Deletion of Irs2 reduces amyloid deposition and rescues behavioural deﬁcits in APP transgenic mice

As impaired insulin signalling (IIS) is a risk factor for Alzheimer’s disease we crossed mice ) over- expressing human amyloid precursor protein (APP), with insulin receptor substrate 2 null which develop insulin resistance. The resulting Tg2576 / Irs2 animals had increased tau phosphorylation but a paradoxical amelioration of A b pathology. An increase of the A b binding protein transthyretin suggests that increased clearance of A b underlies the reduction in plaques. Increased tau phosphorylation correlated with reduced tau-phosphatase PP2A, despite an inhibition of the tau-kinase glycogen synthase kinase-3. Our ﬁndings demonstrate that disruption of IIS in Tg2576 mice has divergent effects on patho- logical processes—a reduction in aggregated A b but an increase in tau phosphorylation. However, as these effects are accompanied by improvement in behavioural deﬁcits, our ﬁndings suggest a novel protective effect of disrupting IRS2 signalling in AD which may be a useful therapeutic strategy for this condition.


Introduction
Alzheimer's disease (AD) is characterised by extracellular plaques, composed predominantly of b-amyloid (Ab) peptides and intracellular neurofibrillary tangles (NFTs), composed of hyperphosphorylated forms of the microtubule associated protein tau. The amyloid cascade hypothesis suggests that generation and aggregation of Ab precedes and indeed promotes tau-related pathology in AD pathogenesis. Altered IIS may be a critical modifier of these processes. For example, IIS regulates both Ab formation and turnover [1][2][3] and tau phosphorylation [4,5] while in humans type 2 diabetes and variation in IIS genes are associated with AD [6][7][8]. Furthermore, in rodent models, brain-insulin resistance increases tau phosphorylation [9] while diet-induced insulin resistance or impaired IGF-1 signalling increases Ab pathology [10]. Conversely, dietary restriction [11], which enhances insulin action, and IGF-1 therapy [12] ameliorate AD pathology. These observations have led to efforts to develop therapies for AD that improve IIS. However, paradoxically, impaired IIS has also been shown to reduce Ab aggregation and toxicity in a Caenorhabditis elegans model with AD-associated pathology in muscle [13] and impaired IIS is associated with increased longevity in a variety of model organisms [14]. To determine whether toxic or protective effects of reduced IIS upon AD pathology predominate in the mammalian brain, we disrupted IIS in a mouse model of Ab deposition by crossing plaque-prone Tg2576 mice with Irs2 À/À mice, a model of type 2 diabetes.
Western blotting Phosphorylated tau. S1 and S2 samples were diluted in equal volumes of 2Â reducing sample buffer (BioRad, UK), heated to 100°C for 5 min, spun and separated on 8% SDS-polyacrylamide gels. Primary antibodies were detected with appropriate secondary antibodies conjugated to fluorophors of 700 or 800 nm, and densitometry performed, using a near infrared Odyssey imager (Licor, UK). Phospho-tau immunoreactivity values were normalized to total tau.
Detection of full-length APP and APP-CTFs. S2 fractions were used for detection of full-length holo-APP and COOH-terminal fragments of APP (APP-CTFs). Holo-APP was detected using a 6% Tris-glycine SDS-PAGE with polyclonal anti-APP antibody 369 (20). APP-CTFs were detected using 4-12% Bis-Tris NuPAGE system (Invitrogen, UK). Levels of full-length APP were normalized to the levels of actin. APP-CTFb and APP-CTFa levels were normalized to fulllength APP.
Other proteins. For all other phospho-proteins, levels were normalized to total protein and non-phosphoproteins to b-actin.

