Oxidative tissue injury in multiple sclerosis is only partly reflected in experimental disease models.

Recent data suggest that oxidative injury may play an important role in demyelination and neurodegeneration in multiple sclerosis (MS). We compared the extent of oxidative injury in MS lesions with that in experimental models driven by different inflammatory mechanisms. It was only in a model of coronavirus-induced demyelinating encephalomyelitis that we detected an accumulation of oxidised phospholipids, which was comparable in extent to that in MS. In both, MS and coronavirus-induced encephalomyelitis, this was associated with massive microglial and macrophage activation, accompanied by the expression of the NADPH oxidase subunit p22phox but only sparse expression of inducible nitric oxide synthase (iNOS). Acute and chronic CD4+ T cell-mediated experimental autoimmune encephalomyelitis lesions showed transient expression of p22phox and iNOS associated with inflammation. Macrophages in chronic lesions of antibody-mediated demyelinating encephalomyelitis showed lysosomal activity but very little p22phox or iNOS expressions. Active inflammatory demyelinating lesions induced by CD8+ T cells or by innate immunity showed macrophage and microglial activation together with the expression of p22phox, but low or absent iNOS reactivity. We corroborated the differences between acute CD4+ T cell-mediated experimental autoimmune encephalomyelitis and acute MS lesions via gene expression studies. Furthermore, age-dependent iron accumulation and lesion-associated iron liberation, as occurring in the human brain, were only minor in rodent brains. Our study shows that oxidative injury and its triggering mechanisms diverge in different models of rodent central nervous system inflammation. The amplification of oxidative injury, which has been suggested in MS, is only reflected to a limited degree in the studied rodent models. Electronic supplementary material The online version of this article (doi:10.1007/s00401-014-1263-5) contains supplementary material, which is available to authorized users.

For direct comparison with animal models, as performed in our present study, we concentrated on those markers, which gave the most distinct and reproducible results in human MS tissue and also showed reproducible and distinct staining with comparable sensitivity to human tissue in inflamed rodent brains (E06 for oxidised phospholipids; p22phox for NADPH oxidase, inducible nitric oxide synthase (iNOS) and iron reactivity with the DAB-enhanced Turnbull blue reaction, see also supplementary material and methods). To ensure comparability of the staining reaction in human and rodent tissue, we always included human MS and/or stroke lesions as positive control, when rodent tissue was stained.
Relevant for our comparison with rodent animal models it has to be emphasized that extensive oxidative injury, reflected by the presence of oxidised lipids and DNA and the expression of --NADPH oxidase, as well as iron accumulation has been already observed in patients with acute MS with a disease duration between 14 days and 2 months [10,11] . and aged (14 months) female Lewis rats. In both groups of animals, disease started 3 to 4 days after T cell transfer with inflammation, which was most pronounced in the lumbar spinal cord. The peak of disease and inflammation was reached on day 6 and declined during the following days. CNS inflammation was characterised by T cell infiltration associated with pronounced microglial activation and peripheral blood monocyte/macrophage recruitment, as shown before in bone marrow chimeric animals [12]. Clinical disease was mainly associated with axonal injury, whereas demyelination was sparse [1]. Animals recovered completely between day 12 and 15 and pathology at this time point revealed only few perivascular inflammatory infiltrates. In the present study, lumbar spinal cord tissue was analysed 6 and 12 days after T cell transfer.
Chronic relapsing EAE in C57BL/6 mice induced by active sensitization with MOG 35-55 peptide: Young female adult C57BL/6 mice were actively immunised with 40 g of rat MOG  peptide in complete Freund's adjuvant. Mice also received an intraperitoneal injection of 200 ng of pertussis toxin (Sigma-Aldrich) on days 0 and 2 post immunization [24]. Clinical disease started at day 10 after immunization and gradually increased over the next 20 days. During the early disease stages (days 10-18), pathology was dominated by perivenous inflammation by T cells associated with microglial activation and macrophage infiltration. Confluent demyelinating destructive lesions developed around day 20 and increased in size and incidence until day 34. Demyelination was associated with extensive axonal injury and loss. For the present study, animals were sampled for pathological analysis on days 21, 27 and 34.

Chronic relapsing EAE in DA rats induced by active sensitization with MOG 1-125 peptide:
Chronic relapsing EAE was induced by active sensitization with 50 g myelin oligodendrocyte glycoprotein (MOG 1-125 ) in complete Freund's adjuvant in female and male dark agouti (DA) rats [22].
In the first bout of the disease, perivenous and subpial inflammation by T cells, macrophages and granulocytes was associated with perivenous and subpial demyelination. In the chronic stage, besides initial perivenous lesions as described above, large confluent plaques of demyelination with loss of oligodendrocytes and a variable extent of axonal injury were observed. Furthermore, within the same animal actively demyelinating lesions were seen side by side with inactive demyelinated plaques. In this model, inflammation is triggered by encephalitogenic MOG-reactive CD4 + T cells.
Demyelination is induced by demyelinating anti-MOG antibodies, which destroy myelin in parallel through complement activation and antibody dependent cellular cytotoxicity [18,19]. For our study, we selected from a large pool of pathological material animals from chronic relapses (40 to 90 days after sensitization), which all contained actively demyelinating lesions (for definition and staging see below) together with early perivenous lesions and inactive demyelinated plaques.

