective lens, a glass, 12 mm diameter, 0.25 mm thick microscope coverslip put on top of the nerve and a drop of deionized water placed on it to immerse the objective lens (20X, Carl Zeiss Microscopy, LD CApochromat, Ref. 421,887–9970).
After imaging of the sciatic nerve in vivo for at least 30 min, 5 μl of 1 mg/ml LPC (Sigma, Ref. N4129) was injected directly into the sciatic nerve using a Hamilton syringe (Ref. 80,930) to induce demyelination. Injection of 5 μl PBS is used as a negative control. After injection, the nerve was placed back to its original location in the body. The skin of the incision was realigned together using a blunt scalpel and stapled along the wound with two clips (Fine Science Tools; 12,020–00). The area around the wound was disinfected again with betadine solution. The anesthesia mask was removed and the mouse was monitored until it had woken up. Motor function of the injected and non-injected hindlimb was followed by lifting the animal by the tail. At 1 week after LPC injection, the non-injected hindlimb showed a normal postural reflex characterized by spreading of leg whereas the injected hindlimb showed abnormal reflexes such as tremors, clasping or retracting of the paw. This behavior has been described and linked to deficient myelination in the PNS in previous studies . The severity of abnormal hindlimb function decreased at 2 weeks after LPC injection and no difference in leg function between the injected and non-injected hindlimb could be observed anymore at 3 weeks after LPC injection.
Saphenous nerve stimulation in living mice
After skin incision and removal of connecting tissue, the saphenous nerve is lifted up and isolated using a plastic strip (Fig. 2a). To stimulate the saphenous nerve, two platinum kapton microelectrodes (World Precision Instruments, PTM23B05KT) were inserted at the posterior side of the plastic strip using a micromanipulator (Fig. 2a). One hook-shaped recording electrode (AD Instruments, MLA 1203) was placed at the anterior side (Fig. 2a). The ground electrode was inserted in the tail of the mouse and the negative electrode in the groin area (Fig. 2b). The mouse was placed under the two-photon microscope (Fig. 2a), a microscope glass coverslip was placed on top of the nerve and the electrodes were connected to a Powerlab 26 T (AD Instruments; ML4856). A drop of deionized water was then placed on top of the microscope glass to immerse the 20× objective lens.
Two-photon image acquisition
All in vivo images were obtained with a two-photon microscope LSM 7 MP OPO (Zeiss, France) coupled to a dark microscope incubator (L S1 Dark, Zeiss) in which the temperature was maintained at 37 °C (Heating Unit XL S, Zeiss, France). Mitochondria images were acquired by time-lapse recording varying from one image every minute to one image every 5 min during 1 h. Each image is a stack at Maximum Intensity (ZEN software, Zeiss) of 10 scans over 40 μm depth. For ATeam imaging, a single track at 850 nm excitation wavelength is used to obtain both the CFP (em. 475 nm) and Venus (em. 527 nm) image at the same time point. For roGFP imaging, the two images were acquired for each time point using alternating tracks at 940 nm and 800 nm. Change of track was set after each stack. For Coherent Anti-Stokes Raman Scattering (CARS) imaging, 2 synchronized laser lines at excitation wavelengths 836 nm and 1097 nm are used simultaneously thanks to the OPO system. Each scan was acquired with constant laser intensity (20% for 940 nm, 10% for 850 nm, 10% for 800 nm, 15% for 836 nm, 4% 1097 nm) at a 512 × 512 pixel resolution and microscope imaging parameters were maintained over all different regions we imaged. Images, acquired with ZEN software (Zeiss), were saved in .czi format.
Data and statistical analysis
We used ImageJ software to analyze the relative ATP or H2O2 levels in mitochondria of peripheral axons. The acquired images for each wavelength were aligned using the Template Matching plugin. We defined a Region of Interest (ROI) encompassing all labeled mitochondria of the same axon and the mean fluorescent intensity on the ROI was measured on both images. These light intensities were then corrected for background light intensity determined as an area within the nerve where no fluorescent signal from the viral probe can be observed. Either the ATeam Venus/CFP ratio or the oxidized roGFP/reduced roGFP ratio was then calculated from the 2 values for mean light intensity.