Immunohistochemistry
Histological staining. Brains were dissected between 11 and 14 months of age, fixed in 10% formalin and paraffin embedded.
Tissue sections of 10 lm thickness were stained with haematoxylin/eosin, Nissl or Gallyas methods or with Congo red. B19 is a rabbit polyclonal raised to adult bovine tau, reacting with all isoforms in a phosphorylation-independent manner [17]. A rabbit polyclonal antibody to human Ab 42 (Bio-Source, Belgium) was used for the detection of Ab.
Immunocytochemistry. Immunohistochemical labelling of brain tissue was performed using the ABC method, as previously described (22). Briefly, tissue sections were treated with H 2 O 2 and incubated with a blocking solution (10% v/v horse serum in TBS-10 mM Tris, 150 mM NaCl, pH 7.4). After overnight incubation with primary antibody, sections were incubated with horse anti-mouse antibodies conjugated to biotin followed by the ABC complex (Vector Labs, Belgium). Peroxidase activity was revealed using diaminobenzidine as chromogen. For immunolabelling with the Ab antibodies, rehydrated tissue sections were pre-treated with 100% formic acid for 10 min before incubation with the blocking solution.
Quantification of Ab staining. Digital images of whole brain sagittal sections taken close to the midline, were analysed with the NIH Image J program: the total area covered by Ab positive deposits was measured using image thresholding and the total cortex surface was measured using manual selection. The area covered by Ab deposits is expressed relative to the total cortex surface.
ac-cording to manufacturers instructions (The Genetics Company, Switzerland).

Quantitative RT-PCR
Total RNA was extracted from frontal cortex using Triazol according to manufacturers instructions (Sigma, UK). Total RNA (1 mg) was reverse transcribed using random hexamers with a Taqman RT reagent kit (Perkin Elmer, UK). Quantitative RT-PCR for transthyretin was performed as previously described [18] and using Taqman Gene Expression assay FAM/TAMRA primers (Applied Biosystems): transthyretin (Ttr) (Mm00443267_m1), Gapdh (Mm99999915_g1). Primers to mouse IDE, were designed using Universal Probe Library (Roche, UK) software. Real time PCRs were performed on a Lightcycler (Roche, UK) using QuantiTect SYBR green reagent (Qiagen, UK).

Behaviour
For contextual conditioning mice at 10-12 months were trained in a conditioning chamber (Med Associates, St. Albans, USA) in a soundproof box-after a 120 s introductory period a tone (80 dB, 3.0 kHz) was presented for 30 s, the last 2 s of which coincided with a foot-shock (0.75 mA). A further two tone/foot-shock pairings were administered at 60 s intervals and after a final 60 s period the mice were returned to their home cage. Twenty-four hours after training the mice were re-exposed to the conditioning chamber for 5 min to test for contextual fear memory. Freezing behaviour (defined as complete lack of movement, except for respiration) was scored for 2 s in every 5 s.

Expression of mutant APP increases tau phosphorylation in the context of Irs2 deletion
Modest increases in tau phosphorylation were seen in 12-15 month-old Irs2 À/À mice. In contrast, in Tg2576/Irs2 À/À mice we detected substantially increased tau phosphorylation indicating that impaired IIS promotes tau phosphorylation which is in turn enhanced by the presence of Ab pathology (Fig. 1). However, we found no change in tau aggregation (data not shown) but did detect reduced expression of the tau-phosphatase PP2a in animals lacking Irs2; this latter observation may, as suggested by others [9], underlie the increased tau phosphorylation seen in Irs2 À/À animals even though the tau-kinase glycogen synthase kinase-3 (GSK-3) was relatively inhibited (Fig. 2).

Deletion of Irs2 reduces amyloid burden in Tg2576 mice
Twelve month old Tg2576 mice displayed the expected numerous, large Congophilic Ab deposits but in contrast, in age-matched Tg2576/Irs2 À/À mice plaque number appeared reduced and plaques were smaller and less intensely labelled. Quantification of extracellular Ab deposits revealed that the area covered by Ab deposits was significantly reduced in brains of the Tg2576/Irs2 À/À mice compared to littermate Tg2576 animals (p = 0.01; Fig. 3).
Measurement of APP metabolites in the temporal cortex of agematched Tg2576 and Tg2576/Irs2 À/À mice, showed no differences in holo-APP, or b C-terminal fragments or in soluble Ab 1-40 and Ab 1-42 levels (Fig. 3). However, there was a significant reduction in insoluble, aggregated Ab 1-40 and Ab 1-42 levels in Tg2576/Irs2 À/À mice compared to Tg2576 littermates, closely reflecting the reduction in Ab load measured by immunocytochemistry. Together these re-sults suggest that altered APP processing does not underlie the alterations in Ab generation.