Acute CD8 + T cell-mediated EAE
Acute demyelinating encephalomyelitis was induced by passive transfer of 3x10 7 hemagglutinin-reactive CD8 + T cells into female transgenic BALB/c mice expressing hemagglutinin specifically in oligodendrocytes [21]. In this model, the onset of disease is variable, ranging from day 9 to day 30 after transfer. At any time point within this time window, new actively demyelinating lesions with partial axonal preservation may develop. Thus, at later time points active and inactive lesions are seen in the central nervous system side by side. Active demyelinating lesions are characterised by inflammation, with dominant infiltrates by CD8 + T cells and numerous ramified phagocytic cells showing a microglial morphology. Active demyelination is associated with infiltration of the tissue with granzyme B and CD8 + T cells, which closely attach to oligodendrocytes.
Oligodendrocyte destruction is mediated through apoptosis, followed by demyelination [21]. In addition, abundant activated microglia and macrophage-like cells are present in the lesions, which may further contribute to tissue injury. In our present study, we selected brain and spinal cord tissue, containing actively demyelinating lesions from mice sampled between 9 and 28 days after T cell transfer.

LPS injection-induced inflammation
Acute demyelinating myelitis was established by focal injection of 0.5 l lipopolysaccharide (LPS; 100 ng/l in saline) into the dorsal column of the spinal cord of adult male Sprague Dawley (SD) rats [5,16], leading to focal inflammation in the dorsal column 12 to 24 hours after LPS injection.
Inflammatory infiltrates at this stage of lesions were composed mainly of granulocytes, as well as some T cells and macrophages. This inflammatory phase was followed by a demyelinating phase giving rise to focal demyelinated lesions with relative axonal preservation. This was associated with massive microglia activation and macrophage recruitment starting around day 5 to 7 after focal LPS injection. The peak of active demyelination occurred between 9 and 12 days after injection. For our studies, tissue derived from the inflammatory (0.8-3 days after injection) as well as the demyelinating phase of the disease (5-15 days after injection) was used. Control animals were injected with saline instead of LPS and sampled likewise.
Profound active demyelination was found 35 days after the initiation of the cuprizone diet, which involved large parts of the corpus callosum, the periventricular white matter and also extended into the deeper cortical layers. In our present study, this stage of active demyelination was sampled for further analysis.
This resulted in an acute inflammatory disease mainly affecting the grey matter of the CNS with pronounced virus antigen expression in neurons, astrocytes and oligodendrocytes. Other animals showed a late-onset type of disease, starting around day 30 after virus infection. Pathologically this was reflected either by a chronic panencephalitis affecting grey and white matter, or a chronic inflammatory demyelinating encephalomyelitis mainly involving the brain stem and the spinal cord white matter. The virus antigen was primarily seen in glia cells at these stages of the disease [27]. In the present study, we analysed animals with active lesions from chronic disease stages [2] .

Control animals
In order to validate the stainings, brain, spinal cord and lymphatic tissues of young and aged Lewis rats, SD rats (untreated and saline injected), DA rats and C57BL/6 mice without inflammatory lesions were analysed.

Selection of experimental animals and lesions:
Oxidative injury in multiple sclerosis is associated with inflammation, microglia activation and active demyelination or neurodegeneration and is most pronounced in brain areas with active lesions or ongoing axonal or neuronal degeneration [6,7,10]. To test its potential pathogenic role for demyelination and neurodegeneration in the respective animal models, we concentrated on lesions with inflammation, mediated by different components of the immune system and on lesions with active demyelination or axonal and neuronal injury. Since in individual animals such lesions are present in variable incidence in the chronic stage of the disease, we selected from a much larger pool of animals and tissue blocks those, which still showed signs of active demyelination (for definition see below) or axonal injury, determined by the presence of axons with accumulation of amyloid precursor protein as a sign of acutely impaired fast axonal transport. As active lesions in these animals occur side by side with inactive lesions this sampling strategy also allowed to compare different lesion stages, such as initial lesions, classical active lesions and inactive lesions. Thus, our selection strategy for further detailed analysis was based on the presence of the entire pathological spectrum of different lesion stages rather than on the time points after disease induction.

