Background
Multiple sclerosis (MS) is a chronic inflammatory disease associated with focal demyelinating lesions in white and gray matter of the brain (BR) and spinal cord (SC) [
1‐
7]. Active lesions characteristic of relapsing remitting MS (RR-MS) contain leukocyte infiltrates and are associated with blood-brain barrier (BBB) leakage, loss of mature oligodendrocytes (OLG), and partial axonal damage, which is balanced by endogenous repair. Remyelination by OLG progenitors cells (OPC) requires clearance of myelin debris, migration of OPC to areas of damaged myelin, and their proliferation and differentiation to newly myelinating cells [
8‐
13]. Irrespective of initially effective treatment options prolonging the onset of severe neurological cognitive and motor impairments in patients with RR-MS, therapeutic interventions fail in patients with primary and secondary progressive (P)-MS. This disease state is characterized by sparse inflammation, lack of remyelination, mitochondrial dysfunction, axonal loss, and CNS atrophy [
1‐
7,
14]. Progress in understanding mechanisms underlying irreversible disease progression has been hampered by the lack of animal models exhibiting the pathology and progressive disability observed in P-MS patients [
15,
16]. Further, while much attention has focused on detrimental functions of pro-inflammatory factors, monocytes, and microglia, less is known about the transition to repair functions by glia and mechanisms shifting the balance to irreversible damage. For example, astrocytes, interferon γ (IFNγ), and tumor necrosis factor (TNF) individually are all known to promote inflammation but also mediate neuroprotective functions in demyelinating disease [
17‐
21].
Astrocytes play a vital role in maintaining CNS homeostasis and display a wide range of functions in response to injury and subsequent disease processes, including demyelination. These range from altering morphology and biochemical metabolism, inducing pro- and anti-inflammatory responses, as well as cytokines guiding peripheral immune cells or resident OPC for repair, directly enhancing or inhibiting remyelination, and forming glial scars [
19]. In MS, activated astrocytes can also take on several morphological states depending on lesion activity [
4,
9,
21,
22], While it is thus apparent that astrocytes display highly dynamic functions guided by their integration of stimuli in the local environment at various stages of disease [
23‐
25], the extent of their transient or permanent adaptation and function once injury or inflammation subsides requires better characterization.
We have recently reported that transgenic mice, in which astrocytes are unable to respond to IFNγ by expression of a dominant negative IFNγR receptor capable of IFNγ binding, but not signaling (GFAPγR1∆), develop a chronic non-resolving form of EAE [
26]. This progressive disease recapitulates some hallmarks of P-MS including increasing morbidity associated with extensive spinal cord white matter demyelination, axonal damage, and astrogliosis in white and gray matter. The inability of GFAPγR1Δ mice to resolve acute EAE was associated with elevated inflammation, including IL-6, and sustained TNF production specifically by macrophages [
26‐
28]. Disease-enhancing activity of TNF is supported by TNF in active MS lesions [
29] as well as elevated TNF in serum and cerebral spinal fluid [
30]. In rodents, TNF in the CNS correlates with EAE disease [
31] and transgenic TNF expression within the CNS leads to demyelinating disease [
32‐
34]. TNF can not only directly induce OLG death [
35‐
42], but also inflict excitotoxic damage to OLG and neurons by modulating the release of glutamate from astrocytes [
35,
39], thereby impairing OPC differentiation [
40]. Consistent with a detrimental role of sustained elevated TNF in GFAPγR1Δ mice undergoing EAE, TNF neutralization initiated at the peak of acute disease ameliorated disease and demyelination, limited CNS cellular infiltrates indiscriminately without affecting peripheral T cell responses, and limited BBB breakdown [
28]. Similar anti-TNF treatment in WT mice did not accelerate EAE disease remission [
28]. Furthermore, in MS patients, anti-TNF therapy failed to mitigate RR-MS and even worsened disease [
43]. These contrasting results and historic setbacks in targeting TNF in MS patients have since been attributed to distinct TNF activities mediated by interactions of soluble (s)TNF and its transmembrane (tm)TNF precursor with their preferential TNFR1 and TNFR2 receptors: sTNF with higher affinity for TNFR1 mediates apoptosis and chronic inflammation [
17]. Conversely, tmTNF with higher affinity for TNFR2 activates genes important for cell survival, resolution of inflammation, and OPC differentiation [
17,
44‐
49]. By promoting remyelination [
17,
44‐
49] as well as T
reg differentiation and survival [
50,
51], TNF/TNFR2 signaling thus acts at multiple levels to enhance protection and repair. Blockade or genetic manipulation of distinct TNF activities in EAE has substantiated the dual pro- and anti-inflammatory activities of TNF in regulating disease activity [
48,
52,
53].
