Angiotensin II AT2 Receptor Oligomers Mediate G-protein Dysfunction in an Animal Model of Alzheimer Disease*

Progressive neurodegeneration and decline of cognitive functions are major hallmarks of Alzheimer disease (AD). Neurodegeneration in AD correlates with dysfunction of diverse signal transduction mechanisms, such as the G-protein-stimulated phosphoinositide hydrolysis mediated by Gαq/11. We report here that impaired Gαq/11-stimulated signaling in brains of AD patients and mice correlated with the appearance of cross-linked oligomeric angiotensin II AT2 receptors sequestering Gαq/11. Amyloid β (Aβ) was causal to AT2 oligomerization, because cerebral microinjection of Aβ triggered AT2 oligomerization in the hippocampus of mice in a dose-dependent manner. Aβ induced AT2 oligomerization by a two-step process of oxidative and transglutaminase-dependent cross-linking. The induction of AT2 oligomers in a transgenic mouse model with AD-like symptoms was associated with Gαq/11 dysfunction and enhanced neurodegeneration. Vice versa, stereotactic inhibition of AT2 oligomers by RNA interference prevented the impairment of Gαq/11 and delayed Tau phosphorylation. Thus, Aβ induces the formation of cross-linked AT2 oligomers that contribute to the dysfunction of Gαq/11 in an animal model of Alzheimer disease.

dysfunction of diverse signal transduction mechanisms, such as the G-protein-stimulated phosphoinositide hydrolysis mediated by G␣ q/11 (4 -8). G␣ q/11 -stimulated signal transduction pathways are important for neuronal communication, synaptic plasticity, and neuronal survival (9,10). Therefore, it is likely that the G␣ q/11 signaling defect of AD patients plays a role in the disease process leading to neurodegeneration and dementia. In agreement with this notion, G-protein dysfunction is directly associated with disease severity of AD patients (8).
Further insight into the pathological role of the G-protein dysfunction of AD patients is lacking, because the underlying mechanism is barely understood. Several observations point to a specific defect at the level of G␣ q/11 : (i) protein levels of G␣ q/11 are not changed (7); (ii) downstream receptor/G-protein-independent phosphoinositide hydrolysis is intact (7); and (iii) receptor-mediated activation of other G-proteins, such as the G␣ i/o proteins, is not affected (5). In view of a putative role in the pathogenesis, we investigated the mechanism accounting for defective G␣ q/11 activation in AD.
G␣ q/11 activation is under control of specific receptors. Although most receptors interacting with G␣ q/11 stimulate guanine nucleotide exchange and thereby trigger G␣ q/11 -mediated signaling, the angiotensin II AT 2 receptor is inhibitory and capable of disrupting receptor-stimulated G␣ q/11 activation under specific conditions (11). Moreover, several studies link the AT 2 receptor to diverse neuronal functions related to behavior (12), apoptosis (13), and memory (14). Focusing on the inhibitory AT 2 receptor as a potential candidate accounting for defective G␣ q/11 activation in AD, we identified cross-linked oligomeric AT 2 receptors that form by a two-step cross-linking process triggered by aggregated A␤.

EXPERIMENTAL PROCEDURES
Functional Assays-Basal and stimulated binding of [ 35 S]GTP␥S (specific activity 1250 Ci/mmol; final concentration 0.5 nM) to G␣ q/11 was determined in the presence of 0.1 M GDP in a volume of 200 l in triplicates with membranes (25 g of protein/point) prepared from human prefrontal cortex specimens and mouse hippocampal tissue as described (15), followed by immunoaffinity enrichment of G␣ q/11 . The method measures specifically the activation of G␣ q/11 , because the applied antibodies cross-react specifically with G␣ q/11 as determined in immunoblot. The assay of [ 3 H]phosphatidylinositol hydrolysis was carried out in triplicates with membranes (100 g of protein/point) of human prefrontal cortex specimens, as * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. 1 To whom correspondence should be addressed: Y17M70, Winterthurerstrasse 190, CH-8057 Zurich, Switzerland. Fax: 41-44-635-6881; E-mail: ursula.quitterer@pharma.ethz.ch. 2 The abbreviations used are: AD, Alzheimer disease; A␤, amyloid ␤; APP Sw mice, transgenic mice expressing human APP695 with the double mutation (K670N/M671L) that was identified in a Swedish family with early onset Alzheimer disease; GTP␥S, guanosine 5Ј-O-(thiotriphosphate); CHAPS, 3-[(3-cholamidopropyl)dimethylammonio]-1-propanesulfonic acid; HA, hemagglutinin; TUNEL, terminal deoxynucleotidyltransferasemediated dUTP nick end labeling; HEK, human embryonic kidney; RNAi, RNA interference. detailed previously (6). Transglutaminase activity was determined by a [ 3 H]putrescine incorporation assay (15). Immunoaffinity enrichment of G␣ q/11 followed by immunoblot detection of co-enriched AT 2 was performed by a method that was described (16).
