Introduction
Multiple sclerosis (MS) is a central nervous system (CNS) chronic inflammatory demyelinating disease which targets oligodendrocytes and myelin. Oligodendrocytes are the CNS cells that stain most robustly for iron [
63], and are also highly susceptible to inflammatory-mediated injury [
73]. Metals are essential for the synthesis, stability, and maintenance of myelin [
14,
29,
36,
63,
64], and are required for normal CNS functioning [
36,
72]. Metal dyshomeostasis causes myelin breakdown in Huntington disease [
4], and hyperzincemia-induced copper deficiency [
51]. This suggests an important role for iron and other metals, such as zinc, in MS pathogenesis. Iron levels in particular must be balanced delicately since loading of cells with iron may lead to the formation of reactive oxygen species and oxidative damage of DNA, lipids, and proteins.
Oxidative stress, dysregulation of metals and metalloproteins in the serum, cerebrospinal fluid or brains of MS patients, and iron liberation in the CNS extracellular space have been linked to the conversion of isolated demyelinating episodes to clinically definite MS, as well as to MS progression via amplification of demyelination and neurodegeneration [
22,
27,
31,
56]. While some magnetic resonance imaging (MRI) sequences have been used to image iron in vivo in MS [
1,
3,
26,
39,
75,
76], histochemistry is the gold standard to localize metals in tissue. However, iron histochemistry detects only nonheme iron [
38], and histochemistry for zinc has poor sensitivity, specificity and is complex [
15,
32]. Each method employs a different chemical reaction, and therefore cannot be used on the same tissue section. X-ray fluorescence imaging (XFI) is element-specific, quantitatively detects all chemical forms of any metal and simultaneously maps multiple metals, therefore addressing the limitations of histochemistry [
45,
46,
48,
49,
52]. This is the first systematic synchrotron X-ray fluorescence study to compare the distribution and quantification of iron and zinc in MS lesions to the surrounding normal appearing white matter (normal appearing WM) and periplaque white matter (periplaque WM) from a given patient, and to assess the involvement of these metals in MS lesion pathogenesis.
Discussion
Our study describes distinct and heterogenous patterns of iron and zinc across different chronic MS plaque types. With few remarkable exceptions these metals do not accumulate in chronic MS lesions. Using cutting edge synchrotron techniques that are sensitive and specific to detect metals [
52], we report for the first time that astrocytes in large astrogliotic regions in a subset of smoldering and inactive plaques accumulate iron, and safely store it as ferrihydrite in ferritin. Furthermore, we provide preliminary insights into the complex and dynamic relationship of cell-specific iron loading and release and its impact on injury and repair over the chronic MS course.
Smoldering plaques represent the most notable example of iron accumulation within MS plaques. As smoldering plaques are only found among progressive MS patients [
19], they are an important plaque type to better understand. Iron has been reported to accumulate in the microglia/macrophages, forming the smoldering rim [
1,
22,
39] of these plaques. We, however, find that not all smoldering lesions contain iron-rich microglia/macrophage rims. This may relate to varying degrees of smoldering activity at the plaque edge. We further show that even when an iron-rich rim is present, its iron content is not higher than the normal appearing WM iron, and is only visible because of the iron-poor rim-adjacent periplaque WM separating the two regions. The distribution of iron differs between these two regions: oligodendrocytes and myelin in normal appearing WM stain most intensely for iron [
8,
63], whereas iron in smoldering rims is localized to dystrophic and normal microglia/macrophages [
22], as well as to astrocytes.
It has been suggested that the destruction of iron-loaded oligodendrocytes in MS active lesions may release iron extracellularly, where it is subsequently picked up by microglia/macrophages. Degeneration of these iron-loaded microglia/macrophages may induce a second wave of iron release, with subsequent axonal iron accumulation, oxidative damage and neurodegeneration [
22]. While it is known that reactive astrocytes incorporate iron [
22], we describe for the first time that reactive astrocytes forming large areas of astrogliosis accumulate iron in a subset of MS patients. These iron-rich gliotic patches may represent another potential protective barrier the brain mounts when faced with degeneration of iron-loaded microglia/macrophages. This is further supported by our finding of astrocytes in close contact with, and incorporating iron-reactive fragments from iron-loaded macrophages.
