Introduction
Neuromyelitis optica (NMO) is a chronic inflammatory disease of the central nervous system, resulting in demyelinating and destructive lesions predominantly in the spinal cord and the optic system [
6,
28]. Recently, auto-antibodies directed against the astrocyte water channel aquaporin 4 (AQP4) have been discovered in NMO patients, which turned out to be a highly specific and sensitive paraclinical diagnostic marker of the disease [
11,
25,
26]. Furthermore, the potential pathogenicity of these auto-antibodies has been shown in in vitro and in vivo experiments [
2,
3,
20,
42,
46]. From these data it is now well established that AQP4 antibodies, when getting access to the central nervous system compartment in vivo, can destroy astrocytes [
8]. It is assumed that the antibodies drive complement-dependent lysis, and that granulocytes and eosinophils recruited into the lesions are major effector cells [
42]. Demyelination and axonal destruction may in part be mediated by excitotoxic mechanisms, which may develop when the excitatory amino acid transporter 2 (EAAT2) is lost from dysfunctional astrocytes [
13,
30]. In addition, loss of AQP4 from astrocytes may disturb water homeostasis and result in brain edema [
12,
14]. However, to what extent these concepts, mainly developed in in vitro models, are also operating in the patient’s lesions in vivo, is less clear. In addition, it remains to be determined, whether similar mechanisms of tissue injury are also relevant for the development of demyelinating lesions in multiple sclerosis patients [
38]. This question has gained further attention, since it has recently been described that about half of all MS patients have circulating autoantibodies against a potassium channel expressed on astrocytic foot processes (Kir 4.1), and that these antibodies may destroy astrocytes in vitro in a complement-dependent manner [
50]. In this study we performed a detailed comparison of astrocyte pathology in relation to demyelination and neurodegeneration in active NMO and MS lesions. As described before [
31,
35,
36,
43], our studies show that astrocyte pathology is unique and highly characteristic in NMO, but that different mechanisms lead to astrocyte destruction, demyelination and neurodegeneration. In contrast, we did not find evidence for antibody or complement-mediated astrocyte injury in MS lesions.
Discussion
The specific features in acute stages of NMO lesions are the loss of AQP4 [
35,
36,
43] and the perivascular or subpial deposition of humoral factors such as immunoglobulin IgG and IgM, or activated complement (C9neo; [
28,
43]). This apparently reflects a humoral immune attack against the glia limitans especially against the astrocyte foot processes by AQP4 reactive auto-antibodies. Based on these observations and experimental evidence it has been suggested that blockade of terminal complement activation, of Fc-receptor-mediated, antibody-dependent cellular cytotoxicity or of the recruitment of neutrophils and eosinophils should be followed as therapeutic strategies in NMO patients with active disease [
42,
47,
53].
In this study, we show that tissue injury in the brain and spinal cord of NMO patients is complex and is mediated by at least two different mechanisms. In the first (lesion type I), active tissue injury is associated with antibody deposition and complement activation on astrocytes and their processes and this process is associated with profound recruitment of granulocytes and some eosinophils. This pattern can be well reproduced in in vivo rodent models, mediated by NMO antibodies in animals with autoimmune encephalomyelitis [
3] as well as after local injection of NMO IgG into the mouse brain together with human complement [
46]. Such lesions are in close topographical relation with inactive destructive lesions (lesion type 2) and we suggest that global overt tissue destruction and the formation of cystic cavities is the direct consequence of complement and granulocyte-induced tissue injury. In addition, the lack of astrocytes expressing the excitatory amino acid transporter (EAAT2) may promote excitotoxicity due to excess glutamate in the extracellular space [
13,
30]. Interestingly, the cystic cavities in early stages of destructive lesions contain a protein-rich fluid, which is highly reactive for GFAP. Thus, liberation of soluble GFAP through necrotic destruction of astrocytes may be the reason for high GFAP levels in the patient’s cerebrospinal fluid during active phases of the disease [
37,
51]. Furthermore, extensive axonal destruction in such necrotizing lesions will result in secondary Wallerian tract degeneration in rostral or caudal segments of the spinal cord (lesion type 3).
A second type of lesions (type 4 and 5) apparently developed in the absence of complement activation and granulocyte infiltration at the site of active tissue injury. This type of tissue injury affected astrocytes leaving other components of the nervous tissue unaffected. It was reflected either by selective loss of AQP4 on astrocytes (lesion type 4) or by profound astrocyte loss associated with degenerative astrocyte alterations at the lesion edge (lesion type 5).
As described before [
35,
36,
43] astrocytes, which are still preserved around NMO lesions, may lose AQP4 from their surface. The functional significance of pure AQP4 loss is currently not clear. It has been suggested that such astrocytes are functionally impaired with respect to control of water homeostasis [
12], although these in vitro results could not be confirmed by others [
44]. In addition, in the human brain and in such NMO lesions, but not in the rodent CNS, astrocytes also express AQP1, which is similarly efficient in regulating water transport in comparison to AQP4 [
14].