Activities of ␣-secretase and ␤-secretase present in hippocampal tissue of 13-month-old nonstressed and stressed APP Sw mice were determined using commercially available secretase kits (R&D Systems). Soluble and aggregated A␤ peptide levels (A␤-  and A␤-(1-42)) were determined in supernatant (SDS-soluble) and the formic acid extract of the pellet (SDS-insoluble) of hippocampal tissue preparations from 13-month-old nonstressed and stressed APP Sw mice using a sandwich enzyme-linked immunosorbent assay (Invitrogen).
Antibodies for Immunoblotting and Immunohistochemistry-The following antibodies were used for immunoblotting, affinity purification, and immunohistochemistry (11,15,16): affinity-purified rabbit/rat polyclonal anti-AT 2 antibodies (raised against an antigen encompassing amino acids 320 -349 of the human AT 2 receptor); affinity-purified rabbit polyclonal anti-AT 2 antibodies (raised against an antigen encompassing amino acids 16 -35 of the human or mouse AT 2 receptor); affinitypurified rabbit/rat polyclonal anti-M 1 antibodies (raised against an antigen encompassing amino acids 231-350 of the human M 1 receptor); affinity-purified rabbit polyclonal antitransglutaminase antibodies (raised against an antigen encompassing amino acids 1-20 of mouse transglutaminase-2); affinity-purified rabbit polyclonal anti-G␣ q/11 antibodies (raised against the C terminus of G␣ q/11 ). Immunoblotting and immunohistochemistry were routinely used to determine and confirm cross-reactivity of the antibodies with the respective proteins (11,15,16).
Protein Detection in Immunoblot-Insoluble, formic acid (70%)-extractable A␤ in the prefrontal cortex specimens indicative of A␤ plaque load was assessed by immunoblot with A␤-specific rat polyclonal antibodies after serial extraction in the presence of protease inhibitors by high salt (150 mM Tris, pH 7.6, 750 mM NaCl, 2 mM EDTA) and detergents (1% Triton X-100 in high salt buffer followed by radioimmune precipitation buffer and 2% SDS in 150 mM Tris, pH 7.6).
Membranes were prepared by sucrose density gradient centrifugation at 4°C, followed by partial enrichment (16). Briefly, membranes (500 g of protein/1.5 ml containing ϳ300 -400 fmol/mg AT 2 ) were solubilized by 20 mM CHAPS in 50 mM Tris, pH 7.4, supplemented with 1 mM EDTA and protease inhibitor mixture (Sigma). Particulate material was removed by centrifugation at 20,000 ϫ g for 15 min at 4°C. The solubilized proteins were precipitated and delipidated by acetone/methanol (12:2; final concentration 83%) for 90 min at 4°C. The precipitate was collected by centrifugation, followed by three washing steps with 0.2 ml of cold acetone. The pellet was dissolved in SDS-sample buffer containing 2% SDS, 5% ␤-mercaptoethanol and 6 M urea for 90 min at room temperature and stored for further use at Ϫ70°C. For immunoblotting analysis, protein samples were separated by SDS-PAGE under reducing conditions and supplemented with urea followed by transfer to polyvinylidene difluoride membranes as described (15). Affinity-purified antibodies or F(ab) 2 fragments of the respective antibodies preabsorbed to human and mouse proteins, respectively, were used for detection of AT 2 , AT 1 receptors, and G␣ q/11 . Applied antibodies were characterized in previous studies (11,(15)(16)(17). Bound antibody was visualized by preabsorbed F(ab) 2 fragments of enzyme-coupled secondary antibodies or by enzyme-coupled Protein A followed by enhanced chemiluminescence detection (ECL Plus).
Receptor Quantification-The number of AT 2 receptors was determined with membranes (50 g of protein/point) suspended in phosphate-buffered saline containing 1% bovine serum albumin and 1 mM EDTA (including protease inhibitor mixture) by incubation for 4 h at 4°C with 50 nM 125 I-labeled F(ab) 2 fragments of affinity-purified AT 2 -specific antibodies (ϳ1 Ci) in a total volume of 100 l in the presence or absence of a 100-fold excess of the unlabeled antibodies to determine nonspecific binding. Receptor-bound antibodies were separated from free antibodies by rapid filtration over glass fiber filters (GF/B; Whatman), followed by washing with ice-cold incubation buffer. Bound radioactivity was measured in a ␥-counter. The binding assay was standardized with affinitypurified recombinant AT 2 receptors expressed in Spodoptera frugiperda (Sf9) cells infected with a recombinant baculovirus encoding AT 2 .
