This study used 0.1% CPZ-feeding for 12 weeks in young adult C57Bl/6 mice to investigate whether a slow and progressive degenerative process can induce histological and proteomic alterations in the axonal tracts and nuclear components of the visual pathway. Although histological changes such as demyelination and glial activation were evident in the corpus callosum, none of these changes were observed in the components of visual pathway. However, multiple alterations in the optic nerve/tract proteome were detected using a sensitive, well-established, quantitative top-down proteomic analysis. Subsequent comprehensive literature and bioinformatic analyses revealed that many of the proteoforms identified in the soluble and membrane sub-proteomes of the optic nerve/tract are involved in structural and metabolic functions. Likely aggregation/oligomerization of some proteoforms was also detected.
Limited histological changes
Demyelination and inflammation in the components of the visual pathway (e.g. visual cortex, pretectal nucleus, lateral geniculate nucleus, superior colliculus, optic tract) are associated with optic neuritis (Martínez-Lapiscina et al., 2014; Sapienza et al., 2016) and these conditions have been investigated in animal studies (Guido, 2018; Huberman and Niell, 2011; Seabrook et al., 2017). In a previous study, we found significant demyelination and gliosis in the visual cortex when mice were fed with 0.2% CPZ for 5 weeks, but not in the optic tract (Sen et al., 2020b). In the current study, the lack of detectable demyelination in the visual cortex might be attributable to the staining method or to the concentration of CPZ used. For example, Nile Red staining detected changes in the myelin components of the corpus callosum following only two days of CPZ-feeding, whereas other traditional histological staining methods (e.g. silver, or Luxol-fast blue) detected quantifiable demyelination only after 4-5 weeks (Hiremath et al., 1998; Sen et al., 2020b; Teo et al., 2021). Additionally, demyelination has been detected in the optic nerve after 12 weeks of CPZ-feeding, but no other part of the visual pathway was assessed in that study (Namekata et al., 2014). In the current study, 0.1% CPZ-feeding was extended to 12 weeks to analyse demyelination and gliosis in optic tract, visual cortex, pretectal nucleus, lateral geniculate nucleus, and superior colliculus under the conditions of a slow, progressive demyelination that is more reminiscent of MS. While demyelination and gliosis in the corpus callosum were observed, no significant changes occurred in any of the nuclei associated with the visual pathway, consistent with previous work by Taveggia et al. (2008). In addition, the temporal response of CPZ may be due to the differential expression profiles of oligodendrocytes in different CNS regions (Butt et al., 1995). For example, mice haplo-insufficient for type III neuregulin-1 (a growth factor that promotes oligodendrocyte and Schwann cell development) showed less myelination in the corpus callosum but no effect on the optic nerve and spinal cord, indicating regional differences in the regulation of oligodendrocyte function (Taveggia et al., 2008). Whether these factors (e.g. neuregulin-1 expression) lead to the heterogeneity of oligodendrocyte distribution in the brain vs spinal cord or in optic nerve (Ornelas et al., 2016), as well as in the response to CPZ, remains untested.
Marked proteomic changes
Demyelination in the optic nerves of MS patients is inferred by changes in visual evoked potentials (Barton et al., 2019; Langwińska-Wośko et al., 2012; Leocani et al., 2018), although early changes (i.e. subtle alterations in vision) are not reported in the clinic until after an MS diagnosis has already been made. Increased abundance of inflammatory cytokines (e.g. tumour necrosis factor-α) and the axonal marker neurofilament light chain in the cerebrospinal fluid (CSF) have been proposed as early biomarkers to diagnose MS and/or optic neuritis in patients (Olesen et al., 2019). In addition, these markers could also then be considered as pre-symptomatic/early indicators of MS risk, since optic neuritis patients often go on to develop MS (Costello, 2016; Kale, 2016). However, there exist issues with these and other canonical proteins identified to date with regard to their capacity to serve as selective biomarkers for MS (Sen et al., 2021).