Moreover, we observe the highest iron accumulating reactive astrogliotic regions in inactive centers of smoldering lesions, further suggesting that reactive astrocytes may play an important role in iron uptake and storage in chronic MS lesions. This is not surprising since upon inflammatory activation astrocytes elevate their capacity to incorporate iron [
42,
54,
73], and their ability to resist iron-dependent oxidative stress [
35]. Immunohistochemistry confirms that the iron-loaded astrocytes within these gliotic patches are in fact immunoreactive for iron storage proteins, consistent with their ability to safely incorporate large amounts of iron in ferritin [
16]. Furthermore, XANES analysis, although limited to two cases, indicates at least half of the iron in both the smoldering rims and high-iron gliotic regions within the inactive centers of these plaques consists of the highly ordered ferrihydrite of ferritin. Surprisingly, a high proportion of this accumulated iron represents goethite, the predominant iron bio-mineralization component of haemosiderin in thalassaemia [
66]. While the mechanism of goethite formation is undefined, it is unlikely that it forms from ferritin’s ferrihydrite [
10,
11]. Iron in goethite is less soluble and released less readily than iron of ferritin [
2,
40,
67]. Because XFI cannot distinguish between cell types, it is conceivable that astrocytes pick up the goethite already biomineralized by macrophages, or they biomineralize the iron picked up from macrophages. In either case an important proportion of iron is stored as goethite that is even less bioavailable and less toxic to cells than iron of ferritin.
The fate of iron-loaded astrocytes in MS lesions is unknown. Astrocytes resist iron overload until their antioxidant potential is exhausted [
35]. It is possible that while astrocytic iron accumulation is protective in the short term, incorporating increasing amounts of free iron over time as a consequence of continuous inflammatory activity with ongoing destruction of iron-loaded oligodendrocytes and macrophages will eventually exhaust the astrocytes’ antioxidant defenses, thereby leading to their death. This may explain the presence of tissue rarefaction observed surrounding high iron astrocytic patches within the inactive center of smoldering plaques. While iron is decreased in these tissue-rarefied regions, its chemistry more closely resembles that of the periplaque WM with half sheltered in ferritin and half associated with heme proteins. The latter may be hemoproteins that are involved in mitochondrial respiration, suggesting a potential metabolic reversal of astrocytes from an iron-storage to a normal phenotype. Alternatively, the heme observed in rarefied gliotic low iron regions may consist of brain globins, which are neuroprotective heme-containing proteins [
5,
53]. Since XANES analysis only interrogates the heme iron, further studies will need to elucidate their exact nature, and contribution.
Iron efflux from astrocytes is known to be important for remyelination [
60]. The iron-induced oxidative destruction of iron-loaded astrocytes in a noninflammatory setting could provide a double benefit: the removal of the glial scar that impairs the migration of oligodendrocyte precursor cells into lesions [
25], and the release of iron crucial for efficient remyelination [
63]. Alternatively, astrocytes can provide iron to oligodendrocytes through the ferroportin-ceruloplasmin system [
22,
41,
70]. However, in the setting of even a minimal inflammatory milieu, as expected in long standing chronic MS [
18], iron liberation from astrocytes would serve to augment oxidative damage rather than promote remyelination.
Age and disease duration may contribute to the development of iron-rich rims or iron-rich gliotic patches observed in a subset of smoldering lesions. All but one of the smoldering lesions with an iron rim and all smoldering plaques with iron-rich cores were observed in younger patients (<50 years old) with less than 15 years disease duration. It is possible that younger patients during earlier disease stages have more active inflammatory disease resulting in more pronounced oligodendrocyte destruction with subsequent iron release compared to older longstanding patients with less active disease [
18]. Alternatively, the variable presence of iron rims may reflect differences in macrophage polarization [
55] or changes in macrophage polarization as a result of iron loading [
39].
While the majority of chronic inactive lesions showed iron loss, one chronic plaque from an elderly patient [
62] demonstrated concentric rings of iron increase and loss. Iron concentration within this lesion resembled the inactive iron-rich gliotic centers observed among smoldering plaques with an accumulation of iron within reactive astrocytes immunoreactive for both FTH and FTL, higher ferrihydrite concentrations in iron-rich versus iron-poor rings, and a concomitant decrease in heme iron. These observations suggest that in some chronic lesions iron safely accumulates in ferritin in astrocytes, but not microglia.
Interestingly, in the periplaque WM of most smoldering and chronic inactive plaques, there is a gradient of iron loss towards the lesion edge, with three smoldering plaques demonstrating a clear ring of iron loss adjacent to the smoldering rim. Iron in the periplaque WM is significantly lower than iron in the normal appearing WM. While iron histochemistry shows that iron-positive myelinated axons are present in the periplaque WM, FTH-immunoreactive oligodendrocytes in the periplaque WM do not stain for iron. This suggests that oligodendrocytes in the periplaque WM are able to dispose of their iron, a finding compatible with previous studies reporting active iron export from periplaque WM oligodendrocytes [
22]. Iron-loaded oligodendrocytes are more vulnerable to cytokine toxicity than iron-depleted oligodendrocytes [
73]. Therefore, this disposal of iron by periplaque WM oligodendrocytes may be a protective, but transient, mechanism oligodendrocytes employ when faced with the advancing front of inflammation [
73]. Microglia may pick up the oligodendrocyte-released iron, and upon activation and in the context of iron loading, increase their release of proinflammatory cytokines, thereby switching from a trophic to a toxic phenotype [
74]. The presence of iron-negative reactive astrocytes in the periplaque WM indicates that CNS may be prepared to deal with the potential iron release that ensues.