Patients-Prefrontal cortex specimens were obtained at autopsy from 10 demented individuals with clinical diagnosis of (probable) AD according to NINCDS-ADRDA criteria (male/ female ratio 4:6, average age 73 Ϯ 2 years, range 63-81 years, postmortem interval 3-10 h) and from 10 nondemented patients (male/female ratio 5:5, average age 71 Ϯ 2 years, range 60 -78 years, postmortem interval 4 -10 h) used as a control group. Informed consent was obtained from all patients and control individuals (or subjects' relatives), and the study was approved by the Ethical Committee, Faculty of Medicine, Ain Shams University.
Transgenic Animals-Transgenic mice (Tg2576; Taconic) used in this study express human APP695 with the double mutation (K670N/M671L; APP Sw ) that was identified in a Swedish family with early onset Alzheimer disease (18). Nontransgenic mice served as a control group for immunoblotting studies.
Lentiviral Vector Production-Vector plasmids were constructed for the production of third generation lentiviruses expressing HA-transglutaminase-2 under control of the cytomegalovirus promoter (Invitrogen). For RNA interference, lentiviral constructs with RNA polymerase II promoter-driven expression of micro-RNAs targeting AT 2 (nucleotides 296 -316) or ␤-galactosidase (nucleotides 1298 -1318) were used. The generation of viral particles pseudotyped with the vesicular stomatitis virus G glycoprotein was performed with a four-plasmid system by transient transfection of 293T cells. Forty-eight hours later, the supernatants were collected. Titers of the lentiviral stocks were determined by transduction efficiency of HT1080 cells. High titer stocks were obtained by ultracentrifugation. For stereotactic injection in mice, high titer lentiviral stocks were used (Ͼ1 ϫ 10 7 transduction units/l).
Immunotherapy with A␤-  Peptide Immunization-Immunotherapy of APP Sw mice with A␤-(1-40) peptide immunization was performed essentially as described (19), starting at 12 months of age. Prior to immunization, A␤-(1-40) (2 mg/ml in phosphate-buffered saline) was incubated for 24 h at 37°C. Immunization was performed with A␤-(1-40) (100 g/injection) or phosphate-buffered saline (control) mixed 1:1 with complete Freund's adjuvant. Booster injections in incomplete Freund's adjuvant followed 2 weeks after the first injection and monthly thereafter up to 18 months. Blood samples were collected prior to immunization and 3 weeks after the initial immunization. Titers of anti-A␤ antibodies were determined by enzyme-linked immunosorbent assay with serum samples from A␤and control-immunized APP Sw mice. A␤-immunized mice, which did not exhibit high titers of anti-A␤ antibodies, were not included in the study.
At 18 months of age, APP Sw transgenic mice were anesthetized and perfused intracardially with 0.1 M sodium phosphate buffer, pH 7.2, followed by 4% (w/v) paraformaldehyde in the same buffer. Brains were removed, fixed Ͼ48 h in 10% neutral buffered formalin, and paraffin-embedded by standard methods. Immunoblotting and functional studies were performed with hippocampal membrane tissue isolated from nonfixed brain samples.
Immunohistochemistry, Radioimmunohistochemistry, and Immunofluorescence-For immunohistochemistry, paraffinembedded sections (8 m, taken at 50-m intervals for analyses, 10 -15 sections/set) were deparaffinized followed by antigen retrieval (15). Immunohistochemical staining of transglutaminase-2, AT 2 receptor, and M 1 receptor were performed with F(ab) 2 fragments of preabsorbed affinity-purified polyclonal antibodies (11,15). Sections were stained for A␤ plaques with monoclonal antibodies cross-reacting with residues 1-12 of the A␤ peptide (clone BAM-10; Sigma). Phosphorylated Tau was visualized with AT8 antibodies (Pierce). Immunohistochemistry with antibodies to MAP2 (microtubule-associated protein 2) was used as a marker of neuronal cell bodies and dendrites (anti-MAP2 antibodies; Sigma). DNA strand breaks were determined in situ applying TUNEL technology (Roche Applied Science). All sections were imaged with a Leica DMI6000 microscope equipped with a DFC420 camera. Plaque burden was analyzed at 18 months of age by computerized quantitative image analysis of immunolabeled hippocampal areas.