Fig. 2. GSK-3 is inhibited and PP2a is reduced in mice lacking
Irs2. The tau-kinase, GSK-3 and the protein phosphatases PP1 and PP2A were examined in wild-type (WT) and in IRS2À/À mice by Western blot. (A) Animals lacking Irs2 showed no change in total GSK-3 protein or in phosphorylation at the Tyr279 in GSK-3a/GSK-3b 216 epitope. However there was a substantial (p < 0.05) increase in GSK-3 phosphorylation at Ser 21 GSK-3a/Ser 9 GSK-3b epitope reflecting relative inhibition of GSK-3 activity in these animals. (B) There was no change in PP1 but a significant increase in PP2A in animals lacking Irs2. (C) Comparing PP2a in hippocampus across all four genotypes confirmed a reduction in PP2a in all animals lacking Irs2 but no effect of APP expression. The mean surface covered by Ab deposits in the cortex was significantly reduced in Tg2576/Irs2 À/À mice compared with Tg2576 mice (n = 32; * p = 0.01, t-test). (D) APP processing was examined by immunoblotting for holo-APP and APP-CTFs in temporal cortex of Tg2576 and Tg2576/Irs2 À/À animals. Holo-APP values were normlised to actin, CTF values were normalised to holo-APP. No differences were found between genotypes. (E) Human Ab 1-40 and Ab 1-42 were measured by ELISA in S1 (soluble) and S2 (detergent soluble) fractions and formic acid extracts of P2 pellets from temporal cortex. Significant reductions in Ab 1-40 and Ab 1-42 were found in formic acid extracts from Tg2576/Irs2 À/À compared to Tg2576 animals (n = 21; * p = 0.0005 [Ab  ] and 0.002 [Ab  ], t-test). Fig. 4. Increase in insulin degrading enzyme and in transthyretin in mice lacking Irs2. Amyloid turnover is regulated by proteolysis through proteases such as insulin degrading enzyme (IDE) and by increased clearance following binding to proteins such as transthyretin; both known to be altered in response to insulin signalling. IDE mRNA was unaltered in IRS2À/À mice (A) but protein levels in the membrane-bound fraction were modestly increased (B). Transthyretin mRNA was increased substantially (3.9fold; p < 0.02; SEM 1.05) in both genotypes lacking Irs2.

Reduced Ab burden is associated with increased expression of transthyretin and altered membrane expression of insulin degrading enzyme
Ab clearance is regulated by both increased proteolysis and through mechanisms dependent on binding to carrier proteins including transthyretin. Both mechanisms have been implicated in IIS-one of the key Ab proteases, insulin degrading enzyme (IDE) or insulysin (38), is involved in both insulin and Ab degradation and transthyretin, recently implicated in Ab proteolysis as well as Ab clearance [19], is elevated in both insulin resistant mice and people with type II diabetes [20,21]. We therefore examined the expression of both genes. IDE mRNA expression was not altered but TTR expression was increased 3.9-fold (p = 0.01) in animals null for Irs2 compared to wild-type animals (Fig. 4A,C). We next examined the protein levels of IDE by western blotting. As it has been recently reported that membrane-bound, but not cytosolic, IDE protein is significantly decreased in brain tissue of individuals at high risk of developing AD (40), we examined IDE in both the soluble and detergent soluble fractions. There was a significant increase in membrane associated fraction in Tg2576/Irs2 À/À animals (mean 1 (SD 0.4) vs. 1.6 (SD 0.6); t-2.55; p < 0.05) (Fig. 4B).

Deletion of Irs2 reverses behavioural deficits in Tg2576 mice
To assess the impact of disrupted Irs2 on the hippocampal learning and memory ability of the Tg2576 mice, we tested 10-12 month old mice in contextual fear conditioning, a behavioural paradigm previously shown to be impaired in the Tg2576 model [22]. Oneway ANOVA showed that there was an overall significant difference between the groups (n = 32; F 3, 28 = 4.6, p = 0.01) due to impairment in Tg2576 mutants (post hoc Tukey's p < 0.05 for all groups). Thus, the deletion of the Irs2 gene in the Tg2576 mice is able to rescue the contextual fear deficit suggesting that the reduction of Ab load and/or the inhibition of GSK-3 in these animals reversed the effects of over-expression of human APPsw.