Statistical significance for the effect of time on probe stability was determined using linear regression. The effects of the positive controls and negative controls on probe validation were determined using Student two-tailed T-tests. The effect of nerve stimulation was determined using one-way ANOVA and Dunnett’s post-hoc tests. In addition, the threshold for a responding axon was set at 20% change in probe fluorescence ratio. These values for MFN2 mutant mice were obtained using MFN2 mutant and control littermates and the experiments were performed in a blinded way (surgery, imaging and image analysis was performed before knowing the mouse genotype). The differences between MFN2 mutants and wild-type mice were determined using either Student two-tailed T-tests or two- way ANOVA with Sidak post-hoc tests. Differences between internodes and node of Ranvier mitochondria were determined using paired two-tailed T-tests. The effect of demyelination was determined using paired two-tailed T-tests.
Validation of the mito-roGFP-Orp1 and mito-ATeam probes
We used ATeam, which is a genetically-encoded fluorescence probe based on fluorescence resonance energy transfer (FRET), to detects ATP . This probe consists of the cyan fluorescent protein (CFP) mseCFP, the yellow fluorescent protein (YFP) variant monomeric Venus, both linked at the ε-subunit of the Bacillus subtilis FoF1-ATP synthase. In the presence of ATP, the ε-subunit retracts to bring the two fluorescent proteins close to each other, thereby increasing FRET efficiency. The conformational change is reversible. Relative ATP levels are measured using the Venus/CFP fluorescence ratio.
Redox sensitive GFPs (roGFPs) harbor an engineered dithiol/disulfide switch on their surface, which determines their wavelength of excitation . RoGFP2 has been converted into an H2O2 specific probes by fusion with the microbial H2O2 sensor Oxidant Receptor Peroxidase 1 (Orp1) . The reaction is reversible via reduction by cellular thioredoxin (Trx) or GRx/GSH . Relative H2O2 levels are measured using the ratio of emitted fluorescence lights when excited at 800 nm (oxidized form) or 940 nm (reduced form). All experiments were done in the same imaging conditions and analysis. As both probes are ratiometric, the measured values are independent of the number of mitochondria or of the absolute fluorescence intensity of the probe. Both probes were targeted to the mitochondrial matrix.
Then the ratio of the fluorescent intensity of oxidized/reduced roGFP-Orp1 (R) was followed over time (Fig. 1c), showing a small and slow increase of the ratio (0.206 ± 0.002 per hour) probably due to air oxidation. Addition of reducer DTT on the nerve resulted in a significant decrease of the fluorescence ratio, while injection of oxidizer diamide resulted in a significant increase of fluorescence ratio (Fig. 1d). These results show that the dithiol/disulfide switch of mito-roGFP-Orp1 is an efficient and reliable indicator of relative redox changes in axonal mitochondria of mouse peripheral nerves in vivo. The fluorescence ratio Venus/CFP of mito-ATeam (R) was very stable over time (Fig. 1e). Injection of 0.4 M and 1 M ATP into the sciatic nerve in an increase of ATeam fluorescence ratio (Fig. 1f), showing this probe is also functional in vivo.
The effect of nerve stimulation on ATP and H2O2 production
ATP and H2O2 production are altered in CMT2A neuropathic mice
Similar to ATP, we found no difference between controls and MFN2R94Q mice in the basal H2O2 levels (Fig. 4d). When nerves were stimulated, mitochondrial H2O2 significantly increased after each stimulation period (Fig. 4e). However, this increase was delayed compared to controls and H2O2 levels remained high for longer time (Fig. 4f), showing that axonal mitochondria produce more deleterious H2O2 than control mice in response to axonal activity. Since ROS production in the mitochondrial matrix reflects the oxidative phosphorylation process [7, 11], these data indicate that oxidative phosphorylation is decoupled from ATP production in mitochondria of these neuropathic mice.
ATP and H2O2 production in nodes of Ranvier and internodes
ATP and H2O2 levels in mitochondria of demyelinated axons
We therefore measured the fluorescence ratio for these probes in non-stimulated axonal mitochondria during the process. One week after LPC injection, when axons are completely demyelinated, ATP level was slightly lower than before demyelination, but not significantly (Fig. 6b), while H2O2 level peaked (Fig. 6c). ATP level was significantly lower in axonal mitochondria at 2 weeks after LPC injection, while H2O2 level was back to control and pre-demyelination level (Fig. 6b-c). Finally, 3 weeks after injection, when remyelination is completed, ATP an H2O2 levels were back to control and pre-demyelination levels (Fig. 6b-c). No change either in ATP or H2O2 levels were detected at 1 week after injection with PBS in control mice (Fig. 6b-c). The increase in ROS did not result from a general oxidative stress in the demyelinated nerve, since CellROX dye did only show a minor fluorescent signal, except at the LPC/PBS injection site where we did not image our probe (Additional file 2). Moreover, the CellROX signal did not show any significant difference between LPC and PBS injected nerves, whereas we did measure differences in mito-roGFP-Orp1 signal. Taken together, these data indicate that the myelin sheath has a profound impact on axonal mitochondria physiology. Moreover, as ATP level decreased during myelin break-down while ROS level increased this suggests that demyelination induces the decoupling of ATP production from oxidative phosphorylation.