Assessment of the site and area of lentivirus infection was performed with paraffin sections obtained from brains of 17-month-old APP Sw mice 5 weeks after injection into the CA3 area of 2 l of a lentivirus preparation (ϳ1 ϫ 10 7 transduction units/l) encoding HA-transglutaminase-2. Lentivirus-driven expression of HA-transglutaminase-2 was visualized by radioimmunohistochemistry applying anti-HA antibodies (Santa Cruz Biotechnology, Inc., Santa Cruz, CA), followed by 125 Ilabeled secondary antibodies (specific activity ϳ3000 Ci/mmol) and autoradiography.
Immunofluorescence staining of phosphorylated Tau, AT 2 , and M 1 receptors was performed with cryosections (8 m) of postfixed and frozen brains obtained from 13-or 15-month-old APP Sw mice subjected to stress for 4 weeks (short term) or 12 weeks (long term), respectively. For co-localization of AT 2 and M 1 , affinity-purified rat anti-AT 2 antibodies and rabbit anti-M 1 antibodies were applied (dilution 1:4000), followed by second-ary antibodies labeled with Alexa Fluor 488 and Alexa Fluor 546 (dilution 1:5000; Molecular Probes) and counterstaining with 4Ј,6-diamidino-2-phenylindole. Sections were imaged with a Leica DMI6000 microscope equipped with a DFC350FX camera.
Cerebral Injection of Lentiviruses and A␤ Peptide-Expression of transglutaminase-2 in the hippocampus of 16-monthold APP Sw transgenic mice was induced by stereotactic injection of 2 l of a transglutaminase-encoding lentivirus under control of the cytomegalovirus promoter into the CA3 area of the hippocampus (anteroposterior Ϫ1.58 to Ϫ2.3 mm; lateral 1.8 -2.8 mm; dorsoventral Ϫ2 mm) as described (20). Briefly, anesthetized mice were placed on a Kopf stereotactic apparatus, coordinates were determined according to the Franklin and Paxinos atlas, the skin was delicately bored with a 0.5-ml insulin syringe equipped with a 30-gauge needle, and lentiviruses (ϳ1 ϫ 10 7 transduction units/l) were delivered using a 10-l Hamilton syringe with a 33-gauge blunt tip needle at a rate of 0.25 l/min. The control group of 16-month-old APP Sw transgenic mice received injection of a control lentivirus encoding ␤-galactosidase. A total of 12 APP Sw transgenic mice/group were injected unilaterally. Twenty days after the lentivirus injection, mice were anesthetized, brains were removed after intracardial perfusion, and injected hemispheres (or brains) of mice were fixed and paraffin-embedded. Immunoblotting and G␣ q/11 activity measurement were performed with nonfixed hippocampal tissue isolated from the injected brain hemispheres.
For down-regulation of AT 2 expression, a lentivirus that delivers a micro-RNA targeting the coding region of AT 2 (nucleotides 296 -316) by RNA interference (RNAi) was injected into the CA1 area of the hippocampus of 12-monthold APP Sw mice as detailed above. The control group received injection of a control lentivirus targeting ␤-galactosidase (nucleotides 1298 -1318). After 1 week of recovery, mice were subjected to the stress paradigm for 4 weeks. Five weeks after the lentiviral injection, behavioral analysis, biochemical analysis, and immunohistochemistry were performed as described.
To assess the effect of A␤ on AT 2 oligomerization in vivo, the indicated amounts of the A␤-(1-40) peptide in 2 l were microinjected under anesthesia into the CA1 area of the hippocampus of 13-month-old nontransgenic mice as described (21,22). Vehicle was used as control. Seven days after the injection, biochemical analysis of AT 2 receptors was performed by immunoblot.
Induction of Stress-To analyze the effect of stress, 12-month-old APP Sw transgenic mice were distributed into two subgroups. One subgroup was housed in normal conditions (nonstressed), and the other subgroup was subjected to stress (stressed) as described (23). The following stimuli were administered each week in a random order for 4 weeks: 3 h of 45°cage tilt, two times a week; soaked cage for 12 h, once a week; water deprivation for 14 h, once a week; lights on for 9 h during the dark phase, once a week; noise in the room for 3 h, three times a week; flashing light for 30 min, three times a week. Stressed mice with a significant decrease in sucrose preference (Յ50% of sucrose consumption compared with nonstressed controls) were included in the study. Stressed mice also exhibited a significant increase in plasma corticosterone levels (11.4 Ϯ 0.6 g/dl of stressed APP Sw mice and 5.2 Ϯ 0.5 g/dl of nonstressed APP Sw mice; n ϭ 12 mice/group). After 28 days of stress, 4 h before the beginning of the dark phase, brains from stressed and nonstressed APP Sw mice were removed and processed for immunoblotting, functional studies, and immunohistochemistry as described above. Twelve animals of each subgroup were used for behavioral studies.
