Modulation of inflammation in transgenic models of Alzheimer’s disease

Genetic manipulation of several immune and inflammatory pathways in mouse models of AD has been carried out during the past decade to explore how increasing or decreasing neuroinflammation may affect AD progression (see Table 1). Unfortunately, most of these reports have focused only on the effect on amyloid deposition and there is a general lack of cognitive and longitudinal live imaging studies. These investigations have provided some indications to potential mechanisms by which inflammation may trigger changes in AD pathology. However, there has been some variability in the results obtained from these studies, which are largely dependent upon in which transgenic mouse model the studies have been carried out. For example, deletion of inducible nitric oxide synthase (iNOS) in an APP/PS1 background resulted in different outcomes on Aβ load compared to iNOS knockout in the Tg2576 mouse model [17, 18]. In general it is expected that overexpression of pro-inflammatory mediators will enhance progression of the disease and therefore treatments should follow an anti-inflammatory approach. For example, blocking signaling of the pro-inflammatory cytokines IL-12 and IL-23 via ablation of the common subunit p40 in APP/PS1 mice has been shown to reduce glial activation and amyloid burden [19]. Furthermore, IFNγ signaling loss in APP mice knockout for IFNγ receptor type I (GRKO mice) reduced gliosis and amyloid plaques in Tg2576 mice [20]. Interestingly, a significant reduction in the number of BACE1-positive astrocytes was seen in APP/GRKO mice as compared with APP littermates. In line with this, deletion of TNFRI in APP23 mice has been reported to reduce BACE1 protein levels and activity as well [21]. These studies in animal models support our in vitro observations, which showed that inflammation enhances BACE1 expression [22, 23].

Another potential way by which inflammation may contribute to AD pathology is by increasing Aβ aggregation. Nitration of Aβ has been shown to accelerate its aggregation and was detected in the core of Aβ plaques of APP/PS1 mice and AD brains. Studies carried out in nitric oxide synthase (NOS)2 knockout mice have shown strongly decreased 3NTyr(10)-Aβ, overall Aβ deposition and cognitive dysfunction in APP/PS1 mice [18].

A recently studied and significant factor in immune response is the NLRP3 inflammasome, which is a multiprotein oligomer consisting of caspase 1, PYCARD, NALP and sometimes caspase 5. It is upregulated in response to the stimulation of macrophages with pathogen-associated molecule patterns. APP/PS1 mice crossed with NLRP3−/− mice have rescued spatial memory, synaptic plasticity and a reduction in Aβ load when compared with age-matched APP/PS1 mice. These changes were associated with an increase in microglial phagocytic activity and increased insulin degrading enzyme [39]. APP/PS1 mice deficient in NLRP3 show increased M2 phenotype markers: FIZZ1, arginase-1, and IL-4, with reduced NOS2 expression. Complementary results were also shown with deletion of caspase 1, an important effector enzyme, in APP/PS1 mice [39].

Yet, intriguingly, many studies that induce an inflammatory state by administration of lipopolysaccharide (LPS) or IL-1β lead to a decrease in Aβ burden. This effect has been associated with enhanced microglial activation and subsequent Aβ clearance [27, 28, 40‐42]. This is a seemingly artificial method of activation of microglia, however, as acute administration of these strongly activating factors does not mimic the chronic inflammation shown in AD and, as IL-1β and LPS induce memory impairments in rodents [43, 44], they could never be thought of as a viable therapy.

In summary, studies in APP models have shown that inflammation may potentiate AD pathology in APP transgenic mice by increasing Aβ generation, aggregation and by affecting its clearance.

There is a scarcity of studies directly assessing the effect of inflammation in tau models of AD. The few that have been conducted have led to some intriguing results that suggest there may be immune responses to APP processing and tau hyperphosphorylation. While APP or APP/PS1 models do not develop neurofibrillary tangles, they do show increased tau phosphorylation [45, 46]. Products of inflammation, such as pro-inflammatory cytokines, can change the substrate specificity of kinases/phosphatases leading to tau phosphorylation at pathological sites [47].