Although it has been shown that PNS axonal mitochondria, like most mitochondria, produce both ATP and ROS, the regulations of their production are still not well understood in vivo and in situ. We used a combination of viral delivery of mitochondria-targeted fluorescent probes to active peripheral neurons and two-photon and CARS live imaging in mice to fill that gap. The mito-ATeam and mito-roGFP-Orp1 fluorescent probes had been previously validated in vitro  and in vivo  in other systems but we also validated them in vivo in our system. We did not attempt to measure absolute values of ATP or H2O2 with each probe fluorescence in vivo because firstly, this was difficult as probes are targeted to mitochondrial matrix and secondly, it was not necessary as we aimed to compare different genotypes and conditions in particular along time. So only relative values obtained in the same imaging conditions are shown here. We found that the range of detection was large enough to cover the changes occurring in axonal mitochondria in vivo. Moreover the probe sensitivity was sufficient to observe variations in real-time using a 5 min delay between measures. This delay was partly imposed by relative low speed of our imaging system (1 min to create 10 scans of 200μm2 over 40 μm depth for each wavelength), but also by the intrinsic limits of our set-up: electrical stimulation induced slight contractions of saphenous nerve surrounding muscles which made the pictures blurred during the stimulation. The probes alterations were also reversible fast enough to observe changes in both directions.
Mito-ATeam was stable, but mito-roGFP-Orp1 showed a slight increase of its fluorescence ratio over time under our experimental settings. These results are not explained by photobleaching or degradation of the probe, since the total fluorescent signal does not decrease. This change in ratio of fluorescence may be caused by air oxidation . However, buffered artificial cerebrospinal fluid as an incubation medium for the saphenous nerve did not prevent the probe oxidation. Nonetheless, this slow oxidation of mito-roGFP-Orp1 had negligible effects on our experiments, since it was too slow to cause significant changes in our relatively short term imaging.
Our probes gave us an instant read-out of ATP and H2O2 levels in axonal mitochondria. Since ATP export and ROS scavenging exist in mitochondria, these levels are in absolute the result of ATP and ROS production minus ATP export and ROS scavenging. However, during time-lapse imaging over a relatively short period (minutes), we considered that ATP export and ROS scavenging were unlikely to be the main players of the changes we observed. Indeed, ADP/ATP translocase, the main exporter of ATP, mediates a relative steady-state ADP/ATP exchange rate [25, 47] and ROS scavenging occurs through antioxidant enzymes that are always heavily expressed within mitochondria . So, in these conditions, we believe that these fluorescent probe changes reflect mainly changes in ATP and ROS production.
Healthy mitochondria produce ATP through oxidative phosphorylation and as a consequence, ROS levels also increase. So in healthy mitochondria ATP and ROS productions are usually correlated in the same direction, meaning they are coupled. When changes in mitochondrial ATP and ROS productions are not correlated in the same direction (i.e. decreased ATP, but increased ROS or the reverse) we defined this as a decoupling. Axonal activity induced stereotypic changes in ATP and H2O2 production by axonal mitochondria. Shortly after the stimulation, production of ROS overcomes the scavenging capacities of antioxidant enzymes resulting in a peak, and then ATP increased for several minutes. This coupling was stereotypic, because it occurred almost identically after both stimulations we did. The delay between ROS production and ATP peak production may be due to the requirement for a significant proton gradient to be generated before ATP production can detected by our probe. On the level of a single mitochondrion, this delay may be extremely short but it may be significantly longer over a large population of mitochondria as we are looking at.
After the first stimulation, not all axonal mitochondria produced more ATP, suggesting that either our stimulation protocol was not mobilizing enough axons or axonal mitochondria were deactivated, similar to some ion channels. This limitation was lifted by the second stimulation. So, it could also imply that actually our ATP probe may not be sensitive enough to detect a relatively small amount of ATP and this would delay the detection of ATP production in mitochondria until the production reaches a certain level.