Behavioral Studies-We used the standard water maze task (hidden platform) to test for spatial memory (24). Testing involved four trials per day over 10 days starting after the stress period. On the day following the 10 days of acquisition testing, memory retention was determined in a single 60-s probe trial, for which the submerged platform was removed. All behavioral studies were performed essentially as described (25).
Animal experiments were reviewed and approved by the committees on animal research at the Universities of Hamburg and Cairo and were conducted in accordance with National Institutes of Health guidelines.
Statistics-Unless otherwise stated, data are expressed as mean Ϯ S.E. To determine significance between two groups, we made comparisons using the unpaired two-tailed Student's t test. p values of Ͻ0.05 were considered significant.

Impaired G␣ q/11 -stimulated Signaling on Prefrontal Cortex
Specimens of AD Patients-Prefrontal cortex specimens of AD patients with significant levels of "insoluble" amyloid ␤ (A␤) characteristic of A␤ plaque load (Fig. 1A) were applied to analyze the G-protein signaling defect. Specimens of AD patients were characterized by impaired G q/11 -stimulated signaling, as revealed by significantly reduced phosphoinositide hydrolysis under basal conditions or upon stimulation with the muscarinic receptor agonist carbachol and the G-protein activator GTP␥S or AlF 4 Ϫ (Fig. 1B). By contrast, direct stimulation of phospholipases C by calcium was not different between AD cases and nondemented controls (Fig. 1B). Post-mortem time (Յ10 h) did not significantly affect the phosphoinositide signal (data not shown). These findings with prefrontal cortex specimens from AD patients confirm previous data of defective G␣ q/11 -stimulated phosphoinositide signaling in AD (4 -8).
Oligomeric AT 2 Receptors in AD Patients-We hypothesized that the neuronal angiotensin II AT 2 receptor could be involved in the G␣ q/11 signaling defect of AD patients, because the AT 2 receptor is one of the few inhibitory receptors capable of disrupting receptor-stimulated G␣ q/11 activation under specific conditions (11). Moreover, several studies link the AT 2 receptor to cerebral functions related to behavior, neuronal apoptosis, and memory (12)(13)(14). To determine the AT 2 receptor protein in immunoblot, we applied AT 2 -specific antibodies (11). The AT 2 -specific antibodies cross-reacted with the AT 2 receptor of ϳ65 Ϯ 5 kDa of receptor-transfected human embryonic kidney (HEK) cells, whereas AT 2 was absent in the mock con-trol (Fig. 1C, left). AT 2 receptors expressed in HEK cells appeared predominantly as a monomer of ϳ65 kDa, even when receptor expression levels were Ͼ5 pmol/mg protein (Fig. 1C,  left). In contrast to the monomeric AT 2 receptor of HEK cells, the AT 2 protein on AD specimens showed significant levels of a high molecular mass AT 2 receptor of ϳ250 kDa under ureasupplemented reducing conditions of SDS-PAGE (Fig. 1C, left versus right). In addition to the high molecular mass AT 2 receptor, the monomeric AT 2 receptor of ϳ65 Ϯ 5 kDa and minute amounts of an SDS-stable dimeric form of ϳ130 kDa were detected (Fig. 1C, right). The high molecular mass AT 2 receptor was completely absent on specimens of nondemented control individuals (Fig. 1C, right). Preabsorption of the antibodies with the immunizing antigen abolished the specific interaction confirming antibody specificity (Fig. 1C, right). As a control and in agreement with previous results (7), protein levels of membrane-bound G␣ q/11 were similar between the two groups (Fig.  1C, bottom). Post-mortem time (Յ10 h) and gender did not significantly affect AT 2 and G␣ q/11 levels (data not shown). Together, these findings demonstrate a high molecular mass oligomeric AT 2 receptor protein in AD patients that is absent in nondemented individuals.
In contrast to the AT 2 receptor, the AT 1 receptor on prefrontal cortex specimens of AD patients appeared as a pure monomer when detected in immunoblot with AT 1 -specific antibodies (Fig. 1D). Thus, AT 2 receptors are specifically altered in AD, whereas the closely related AT 1 receptor is not affected.