In this study, optic nerve/tract tissue was used to investigate whether CPZ-induced changes in the proteomic profile could lead to the identification of candidate proteoform biomarkers and/or some indication of initiating pathological mechanism(s). Top-down proteomic analysis revealed changes in the abundance of at least 75 proteoforms (Table 1) in the optic nerve/tract tissue, using stringent criteria to ensure only the most robust identifications. To understand the function of these proteoforms and their relevance to demyelinating CNS conditions such as MS or its animal models, a comprehensive literature search was carried out using the PubMed search engine. The optic nerve/tract proteome changes identified indicate a substantial similarity with those previously identified in EAE studies (Table 1). In part, this similarity is likely attributable to the large number (15) of proteomic studies carried out using the EAE model (Alt et al., 2005; Dagley et al., 2014; Farias et al., 2012; Fazeli et al., 2013; Fazeli et al., 2010; Gonzalez et al., 2019; Hasan et al., 2019; Jain et al., 2009; Jain et al., 2012; Jastorff et al., 2009; Linker et al., 2009; Liu et al., 2007; Raphael et al., 2017; Stoop et al., 2012; Vanheel et al., 2012), whereas only 6 studies used the CPZ model (Martin et al., 2018; Oveland et al., 2018; Partridge et al., 2016; Sen et al., 2019a; Szilagyi et al., 2020; Werner et al., 2010). Importantly, none of these studies investigated the optic nerve/tract proteome but used various other biological samples including CSF, brain, spinal cord and tears. Thus, the current study is the first to investigate changes in the optic nerve/tract proteome. Although not directly comparable, the proteomic changes found here did in part overlap with previously identified changes in the abundance of certain canonical proteins; the major difference here was that use of a top-down approach enabled resolution and identification of critical proteoforms. Thus, several of the proteoform changes in optic nerve/tract correlated with canonical protein changes previously seen in other samples, including cerebellum (Hasan et al., 2019), spinal cord (Fazeli et al., 2013), CSF (Liu et al., 2009), tears (Salvisberg et al., 2014), and blood, reported in CPZ, EAE and MS studies (Berge et al., 2019; Partridge et al., 2016; Raphael et al., 2017). Notably, by separately analysing the soluble and membrane sub-proteomes, and assessing proteoforms rather than canonical proteins, we have been able to gain more detailed information not available in previous studies. For example, neurofilament medium chain was found to increase in abundance in cerebral tissue from mice fed with 0.2% CPZ (Szilagyi et al., 2020) and spinal cord from EAE mice (Farias et al., 2012; Hasan et al., 2019); the data here conclusively identify a decrease of one proteoform in the SP fraction of the optic nerve/tract tissue but a corresponding larger increase in a distinctly different proteoform in the MP sub-proteome (Table 1). While this clearly highlights the essential need to resolve and identify critical proteoforms rather than canonical protein sequences, as well as the potentially serious ramifications of not doing so, the findings may also indicate differential expression/abundance of neurofilament isoforms among different CNS regions, as likely already seen in the hippocampus and cortex (Mesulam and Geula, 1991; Nakamura et al., 1992). Moreover, Szilagyi et al. (2020) fed 8-week male C57Bl/6 mice 0.2% CPZ for four-weeks (samples analysed at 12 weeks), whereas in the current study, mice were fed 0.1% CPZ for 12 weeks, suggesting that prolonged feeding of CPZ may alter the abundance and localization of different neurofilament proteoforms. Furthermore, two other studies (Farias et al., 2012; Hasan et al., 2019) used spinal cord from EAE mice suggesting that differential disease induction (autoimmune in EAE vs metabolic changes in CPZ) may also alter the abundance of neurofilament proteoforms in the CNS. Notably, an increase of neurofilament light chain in serum has been considered as a potential pre-symptomatic biomarker of neurodegeneration in MS (Bjornevik et al., 2019; Varhaug et al., 2019); however, comparable changes also appear in other neurodegenerative disorders as well as in cases of neurotrauma raising questions as to the selectivity of this as a biomarker for MS, unless specific proteoforms are found to be altered in the different conditions (Sen et al., 2021).