The difficulties in targeting TNF as a therapeutic strategy to limit demyelination and favor remyelination thus clearly depend on the interplay of both TNF forms with their respective receptors on distinct cell types. The effectiveness of anti-TNF mAb therapy on EAE progression in our GFAPγR1Δ mouse model without evidence of detrimental effects supported the notion that the TNFR2 pathway may be overwhelmed once a pathological threshold is reached. Herein, we examined how TNF blockade during acute EAE in GFAPγR1Δ mice affects both de- and re-myelination. Distinct from a reduction in demyelinating lesion size during EAE disease remission in WT mice, lesion size increased during progressive EAE in GFAPγR1Δ mice, suggesting ongoing demyelination overrides any efforts of remyelination. Anti-TNF treatment significantly reduced lesion size coincident with reduced astrocyte and macrophage/microglia reactivity within demyelinating lesions. Moreover, anti-TNF mAb not only limited OLG apoptosis, but also promoted OPC differentiation within lesioned areas in GFAPγR1Δ mice. The implication that sTNF/TNFR1 mediated OLG death and inhibition of OPC differentiation dominates over TNFR2-mediated repair during disease progression in GFAPγR1Δ mice was supported by reduced TNFR2 relative to TNFR1 mRNA expression compared to remitting EAE in WT mice. Importantly, TNF blockade further limited expression of the vasoconstrictive peptide endothelin 1 (ET-1) [
54], which was vastly upregulated in lesions during progressive EAE in GFAPγR1Δ mice. Taking into account the inhibitory effects of ET-1 on OPC maturation [
55], our data demonstrate that TNF blockade limits progressive demyelination by reducing OLG death and increasing OPC differentiation, thereby promoting repair during progressive EAE. These dual effects support TNF as a potential target for therapeutic approaches in progressive forms of MS.
Discussion
TNF blockade is therapeutic in several progressive autoimmune diseases such as rheumatoid arthritis, ankylosing spondylitis, and inflammatory bowel disease. However, despite strong preclinical data supporting TNF neutralization as a therapeutic for MS, relapsing remitting patients treated with lenercept, a soluble TNFR1 IgG fusion protein, developed increased relapse rates and worsening of neurological functions [
43]. Similarly, treatment of two rapidly progressing MS patients with anti-TNF mAb exacerbated disease consistent with augmented lesion numbers and lymphocyte infiltration [
75]. Conversely, treatment of primary and secondary P-MS patients with the TNF inhibitor pirfenidone improved clinical disability [
76,
77]. These opposing results in MS may be attributed to the dual functions of TNF engagement of TNFR1 and TNFR2 in CNS inflammation and demyelination, as well as disease severity. EAE studies in WT mice demonstrated that TNF blockade in early stages of disease ameliorates or delays disease onset [
78], while it worsens disease when initiated at the peak of acute EAE [
46]. The implication that TNF contributes to tissue damage through TNFR1 at early disease stages but conversely exerts protective effects through TNFR2 during chronic EAE was confirmed in respective TNFR knockout mice [
48,
52,
53]. However, distinct from these studies, neutralization of both TNF forms in GFAPγR1Δ mice ameliorated disease; moreover, reduced demyelination coincided with reduced BBB permeability, reduced CNS cellular infiltration indiscriminate of cell type, and enhanced anti-inflammatory responses [
28]. These findings suggested that detrimental TNF signals can override protective TNF/TNFR2 effects in some clinical settings during progressive disease. This was supported by continued lesion expansion and destruction in GFAPγR1Δ mice during chronic EAE, contrasting decreasing lesion size in WT mice. Importantly, anti-TNF treatment not only constrained lesion expansion, but also decreased lesion size compared to the acute phase, suggesting repair.