Sequestration of G␣ q/11 by AT 2 Receptor Oligomers in AD Patients-The oligomeric AT 2 receptor on prefrontal cortex specimens of AD patients bound specifically to G␣ q/11 in the absence of agonist as determined by co-enrichment, whereas AT 2 receptors of control individuals did not interact significantly with G␣ q/11 (Fig. 1E). The interaction of oligomeric AT 2 with G␣ q/11 in AD was paralleled by G␣ q/11 dysfunction as detected by significantly reduced G␣ q/11 activation either under basal conditions or upon stimulation with carbachol (Fig.  1F). These findings suggest that the AT 2 oligomer of AD patients could act as a dominant negative receptor that inhibits G-proteins by sequestration (26).
A␤ Triggers AT 2 Oligomerization in Vivo-Since the previous experiments suggested an involvement of AT 2 oligomers in the G-protein dysfunction of AD patients, we investigated the possibility of a causal relationship between AT 2 oligomers and aggregated A␤ as a prominent feature of AD. Cerebral microinjection of (aggregated) A␤ sufficient to induce symptoms of cognitive impairment and neuronal degeneration (21,22) into the hippocampus of nontransgenic mice triggered the sequential dimerization/oligomerization of AT 2 receptors in a dosedependent manner (Fig. 2A). This experiment provides strong evidence for a direct involvement of A␤ in inducing AT 2 oligomers in vivo.
Formation of AT 2 Oligomers in an Animal Model of AD-To further analyze the formation of AT 2 oligomers in vivo, we chose APP Sw transgenic mice as an animal model of AD (18). These mice develop A␤ plaques with increasing age (18); however, major symptoms of clinical AD (i.e. neurodegeneration and memory impairment) are not profound in APP Sw transgenic mice (25,27). To enhance the A␤-dependent neurode-generative process as a typical feature of AD, stress was used as an environmental factor that is known to promote the progression of AD in patients and mice (28 -30). Stress enhanced the formation of aggregated SDS-insoluble A␤ and SDS-soluble A␤ in APP Sw transgenic mice (Fig. 2, B and C). Stress also marginally promoted the amyloidogenic processing of A␤ by decreasing the activity of ␣-secretase and increasing the activity of ␤-secretase (Fig. 2, D and E). The stress-induced increase in aggregated A␤ species was accompanied by the induction of AT 2 oligomers (Fig. 2F). In contrast, the AT 2 receptors of nonstressed APP Sw transgenic mice with significantly lower levels of aggregated A␤ appeared as a monomer and dimer (Fig. 2F). These findings are complementary to the sequential induction of AT 2 dimers/oligomers by microinjected A␤ in nontransgenic mice (cf. Fig. 2A).
Progressive Neurodegeneration in Stressed APP Sw Mice-In view of the increased A␤ generation and the induction of AT 2 oligomers by stress, we asked whether the characteristic ADrelated neurodegenerative process was also enhanced in APP Sw mice by stress. Four weeks of stress led to increased symptoms of dendritic degeneration in the hippocampal CA1 area of 13-month-old APP Sw transgenic mice as revealed by immunohistochemistry with MAP2-specific antibodies (Fig. 3A). Stress also induced DNA-strand breaks, as determined by a TUNEL assay (Fig. 3B). Tau phosphorylation as a marker of degenerating neurons (31) was also significantly higher in the hippocampal CA1-CA3 area of stressed APP Sw mice, as determined by immunohistochemistry with AT8 antibodies (Fig. 3C). Quantification by 125 Ilabeled AT8 antibodies confirmed the increase in phosphorylated Tau levels in the CA1 area of stressed APP Sw mice (Table 1). Concomitantly to neurodegenerative symptoms, APP Sw transgenic mice with stress-enhanced A␤ aggregation showed significant impairment of a typical CA1-mediated learning task, as measured by the water maze acquisition (Fig. 3D) and memory retention test (Fig. 3E).
