Axonal response of mitochondria to demyelination and complex IV activity within demyelinated axons in experimental models of multiple sclerosis

Abstract Aims Axonal injury in multiple sclerosis (MS) and experimental models is most frequently detected in acutely demyelinating lesions. We recently reported a compensatory neuronal response, where mitochondria move to the acutely demyelinated axon and increase the mitochondrial content following lysolecithin‐induced demyelination. We termed this homeostatic phenomenon, which is also evident in MS, the axonal response of mitochondria to demyelination (ARMD). The aim of this study is to determine whether ARMD is consistently evident in experimental demyelination and how its perturbation relates to axonal injury. Methods In the present study, we assessed axonal mitochondrial content as well as axonal mitochondrial respiratory chain complex IV activity (cytochrome c oxidase or COX) of axons and related these to axonal injury in nine different experimental disease models. We used immunofluorescent histochemistry as well as sequential COX histochemistry followed by immunofluorescent labelling of mitochondria and axons. Results We found ARMD a consistent and robust phenomenon in all experimental disease models. The increase in mitochondrial content within demyelinated axons, however, was not always accompanied by a proportionate increase in complex IV activity, particularly in highly inflammatory models such as experimental autoimmune encephalomyelitis (EAE). Axonal complex IV activity inversely correlated with the extent of axonal injury in experimental disease models. Conclusions Our findings indicate that ARMD is a consistent and prominent feature and emphasise the importance of complex IV activity in the context of ARMD, especially in autoimmune inflammatory demyelination, paving the way for the development of novel neuroprotective therapies.


INTRODUCTION
Axonal loss is a cardinal neuropathological feature of multiple sclerosis (MS) [1,2]. Axonal injury is most prominently observed in actively demyelinating regions of MS, and a growing body of evidence implicates a state of energy failure in the degeneration of demyelinated axons [3]. The bioenergetic need (in terms of ATP) of demyelinated axons is thought to be increased, due to the redistribution of ion channels [4]. In keeping with this increase in energy demand of demyelinated axons, we recently identified a neuronal compensatory mechanism where mitochondria move from the cell body to the acutely demyelinated axon, increasing the axonal mitochondrial content and therefore the energy-producing capacity [5]. We termed this homeostatic mechanism the axonal response of mitochondria to demyelination (ARMD) [5].
Increased axonal mitochondrial content, reflecting ARMD, has been reported in MS and in a limited number of experimental disease models [6][7][8]. In MS, the increase in mitochondrial content of nondegenerated demyelinated axons was accompanied by an increase in activity of mitochondrial respiratory chain complex IV (cytochrome c oxidase or COX), where oxygen is reduced to generate ATP aerobically [9][10][11]. At the edge of chronic active MS lesions, where acute axonal injury is most prominent, axonal complex IV activity inversely correlated with the extent of inflammation [9]. In highly inflammatory demyelinating environments, such as experimental autoimmune encephalomyelitis (EAE), axonal mitochondrial function is compromised, perhaps by nitric oxide, an inhibitor of complex IV, and mitochondria may be damaged by post-translational modification of respiratory chain complex subunits due to nitration [12,13]. Furthermore, axonal mitochondrial transport may be perturbed by inflammation, as evident in EAE [14,15]. However, it is not known whether ARMD is consistently evident in experimental disease models, whether it always leads to a corresponding increase in mitochondrial respiratory chain complex IV activity in demyelinated axons, nor how any differences in ARMD might relate to axon degeneration.
Against this background, we used nine experimental disease models relevant to MS (highly inflammatory models and nonautoimmune models) and quantified the mitochondrial content, as well as mitochondrial respiratory chain complex IV activity of both myelinated and demyelinated axons. Complex IV is made up of