Previous MS proteomic studies identified an increase in the abundance of heat shock protein 90-beta (HSP90β) in tears (Salvisberg et al., 2014) and in peripheral blood mononuclear cells (De Masi et al., 2009). However, neither of these studies provided evidence of full length, intact species. On the contrary, the current study identified a reduction of HSP90β in the optic nerve tissue. Perhaps the increased abundance of heat shock protein in tears and the circulation is a breakdown product of HSP90β from other CNS regions such as the optic nerve. This may lead to a reduction in the abundance of HSP90β in the optic nerve as observed in the current study. Of note, HSP90β exerts two potential neuroprotective roles in the CNS tissue: firstly, it prevents protein misfolding and aggregation by its chaperone activity and secondly, it inhibits multiple steps in the apoptosis process (Didonna and Opal, 2019; Lanneau et al., 2007; Mosser and Morimoto, 2004). In MS patients, these proteins (or, likely, proteoforms thereof) are overexpressed in neuronal cells and oligodendrocytes around demyelinated lesions, seemingly to protect these cells from degeneration (Cwiklinska et al., 2010; Turturici et al., 2014). Other possible reasons for the differences in the trends of protein abundance is the use of different analytical techniques, experimental model or sample analysed. For example, Szilagyi et al. (2020) and Hasan et al. (2019), respectively, used two different labelling variations of bottom-up proteomic analysis; one identified ~190 canonical proteins that appeared to change in abundance in the corpus callosum of CPZ-fed mice (Szilagyi et al., 2020), while the other identified ~1900 canonical protein changes in CNS samples from EAE mice (Hasan et al., 2019). In contrast, Farias et al. (2012) used a top-down (2DE) proteomic approach with spinal cords from EAE mice and identified alterations in 35 proteoforms (although only theoretical MW and pI were reported). Using tear samples from MS patients, Salvisberg et al. (2014) identified 42 canonical proteins differing in abundance between MS and control patients. Thus, as is often the case, bottom-up (i.e. shotgun) studies detected more apparent changes in the abundance of canonical proteins relative to proteome changes identified using a top-down (2DE) approach (De Masi et al., 2009; Farias et al., 2012; Hasan et al., 2019; Salvisberg et al., 2014; Szilagyi et al., 2020). Does this imply differences in sensitivity of the methods? This seems unlikely as bottom-up peptide analysis generally uses less stringent sequence coverage than top-down, and only identifies canonical proteins by inference to amino acid sequences. Therefore, comparable to its correlate transcriptomics, the identification of a large number of potential canonical proteins is expected using the bottom-up method. In contrast, the top-down approach uses more stringent criteria to detect changes in intact proteoforms (i.e. a more selective analysis). Therefore, top-down is likely to yield more reliable and focused data (i.e. changes in abundance of relevant species rather than total changes in a canonical protein sequence that likely represents many dozens of proteoforms) (Aebersold et al., 2018; Coorssen and Yergey, 2015; Oliveira et al., 2014; Zhan et al., 2019). Thus, the likelihood of identifying a critical change relevant to underlying molecular mechanisms or the identification of a highly selective biomarker lies in the routine, high-resolution assessment of proteoforms (Sen et al., 2021). Our previous detailed review of proteomic studies into MS found that at least nine proteoforms (of septin, tubulin, complement, glial fibrillary acidic protein, protein disulfide isomerase, calreticulin, hexokinase, aconitate hydratase and dynamin 1) consistently changed in abundance in both MS and animal models (Sen et al., 2021). Off these, five proteoforms (of septin, glial fibrillary acidic protein, aconitate hydratase, protein disulfide isomerase and tubulin) were also identified in the current study - suggesting that these may be potential early biomarkers. Opposite trends in abundance (i.e. increase or decrease) in different biological samples may also be attributed to the differential expression of proteins and distribution of proteoforms in these different samples, which depend on tissue function and the magnitude of pathological changes. For example, the spinal cords of EAE mice (i.e. the most pathologically affected CNS region in EAE) showed changes in the abundance of 1357 (uncategorised proteins were not considered) canonical proteins whereas only ~50 protein changes were found in brain stem and cerebellum (Hasan et al., 2019). The literature search was thus extended to include proteome changes in eye disorders to investigate their potential relevance to optic nerve/tract proteomic changes in CPZ-fed mice. Only three identifications (transitional endoplasmic reticulum ATPase, neurofilament light polypeptide and haemoglobin subunit alpha) were consistent with those reported in CSF samples from neuromyelitis optica (Bai et al., 2009) and optic neuritis (Olesen et al., 2019), and in tear samples from patients with dry eye disease (Jung et al., 2017). These findings suggest a link between eye diseases and the CPZ-induced optic nerve/tract proteome changes that requires further study, particularly with regard to likely early changes prior to an MS diagnosis.