On the attempts we did for further stimulations, sometimes the ATP level was maintained at high range and sometimes mitochondria did not respond anymore. The reason for these variations in the long term is unknown and we did not pursue this further. However, taken together, this suggests that ATP production in axonal mitochondria is positively correlated with the level of axonal activity and can eventually reach a plateau in our conditions. Due to technical reasons we could not measure H2O2 levels in mitochondria after several stimulations. However, looking at the first two stimulations, we conclude that H2O2 production in mitochondria is also positively correlated with axonal activity but not cumulative.
It would be interesting to look further for the molecular mechanisms that allow mitochondria to detect axolemma depolarization or ion channels activity on a very short period of time (around 1 min in our detection capacities). This fast kinetic suggests the role of ions, such as calcium ions, which are mobilized during AP firing at the node of Ranvier , or kinases. A mechanism involving calcium has been shown to anchor mitochondria at the node of Ranvier , suggesting that mitochondria are associated with the axolemma there. Consistently, we found nodal mitochondria to be more metabolically active than internodal ones. However, our data following stimulations were recorded on mitochondria independently of their location on axons. As the node is much smaller than the internodes (1 μm versus 600 μm in average in the mouse nerve) the probability we recorded nodal mitochondria is extremely small. So the molecular mechanism that mobilizes mitochondria after AP firing has to exist also in internodal mitochondria.
Looking at the known origin of ROS in mitochondria, our data suggest that the H2O2 production we detected in axonal mitochondria reflects the loading of the mitochondrial matrix with protons before the production of ATP. However, since the H2O2 production drops very quickly and, contrary to the ATP production, it is not cumulative, a more complex scheme has to be taken in consideration. Indeed, H2O2 is formed by SOD enzyme from O2•−, which directly results from the ETC. Moreover, H2O2 is then changed in H2O and O2 by several reducing enzymes, such as mitochondrial GPx, so actually the amount of H2O2 we detected is the result of equilibrium between O2•− production and SOD and reducing enzymes’ activities. As these enzymes are highly expressed in mitochondria , they probably do not constitute a limiting factor and the H2O2 levels we detected still reflect axonal mitochondria metabolic activation. However, H2O2 levels drop quickly, probably because reducing enzymes are removing it quickly. In addition, these enzymes’ activities change following fast post-translational modifications such as phosphorylation  or environmental changes . While this may explain why H2O2 levels are not incremental following several stimulations, further investigation will be needed to clarify underlying mechanisms.
Several CNS and PNS neuropathies are linked to mitochondrial dysfunctions in neurons [19, 54, 67]. Our data show that the electrical activity of neurons directly regulates the physiology and the respiration of axonal mitochondria. As axonal mitochondria are the most abundant among neuronal mitochondria, this suggests that axonal mitochondria dysfunctions may severely impact neuronal physiology and survival. We evaluated this hypothesis using a couple of peripheral neuropathy mouse model, such as a model of CMT2A disease expressing mutant MFN2R94Q . While in vitro data had shown increased H2O2 levels , but no change in ATP production  in neuronal mitochondria of these mice, in vivo we did not measure significant differences in these parameters between resting MFN2R94Q mice and control mice. A first explanation for this discrepancy is that cultured cells are living in an environment that is significantly different from the in vivo conditions. This is especially true in terms of metabolism . Another explanation is that mitochondria physiology is altered, but this can only be seen when mitochondria are challenged. Environmental changes in culture may constitute this challenge, while more physiologically relevant challenges have to be found in vivo. Our electrical stimulations were physiologically meaningful challenges for axons and, similar to control mice, we observed ATP levels increasing in several axons of mutant mice, but not all, after a first nerve stimulation. However, while control mice mitochondria increased their ATP production further following a second stimulation, no increase was observed in mutant mice mitochondria. We have shown that Mitochondria Associated Membranes (MAM), which mediate interactions between the endoplasmic reticulum and mitochondria , are impaired in these mutant mice. So we suggest that an impaired Ca2+ uptake by mitochondria through MAM  could alter the ATP production as it has been shown that calcium is required for the TCA cycle . However, mutant mice mitochondria were still able to produce a large amount of H2O2 following both stimulations, suggesting the ETC and therefore the TCA cycle are working properly. Yet this H2O2 production was completely abnormal: it was largely delayed and was sustained for over at least 10 min. As discussed before the levels of H2O2 result from an equilibrium between several parameters, so it is difficult to explain this abnormal production. Nevertheless, a relationship between mitochondrial shape and ROS has been shown in vitro . It has been suggested that smaller mitochondria result in intra-mitochondrial redistribution of cytochrome c and the pro-aging molecule p66shc that acts trough ROS upregulation [20, 30]. Smaller fragmented mitochondria, such as MFN2 mutants , correlate with a malfunctioning antioxidant system and an increased production of ROS [20, 52]. In any case, our data show that an altered MFN2 function lead to the dysfunction of axonal mitochondria only when axons are firing. This dysfunction is strongly deleterious as mitochondria don’t produce more ATP, but show a sustained production of deleterious ROS. The functional decoupling between ATP and H2O2 production , occurring in the most abundant mitochondria of the neurons and in conditions of normal firing activity, is likely to be a significant cause of the neuronal dysfunction in this peripheral neuropathy.
