Here we demonstrate that the TNF-α-lowering agent, 3,6′-dithiothalidomide, ameliorates key aspects of neuroinflammation in multiple acute and longer term CNS rodent models. Importantly, several of these models emulate specific cardinal characteristics of AD, and highlight the complex cyclic interaction among the synthesis of TNF-α, the development of neuroinflammation and impact on disease progression, inducing its advancement. Our data suggest that breaking this cycle by lowering TNF-α generation and neuroinflammation can favorably impact AD, as assessed at both a behavioral and biochemical level, even late during the disease course. Our studies hence reinforce the significant role of neuroinflammation in AD and other degenerative neurological disorders, and highlight the potential for targeting TNF-α.
TNF-α has been implicated in the pathogenesis of a wide number of neurological disorders, including AD, PD, stroke and head trauma [5–12]. Indeed, TNF-α levels have been found to be elevated within the CSF of AD patients by as much as 25-fold , in line with substantial elevations in TNF-α synthesis that were rapidly induced in RAW 264.7 cells and animals challenged with LPS (Figure 1A,E,F,G). Studies in subjects with mild cognitive impairment (MCI) that progress to develop AD suggest that increased CSF TNF-α levels are an early event, and their rise correlates with disease progression . In accord with this, Janelsins and colleagues  noted an elevated expression of TNF-α transcripts within the entorhinal cortex of 3xTg-AD mice at 2 months, prior to the appearance of amyloid and tau pathology, and this increase correlated with the onset of cognitive deficits in these mice . These studies, together with others demonstrating that (1) TNF-α polymorphisms that elevate TNF-α production may increase AD risk, particularly in patients carrying one or more apolipoprotein E ϵ4 alleles [46–49], and that (2) genetic ablation of TNF-α receptor 1 (TNFR1) in APP23 AD mice  or a selective lowering of soluble TNF-α brain levels in 3xTg-AD mice  reduces AD progression, support the concept of TNF-α inhibition strategies for treatment of AD [10, 51, 52].
Protein-based TNF-α inhibitors (etanercept and infliximab) that can effectively regulate circulating TNF-α levels by binding them  have provided a means to initially target brain TNF-α in AD, and perispinal etanercept administration followed by Trendelenburg positioning in a small prospective open-label pilot study has been reported to provide a rapid onset of cognitive improvement . TNF-α levels can also be regulated at the level of synthesis, which is tightly controlled at the level of mRNA stability to facilitate rapid responses to exogenous and endogenous stimuli , as occurs with LPS and Aβ challenge, respectively, and hence is amenable to regulation by small-molecular-weight drugs. The presence of adenylate-uridylate-rich elements (AREs) within the 3′-UTR of TNF-α mRNA supports the potential for post-transcriptional repression, targeting it for rapid degradation or inhibition of translation. This is mediated through interactions with RNA-binding proteins (RBPs), epitomized by HuR, which binds and stabilizes ARE-containing transcripts and conveys them to translational machinery to upregulate protein synthesis and, conversely, by tristetraprolin, which aids the acceleration and degradation of bound mRNAs [56–59].
Exogenous signals, as arise from exposure to bacterial proteins, potently induce inflammatory responses within the CNS in a manner mimicked by RAW 264.7 cells challenged with LPS . The resulting elevation in TNF-α derives from an LPS-mediated increase in the half-life of TNF-α mRNA, allowing release of its translational repression. By contrast, translational blockade can be induced by small-molecular-weight compounds, such as thalidomide, which induce a shortening of the TNF-α mRNA half-life . In this regard, 3,6′-dithiothalidomide is a more potent TNF-α-lowering thalidomide analogue that acts at the level of the 3′UTR of TNF-α [27, 60]. As is evident in Figure 1, LPS challenge to RAW 264.7 cells or animals activated toll-like receptors (TLR), and induced the generation of TNF-α and nitrite, a stable surrogate marker of highly unstable nitric oxide production. APP levels also were elevated, in line with prior studies  that additionally describe rises in interleukin (IL)-1, 6 and 12, and cyclooxygenase-2 . 3,6′-Dithiothalidomide dose-dependently lowered LPS-induced TNF-α, nitrite and APP levels in the absence of cellular toxicity in RAW 264.7 cells, in contrast to thalidomide, which proved ineffective at concentrations up to 30 μM (not shown), but has been reported to lower APP levels in PC12 and SH-SY5Y neuronal cell lines . This action of 3,6′dithiothalidomide effectively translated to both a systemic LPS challenge in vivo, lowering systemic and central TNF-α expression (Figure 1E-G), as well as to a central LPS challenge.