Aggregation and oligomerisation proteoforms
The current study detected changes in 12 molecular chaperone proteoforms (Table 1). These changes are not unique to optic nerve/tract as they were reported previously in other CPZ (Partridge et al., 2016; Szilagyi et al., 2020), EAE (Hasan et al., 2019; Jain et al., 2012; Vanheel et al., 2012) and MS studies (De Masi et al., 2009; Salvisberg et al., 2014). Decreased molecular chaperone proteoforms have been linked to protein aggregation (Ciechanover and Kwon, 2017; Liberek et al., 2008), which can contribute to neurodegeneration and demyelination associated with human degenerative diseases (David and Tayebi, 2014; Soto and Pritzkow, 2018). It has been shown that the increased expression of chaperones (e.g. heat shock proteins) in astrocytes and neurons inhibits apoptosis of these cells in rat spinal cord (Chang et al., 2014). Moreover, optic nerve regeneration and retinal ganglion cell survival in zebrafish was promoted by HSP70 and these processes were reduced when the HSP70 was inhibited (Nagashima et al., 2011). Another study demonstrated that neurite growth of rat retinal cells increased following application of exogenous HSP αB-crystallin (Wang et al., 2012). Chaperones also modulate the cytoskeleton of neuronal cells and mediate their regeneration via enhancing intermediate filament assembly (Hirata et al., 2003). These observations indicate that the increased abundance of some chaperone proteoforms (e.g. HSP 70 kDa and HSP 47 kDa) in the present study could reflect the demands for these in the optic nerve/tract in order to reduce structural deformities and protect or regenerate neuronal tissue injured by CPZ exposure.
We (Sen et al., 2019a) and others (Liu et al., 2009) have found evidence of homo-oligomerization of proteoforms. The current data also suggest oligomerization (e.g. an approximate doubling of molecular weight; Table 1) of some species such as tubulin alpha-1C chain (50.6 kDa monomer vs 110.6 kDa experimentally observed), serine protease inhibitor A3K (47.1 kDa vs 95.9 kDa), beta-synuclein (14.1vs 27.0) and charged multivesicular body protein 4b (24.9 kDa vs 58.4 kDa). While oligomerization of tubulin alpha-1C chain, serine protease inhibitor A3K, beta-synuclein or charged multivesicular body protein 4b have not previously been identified in CPZ-fed mice, earlier studies have observed that these proteins do indeed oligomerize (Carrell et al., 2008; Mozziconacci et al., 2008). Notable in the case of beta-synuclein is that it forms hetero-oligomers with alpha-synuclein, and these were found together in the ~27kDa spot (SP12, Table 1). Oligomerization of proteins can lead to the proteins aggregation which is argued to be the cause of many neurological diseases including MS (David and Tayebi, 2014; Michaels et al., 2015). In addition, gel shifts in MW and pI relative to theoretical values (i.e. of the amino acid sequence only) were also observed for some of the identified species (Table 1, Figure 3C) such as neurofilament medium polypeptide (96.0/4.6 vs 117.3/4.7), myelin-associated glycoprotein (70.1/4.9 vs 90.2/4.4), ATP synthase subunit alpha (59.9/9.6 vs 66.3/6.8), cytochrome c oxidase subunit 5B (14.1/8.5 vs 17.5/5.9), consistent with post-translational modifications and thus the identification of select proteoforms (Rabilloud and Lelong, 2011; Sen et al., 2019a).
Structural proteoforms
In the current study, major changes in cytoskeleton proteoforms such as actin and tubulin were detected in the optic nerve/tract. The alterations of these structural proteoforms in this slow progressive demyelinating model may reflect early changes in cellular structure as a result of CPZ-feeding. Importantly, the present study identified an increase in the abundance of intermediate filament proteins such as glial fibrillary acidic protein and vimentin, suggesting that astrocytes are activated in the optic nerve/tract (Sofroniew and Vinters, 2010). However, histological examination of glial fibrillary acidic protein in the optic tract did not detect any significant difference in glial staining intensity relative to the Ctrl group. This suggests that 2DE is more sensitive in revealing early proteoform changes, perhaps due in part to the larger amount of sample analysed relative to tissue sections. These proteomic data are thus indicative of notable structural disturbances in the optic nerve/tract following CPZ-feeding that might contribute to conditions such as MS (Love, 2006).