Key Points
• The mitochondrial content of demyelinated axons in animal models is significantly greater than myelinated axons, irrespective of the mode of experimental demyelination.
• The increased axonal mitochondrial content following demyelination is consistent with the recently reported axonal response of mitochondria to demyelination (ARMD), which can be enhanced to protect acutely demyelinated axons.
• In lysolecithin-induced focal lesions, the increased mitochondrial content of demyelinated axons is reflected at the level of complex IV activity, whereas highly inflammatory models such as experimental autoimmune encephalomyelitis (EAE) show a relative lack of complex IV activity within demyelinated axons.
• Complex IV activity of demyelinated axons inversely correlates with the extent of axonal injury in animal models. multiple subunits encoded by both nuclear DNA and mitochondrial DNA. Complex IV deficiency may be due to the loss of subunits, following mitochondrial DNA mutations, and/or modification of subunits by reactive oxygen species in demyelinating regions. Therefore, we assessed complex IV subunit-I (COX-I), which is encoded by mitochondrial DNA and forms a key part of the catalytic core of complex IV, in axonal mitochondria to gain insight into the potential cause(s) of complex IV deficiency in axons. Finally, we correlated complex IV activity with axonal injury in demyelinating lesions. We found ARMD to be a consistent feature of all nine animal models. However, axonal complex IV activity was only increased significantly following lysolecithin-induced demyelination. In all experimental disease models, COX-I was intact in complex IV deficient axonal mitochondria indicating that the complex IV deficiency in animal models is not due to mtDNA deletions, which leads to the loss of this subunit. The inverse correlation between complex IV activity within demyelinated axons and the extent of axon degeneration suggests a crucial role for complex IV activity in meeting the increased energy demand of the acutely demyelinated axon. This study advances our previous characterisation of ARMD, by including axonal complex IV activity in multiple animal models (highly inflammatory models, such as EAE, and nonautoimmune models), and highlights the need to protect complex IV activity of mitochondria in demyelinated axons for neuroprotection.

Experimental disease models
Snap frozen mouse, rat and marmoset CNS tissue of nine disease models and an equal number of age-matched controls (except for marmoset EAE), were generated, as indicated in Table 1. Tissue from peak demyelination time points, indicated in Table 1, was used in this study. National laws on the principles of laboratory animal care were followed and institutional ethical review committee approvals were obtained for all studies.

Triple label immunofluorescence histochemistry
Cryosections (15 μm thickness) were prepared from the entire snap frozen brain (for sagittal) and alternative 5 mm blocks of the whole spinal cord (for longitudinal sectioning). Cryosections, selected based on the presence of inflammatory infiltrates in haematoxylin/eosin staining in adjacent sections, were air dried for 60 min and fixed in cold 4% paraformaldehyde (PFA) before antigen retrieval was carried out using boiling ethylenediaminetetraacetic acid (EDTA) pH 8.0 for 5 min. The fixed cryosections were then washed and normal goat serum solution (1%) was applied for 30 min at room temperature for blocking. Primary antibodies against either neurofilament, myelin basic protein (MBP) and porin (for mitochondrial content) or neurofilament, complex IV subunit-I and complex II 70 kDa (for subunit analysis) were applied in TBS for 90 min at room temperature. Appropriate isotypespecific secondary antibodies (Life Technologies) that are directly conjugated with selected fluorochromes (FITC, rhodamine and Cy5) were applied following three washing steps thereafter. The sections were then washed and mounted using Vectashield with DAPI and stored at À20 C until required for confocal microscopy.

Sequential complex IV histochemistry and triple label immunofluorescence histochemistry
Axonal complex IV activity and complex IV subunit-I confirmation of myelinated and demyelinated axons when antibodies against myelin were not used in the triple staining protocol [17].
Sequential complex IV histochemistry and immunofluorescent histochemistry brightfield images of complex IV activity and immunofluorescent labelling of axons as well as mitochondria that lack complex IV activity, due to the blocking of immunolabelling by the deposits of complex IV histochemical reaction, were obtained using the Apotome microscope.
In adjacent sections, complex IV activity and immunofluorescent labelling of APP and synaptophysin were similarly captured. Images were taken using a x63 oil lens and brightfield and FITC, TRIC and Cy5 channels were imaged sequentially.

