1 Introduction
2 Tumor Necrosis Factor-a
2.2 Role of TNF in CNS Disease
The potent cytotoxic effects implicate TNF in the pathophysiology of inflammatory and autoimmune diseases. TNF levels in the brain and/or cerebrospinal fluid (CSF) become elevated in a wide range of CNS disorders, including ischemia (Feuerstein et al., 1994; Liu et al., 1994), trauma (Goodman et al., 1990), multiple sclerosis (Hofman et al., 1989; Selmaj et al., 1991; Sharief and Hentges, 1991; Rieckmann et al., 1995; Raine et al., 1998), Alzheimer’s disease (AD)(Fillit et al., 1991;
Paganelli et al., 2002; Alvarez et al., 2007), and Parkinson’s disease (PD)(Boka et al., 1994; Mogi et al., 1994, 1999, 2000; Hirsch et al., 1998; Bessler et al., 1999;
Hunot et al., 1999; Hasegawa et al., 2000; Nagatsu et al., 2000a, b; Nagatsu and Sawada, 2005). While there is no evidence to support a role for TNF in the etiology of any of these diseases, it has been suggested that TNF-dependent mechanisms act to modify, and typically accelerate disease progression.
Polymorphisms in inflammatory genes for interleukin 1α, interleukin 1β, inter-leukin-6, TNF, α 2-macroglobulin, and α 1-antichymotrypsin have been investi-gated both for CNS disease association and as markers of disease progression. In particular, the TNF gene has received much attention possibly because it resides within the MHC gene cluster. Individual studies have reported polymorphisms in the TNF promoter that may affect susceptibility to different CNS diseases includ-ing cerebral malaria (Knight et al., 1999), stroke (Karahan et al., 2005), vascular dementia (McCusker et al., 2001), Alzheimer’s disease (McGuire et al., 1994, 1999; Knight et al., 1999; Perry et al., 2001a, b), and Parkinson’s Disease (Nishimura et al., 2001). Several studies suggest that heritable differences in TNF production could be linked to certain HLA haplotypes within human populations that are over-represented in several autoimmune and inflammatory diseases (Wilson et al., 1993). However, consistent and reliable results on the functional signficance of the genetic variations have not been obtained. It is therefore possible that many of the reported associations between TNF alleles and susceptibility to disease merely reflect linkage with MHC genes or chance association.
Much of our understanding of TNF and its receptors in the CNS has come from detailed evaluation of the phenotypes of a number of genetic mouse models developed in the last 15 years and their phenotypes after exposure to various toxins or pathogens
Cytokines in CNS Inflammation and Disease 63
(Table 1) (reviewed in Probert and Akassoglou, 2001; Corti and Ghezzi, 2004;
Kollias, 2005). Consistent with a role of TNF in modulating synaptic plasticity, hippocampal brain slices from TNFR-deficient mice do not display long-term depression induced by low-frequency stimulation of Schaffer collateral axons (Albensi and Mattson, 2000). Whole animal studies in which TNF knockouts were compared to normal animals indicated TNF deficient animals performed better in spatial memory and learning tasks (Morris Water Maze) (Golan et al., 2004).
Conversely, two mouse lines overexpressing hTNF show significant impairment in spatial learning (Aloe et al., 1999b). One obvious caveat in the interpretation of these studies is that the ‘substrate’ of learning and memory is not the same in knockout and wild-type mice since TNF deficiency affects hippocampal develop-ment. At pathophysiological levels, TNF has been shown to have inhibitory effects via the p38 mitogen activated kinase pathway on hippocampal long-term potentia-tion (LTP) (Cunningham et al., 1996; Butler et al., 2004), a long-lasting increase in synaptic efficacy involving glutamate receptor activation and increased intracellular calcium levels and thought to be an underlying mechanism of learning and memory formation (Bliss and Collingridge, 1993). Elevated levels of TNF, also through a p38 MAPK-dependent pathway, may further contribute to LTP impairment through upregulation of RGS7 (a regulator of G-protein signaling) expression (Benzing et al., 2002). Lastly, studies using TNF over-expressing mice demonstrated an indirect role of TNF in influencing survival of basal forebrain cholinergic neurons via direct regulation of the levels of nerve growth factor (NGF) (Aloe et al., 1993), a key survival factor for this and other neuronal populations. Genetic ablation of TNF or TNF receptors in rodent models of PD, which show neurotoxin induced loss of dopaminergic neurons yielded variable results (Table 1) (Bruce et al., 1996;
Albensi and Mattson, 2000; Rousselet et al., 2002; Ferger et al., 2004; Leng et al., 2005; Sriram et al., 2006a, b). However, because lack of TNF signaling during development results in arrested dendritic cell development (Pasparakis et al., 1996) and stunted microglial responsiveness in adult animals (Sriram et al., 2006b) it is difficult to implicate TNF directly in neurodegeneration based on these studies.