Notably, actin and tubulin are often used as ‘house-keeping’ loading controls in assays such as western blots, in the very risky and unfounded hope that they do not change under the conditions of the experiment (Zhang et al., 2012). The marked changes in abundance of these structural proteoforms in the current study argue strongly against this practise and are consistent with other reports cautioning against this (Eaton et al., 2013; Nie et al., 2017). Changes in the abundance of such inappropriately named ‘house-keeping’ proteins have also been shown in other proteomic studies, including CPZ (Sen et al., 2019a; Werner et al., 2010), EAE (Farias et al., 2012; Fazeli et al., 2010; Hasan et al., 2019) and MS (De Masi et al., 2009; Dumont et al., 2004; Hammack et al., 2004; Liu et al., 2009). Therefore, total protein concentrations must be assessed in each sample and equal concentrations used for analysis (Almuslehi et al., 2020; Hu et al., 2016; Noaman and Coorssen, 2018; Sen et al., 2019a).
Metabolic proteoforms
Another key finding in this study was the detection of changes in the abundance of numerous proteoforms associated with metabolic and mitochondrial functions in the optic nerves/tracts. Metabolic dysregulation can lead to demyelination in the CPZ model (Caprariello et al., 2018; Sen et al., 2019a; Teo et al., 2021; Werner et al., 2010), and is thus hypothesized as an early dysfunction leading to MS. Despite finding changes in metabolic proteoforms, no demyelination was detected in the optic nerve/tract. This may be attributed to the highly compact myelin structure in the optic nerve tissue which increased the intensity of silver staining (i.e. resulting in saturation) and thus may have limited the detection of demyelination (Sen et al., 2020a). This interpretation is supported by observations in the diphtheria toxin model, in which alterations in axonal structure were observed by ultrastructural analysis despite no overt demyelination (Pohl et al., 2011). Likewise, reduction of myelin basic protein in the optic nerve may result from metabolic turnover (Namekata et al., 2014). Previously, we reported that ~50% of proteoforms that changed in abundance in whole brain samples were metabolic (Sen et al., 2019a) whereas here, in the optic nerve/tract, metabolic proteoforms constituted only ~24%. Although an indirect comparison, this suggests less metabolic disturbance in the optic nerve/tract tissue relative to the brain, and thus demyelination is readily evident in the corpus callosum but not in the optic nerve/tract. Additional proteomic studies are thus required for direct comparison (optic nerve/tract vs corpus callosum) to investigate the threshold of changes in metabolic proteoforms that are necessary for demyelination in optic nerve.
It might be argued that the dosage of CPZ (e.g. 0.2 vs. 0.1%) plays a significant role in changing the profile of metabolic proteins. This seems less likely since a comparable level of demyelination occurred by feeding mice with 0.2% CPZ for 5 weeks or with 0.1% for 12 weeks (Sen et al., 2019a). Likewise, feeding mice with either 0.1% or 0.2% CPZ for 2 weeks yielded comparable demyelination and glial activation in the corpus callosum (Almuslehi et al., 2020). Therefore, it is expected that changes in the abundance of metabolic proteoforms in the optic nerve/tract following 0.1% CPZ-feeding for 12 weeks may be comparable to those seen in 0.2% CPZ-feeding for 5 weeks. Nonetheless, this does not rule out progressive but localized effects of CPZ in different areas of the CNS (including the optic nerve/tract). This study also detected changes in six proteoforms identified in our previous proteomic analysis of brain (i.e. creatine kinase U-type, neurofilament light polypeptide, glial fibrillary acidic protein, ATP synthase subunit alpha, aconitate hydratase, charged multivesicular body protein 4b) (Sen et al., 2019a). Among these, only 3 (creatine kinase U-type, ATP synthase subunit alpha and aconitate hydratase) are recognized to be directly involved in metabolism. Creatin kinase exerts a variety of bioenergetic and neuroprotective properties in CNS and retinal neurons including buffering and stabilization of intracellular energy reserves, neutralizing calcium ion fluxes, inhibition of mitochondrial permeability, and counteracting intracellular oxidative stress (Beal, 2011; Sia et al., 2019). ATP synthase subunit expression