Unlike the increases in Aβ pathology shown in PDAPP J20 [48], suppression of inflammation by overexpressing the complement inhibitory factor sCrry in P301L tau Tg mice resulted in reduced tau pathology [49]. This suggests that acute activation of the complement activation pathway is detrimental in tau mice. The 3×Tg-AD mouse model has also been used to test the effect of manipulating inflammation on tau pathology. Acute activation of the immune response in 3×Tg-AD mice by LPS treatment induced tau hyperphosphorylation via a Cdk5-dependent mechanism [50]; however, no changes were detected in amyloid pathology. In line with this, viral infection-induced acute or chronic inflammation in 3×Tg-AD mice significantly exacerbated tau pathology and led to impairments in spatial memory. In this case, tau phosphorylation was increased via a glycogen synthase kinase-3β-dependent mechanism [51]. Other approaches to induce inflammation in the brain of the 3×Tg-AD mice such as by controlled cortical impact traumatic brain injury caused acute intra-axonal Aβ accumulation and increased phospho-tau [52]. Interestingly, and in contrast to that which is observed for transgenic mice overexpressing only APP, IL-1β overexpression in 3×Tg-AD mice resulted in increased tau phosphorylation, associated with higher p38 MAPK and GSK3β activity with reductions in Aβ load [26].

However, other studies have shown controversial results in this 3×Tg-AD model, reporting reduced tau phosphorylation after delivery of IFNγ (rAAV1-IFNγ) via recombinant adeno-associated virus vector [53]. In addition, disruption of TNFα signaling in 3×Tg-AD exacerbated amyloid and tau pathology [24, 54].

With these last few exceptions, the results of modulation of inflammation in tau transgenics suggest that increased neuroinflammation leads to enhanced phosphorylation of tau, and this effect is not necessarily dependent on increased Aβ generation.

The microglial/macrophage response is a key mediator of the immune response in the brain. Microglia can be activated following exogenous or endogenous stimulation by a variety of receptors. Stimulation of these receptors can induce activation of microglia into a ‘classical (M1)’ or ‘alternative (M2)’ phenotype. That microglia play a significant role in eliciting inflammation and clearing toxic products and damaged tissue cannot be disputed, but their direct role in disease progression is unclear. Near complete ablation of microglia, by crossing either APP23 or APP/PS1 mice with CD11b-TK mice, did not show differences in plaque formation and only a very small reduction in diffuse Aβ in the APP23/CD11b-TK model [55], suggesting more subtle approaches to study their role are necessary.

A number of recent reviews have highlighted the current literature trends and debated the seemingly contradictory results relating to microglial involvement in AD [56‐58]. The activation state of microglia and their ability to phagocytose and clear amyloid in the brain seems to be a significant, but contentious, factor. Microglia and macrophages express a number of different receptors that can promote phagocytosis and clearance of Aβ that have been targeted. These include complement receptors, scavenger receptors, and cytokine/chemokine receptors that are associated with pathogen recognition (Table 2). These data can often seem incompatible and contradictory in many cases and yet yield some significant therapeutic targets and emphasize the multi-faceted and heterozygous nature of microglial response in AD from the beginning of the disease throughout its progression. Specific manipulation of signaling factors associated with a shift to the M2 phenotype is reported to promote clearance of Aβ and ameliorate other symptoms, as microglia exhibit a more anti-inflammatory, phagocytic phenotype. For example, suppression of fractalkine signaling, a negative regulator of microglial activation, is successful in reducing amyloid plaque burden and neuronal loss [59‐62]. In mouse models of other neurodegenerative disease such as Parkinson’s disease or ALS (amyotrophic lateral sclerosis), lack of CX3CR1 causes widespread neuron loss [63], suggesting that the microglial activation profile seen here is an AD-specific effect. However, as APP mouse models do not exhibit significant neuron loss it is difficult to conclude if this is a tau-specific effect or relevant to human AD.

In addition, targeting of the phagocytic phenotype of microglia has shown some promising results in AD mouse models. The complement pathway has been extensively studied in relation to AD and reports suggest that upregulating complement factors may target inflammatory processes by promoting migration and phagocytosis of inflammatory cells [48, 71, 75].

Microglia and macrophages express a number of receptors that can promote clearance of Aβ, such as scavenger receptor class A1 (Scara1) and class BI (Scarb1). Knockout models for Scarb1[65] and Scara1[64] have shown alterations in Aβ load.