Definition of lesion stages and activity of demyelination and neurodegeneration:
For classification of lesional activity, we applied the criteria extensively defined in MS lesions [4,15]. Actively demyelinating lesions were identified by the presence of abundant numbers of macrophages containing myelin degradation products. In early active lesions, the myelin degradation products showed immune-reactivity for all myelin proteins, including myelin oligodendrocyte glycoprotein. Furthermore, early active lesions were surrounded by a rim of microglia activation, only few of them containing sparse myelin protein-reactive intra-cytoplasmic granules (initial lesions, [15]). These lesions showed initial stages of demyelination and oligodendrocyte apoptosis (prephagocytic lesions; [3]). Late active lesions contained macrophage inclusions reactive for major myelin proteins, such as myelin basic protein or proteolipid protein. In later stages of lesion development, myelin proteins were degraded in macrophages, which were still abundantly present in the lesions but only contained lipid remnants of myelin. Completely inactive lesions were sharply demarcated from the surrounding normal appearing white matter and showed complete demyelination and astrocytic scaring in the absence of macrophages with myelin degradation products. Acute axonal injury, reflected by APP positive axons, was profound in early and late active lesions but only sparse in the inactive plaque center.
In patients, who died during the progressive stage of the disease, such classical active lesions were rare. The majority of lesions were chronic inactive plaques. In addition, however, a variable number of lesions showed changes, suggesting a slow expansion of pre-existing lesions [14]. Such lesions were characterised by a dense rim of activated microglia at the lesion edge together with some cells with macrophage phenotype, which contained early myelin degradation products. The number of such macrophages with early myelin degradation products was variable between different plaques, apparently reflecting the degree of ongoing demyelinating activity [14,15].
Lesion staging within the multiple sclerosis tissue blocks was first performed in studies on inflammation [9] and demyelination [14] by the first authors and the senior author of the respective publications. In the subsequent studies on oxidative injury and iron deposition, the classification of the activity stage of lesions was re-evaluated by the principal investigators (LH and SH), since the exact areas of activity may change in some blocks, when multiple serial sections are taken. Thus, each lesion was classified by at least 4 to 5 investigators. Active inflammatory demyelinating lesions in experimental rodent models showed a lesion architecture, which was comparable to that in human MS lesions, and thus allowed to apply the same criteria for lesional staging. The respective blocks were first selected according to lesional activity documented in the neuropathological protocols from the original studies. Then, new sections were cut and the original classification was confirmed by CS and HL in the present study.

Oxidative damage
In vitro oxidation of native fresh frozen tissue sections Although the antibody recognising the E06 epitope (oxidised phospholipids) had been used before in rodent tissue, we tested if the epitope could be produced and detected with similar sensitivity and specificity in rodent tissue in our experimental settings (Supplementary Fig.1). Wildtype Lewis rats were perfused with phosphate buffer and brains were snap-frozen in pre-cooled isopentane. Sections (7 µm) were cut and stored at -20 °C. For producing the E06 epitope, we used iron-induced oxidative damage in vitro. For Fe 2+ /ascorbate and H 2 O 2 -induced lipid peroxidation [20,23], the sections were treated over night at 37 °C with 20 µM FeSO 4 (Merck) and 1 mM ascorbate (Sigma Aldrich) and 500 µM tbH 2 O 2 (Luperox Sigma Aldrich). Prior to staining for E06, slides were fixed with acetone/PFA and the staining was performed as described in the materials and methods part of the main manuscript.

Iron
Validation of DAB-enhanced Turnbull blue staining (TBB) with ferrozene assay For methodological reasons, we compared the iron content of rat cerebella at different ages using two different techniques: the DAB-enhanced Turnbull blue staining (TBB) as an established histological detection of iron [17] and the ferrozene assay [8] for biochemical iron quantification of fresh tissue. We stained cerebella of rats from the age of 2 to 20 months using the TBB protocol and analysed the optical density of pictures taken from the dentate nucleus, which shows iron accumulation in the TBB staining with increasing age. Similarly, we quantified iron of whole rat cerebella using the ferrozene assay. We found a linear correlation (r = 0.485, p < 0.001) when we compared the optical density of TBB stained slides with the iron concentration in cerebella ( Supplementary Fig. 2).

Iron quantification
Rats aged from 2 to 20 months were sacrificed and perfused with phosphate buffer.
Cerebella were homogenised in sodium potassium phosphate buffer (5 mM, pH = 7.4). The iron content of rat cerebella was determined using the colorimetric ferrozene assay [8]. Iron was released from proteins and stable iron complexes by acidic permanganate treatment at 60 °C for 2 hours.
Subsequently, iron was reduced by ascorbic acid to Fe 2+ and quantitatively complexed by ferrozene.
In order to catch interfering copper ions, neocuproine was added to the reaction. Absorption was measured at 540 nm using a Powerwave X-340 plate reader. Data were normalised to protein concentration determined by total protein measurement using a Nanodrop 2000 spectrophotometer. Iron accumulation in rat brain tissue detected by the DAB-enhanced Turnbull reaction correlated significantly with the iron content in the tissue determined by biochemical analysis of total iron content ( Supplementary Fig.2).

Supplementary Figures
Supplementary Figure 1: Optical density of E06 after in vitro oxidation. The epitope of E06 could be detected in rat tissue after in vitro oxidation of frozen native brain sections using iron containing free-radical generating systems. After staining for E06, the optical density of oxidised phospholipids in iron-treated sections was higher than in control sections (n = 6, p ≤ 0.001, indicated by ***)