Administered to brain, LPS reliably induces chronic neuroinflammation associated with the activation of microglia, which is allied to impaired hippocampal-dependent spatial cognitive function [17, 18, 21]. In the present study this was achieved by the slow continuous infusion of a low dose of LPS into the fourth ventricle of the brain, which produced microglia activation within the hippocampus and, importantly, induced abnormal Arc expression in response to a simple behavioral task (Figures 2 and 3). The activity-regulated, cytoskeleton-associated IEG Arc is a key regulator of protein synthesis-dependent forms of synaptic plasticity, which are fundamental to memory formation [37, 64, 65]. In healthy brain, Arc protein functions in a transient manner, and its abnormally elevated sustained expression, as occurs during neuroinflammation, may generate synaptic noise and thereby impair long-term memory formation . Co-administration of systemic 3,6′-dithiothalidomide, which has been reported to readily enter the brain (brain/plasma ratio 1.34)  and reversed an acute LPS-mediated increase in TNF-α expression (Figure 1G), fully inhibited LPS-induced activation of microglia and the resulting altered coupling of neural activity with de novo synthesis of Arc (Figures 4 and 3, respectively). A similar dose of 3,6′-dithiothalidomide has recently been described to normalize the expression of Arc and to restore the acquisition and consolidation of spatial memory impairments in a fully established model of neuroinflammation , in which LPS was administered to rodents for a full 28 days prior to drug treatment (rather than in parallel with drug treatment, as in the present study). In this study by Belarbi et al. , 3,6′-dithiothalidomide normalized LPS-induced elevations in brain TNF-α expression, but not IL-1β, and additionally normalized the expression of specific genes involved within the TLR-mediated signaling pathways (in particular, TLR2, TLR4, Hmgb1 and IRAK1) that are established to lead to elevated TNF-α expression .
Aβ, particularly in the form of soluble oligomeric assemblies or Aβ-derived diffusible ligands (ADDLs) [67–69], has been described to target synapses, induce neuronal dysfunction and impair cognition. Its administration into the lateral ventricle of mice has been widely used to model neuroinflammation and induce these AD-related impairments [32, 70–72], which in the present study resulted in the activation of microglia and neuronal degeneration within the DG, accompanied by a learning impairment in the Morris Water Maze paradigm (Figure 6). Pretreatment of animals with 3,6′-dithiothalidomide markedly inhibited each of these aspects and, together with our prior studies, suggested that the agent could prove of value in Tg models of AD that, like the human condition, increasingly develop neuroinflammation during disease progression [23, 24]. This hypothesis was tested in two cohorts of 3xTg-AD mice of 10 and 17 months age, chosen to represent times that in our specific line coincided with the pre- and post-development of amyloid plaques and neurofibrillary tangles, as the presence of activated microglia in close proximity to amyloid plaques is a cardinal feature of AD-afflicted brain .