Additionally, Toll-like receptors (TLRs) and their co-receptors including MD-2, CD14, and CD36 [90] are of great importance for the recognition of pathogens in the body and participate in the response of microglial cells to fibrillar forms of Aβ [91]. Deletion of CD14, which acts as a co-receptor for LPS along with TLR2 and TLR4, in APP/PS1 mice reduced total microglial numbers, in particular CD45-positive microglia, attenuated AD pathology whilst also increasing the expression of TNF-α and IL-10, suggesting an induction of a shift of activation of microglia towards the M2b state [76]. On the other hand, TLR2 deficiency accelerated spatial and contextual memory impairments, which correlated with increased levels of Aβ(1–42) and transforming growth factor-β in the brain of APP/PS1 mice [87]. An essential adaptor protein for all TLR signaling, with the exception of TLR3, is the myeloid differentiation primary response protein 88 (MyD88). Decreasing the expression of MyD88 in APP/PS1 mice led to exacerbation of spatial memory deficits, increases in Aβ, reduced expression of the fractalkine receptor CX3CR1 and increased levels of APOE (Apolipoprotein E) together with reduced astrocyte and microglial activation [85, 86]. These data indicate that TLR2 and TLR4 may be involved in Aβ clearance in vivo and hence provide neuroprotection in AD [92]. They also suggest that targeting specific glial activation states may prove fruitful in future clinical studies.

CD33 gene and TREM2, which are expressed in microglia, have been recently identified as genetic risks factors for AD [93‐96]. It was reported that CD33 is able to inhibit the uptake and clearance of Aβ42 in microglial cell cultures. This was confirmed by in vivo results showing that brain levels of insoluble Aβ42 as well as amyloid plaque burden were markedly reduced in APP(Swe)/PS1(ΔE9)/CD33(−/−) mice. Therefore, CD33 inactivation appears to mitigate Aβ pathology [77]. On the other hand, hypothesizing that the TREM2 risk variants impair TREM2 function, these new genetic studies suggest that reduced function of TREM2 causes reduced phagocytic clearance of amyloid proteins or cellular debris and thus impairs a protective mechanism in the brain [94, 96].

There are a number of studies that attribute the clearance of amyloid in mouse models to infiltrating monocytes or perivascular macrophages [97‐100]. This is due to the evidence showing a reduced efficiency of microglia with age [101] and bacterial and viral infections [102]. However, the role of these peripheral monocytes in neurodegeneration remains unclear. One important aspect is the contribution of monocytes to resident macrophages, which is highly tissue-dependent and has been shown not to be relevant for brain microglia. However, recently it was suggested that, irrespective of their origin, macrophages/microglia can self-renew by local proliferation similar to that of stem cells [103]. In fact, in animal models of prion disease it has been demonstrated that microglial proliferation is a major component in the evolution of chronic neurodegeneration [104].

Many models that show peripheral monocytic infiltration use whole body irradiation which damages the blood–brain barrier itself, induces peripheral immune activation and can facilitate infiltration. Using this approach, it was recently published that microglia-depleted brain regions of CD11b-TK transgenic mice are repopulated with new Iba1-positive cells within 2 weeks, creating a niche for myeloid cells [105]. However, using the technique of parabiosis (in which two mice share vasculature), GFP (Green Fluorescent Protein) -labelled monocytes from one mouse are not seen to infiltrate the brain of the other mouse, except following irradiation and bone marrow transplantation, which would suggest a pre-existing disease state is necessary in the brain for significant infiltration to occur [102, 106]. In line with this, recent data provide strong evidence that the engraftment of myeloid cells in the brain parenchyma of AD transgenic mice does not occur normally during disease progression, but requires prior central nervous system conditioning to sufficiently attract bone marrow cells [102]. These studies also highlight the importance of the chemokine receptor CCR2 in monocyte migration as the infiltrating cells following irradiation are characterized as CCR2⁺. Interestingly, deletion of CCR2 in Tg2576 mice increased Aβ accumulation and reduced microglial recruitment into the brain, in particular phagocytic macrophages [67]. In agreement with this, another study showed that restriction of CCR2 deficiency to perivascular myeloid cells drastically impaired Aβ clearance and amplified vascular Aβ deposition, while parenchymal plaque deposition remained unaffected [102].