is upregulated during ocular hypertension and it is associated with increased ATP concentration in the retinal ganglion of rats (Kanamoto et al., 2019). An in vivo study showed that aconitate hydratase activity increased in optic nerve tissue one day following traumatic injury (Cummins et al., 2013). According to these observations, the increased abundance of creatine kinase U-type, ATP synthase subunit alpha and aconitate hydratase in the optic nerve/tract following CPZ-feeding is indicative of mitochondrial dysregulation. The main cluster of protein-protein interactions in our previous study was identified as metabolic (with malate dehydrogenase, succinate dehydrogenase, aspartate aminotransferase and oxoglutarate dehydrogenase protein (Sen et al., 2019a)), whereas these proteoforms were not identified in the current work. This suggests that minimal changes in the complement of proteoforms, and their interaction with key metabolic proteoforms (e.g. malate dehydrogenase, aconitate hydratase), are required to initiate metabolic disturbances that can induce downstream effects (e.g. demyelination of optic nerve/tract) in the CPZ-fed mice.
Biological functions and interactions
Importantly, from the bioinformatic (UniProt, PANTHER and STRING) and literature (PubMed) analyses, complex linkages among identified proteoforms were indicated (Figure 4). For example, the literature search linked proteoforms to diverse functions including structural, metabolic, and axonal, suggesting CPZ-induced changes at multiple functional levels. Characterizing such potential interactions is important to understanding the underlying dysregulation of biological processes (Berggard et al., 2007; Sen et al., 2019a). Previous studies have shown that 50-80% of proteins undergo protein-protein interactions (Asgarov et al., 2021; Berggard et al., 2007; Dagley et al., 2014; Sen et al., 2019a). These observations are consistent with the idea that while proteins/proteoforms function as monomers, they also interact to form complexes in order to exert their molecular actions (Berggard et al., 2007; De Las Rivas and Fontanillo, 2010; Keskin et al., 2016; Sen et al., 2019a). Consistent with these observations, the current data suggest that ~90% of the identified proteoforms were interconnected, indicating molecular cross-talk (Asgarov et al., 2021; Sen et al., 2019a). For example, myelin-associated glycoprotein, a molecule located in the axonal plasmalemma (inner aspect of the myelin sheath) of oligodendrocytes, is said to interact with axonal neurofilament microtubule proteins, leading to phosphorylation of neurofilament microtubules and modulation of axonal diameter (Nguyen et al., 2009). Another example is alpha-internexin, a neuronal protein implicated in neurodegenerative diseases, that cannot exert its effects independently but functions in association with other neurofilament proteins such as neurofilament medium chain, and this interaction is necessary for the axonal transport of neurofilament medium chain in CNS and optic nerve axons (Yuan et al., 2006). Overall, the protein-protein interaction analysis suggested that alterations in the abundance of the identified proteoforms in optic nerve/tract are likely interrupting molecular cross-talk amongst these species, thereby disrupting associated biochemical reactions and perhaps thereby contributing to disorders such as MS.
Limitations and future work
Despite our best efforts to minimize the experimental variables, we acknowledge certain drawbacks in the current study. Firstly, since 0.2% CPZ-feeding for 12 weeks can reduce visual function in mice (tested using multifocal electroretinograms (Namekata et al., 2014)), it would be important to correlate molecular changes with visual status in future studies. Secondly, this study relied on only one time point (i.e. 12 weeks) of CPZ-feeding. Therefore, future studies should use a temporal analysis to determine the earliest point at which proteoform changes occur as they may identify the triggers for the cascade of molecular alterations that we identified here after 12 weeks of slow demyelination. Finally, the bioinformatic analyses employed (e.g. PANTHER, STRING) are based primarily on literature reports concerning canonical proteins rather than specific proteoforms or oligomers, and this may also influence interpretation of the data in terms of canonical versus actual proteoform changes.