2.2.1 Multiple Sclerosis
Given that TNF was known to regulate immune function, its role in the autoimmune dysregulation characteristic of multiple sclerosis has been extensively investigated.
TNF and its receptors are upregulated in active MS lesions and TNF levels in the CSF of MS patients correlate with disease severity (Hofman et al., 1989; Selmaj et al., 1991; Sharief and Hentges, 1991; Raine et al., 1998). Strong evidence that TNF is important in the MS disease process was derived from experimental rodent models of MS. In particular, TNF blockade was shown to prevent or treat the devel-opment of experimental autoimmune encephalomyelitis (EAE) in rodents (Ruddle et al., 1990; Baker et al., 1994; Selmaj et al., 1995; Korner et al., 1997). As indicated in Table 1, a role for TNF in the induction phase of EAE via modulation of leuko-cyte traffic into the CNS parenchyma (Korner et al., 1997; Kassiotis et al., 1999)
64 M.G. Tansey and T. Wyss-Coray
Table 1 CNS phenotypes resulting from genetic modulation of TNF pathway genes (listed are only models with CNS phenotypes)
Mouse modela
Transgene
(promoter/target cell)b
TNF receptor
statusc CNS phenotype TNF tg Overexpression of mTNF
(GFAP/astrocytes)
Wildtype Lymphocytic meningoencephalomy-elitis and paralysis (Campbell et al., 1997)
↓ Kainate-induced seizures (Balosso et al., 2005)
Tg6074 Overexpression of mTNF-hβ-globin
(TNF/subset of neurons)
Wildtype ↑ Inflammation, oligodendrocyte apoptosis, demyelination: model for chronic MS (Probert et al., 1995)
↑ Grooming in the novel object investigation test
↓ Rearing in novel olfactory cues test; delayed passive avoidance acquisition
↑ Thermal response in hot-plate test (Fiore et al., 1996)
Altered cholinergic neuron survival due to ↓ NGF (Aloe et al., 1999a); impaired learning and memory (MWM) (Aloe et al., 1999b)
Tg6074 × TNFR1 ko
Same as above TNFR1 ko None compared to above (Akassoglou et al., 1998) Tg6074 ×
TNFR2 ko
Same as above TNFR2 ko « From Tg6074 with wildtype TNFRs (Akassoglou et al., 1998) MBP-TNF Overexpression of mTNF
(MBP/oligodendrocytes)
Wildtype No spontaneous pathology but ↑ EAE with MBP adjuvant pro-gressing to chronic demyelina-tion w/macrophage and microglia activation (Taupin et al., 1997) TgK742 Overexpression of hTNF
(NF-L/neurons)
Wildtype Meningeal inflammation (Akassoglou et al., 1997) TgK21 Overexpression of htmTNF
(GFAP/astrocytes)
Wildtype ↑ Inflammation, oligodendrocyte apoptosis, demyelination: model for acute MS (Akassoglou et al., 1997, 1999)
TgK21 × TNFR1 ko
Same as above TNFR1 ko No CNS pathology compared to TgK21 with wildtype TNFRs (Akassoglou et al., 1997, 1999) TgK21 ×
TNFR2 ko
Same as above TNFR2 ko « From TgK21 with wildtype TNFRs (Akassoglou et al., 1997, 1999) TNF ko Constitutive deletion of
TNF gene expression (all cells)
Wildtype Resistant to MPTP toxicity (Ferger et al., 2004)
TNFR1 ko Constitutive deletion of TNFR1 (all cells)
Wildtype TNFR2
Not resistant to MPTP (Rousselet et al., 2002; Sriram et al., 2002;
Leng et al., 2005)
(continued)
Cytokines in CNS Inflammation and Disease 65
and a role for TNFR1 in demyelination were demonstrated (Probert et al., 2000) using TNF genetic models. Similarly, an important role for tmTNF and its pre-ferred receptor TNFR2 in oligodendrocyte precursor proliferation and remyelina-tion was demonstrated using TNF genetic models in the cuprizone toxin model of MS (Arnett et al., 2001). In fact, these data offered a mechanistic explanation for the unfortunate failure of lenercept, an Fc-fused p55/TNFR1, in phase I clinical trials with MS patients whose symptoms worsened between bouts of relapsing-remitting episodes due to the lack of TNF-mediated remyelination (Wiendl and Hohlfeld, 2002).
2.2.2 Traumatic Brain Injury (TBI)
A modulatory role for TNF following traumatic brain injury as well as the impact of TNF neutralization on behavioral deficits has been extensively investigated in a number of different models, in particular fluid percussion injury (Fan et al., 1996;
Knoblach et al., 1999; Vitarbo et al., 2004; Marklund et al., 2005), controlled corti-cal injury (Scherbel et al., 1999), and spinal cord injury (Harrington et al., 2005) with a number of different outcomes depending on the timing of the intervention.