In vivo and in vitro, TNF/TNFR1 signaling leads to neuron and OLG death and increased chronic inflammation, whereas TNF/TNFR2 interactions increase cell survival, anti-inflammatory signals, and remyelination [
44,
46‐
49]. Analysis of the mechanisms underlying increased lesion size during EAE in GFAPγR1Δ mice indicated sustained TNF both enhanced mature OLG death and inhibited OPC differentiation into myelinating OLG. Despite an increase in differentiated OLG during progressive EAE relative to the acute phase in GFAPγR1Δ mice, their numbers remained significantly lower than in WT mice. Notably, these differences could not be attributed to impaired OPC recruitment. Thus, although present, newly differentiated OLG in demyelinated areas during progressive EAE were insufficient to repair the extensive tissue damage. Importantly, anti-TNF treatment limited OLG death and also promoted remyelination independent of TNFR2. Despite the established influence of TNFR1 and TNFR2 signaling, the relative temporal regulation and dynamics of their function in various stages of demyelinating disease are less well defined. Our analysis of TNFR mRNA expression surprisingly revealed that TNFR2 but not TNFR1 mRNA levels were significantly reduced in the CNS of GFAPγR1Δ compared to those of WT mice during chronic EAE, although no differences in TNFR1 or TNFR2 mRNA were evident in naïve GFAPγR1Δ mice. Moreover, reduced TNFR2 mRNA expression in FACS-purified OLG from GFAPγR1Δ compared to WT mice suggested an imbalance in TNFR1 relative to TNFR2 signaling during progressive EAE counteracts repair, which may explain why neutralization of both TNF forms confers protection in GFAPγR1Δ but not WT mice. Increased mature OLG death observed in the GFAPγR1Δ mice during chronic EAE may be partially responsible for differential TNFR1 and TNFR2 expression. However, numerous other cells express TNFR2, including microglia, macrophages, and regulatory T cells (Treg). As TNFR2 signaling on these cells has been shown to contribute to repair via enhancing clearance of debris and Treg suppressor function [
47,
58], a potential role of these cells awaits future analysis. Irrespectively, reduced TNFR2 expression in GFAPγR1Δ mice implies sustained elevated TNF over time preferentially signals through TNFR1 to mediate OLG apoptosis. These findings are reminiscent of the therapeutic effects of the TNF inhibitor pirfenidone in P-MS patients [
76,
77] and suggest that dysregulated TNFR2 expression may contribute to P-MS.
Impaired OPC differentiation may also reside in TNFR2-independent functions mediated by ET-1. ET-1 has been demonstrated to indirectly inhibit OPC maturation by upregulating expression of jagged-1 in reactive astrocytes, which binds to notch-1 in OPCs [
55]. While TNF induces ET-1 in a variety of cell types [
71,
72,
79], specific induction in astrocytes has not been reported to our knowledge. The ability of ET-1 to promote TNF production [
80] and TNF to induce ET-1 production [
81] implicates an autonomous self-enhanced loop exacerbating OLG death and blocking repair. This scenario is indeed supported by a vast increase in ET-1 expression coincident with increased and sustained macrophage-derived TNF during disease progression in GFAPγR1Δ mice. The dramatic reduction in ET-1 by TNF blockade supports TNF in promoting ET-1 production. Similar to MS, ET-1 primarily localized to white matter lesions in both subsiding and progressive EAE. However, while ET-1 primarily colocalized with astrocytes during subsiding EAE in WT mice, it colocalized with both astrocytes and myeloid cells during progressive EAE in GFAPγR1Δ mice, implicating microglia/macrophages as an additional potent source of ET-1 during exacerbated demyelination. Whether ET-1 is primarily produced by monocyte-derived macrophages or microglia remains to be determined. Nevertheless, the increased population of monocyte-derived macrophages in the CNS of GFAPγR1Δ mice [
28] implicates this population as the more prominent source analogous to TNF.