Analogously to AD patients (32), Tau phosphorylation preceded neuronal loss in the CA1 area of APP Sw mice (Fig. 3F). Quantification with neuron-specific NeuN antibodies showed a decrease of neuronal cell bodies by ϳ22% in the CA1 area of stressed APP Sw mice compared with nonstressed mice (Table  1). Intraneuronal accumulation of phosphorylated Tau was evident in the remaining CA1 neurons (Fig. 3G). In agreement with a potential role of AT 2 oligomers in Tau-positive degenerating neurons, immunofluorescence revealed prominent AT 2 receptor membrane localization in CA-1 neurons that were  1-42)) in the hippocampus of 13-month-old APP Sw mice compared with nonstressed APP Sw mice (Non-stressed). Data represent mean Ϯ S.E., n ϭ 4. **, p Ͻ 0.001; *, p Ͻ 0.05. D and E, stress decreased the activity of hippocampal ␣-secretase in APP Sw mice (D) and led to increased activity of hippocampal ␤-secretase (E). Data are given as percentage of control (i.e. the ␣-secretase or ␤-secretase activity of nonstressed mice, respectively (i.e. 100%)) and represent mean Ϯ S.E., n ϭ 4. *, p Ͻ 0.01. F, formation of AT 2 oligomers in the hippocampus of stressed APP Sw mice as determined in immunoblot of partially enriched AT 2 receptors from hippocampal membranes of stressed APP Sw transgenic mice compared with nonstressed APP Sw transgenic mice applying F(ab) 2 fragments of affinity-purified AT 2 -specific antibodies preabsorbed to mouse proteins.
AT8-positive (Fig. 3H). Thus, stress induced AT 2 oligomerization and triggered a pathological sequence of neurodegenerative events in APP Sw mice that resembled the AD pathology of patients.

Stress Induces G-protein Sequestration by AT 2 Receptor Oligomers and G␣ q/11 Dysfunction in APP Sw
Mice-Similarly to AD patients (cf. Fig. 1E), AT 2 oligomers of stressed APP Sw mice sequestered G␣ q/11 , whereas AT 2 receptor monomers and dimers did not interact with G␣ q/11 (Fig. 4, A and B).
We next analyzed whether the induction of AT 2 oligomers led to G␣ q/11 impairment. To assess G␣ q/11 dysfunction, we chose the G(␣) q/11 -coupled muscarinic M 1 receptor, because impaired G-protein coupling of M 1 correlates with disease severity in AD (8). The muscarinic agonist carbachol stimulated the activation of G␣ q/11 proteins in the hippocampus of APP Sw mice (Fig. 4C, columns 1 and 2), and this signal was predominantly mediated by the M 1 receptor as evidenced by inhibition with the M 1 -selective antagonist MT-7 (Fig.  4C, columns 3 and 4). The activation of G␣ q/11 under basal conditions and after M 1 stimulation was significantly reduced in stressed APP Sw mice with G␣ q/11 -sequestering AT 2 oligomers compared with nonstressed APP Sw mice (Fig. 4D). Thus, APP Sw mice with stress-enhanced neurodegeneration displayed G-protein-sequestering AT 2 oligomers and G␣ q/11 dysfunction.
To assess the direct relationship between G␣ q/11 -sequestering AT 2 oligomers and impaired M 1 -stimulated G␣ q/11 activation, we determined the localization of M 1 and AT 2 . Immunohistochemistry revealed hippocampal localization of M 1 receptors in the CA1-CA4 area and the dentate gyrus of APP Sw mice with stress-enhanced neurodegeneration (Fig. 4E, left). In contrast, the AT 2 receptor was only prominent in those regions of the hippocampus that also displayed Tau phosphorylation (i.e. CA1-CA3) (Fig. 4E, right; cf. Fig. 3C). Pre-  (original magnification, ϫ400). B, DNA strand breaks in the CA1 area of the hippocampus of stressed (4 weeks) APP Sw transgenic mice compared with nonstressed APP Sw transgenic mice, as determined by in situ TUNEL labeling (original magnification, ϫ400). C, stress (4 weeks) increased Tau phosphorylation in the CA1 (and CA2-CA3) area of the hippocampus of APP Sw mice (13 months old) as assessed by immunohistochemistry with AT8 antibodies (original magnification, ϫ16). D, significantly impaired learning of the Morris water maze acquisition task of stressed (4 weeks) APP Sw transgenic mice compared with nonstressed APP Sw transgenic mice. Acquisition was measured by latency to find a stationary submerged platform over 10 training blocks. Data represent mean Ϯ S.E., n ϭ 12 mice/subgroup. *, p Ͻ 0.01; **, p Ͻ 0.0008. E, reference memory retention of stressed (4 weeks) and nonstressed APP Sw transgenic mice was determined during a single 60-s probe trial (without platform) following completion of 10 blocks of acquisition training. Time spent in each of the four quadrants is indicated for each group Ϯ S.E. (n ϭ 12; *, p Ͻ 0.01; analysis of variance with Dunn's multiple comparison test). F, long term stress for 12 weeks (stressed, long-term) triggered neuronal loss in the hippocampal CA1 area of 15-month-old APP Sw mice (right) compared with nonstressed mice (left), as determined by Nissl staining of paraffin sections (original magnification, ϫ16). G, phosphorylation of Tau in CA1 neurons of stressed APP Sw mice (long term stress for 12 weeks) was determined by immunofluorescence with AT8 antibodies followed by visualization of bound antibodies with Alexa absorption of the antibodies confirmed antibody specificity (Fig. 4E, lower panels).