Table 1 (continued)
TNFR2 ko Constitutive deletion of TNFR2 (all cells)
Wildtype TNFR1
Not resistant to MPTP (Rousselet et al., 2002; Sriram et al., 2002; Leng et al., 2005); pro-longed kainate-induced seizures (Balosso et al., 2005)
TNFR1 × TNFR2 dko
Constitutive deletion of both TNFR1 and R2 (all cells)
None present Partially resistant to MPTP toxicity (Sriram et al., 2002); Not resistant to MPTP toxicity;
abnormal dopamine metabolism (Rousselet et al., 2002)
↑ Vulnerability to focal cerebral ischemia and excitotoxicity;
Defective LTD (Bruce et al., 1996; Albensi and Mattson, 2000; Guo et al., 2004);
prolonged kainate-induced seizures (Balosso et al., 2005) Tg transgenic; CNS central nervous system; MS multiple sclerosis; EAE experimental autoimmune encephalomyelitis; MBP myelin basic protein; MPTP 1-Methyl-4-phenyl-1,2,3,6-tetrahydropyridine;
LTD long-term depression; NF-L neurofilament L; NGF nerve growth factor; ko knockout; dko double knockout; TACE TNF alpha-converting enzyme; MWM Morris water maze
aDetails on mouse models including expression pattern and levels, the presence of mutations, etc.
are described in the cited papers
bTransgene or targeted deletion of indicated form (m mouse or h human) of TNF or TNF receptors.
cStatus of TNF receptors in mouse model
«, no change; ↑, increased; ↓, decreased
66 M.G. Tansey and T. Wyss-Coray
2.2.3 Stroke (Cerebral Infarction) and Excitotoxic Injury
TNF can potentiate glutamate excitotoxicity directly by upregulating expression of both NMDA (Zou and Crews, 2005) and AMPA on synapses (Hermann et al., 2001;
Beattie et al., 2002; Leonoudakis et al., 2004) and indirectly by inhibiting glial glutamate transporters on astrocytes (Choi, 1988). However, in animal models, the importance of TNF in hippocampal repair after ischemic injury is supported by the finding that stroke-induced hippocampal neurogenesis can be abolished in the pres-ence of a TNF neutralizing antibody (Heldmann et al., 2005), presumably due to inhibition of TNFR2 signaling by tmTNF. Consistent with this finding, neuronal damage caused by focal cerebral ischemia and epileptic seizures were exacerbated in mice lacking both TNF receptors, suggesting that TNF serves a neuroprotective function in hippocampus under ischemic and excitotoxic conditions (Bruce et al., 1996; Guo et al., 2004). Similarly, intrahippocampal injection of TNF or transgenic targeted overexpression of TNF in astrocytes has been shown to inhibit susceptibil-ity to kainate-induced seizures whereas mice lacking TNFR2 receptors (or both TNFR1 and R2) displayed increased susceptibility and prolonged seizure activity, suggesting that the protective effect of TNF is mediated by TNFR2 (Balosso et al., 2005). Data from these studies suggests that use of drugs that target TNF pathways to treat stroke or traumatic brain injury may have deleterious effects on hippocam-pal repair and neurogenesis.
2.2.4 Alzheimer’s Disease
Interest in identifying polymorphisms in the TNF or TNF receptor genes linked to AD was largely fueled by the presence of this cytokine at amyloidogenic plaques and by results of genome-wide screening of families affected with late-onset AD.