The relevant CNS cell types responding to ET-1 have been under investigation [
54,
69,
82]. After CNS injury, the ET-1 receptor ENDR
B is upregulated in glial cells and specifically in reactive astrocytes during demyelination [
83]. During lysolecithin-induced demyelination, ET-1 signals through EDNR
B in reactive astrocytes, thereby mediating inhibitory jagged-1/notch-1 signaling in OPCs [
55,
74]. Moreover, loss of EDNR
B in astrocytes, but not OPC, promotes remyelination [
74]. Our data confirm EDNR
B expression predominantly in reactive astrocytes in subsiding WT EAE. GFAPγR1Δ mice undergoing progressive EAE exhibited vastly increased EDNR
B expression, which was also mostly associated with astrocytes, although modest macrophage/microglia colocalization was also evident. Anti-TNF treatment of GFAPγR1Δ mice resulted in reduced astrocyte and macrophage/microglial activation coincident with reduced EDNR
B expression. As little is known about the role of ET-1 and EDNR
B expression by myeloid cells in CNS pathology, further analysis is essential to delineate the contribution of ET-1/EDNR
B in myeloid cells versus astrocytes in this model. Similar EDNR
A mRNA expression in naïve and EAE mice (data not shown) suggested EDNR
A does not play a role in progressive EAE. Overall, our study supports an inhibitory role of ET-1, EDNR
B, and astrocytes in limiting repair.
ET-1 has several other functions contributing to CNS damage, including breakdown of BBB integrity, thereby promoting CNS entry of peripheral inflammatory cells [
82]. It also induces astrocytic production of CCL2, a chemoattractant for inflammatory monocytes, as well as IL-1β, IL-6, and reactive oxygen species, factors all contributing to CNS cellular damage and dysfunction [
80]. Lastly, it inhibits CX3CL1, which confers neuroprotection in several demyelinating models [
84]. Progressive EAE in GFAPγR1Δ mice is indeed associated with increased BBB permeability and elevated indiscriminate leukocyte infiltration compared to remitting WT mice, which are both alleviated by TNF blockade [
28]. On the other hand, ET-1 has been shown to induce transforming growth factor β (TGFβ) [
85,
86] and an M2 phenotype in human macrophages in vitro [
73], known mediators of protection during EAE. Nevertheless, no evidence for increased TGFβ [
26] or M2 macrophage/microglia during progressive EAE dismiss ET-1 mediated anti-inflammatory effects in our model.
Several mechanisms independent of OLG apoptosis may further contribute to lesion expansion during EAE in GFAPγR1Δ mice. Apoptotic T cells may sustain cytokine release and contribute to tissue debris. During injury or insult, debris clearance triggers tissue regeneration and is an essential component for CNS homeostasis [
87]. Insufficient debris clearance by microglia is highly prevalent in many neurodegenerative diseases and is associated with impaired remyelination. Increased apoptotic T cells observed in GFAPγR1Δ mice during progressive EAE may thus hinder tissue repair. Given the more reactive phenotype of macrophages relative to microglia during progressive EAE [
27,
28], ET-1 and EDNR
B expression mostly restricted to macrophages may additionally sustain pro-inflammatory effects. Overall enhanced lesion activity was supported by significantly increased astrocyte and myeloid cell activation during progressive EAE, which was limited by anti-TNF treatment. Furthermore, in a model of viral-induced demyelination, we have recently uncovered that the anti-inflammatory cytokine IL-10 is critical in limiting expansion of demyelinating lesions by promoting glial mesh formation at lesion borders [
88]. Interestingly, IL-10 is downregulated during progressive EAE, and TNF neutralization increases IL-10 production [
28]. Although IL-10 is known to downregulate TNF production [
89], it is not clear whether TNF suppresses IL-10. Elevated IL-10 after TNF blockade may thus contribute to limiting demyelination during progressive EAE.