In lesions with astrocyte loss, the characteristic hallmark for tissue injury was loss of AQP4 and a pattern of astrocyte dystrophy, which bears similarities to astrocyte clasmatodendrosis. Clasmatodendrosis was initially described in the beginning of the 20th century, as a regressive morphological change of astrocytes with cytoplasmic swelling and vacuolation with beading and dissolution of their dendritic processes [
16]. Recently, clasmatodendrosis has been recognized to be associated with various diseases such as vascular diseases including Binswanger’s leukoencephalopathy [
10,
52] or epilepsy [
19]. It appears to reflect a regressive change of astrocytes, which can be induced by several mechanisms, such as for instance energy deficiency [
16,
19] or autophagic cell death [
45], and which may finally result in astrocyte apoptosis [
19]. In NMO this alteration of astrocytes seems to be directly induced by AQP4 antibodies in the absence of complement. This is supported by our finding that this type of astrocyte injury is associated with intra-cytoplasmic accumulation of AQP4, AQP1 and IgG-positive granules, apparently reflecting endosomal internalization of antibody-opsonized portions of the astrocyte cell membrane. Similar astrocyte changes have been described before in active NMO lesions [
14]. We show further that many of the astrocytes with such alterations have shrunken, condensed and sometimes fragmented nuclei, which contain fragmented DNA. These changes are typical for apoptosis. Thus, there is apparently a pathway of astrocyte destruction in NMO lesions, which is different from complement and granulocyte mediated lysis. Interestingly, this disturbance and loss of astrocyte was not necessarily associated with demyelination nor neurodegeneration, which is in line with previous observations [
41].
In addition we found lesions (lesion type 6) in NMO patients, characterized by complete demyelination and axonal preservation. Astrocytes within these lesions were present, but some showed changes of clasmatodendrosis, described above. It is likely that in these lesions, as in the destructive complement-mediated lesions, oligodendrocyte injury and demyelination follows astrocyte injury and loss. Interestingly, initial alterations of oligodendrocytes and myelin, consistent of oligodendrocyte apoptosis and loss of distal oligodendrocyte processes, which had originally been defined in a subset of active multiple sclerosis lesions [
27], are also seen in actively demyelinating NMO lesions [
5]. It has, thus, been suggested that in NMO as well as in MS, demyelination may follow an initial astrocyte injury [
38]. Astrocytes are important for creating a homeostatic environment for oligodendrocytes. Energy demand of oligodendrocytes and axons is in part provided by astrocytes, transferring lactate through gap junctions to oligodendrocytes [
9]. Loss of gap junction-related connexins that are expressed on astrocytes or oligodendrocytes has been found in active lesions of MS and NMO [
34,
49]. In addition, astrocytes express the excitatory amino acid transporter 2 (EAAT2), which removes toxic glutamate from the extracellular space and can, thus, reduce excitotoxic cell death of neurons and oligodendrocytes [
30]. Since demyelination in a subset of MS lesions is associated with oxidative damage, mitochondrial injury [
29] and a state of virtual hypoxia [
1], it could also be that such lesions are induced in both conditions by additional hypoxia-like mechanisms. In NMO, severe tissue edema may impair blood microcirculation. In addition, oxidative damage and hypoxia-like tissue injury can be amplified by glutamate toxicity, which is expected to occur in such NMO lesions due to the loss of EAAT2 [
13,
30].
Although loss of AQP4 has been described before in actively demyelinating lesions in a subset of MS patients [
32,
33,
49] and was also seen in this study, this was not associated with complement activation, granulocyte infiltration, astrocyte destruction or loss [
35,
43] or astrocyte clasmatodendrosis. AQP4 loss in this condition rather reflected retraction of perivascular astrocyte processes, similar to that present in CNS lesions induced by severe innate immunity-driven inflammation [
49]. The lack of astrocyte loss or complement deposition does not support the view that antibodies against Kir 4.1, which have recently been described in around 47 % of all MS patients [
50], destroy astrocytes in the lesions in a complement-dependent manner.
For clinicians in front of NMO patients, these various patterns must be informative. For Type 1 and 2, we should consider the earliest strategy to remove the humoral factors such as immunoglobulin and complement in the acute stage, probably by plasmapheresis [
55], and to decrease complement activity possibly by anti-C5 antibodies [
40], by preventing granulocyte infiltration [
42,
47]. Alternatively, blocking endogenous auto-antibody binding with aquaporumab, which binds to AQP4 but is unable to activate complement or interact with effector cells [
53] may be beneficial. Inhibition of complement activation or granulocyte infiltration are promising to at least partially block lesion types 1–3, and thus to ameliorate acute disease exacerbations. However, it is unlikely that these treatment strategies have a major effect on the lesion types 4–6, and therefore will not fully prevent disease progression. In contrast, persistent reduction of serum auto-antibody titers, for instance by immunosuppressive drugs or anti-CD20 antibodies [
17], or blockade of AQP4 antibody binding by small molecule inhibitors [
54], may be more effective against the entire spectrum of NMO lesions. Finally, the presence of destructive spinal cord lesions and the resulting Wallerian tract degeneration may explain, why in NMO patients, after the improvement of motor functions of transverse myelopathy, spastic paraparesis and pain with severe girdle sensation tend to persist chronically [
18]. Finally, our current results provide further evidence that NMO is a disease, which is distinctly different from MS, and that astrocyte injury and destruction is the primary event, which occurs in active lesions. However, the mechanisms of tissue injury are complex and effective neuroprotective therapy may have to target several different mechanisms in parallel.