We have also observed iron-negative FTH immunoreactive oligodendrocytes within inactive demyelinated white matter and cortical lesions. While their number is decreased compared to the periplaque and normal appearing WM, these oligodendrocytes are characterized by normal morphology, suggesting they may be capable of surviving in lesions despite the absence of iron.
A lack of iron within oligodendrocytes may impair their remyelinating capacity [
61,
63]. Previous studies have suggested that iron does not accumulate in remyelinated plaques [
22], and indeed iron was decreased in remyelinated lesions when analyzed as a group. However, when examined individually, we found a heterogeneous concentration of iron with some remyelinated plaques displaying increased iron. Oligodendrocytes within these shadow plaques stained robustly for iron, and were characterized by larger nuclei, compatible with an immature phenotype [
30]. At the other end of the spectrum, we observed remyelinated lesions where oligodendrocytes with normal morphology did not stain for iron. These observed differences may be explained by differences between early versus late remyelination. When remyelination is ongoing, an increased amount of iron is required, whereas when remyelination is complete, the iron content may decrease to levels at or below those of the normal appearing WM [
61,
63].
In addition to iron, zinc is important in the structure and compaction of myelin [
14,
29,
64]. Zinc is decreased in most lesions and its distribution parallels that of myelin. Demyelinated white matter lesions are devoid of zinc, and zinc decreases in the periplaque WM showing myelin pallor. Only three white matter lesions had a moderate zinc increase, but their pathologic correlate remains unknown. A single smoldering plaque showed increased periventricular zinc, and two inactive lesions were characterized by rings of zinc at their borders. Cortical lesions showed reduced, but not absent, zinc, likely due to the fact that most cortical zinc is concentrated in neurons [
17]. Zinc was increased in one leukocortical lesion where it localized to corpora amylacea [
65].
We also analyzed the relationship between iron and zinc with age and MS disease duration. As part of normal aging, iron accumulates in microglia and astrocytes [
9], presumably due to a cytokine-induced phenomenon [
13]. In MS, due to the inflammatory environment [
18], iron accumulation is likely accelerated and present already at disease onset. It has been reported that iron decreases in the normal appearing WM of MS patients with increasing disease duration, presumably due to the destruction of iron-loaded oligodendrocytes [
22]. We observed that iron content also decreases within MS plaques with increasing age. In contrast, neither disease duration, nor age was related to zinc content within MS plaques.
Whether there is a role for iron chelation in MS remains controversial [
12,
61,
69] and is based on the premise that iron accumulates in lesions. We have generally observed the opposite, as chronic MS plaques tended to be deficient in iron. Chelation of these minimal amounts of iron may suppress oligodendrocyte metabolic activity and induce cell death [
73]. Chelation of iron from remyelinating plaques could also be detrimental, given the potentially trophic effect of iron and ferritin on oligodendrocytes and myelination [
58,
74]. However, oligodendrocytes are still functional in normal white matter [
9], and survive in chronic and remyelinated lesions, as well as periplaque WM despite showing no stainable iron. Since iron chelation may not negatively affect all oligodendrocytes, it may be considered for the smoldering and chronic MS lesions where iron accumulates in astrocytes, macrophages or microglia. Removing the iron from iron-filled astrocytes could potentially increase their ability to accumulate more iron, thereby prolonging their survival. Iron removal from iron-rich activated macrophages and microglia in smoldering rims may also decrease the release of inflammatory cytokines and protect oligodendrocytes [
74]. Further studies will need to establish if there is a role for iron chelation in MS given its potential detrimental and beneficial effects.
The relationship between iron metabolism diseases and MS is currently unknown. Although genetic hemochromatosis does not increase MS severity [
57], a recently developed hemochromatosis mouse model indicates that brain iron accumulation alters the myelin-related transcriptome [
23,
24]. While this relationship may be elucidated by EAE induction in this model, the relevance to MS remains to be determined.
Our study highlights the limitations of iron histochemistry [
38]. Although iron staining approximates well the XFI iron in iron-rich areas, it grossly underestimates iron in iron-poor lesions, normal appearing WM and periplaque WM. This is explained by the XFI’s high sensitivity and specificity for iron [
52], and by XANES showing that iron-rich areas mostly contain ferric non-heme iron, while iron-poor regions contain more heme iron. Another limitation which impacts both iron histochemistry and XFI is the use of formalin-fixed paraffin-embedded tissue. Formalin fixation is known to leach metals from tissues, although the extent of leaching is debatable [
6,
7,
59]. Paraffin embedding may also accentuate this phenomenon, and further oxidize metals, possibly accounting for why no ferrous iron was detected [
21]. The limited number of cases from which XANES spectra have been collected is another study limitation. Nevertheless, these data are valuable and complementary to ferritin immunohistochemistry, with both approaches indicating that most iron is safely stored in ferritin. Further studies need to establish whether our findings are broadly applicable to iron accumulation in a larger cohort of smoldering and chronic plaques.