Quantification
Triple immunofluorescence histochemistry

Increased mitochondrial content within demyelinated axons indicates ARMD in experimental disease models irrespective of the mode of demyelination
We assessed the mitochondrial content of demyelinated axons in nine different disease models at peak clinical disease or, when a clinical phenotype is not yet present, peak demyelination time point (Table 1) and compared it with myelinated axons in controls (Table 2). Mitochondria within demyelinated axons, selected by neurofilament labelling and lack of myelin basic protein (MBP) immunofluorescence, were identified in confocal images based on immunofluorescent labelling of porin, which is a voltage-gated anion channel (VDAC) expressed in all mitochondria. (Figure 1). The mitochondrial content within F I G U R E 2 Complex IV activity within axons and the detection of complex IV subunit-I relative to complex II 70 kDa in complex IV-deficient axonal mitochondria. Complex IV activity can be localised to the axon by the sequential complex IV histochemistry (bright field images, Ai-Ii) and triple immunofluorescent labelling (Aii-Iii) of neurofilament (green), complex II 70 kDa subunit (red) and complex IV subunit-I (blue) and then by merging the bright field image with triple labelled immunofluorescent image (A-I). This sequential technique immunofluorescently labels the complex IV-deficient mitochondria (labelled with complex II 70 kDa, Aiii-Iiii) and their mitochondrial respiratory chain complex subunits (Aiv-Iiv), as previously described [16]. The grey scale immunofluorescent images of axonal complex II 70 kDa (Aiii-Iiii) and complex IV subunit-I (Aiv-Iiv) are generated by splitting the corresponding triple labelled colour image (Aii-Iii) and clearing the non-axonal mitochondria. As reported previously, the mitochondria with complex IV activity, evident in the bright field images, are not immunofluorescently labelled [16].  Figure 1 and Table 1). The significant increase in mitochondrial content within demyelinated axons of all models arose from increased mitochondrial size and/or increased mitochondrial number (Table 2).

An increase in axonal mitochondrial content is not always accompanied by a corresponding increase in mitochondrial respiratory chain complex IV activity
To determine whether the increased mitochondrial content within demyelinated axons is reflected at the functional level, we assessed complex IV activity of mitochondria at a single axon level using an established technique involving sequential complex IV histochemistry and immunofluorescent labelling of axons in snap frozen serial cryosections ( Figure 2) [16]. This sequential technique labels mitochondria with complex IV activity (in brightfield image). Furthermore, this technique identifies mitochondria that lack complex IV activity (in immunofluorescent images) and permits the determination of the subunit status of mitochondria that lack complex IV activity, as previously described [9,16]. Quantification of complex IV activity within axons revealed a significantly greater area of the demyelinated axons occupied by complex IV active mitochondria in lysolecithin-induced lesions ( Figure 2 and Table 3). In the lysolecithin-induced focal lesions, complex IV active mitochondria showed an elongated morphology and reflected the increased mitochondrial content of demyelinated axons ( Figure 2).
In highly inflammatory models, such as EAE, we did not find a significant increase in complex IV activity within demyelinated axons (Table 3), despite the increase in axonal mitochondrial content upon demyelination ( Table 2). Although complex IV activity had a tendency to increase within demyelinated axons in cuprizone and TMEV models the difference was not statistically significant. As the relative lack of complex IV activity within demyelinated axons, particularly in highly inflammatory models may be due to the loss of complex IV subunits, we proceeded to assess complex IV subunit-I using triple immunofluorescence histochemistry following complex IV histochemistry.

Mitochondrial respiratory chain complex IV subunit-I is preserved in complex IV deficient axonal mitochondria in all experimental disease models
In axonal mitochondria that lack complex IV activity, the extent of complex IV subunit-I (COX-I) labelling was similar in demyelinated axons when compared with myelinated axons, suggesting that the lack of complex IV activity is not caused by the loss of complex IV subunit-I. To confirm that the complex IV subunit is intact in demyelinated axons, we immunofluorescently co-labelled mitochondria and the subunit in serial sections and found a significant increase in complex IV subunit-I within demyelinated axons in all models, reflecting the increased mitochondrial content, compared with myelinated axons (Table 3).