Furthermore, inflammatory IFNγ-secreting Th1 cells and IL-17-secreting Th17 cells have been shown to infiltrate the brain of older APP/PS1 mice [107], supporting the observation of infiltrating T cells in the brain of AD patients [108]. However, the role of these cells in the AD brain is still unknown.

Astrocytes are becoming increasingly recognized as having key immune functions in the brain, and their role in Alzheimer’s disease progression has recently been investigated. Whilst currently falling behind the number of studies that are published assessing microglial function in AD, it is clear that astrocytes have a significant role to play in AD and therefore warrant significant future research.

Attenuation of astrocytic activation via deletion of GFAP and vimentin in APP/PS1 mice exacerbated amyloid plaque load independent of APP processing and Aβ production [89], suggesting that astrocytes are important in amyloid clearance. Yet a previous study has shown that blocking astrocyte activation via AAV-Gfa2 vectors in APP/PS1 mice also attenuates microglia activation, improves cognitive and synaptic function and reduces amyloid load [109]. However these mice were analyzed at a considerably older age (16 to 18 months) when compared with the more recent study (8 to 12 months) which suggests that there may be a significant timing factor involved in targeting the immune response in AD.

Whether astrocytes are promoting amyloid clearance or exacerbating deposition is in debate; α₁-antichymotrypsin (α₁-ACT), an acute-phase protein that is overexpressed by activated astrocytes surrounding the amyloid plaques in human AD brains, has been proven to promote Aβ fibrillization. Confirming this, overexpression of a human transgene by astrocytes in the PDGF-APP_SweInd J9 or PDAPP mouse model promoted Aβ deposition and plaque formation [36, 37]. It also affected tau phosphorylation and p-tau was increased both in single transgenic GFAP-α₁-ACT and in APP-GFAP-α₁-ACT mice [38].

Many inflammatory pathways have been implicated in AD, yet these pathways are not sufficiently well delineated to define those processes and targets that may be pathogenic as opposed to those that may be protective. The finding that treatment with non-steroidal anti-inflammatory drugs (NSAIDs) is associated with a reduced risk and age of onset of AD reinforces the hypothesis that modulating inflammation could have therapeutic efficacy. The beneficial effects of NSAIDs have also been associated with reductions in Aβ generation, since experiments in vitro and in AD models indicate that certain NSAIDs are able to decrease Aβ levels, plaque size and tau phosphorylation [110, 111].

The mechanism by which NSAIDs are protective has yielded controversial results. The initial hypothesis was that NSAIDs may affect Aβ aggregation [112, 113]. Following this, it was suggested that a subset of NSAIDs was affecting the γ-secretase cleavage site and the ratio Aβ40/42 [114, 115]. Some recent studies have shown that treatment of AD mice with a novel NSAID derivative, CHF5074, which has a more selective action on γ-secretase, resulted in modulation of Aβ42 production without affecting C-terminal APP or Notch processing [116‐118]. Chronic treatment in Tg2576 mice ameliorated memory deficits and loss of dendritic spine density together with a reduction in Aβ load, activated microglia and neuronal cell death [119]. Another potential target of NSAIDs is COX-1 [120]. It was recently reported that treatment of 3×Tg-AD mice with the COX-1 selective inhibitor SC-560 improved spatial learning and memory, and reduced amyloid deposits and tau hyperphosphorylation. SC-560 also reduced glial activation and brain expression of inflammatory markers [121]. Certain NSAIDs are also agonists for peroxisome proliferator-activated receptor (PPAR)γ and have been shown to reduce BACE1 [22, 122]. However, PPARγ activation can affect the transcription of other proteins involved in AD as well (see section below).

However, clinical trials have failed to reproduce the beneficial effects of NSAIDs in AD patients. The success of NSAIDs clinically is likely to be dependent on the stage of the disease at which the medication is started as well as the duration of the treatment [111], since their benefit seems to be towards a preventive effect rather than a therapeutic option. Interestingly, clinical trials with anti-inflammatory drugs such as trifusal in MCI (mild cognitive impairment) patients have shown a significant lower rate of conversion to dementia that is likely to be clinically relevant [123].