Discussion
Considerable evidence implicates insulin resistance in the pathogenesis of AD and underlies current efforts to treat AD by improving insulin sensitivity. However, we find that disrupting Irs2 in Tg2576 mice results in improvement of both Ab plaque burden and behaviour despite an exacerbation of tau phosphorylation and the presence of insulin resistance. Previously, over-expression of APP in mice was shown to induce transthyretin expression together with evidence of increased IIS [23]. It was suggested that these are protective mechanisms resulting in the absence of the amyloid cascade in mice. Subsequently, and supportive of this hypothesis, transthyretin protein was shown to prevent Ab toxicity in vitro [24,25] and in mice over-expressing APP, neutralisation with antibody or deletion of the TTR gene both enhance pathology [26,27]. In addition to these roles, transthyretin has been shown recently to be, like IDE, an Ab protease [19] and transthyretin protein levels in CSF are decreased in AD [28,29]. Our data demonstrates a substantial increase in transthyretin expression accompanying the similarly substantial amelioration of plaque pathology, Ab fibrillisation and behavioural deficits; in line with a protective effect of transthyretin.
In marked contrast to the potentially beneficial reduction in Ab pathology, we found increased tau phosphorylation in Tg2276 mice with disruption of Irs2. However we found no evidence of tangle formation, consistent with previous observations [9]. The increase in tau phosphorylation was observed at many but not all epitopes examined. The most pronounced changes were at epitopes positioned at the 396/404, and 235 and 231 sites, sites known to be phosphorylated in AD [30]. However, the AT8/TAU1 epitope covering Ser199/Ser202/Thr205, a key GSK-3 site which is also highly phosphorylated in AD brain, was unaffected in the Tg2576/Irs2 À/À animals. When we examined tau-kinase activity in Irs2 À/À mice we found no increase in the activities of GSK-3, like others previously [9], or in the GSK3 and CDK5 substrate CRMP-2 (data not shown). These data are consistent with the pattern of tau phosphorylation changes we observed, in particular the absence of an increase in phosphorylation at the key GSK-3 sites-Ser199/ Ser202/Thr205. However we did find a highly significant decrease in the tau-phosphatase PP2A in Irs2 À/À mice, suggesting that the mechanism of effect might be mediated not by an increase in kinase activity but by a decrease in the activity of this phosphatase.
The relative role of amyloid versus tau pathologies in influencing neuronal dysfunction and cognitive impairment has been of considerable interest and indeed controversy. The generation of a mouse model with both decreased Ab aggregation and deposition but increased tau phosphorylation permitted us to directly address this question. Using a standard paradigm of hippocampal dependent learning, contextual fear conditioning, we observed a complete reversal of behavioural deficits in the context of Irs2 deletion. Interestingly it has been reported that the Ab induced impairment in LTP, known to be present in the Tg2576 animals [22], is reversed by insulin [31]. The mechanism whereby insulin might restore LTP is not known but one promising candidate is GSK-3 as we and others have recently demonstrated that GSK-3 inhibition is essential for LTP [32,33]. In the Tg2576/Irs2 À/À animals the reduction in Ab and the relative inhibition of GSK-3 might both, together or separately, contribute to the reversal of the behavioural phenotype.
Although much of the current literature suggests that insulin resistance is an aetiological factor in AD, we have recently demonstrated that mice lacking Irs1 have increased lifespan and reduced age-related pathology [34] and deletion of Irs2 in the mouse brain increases longevity [35]. In C. elegans, abrogating IIS protects against a range of proteotoxic neuropathologies, including Ab toxicity [13,36]. Our new findings demonstrate that this is also the case for mammals with specific disruption of Irs2 and suggests that for therapeutic manipulation of this pathway to be beneficial in the treatment of AD an increased understanding of the complex signalling and gene expression mechanisms downstream of IIS will be required.