Looking at the node of Ranvier of MFN2 mutant mice, we found that, similar to control mice, nodal mitochondria were more metabolically active than internodal ones. As MFN2 mutant mitochondria do not increase ATP production in active axons, this suggests that nodal mitochondria are intrinsically more active even in absence of firing. This is consistent with the fact that motile mitochondria slow down in the nodes of Ranvier  and that they have a different morphology . Moreover, nodal mitochondria may participate to axonal degeneration since this process starts in nodes of Ranvier due to high ROS levels . It would have been interesting to investigate changes occurring in ATP and H2O2 production of nodal mitochondria after electrical stimulation but this was impossible due to the small size of the node and the change of focus that occurs following stimulations.
The myelin sheath that covers axons appears to be an important regulator of axonal metabolism and of mitochondria physiology. Indeed in MS, demyelination alters axonal mitochondria [45, 67, 73] and we show here that nodes of Ranvier devoid of myelin house more metabolically active mitochondria. To go further, we therefore investigated the impact of the loss of myelin on axonal mitochondria. LPC was the right tool because it induced demyelination without destroying axons [3, 42, 57]. LPC-induced demyelination resulted in a decrease of ATP levels in axonal mitochondria. This decrease was strongest after 2 weeks and was reversed by 3 weeks, suggesting that ATP production within neurons is impaired until complete remyelination. This decrease was unexpected, since AP propagation in demyelinated axons is more energy demanding: the nodal machinery and the Na+/K+ ATPase are diffused all along the axolemma and all the energy-effective benefits of myelination are lost . In demyelinated area of brains affected by MS, an upregulation of ETC complexes protein expression and more complexes I and IV activity  was observed. Moreover, demyelinated axons showed more mitochondria , which is consistent with a higher energy demand. Therefore, next to a defect in ATP production, an increased export rate of mitochondrial ATP by ADP/ATP translocase  may also participate to the decreased mitochondrial ATP levels in demyelinated axons.
Nevertheless, H2O2 levels strongly increased in axonal mitochondria of fully demyelinated axons in absence of general oxidative stress in the tissue. This decoupling of mitochondrial oxidative phosphorylation and ATP production is in accordance with previous studies that reported a decreased ATP production, together with increased ROS levels in demyelinating diseases . This suggests that actually, mitochondrial ETC complexes are more active and produce more ROS, but this may not lead to more ATP production. Such a mechanism has been shown in particular conditions where the goal of mitochondrial activity is to produce heat. Uncoupling proteins such as UCP2 permeabilize the inner mitochondrial membrane to dissipate the proton gradient . This may not be the sole utility of uncoupling as this process has been shown to occur in several neurodegenerative diseases . However, so far we failed to detect unusual UCP2 expression in demyelinated nerves. The high production of H2O2 during demyelination is probably not without consequences as ROS can alter the function of the ATP synthase . This may explain why ATP production remains weak in mitochondria of axons that are being remyelinated 2 weeks after LPC injection, despite that ROS production is back to normal. This suggests that during demyelination, axons may recruit more mitochondria , or use another metabolism, such as aerobic glycolysis, in order to cover their need in ATP.