The pre-pathological upregulation of TNF-α and associated enhancement of activated microglia have been reported in the 3xTg-AD mouse model [23, 25], and it has been postulated that these activated immune cells are key in the process of clearing extracellular Aβ . A potential consequence of heightened Aβ exposure, however, is microglia TLR4 stimulation and a resultant upregulation of cytokine production and release . TNF-α as well as IL-1β can correspondingly elevate Aβ generation by stimulating γ-secretase activity [24, 75], potentially spawning a self-propagating positive feedback loop of Aβ induction of inflammation and TNF-α signaling that, in turn, may provoke further Aβ generation [5, 6, 9, 10, 62]. In our study, in accord with the literature , activated microglia were markedly elevated in old versus adult vehicle-treated 3xTg-AD mice (Figure 8E), which additionally presented with a significant elevation in brain Aβ1–42 and phosphorylated tau levels, a decline in total tau and a trend towards elevation of APP levels (Figure 7). A substantial accumulation of extracellular amyloid plaques was clearly evident within the cerebral cortex and hippocampus of old versus adult 3xTg-AD mice, which was accompanied by deficits in learning and memory, as assessed within the Morris Water Maze paradigm (Figure 8). The administration of 3,6′-dithiothalidomide to old 3xTg-AD mice reversed each of these parameters, significantly reducing Aβ1–42, phosphorylated tau and APP levels, lowering levels of activated microglia and fully ameliorating memory deficits (Figures 7 and 8), which were accompanied by an elevation in synaptic protein markers (Figures 7E and F). These drug-induced changes are in line with studies by McAlpine and colleagues , demonstrating that blockade of TNF-α signaling (either by viral vector-mediated expression of TNFR constructs or by crossing 3xTg-AD mice with TNFR1 knockout mice) significantly suppressed AD pathology. Importantly, our studies additionally demonstrate that cognitive deficits that accompany the classical pathology of AD appear to be reversible, at least in the 3xTg-AD mouse model.
A caveat with this 3xTg-AD mouse model, like all such models, is that it provides a partial model of the human disease. APP and tau expressions (specifically, human APPSwe and human tauP301L) are driven in the 3xTg-AD model by the unnatural mouse Thy1.2 regulatory element . Hence, the possibility that some actions of 3,6′-dithiothalidomide may be mediated via suppression of this unnatural transgene promoter cannot be ruled out. Importantly, however, the action of 3,6′-dithiothalidomide to favorably lower APP levels as well as neuroinflammation in cellular studies (Figure 1) occurred in cells controlled by their natural endogenous regulatory elements, and wt rodents were used in all other studies.
TNF-α has been shown to regulate numerous cellular processes, not only inflammation and cell death, but also cellular differentiation and survival, and achieves this by binding and activating two cognate receptors, TNFR1 (p55) and TNFR2 (p75) . TNFR1, expressed ubiquitously including on neurons, astrocytes and microglia, possesses an intracellular death domain and contributes to neuronal dysfunction and death following activation by soluble TNF-α , whereas TNFR2, principally expressed on cells of hematopoietic origin but also on neurons, has been associated with cell survival [76, 78–80] and chiefly responds to membrane-bound TNF-α [81, 82]. The engagement of homotrimeric TNF-α to either receptor can activate three major signaling pathways: an apoptotic cascade initiated via the TNF-α receptor-associated death domain, a nuclear factor kappa B (NFκB) signaling pro-survival pathway implemented via NFκB-mediated gene transcriptional actions, and a JNK (c-Jun N-terminal kinase) cascade involved in cellular differentiation and proliferation that is generally pro-apoptotic . In large part, although the contrasting pro-survival versus death-inducing actions of TNF-α plausibly rely on the TNF-α receptor subtype activation, the target cell types involved and their expression ratio of TNFR1/2 and associated coupling proteins, the temporal levels of available soluble and membrane-bound TNF-α , and the scale and duration of neuroinflammation combine in determining the eventual physiological consequences of TNF-α receptor activation [5, 6, 9, 10, 62]. Consequent to the diverse actions of TNF-α and the influence of the brain microenvironment in which they occur, it is hence not always clear under which conditions TNF-α promotes beneficial versus deleterious neuronal actions, and this, in large part, accounts for how an initially pro-survival response may develop into a pro-apoptotic one.
Under appropriate conditions TNF-α signaling, primarily via TNFR2, can mediate homeostatic actions, epitomized by its role in AMPA receptor surface expression and synaptic scaling to impact LTP , as well as neuroprotective ones [9, 10]. The genetic ablation of TNFR1 and -R2 in 3xTg-AD mice has been described to increase the progression of AD pathology . Furthermore, TNF-α has a reported role in hippocampal development and function  and, with the expression of both TNFR1 and -R2 on neuronal progenitor cells, it can modulate neurogenesis within the hippocampal neurons under pathological conditions [86–89]. The finding that adult 3xTg-AD mice were not detrimentally impacted by 3,6′-dithiothalidomide suggests that such homeostatic actions of TNF-α signaling were largely unimpaired, although clearly substantial classical preclinical toxicological studies are warranted before the agent can be considered for clinical use.