Mitochondrial complex IV activity in demyelinated axons inversely correlates with axonal injury in experimental disease models
To assess the relationship between our findings and axonal damage, we correlated axonal mitochondrial parameters (content and complex IV activity) with the extent of axonal injury in all nine models, as indicated by the density of APP and synaptophysin positive elements ( Figure 3). The greatest extent of axonal injury was observed in highly inflammatory models, such as EAE, whereas lysolecithin induced The association between complex IV activity within demyelinated axons and the extent of axonal injury. The density of axonal injury, judged by amyloid precursor protein (APP, A) and synaptophysin (Ai) labelling, varies considerably between the disease models (Kruskal-Wallis p < 0.001). There is a significant inverse correlation between complex IV activity within demyelinated axons and axon degeneration, judged by the density of APP (B, r 2 = 0.421, p = 0.048) as well as synaptophysin (Bi, r 2 = 0.561, p = 0.020) labelling. In contrast, a significant correlation is not found between mitochondrial content in demyelinated axons and the density of APP (C, r 2 = 0.040, p = 0.604) and synaptophysin (Ci, r 2 = 0.147, p = 0.308) labelling. Sequential complex IV histochemistry and immunofluorescent labelling of APP shows that a subset of APP and synaptophysin labelled structures contains mitochondria with complex IV activity in all nine disease models (D, synaptophysin positive structures lacking complex IV activity are shown in T-reg depleted EAE lesion, arrowheads). Each data point in B-Bi and C-Ci indicates the mean value of a single model. From left to right in B and C, data points are in the following order: LPC, cuprizone, LPS, EAE biozzi, marmoset EAE, EAE TCR, TMEV, Treg EAE and rat EAE. From left to right in Bi and Ci, data points are in the following order: LPC, cuprizone, EAE TCR, EAE biozzi, TMEV, rat EAE, marmoset EAE, LPS and Treg EAE. The bar charts indicate the mean plus standard deviation.
lesions contained the least extent of axonal injury ( Figure 3A-Ai). At the level of complex IV activity, there was a significant inverse correlation between the mean complex IV activity within demyelinated axons and the density of APP-positive elements ( Figure 3B). Synaptophysin positive elements also showed a significant inverse relationship with axonal complex IV activity (Figure 3Bi). However, not all APP-positive and synaptophysin-positive elements were complex IV deficient ( Figure 3D). In contrast to complex IV activity, we did not detect a F I G U R E 4 The role of complex IV in ARMD. Consistent with ARMD, demyelinated axons show increased mitochondrial content compared with myelinated axons (green and red mitochondria), irrespective of the mode of experimental demyelination (as depicted for the lysolecithin model and EAE). At the level of complex IV, lysolecithin-induced lesions show a significant increase in mitochondria with complex IV activity (green mitochondria) compared with myelinated axons, reflecting the increased axonal mitochondrial content. In contrast, the increased mitochondrial content of demyelinated axons in EAE is not accompanied by a corresponding increase in complex IV activity. Instead, demyelinating axons in EAE are abundant in complex IV deficient mitochondria (red). Intact COX-I in complex IV deficient axonal mitochondria suggests that the complex IV deficiency is due to inhibition by nitric oxide and subunit modification by reactive oxygen species (ROS). The lack of complex IV activity in demyelinated axons in EAE, despite the increased mitochondrial content, is associated with a substantially greater extent of axon injury (axonal ovoids). This raises the possibility that complex IV is important for the health of acutely demyelinated axons.
significant correlation between mitochondrial content of demyelinated axons and axonal damage ( Figure 3C,Ci).