PPARγ is a nuclear receptor that regulates the transcription of pro-inflammatory genes, such as IL1β and iNOS. Activation of PPARγ is therefore able to inhibit the inflammatory response, and acute and chronic treatment with the PPARγ agonist pioglitazone in APPV717I and Tg2576 mice resulted in a reduction in the number of activated microglia [122, 124]. In addition, our group found that PPARγ activators decrease total Aβ levels under inflammatory conditions by affecting BACE1 transcription [6, 22, 23]. On the other hand, it was shown in neuronal cells that ibuprofen is able to suppress RhoA activity through PPARγ activation, promoting neurite elongation [125]. Therefore PPARγ activation could be beneficial in AD at several levels.

Other groups have suggested that PPARγ may affect Aβ clearance and degradation. It was recently demonstrated that PPARγ activation induces lxrα, apoe, and abca1 expression, promoting Aβ clearance by both microglia and astrocytes [126]. Furthermore, PPARγ can stimulate Aβ phagocytosis by the upregulation of scavenger receptor CD36 expression. It has also been shown that combined treatment with agonists for the heterodimeric binding partners of PPARγ, the retinoid X receptors (RXRs), showed additive enhancement of the Aβ uptake that was mediated by RXRα activation [127].

Treatment with PPAR agonists has also shown benefits in tau models. Treatment with the pan-PPAR agonist bezafibrate significantly decreased tau hyperphosphorylation and caused behavioural improvement, as evidenced by reduced hyperactivity and disinhibition in P301S mice [128]. In addition, 3×Tg-AD mice treated with pioglitazone for 4 months showed improved learning, decreased hippocampal Aβ and tau deposits, and enhanced short- and long-term plasticity [129].

Clinical trials with PPARγ activators have been more successful than those with NSAIDs. A randomised study with pioglitazone (a typical PPARγ agonist) showed significantly increased memory scores in treated patients [130]. Another PPARγ agonist, rosiglitazone, has been trialled with inconsistent results, due to its lack of permeability in the brain and its differential effects depending on the APOE (Apolipoprotein E) ϵ4 genotype [131].

Minocycline, a tetracycline derivative, has potent anti-inflammatory, anti-apoptotic, and neuroprotective properties. In many cases, the neuroprotective properties of minocycline have been attributed to inhibition of caspases. In primary cortical neurons, minocycline was shown to reduce caspase-3 activation and lowered generation of caspase 3-cleaved tau fragments [132]. Recently, minocycline was shown to protect against Aβ-induced cell death and prevent fibrillization of Aβ in vitro[133], reduce iNOS levels [134], prevent Aβ deposition and cognitive decline in APP transgenic mice [134, 135] by reducing BACE1 levels [134], inhibit neuronal death and attenuate learning and memory deficits following administration of Aβ in rats [136, 137]. In addition, treatment of a tau model with minocycline resulted in reduced levels of tau phosphorylation and insoluble tau aggregates [132].

Another potential mechanism of action of minocycline has been related to the inhibition of microglial activation. Administration of minocycline in animal models of ALS attenuated the induction of the expression of M1 microglia markers during the progressive phase, whereas it did not affect the transient enhancement of expression of M2 microglia markers during the early pathogenesis phase [138]. This study suggests that minocycline may selectively inhibit the microglia polarisation to a proinflammatory state.

TNFα is upregulated in AD and it has been found to increase in a stage-specific manner in the APP_SWE/PS1dE9 mouse model [139]. Interestingly, anti-TNFα treatment with the antibody against TNFα, infliximab, reduced Aβ and tau phosphorylation in transgenic mice. In addition, infliximab increased the number of CD11c-positive dendritic-like cells and the expression of CD11c, suggesting that the CD11c-positive dendritic-like cells might contribute to the infliximab-induced reduction of AD-like pathology [140].

The TNFα inhibitor thalidomide has been found to have abilities against tumour growth, angiogenesis, and inflammation. Chronic administration of thalidomide in APP23 and 3×Tg-AD mice resulted in a dramatic decrease in the activation of both astrocytes and microglia, Aβ load, plaque formation and tau phosphorylation [141, 142]. Furthermore, a significant decrease in BACE1 level and activity was also found [141]. However, it is not expected that this type of treatment will be beneficial for tau pathology, according to the results published in TNFRΙ knockout mice.