While a few individual studies find associations between polymorphisms in the TNF gene or the TNFR2 receptor gene with late-onset AD in families with no indi-viduals possessing the APOE ε4 allele (Collins et al., 2000), others find no signifi-cant associations of three polymorphisms in the TNFR1 gene to AD (Perry et al., 2001b). Meta-analyses of genetic association studies will be required to assess overall genetic effect of genetic susceptibility loci and other cytokine genes on AD risk (Cacabelos et al., 2005; Wyss-Coray, 2006). Since polymorphisms in cytokine genes have already been linked to peripheral inflammatory disorders, such as juve-nile rheumatoid arthritis, myasthenia gravis, and periodontitis, associations between cytokine gene polymorphisms and several chronic degenerative diseases may even-tually be demonstrated (McGeer and McGeer, 2001). Dysregulated levels of TNF and other cytokines has been reported in AD patients and mouse models of AD, raising the possibility that they have disease-modifying effects and could be tar-geted in therapies. Higher serum TNF and TNF/IL-1β ratio have been detected in patients with severe AD compared to mild-moderate AD (Paganelli et al., 2002) whereas other studies have found no significant differences between studied groups (Blasko et al., 2006). Further investigations are warranted to validate these findings
Cytokines in CNS Inflammation and Disease 67
and assess their functional significance. In mouse models of AD-like pathology, elevated TNF and MCP-1 transcript levels were reported in entorhinal cortex of 3-month old 3 × Tg AD mice (Janelsins et al., 2005) coincident with accumulation of intraneuronal Aβ in these brain regions (Billings et al., 2005). Since these mice carry three transgenes encoding mutant proteins linked to Familial Alzheimer’s Disease (FAD), these findings suggest that a sensitized genetic background may trigger an early chronic neuroinflammatory response that may involve (but not be limited to) TNF-dependent JNK activation leading to increased γ-secretase activity and enhanced progression of AD-like plaque and tau pathology (Janelsins et al., 2005). Indeed, chronic exposure to systemic lipopolysaccharide (LPS) was shown to hasten pathology in these mice (Kitazawa et al., 2005). In other mouse models of AD-like pathology such as the Tg2576 mice, elevated TNF levels are detectable around amyloid plaques (Mehlhorn et al., 2000; Munch et al., 2003) and exposure to systemic LPS worsens their pathology (Sly et al., 2001). Taken together, these findings strongly suggest that TNF may be an important modulator of AD-associated pathology via multiple cellular mechanisms including modulation of microglial phenotypes and regulation of the various proteases that process APP through TNF-dependent molecular signaling pathways. Consistent with this idea, a prospective, single-center, open-label, clinical pilot study involving 15 patients with mild-to-severe AD receiving twice weekly perispinal extrathecal administration of the TNF inhibitor etanercept for 6 months reported improved cognitive performance with treatment (Tobinick et al., 2006).
These promising findings will need to be confirmed and investigated further in randomized, placebo-controlled clinical trials.
2.2.5 Parkinson’s Disease
The levels of several cytokines, including TNF, IL-1β, and IFNγ are significantly increased in the substantia nigra pars compacta (SNpc) of PD patients compared to normal controls (Hirsch et al., 1998), particularly in the area of maximal destruc-tion where the vulnerable melanin-containing dopamine-producing neurons reside.
Although the genes for various cytokines, chemokines and acute phase proteins have been surveyed and individual reports demonstrate that certain single nucle-otide polymorphisms in the TNF promoter that drive transcriptional activity are over-represented in a cohort of early onset Parkinson’s Disease patients (Nishimura et al., 2001), these findings have not been confirmed in replicative studies or reported in other populations. Once these become available, meta-analyses of mul-tiple such association studies will be needed to assess the overall genetic effect of TNF gene polymorphisms.
In experimental models of PD, significantly elevated levels of TNF mRNA and protein can be detected in the rodent midbrain substantia nigra within hours of in vivo administration of two neurotoxins widely used to model parkinsonism in rodents, 6-hydroxydopamine (6-OHDA) (Nagatsu et al., 2000b) and 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine (MPTP) (Rousselet et al., 2002; Sriram et al., 2002; Ferger et al., 2004). Consistent with a role of TNF in contributing to
68 M.G. Tansey and T. Wyss-Coray
dopaminergic neuron death in chronic parkinsonism, plasma TNF levels were shown to remain elevated in MPTP-treated non-human primates 1 year after admin-istration of the neurotoxin (Barcia et al., 2005). In addition, mice deficient in TNF or both TNF receptors have been reported to have altered dopamine metabolism and reduced survival of dopaminergic terminals (Rousselet et al., 2002) or reduced sensitivity to MPTP-induced neurotoxicity (Sriram et al., 2002; Ferger et al., 2004) (Table 1). Additional evidence that inflammation, and possibly TNF, is involved in nigral DA neuron degeneration comes from two recently developed endotoxin models of PD. In the first model, chronic low dose lipopolysaccharide (LPS) infu-sion into SNpc of rats results in delayed, selective and progressive loss of nigral DA neurons (Gao et al., 2002b). In the second model, exposure of pregnant rats to LPS and thus, in utero exposure of embryos to the endotoxin, caused a loss of DA neurons in postnatal brains (Ling et al., 2002). Most importantly, chronic infusion of dominant negative TNF inhibitor proteins into SNpc of adult rats protected nigral DA neurons from LPS and 6-OHDA induced degeneration (McCoy et al., 2006).
Given that TNF receptors are expressed in nigrostriatal dopamine neurons (Tartaglia et al., 1993; Boka et al., 1994) and these neurons are selectively vulnerable to TNF-induced toxicity (Aloe and Fiore, 1997; Ling et al., 1998; McGuire et al., 2001;
Gayle et al., 2002; Carvey et al., 2005), these early genetic studies and the more recent chronic inflammation models of PD strongly implicate TNF-dependent mechanisms and downstream targets in neurotoxin- and endotoxin-induced loss of nigral DA neurons and suggest that high TNF levels in the midbrain may increase susceptibility for PD in humans.