DISCUSSION
We recently reported a homeostatic mechanism in neurons, termed ARMD, where mitochondria move from the cell body to the axon upon lysolecithin-induced focal demyelination [5]. In the present study, we show that ARMD is a consistent feature of demyelinated axons in nine experimental disease models. Additionally, we assessed axonal complex IV activity and found that the increased axonal mitochondrial content in demyelinated axons is not always reflected by an increase in complex IV activity, particularly in the highly inflammatory EAE models studied. as well as in MS [6,7,9,10]. In this study, we show that the axonal mitochondrial content consistently increases, irrespective of the mode of demyelination, due to increased size and/or number of axonal mitochondria. The lack of increased mitochondrial size in inflammatory demyelination may be due to mitochondrial fragmentation, as previously reported in EAE, as well as decreased fusion of mitochondria in demyelinated axons. Whether mitochondrial fusion occurs within demyelinated axons in EAE, before mitochondrial fragmentation, needs to be investigated using live imaging techniques. While the increased axonal mitochondrial content has been proposed as a pathogenic mechanism, a study that prevented the increase of mitochondrial content in demyelinated axons by disrupting mitochondrial docking established that ARMD is a homeostatic and protective mechanism [8]. Furthermore, recent studies that enhanced ARMD, by over-expressing PGC1α in neurons as well as pharmacologically targeting PGC1α to increase mitochondrial biogenesis, showed protection of acutely demyelinated axons in EAE and lysolecithin-induced lesions [5,18]. Our current findings in nine experimental models establish that homeostatic ARMD is a consistent phenomenon, irrespective of the mode of demyelination.
Whether the increased mitochondrial content of demyelinated axons is reflected at the level of mitochondrial complex IV activity had not been studied in experimental disease models. In this study, we show that complex IV activity of demyelinated axons is dependent on the mode of experimental demyelination (Figure 4). In lysolecithininduced lesions, the significant increase in complex IV activity within demyelinated axons is comparable to the extent of complex IV activity within dysmyelinated axons in shiverer mice, where autoimmune inflammation is also not modelled [19]. In contrast, autoimmune inflammation (EAE) did not lead to a significant increase in axonal complex IV activity despite the increase in axonal mitochondrial content ( Figure 4). The lack of complex IV activity in axonal mitochondria that have responded to demyelination in EAE may be due to a number of potential molecular mechanisms. First, a previous study has shown that modification of complex IV subunits occurs in EAE, although this was not localised to particular cellular structures [12]. Second, an excess of nitric oxide which is known to compete with oxygen and inhibit complex IV is evidenced by the increase in iNOS in acute EAE [5,13] Nitric oxide is also implicated in axon degeneration due to the correlation between iNOS expression and acute axonal injury [20].
Based on the relatively low expression of iNOS in LPS lesions and cuprizone model, relative to EAE that we previously reported, inhibition of complex IV by nitric oxide is likely to be only one of several molecular mechanisms of axonal complex IV deficiency [3,5,21]. In TMEV-induced demyelination, the relative sparing of complex IV activity at 41 days may be due to weak expression of iNOS and relative lack of reactive oxygen species at this stage, as previously reported [22,23]. Third, reactive oxygen species damage axonal mitochondria and disrupt axonal mitochondrial transport in EAE [15,24].
The differential complex IV activity within axons in toxic demyelination and autoimmune inflammatory demyelination raises the possibility that inflammation damages complex IV in demyelinating models.
Our findings of intact COX-I in axons indicate that the lack of complex IV activity in experimentally demyelinated axons in EAE is not due to mitochondrial DNA (mtDNA) mutations or loss of transcripts. These observations in EAE are supported by previous studies of transcripts of mitochondrial respiratory chain subunits and mtDNA, both of which were unaltered in experimental models [5,25]. A limitation of this study is that all the experimental disease models that represent complex IV deficiency in axons, at a mechanistic level are only due to inflammation-induced complex IV deficiency, rather than the irreversible complex IV loss due to mitochondrial DNA mutations that are found in MS [5,26]. neurons is inducibly knocked out [5]. This indicates that complex IV deficiency in axons plays a role in axon degeneration through the loss of function or lack of energy rather than due to a toxic gain of function. Unlike in COX10Adv mice, where the complex IV deficiency and the loss of subunits are irreversible, complex IV deficiency due to inhibition, for example, by nitric oxide, and modification of subunits by reactive oxygen species is potentially reversible. This reversibility stems from the ability of neurons in experimental disease models to generate healthy mitochondria as they do not show mtDNA mutations or loss of nuclear DNA encoded transcripts [5,25]. In these experimental disease models, newly generated mitochondria and their movement from the cell body to the axon may replace damaged mitochondria and restore axonal energy production. In MS, complex IV deficiency of axons is caused by multiple mechanisms, including nitric oxide-mediated inhibition of complex IV, inflammation-related direct damage to complex IV as well as the chronic nature of oxidative injury leading to mtDNA deletions. Besides complex IV deficiency, other mechanisms lead to axon degeneration in the context of demyelination, as reflected by the presence of complex IV activity in a subset of degenerating axons in our animal models. However, the occurrence of potentially reversible complex IV deficiency due to inflammationinduced mitochondrial damage, alongside the irreversible complex IV deficiency due to mitochondrial DNA mutations in MS, offers therapeutic potential [9,26].

CONCLUSIONS
In summary, we show that ARMD is a consistent feature of a wide range of experimental disease models and highlight the importance of complex IV activity for the health of acutely demyelinated axons. Our findings suggest that the enhancement of ARMD, which we reported as a neuroprotective strategy for MS, may be further optimised by limiting complex IV deficiency, especially in highly inflammatory environments.  Table 2, and complied with local animal research and ethics rules.

CONSENT FOR PUBLICATION
Not applicable.

PEER REVIEW
The peer review history for this article is available at https://publons. com/publon/10.1111/nan.12851.

DATA AVAILABILITY STATEMENT
All data generated or analysed during this study are included in the published article in graphic format and tabulated.