While very little amount of data exists on the homeostasis of axonal mitochondria in demyelinated axons of the PNS, in the CNS, demyelination in progressive MS results in axonal mitochondria dysfunction and ROS increase . The data we collected in PNS myelinated axons on ROS production by axonal mitochondria are consistent with this. This suggests that, despite the large difference that exists between PNS and CNS myelinating glia, their similar role in segregating axonal firing machinery at nodes of Ranvier is essential for mitochondria homeostasis in axons. Nevertheless, the electrical isolation of the axon and the formation of the node of Ranvier is not the only function of the myelin sheath as it also largely participates to the metabolic homeostasis of the axon through the lactate shuttle process [28, 76]. This is particularly true for the CNS myelin . In addition, MS disease is significantly different from peripheral nerve diseases, in particular because of the important role played by CNS glial cells astrocytes and microglia. So, it would be essential to confirm the data we obtained in PNS myelinated axons in CNS myelinated axons in vivo, in order to definitely conclude on the alteration of ATP and ROS production by axonal mitochondria in demyelinated lesion of progressive MS.
Perturbations of ATP and ROS production in mitochondria are involved in several neurodegenerative diseases. However, the dynamics of mitochondrial ATP and ROS production in healthy and diseased axons, which represent the main volume of neurons, are unknown. Combining several in vivo techniques, we characterized the dynamics of mitochondrial ATP and H2O2 production during action potential firing in living peripheral nerve axons. In mouse models of CMT2A and demyelinating neuropathies, we showed that ATP and ROS levels change in an opposite way suggesting a decoupling of their productions in mitochondria. Together, these data provide a novel insight into the role of axonal mitochondria in the production of ATP and ROS under physiological and pathological conditions.
We thank the MRI imaging platform, supported by the French National Research Agency (ANR-10-INBS-04), in particular Hassan Boukhaddaoui, and the Animal Facility of the INM. We also would like to thank Volker Baeker for the image analysis protocol.
This work has been supported by the European Research Council grant (FP7-IDEAS-ERC #311610) to N.T., the Association Française contre les Myopathies AFM-Téléthon research grant (#20044) to N.T. and R.C., The Neuromuscular Research Association Basel grant to R.C., by the Swedish StratNeuro program grant to R.C., by the Swedish Research Council grant (#2015–02394) to R.C., and by the Labex EpiGenMed ANR-10-LABX-12-01 to G.H.
Availability of data and materials
The datasets used and/or analysed during the current study available from the corresponding author on reasonable request.
GvH conducted the experiments. GvH and NT designed the experiments and wrote the paper with contributions from GC and RC. MD and PQ contributed to the design of the multiphoton imaging and nerve stimulation experiments. JB and BG contributed to the production and validation of the viral vectors. The data were interpreted by GvH, NT, MD and GC and NT supervised the project. All authors read and approved the final manuscript.
Ethics approval and consent to participate
All animal experiments were conducted in accordance with the French Institutional and National Regulation CEEA-LR-11032.
Consent for publication
The authors declare that they have no competing interests.
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
- 9.Bernard-Marissal N, van Hameren G, Juneja M, Pellegrino C, Louhivuori L, Bartesaghi L et al (2019) Altered interplay between endoplasmic reticulum and mitochondria in Charcot–Marie–tooth type 2A neuropathy. Proc Natl Acad Sci Available from: https://www.pnas.org/content/early/2019/01/17/1810932116
- 43.Li X, Fang P, Yang WY, Chan K, Lavallee M, Xu K et al (2017) Mitochondrial ROS, uncoupled from ATP synthesis, determine endothelial activation for both physiological recruitment of patrolling cells and pathological recruitment of inflammatory cells. Can J Physiol Pharmacol 95:247–252CrossRefGoogle Scholar
- 51.Mytskaniuk V, Bardin F, Boukhaddaoui H, Rigneault H, Tricaud N (2016) Implementation of a coherent anti-stokes Raman scattering (CARS) system on a Ti:Sapphire and OPO laser based standard laser scanning microscope. J Vis Exp 113. https://doi.org/10.3791/54262
- 54.Palau F, Estela A, Pla-Martín D, Sánchez-Piris M (2009) The role of mitochondrial network dynamics in the pathogenesis of Charcot-Marie-tooth disease. In: Espinós C, Felipo V, Palau F (eds) Inherit neuromuscul dis transl pathomechanisms ther. Springer Netherlands, Dordrecht, pp 129–137. https://doi.org/10.1007/978-90-481-2813-6_9 CrossRefGoogle Scholar
- 64.Sedel F, Bernard D, Mock DM, Tourbah A (2016) Targeting demyelination and virtual hypoxia with high-dose biotin as a treatment for progressive multiple sclerosis. Oligodendrocytes Health Dis 110:644–653Google Scholar
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.