Hippocampal Co-localization of AT 2 and G␣ q/11 -coupled M 1 Receptors-We next investigated the co-localization of AT 2 and M 1 . Immunofluorescence revealed extensive colocalization of AT 2 and M 1 receptors in CA1 neurons of the hippocampus of stressed APP Sw mice (Fig. 5A). Notably, all neurons that expressed AT 2 were also positive for M 1 (Fig.  5A). As a control, preabsorption of the antibodies with the immunizing antigen abolished the specific staining (Fig. 5B).
The expression levels of hippocampal AT 2 and M 1 receptors were comparable (63 Ϯ 4 fmol/mg protein and 91 Ϯ 7 fmol/mg protein, respectively; n ϭ 4 mice/group; data not shown), and stress did not alter the total number of M 1 receptors (91 Ϯ 7 and 89 Ϯ 5 fmol/mg protein of stressed and nonstressed APP Sw mice, respectively; n ϭ 4 mice/group; data not shown). Thus, AT 2 and M 1 receptors are co-localized in the hippocampal CA1 area of APP Sw mice with impaired M 1 -stimulated G␣ q/11 activation.

Down-regulation of AT 2 Oligomers by RNA Interference Delayed the Development of G␣ q/11 Dysfunction and Tau
Phosphorylation-To explore whether G␣ q/11 -sequestering AT 2 receptors were directly involved in the impairment of G␣ q/11 , we specifically down-regulated the expression of hippocampal AT 2 in stressed APP Sw mice by RNA interference. Injection of a lentivirus into the CA1 area targeting the AT 2 coding sequence by a micro-RNA led to a substantial down-regulation of AT 2 expression, as assessed by immunohistochemistry, whereas the expression of M 1 was not affected (Fig. 6, A and B). Concomitantly, the amount of AT 2 receptor oligomers was substantially reduced (Fig. 6C). Upon down-regulation of AT 2 , the development of G␣ q/11 dysfunction was prevented (Fig. 6D). Complementary to previous findings (33), intact M 1 receptor signaling upon downregulation of AT 2 was accompanied by significantly lower levels of phosphorylated Tau as a marker of degenerating neurons compared with control mice with impaired M 1 receptor signaling (Fig. 6E).
Down-regulation of AT 2 receptor(s) (oligomers) also prevented the stress-induced decrease in ␣-secretase activity (Fig.  6F) and delayed the generation of SDS-soluble and -insoluble A␤ (Fig. 6, G and H). The latter effects of AT 2 oligomers seem to be directly linked to G␣ q/11 dysfunction, because intact G␣ q/11stimulated signaling is known to mediate activation of ␣-secretase (34,35). Together, these findings provide strong evidence FIGURE 4. Stress induces G-protein sequestration by AT 2 receptor oligomers and G␣ q/11 dysfunction in APP Sw mice. A, oligomeric AT 2 receptors interact specifically with G␣ q/11 as determined by immunoaffinity enrichment of G␣ q/11 and detection of co-enriched AT 2 in immunoblot (AP: anti-q/11; IB: anti-AT2). Co-enrichment was performed with hippocampal membranes of four APP Sw transgenic mice with stress-enhanced neurodegeneration compared with four nonstressed mice. The lower panel shows equal enrichment of G␣ q/11 from both groups (AP/IB: anti-q/11). B, G␣ q/11 is not significantly co-enriched with hippocampal AT 2 receptors of nonstressed APP Sw mice (lane 1), whereas AT 2 of stressed APP Sw mice interacts with G␣ q/11 (lane 2), as determined by immunoaffinity enrichment of AT 2 and immunoblot detection of co-enriched G␣ q/11 (AP: anti-AT2; IB: anti-q/11). The right panel shows equal enrichment of AT 2 receptors from nonstressed (lane 3) and stressed (lane 4) APP Sw mice (AP/ IB: anti-AT2). C, the M 1 -selective antagonist MT-7 (100 nM) inhibited the carbachol-stimulated (Cch) activation of G␣ q/11 in the hippocampus of APP Sw mice. Data represent mean Ϯ S.E., n ϭ 4 mice/group. *, p Ͻ 0.02; **, p Ͻ 0.0001. D, reduced hippocampal activation of G␣ q/11 of stressed APP Sw mice compared with nonstressed APP Sw mice. Data represent mean Ϯ S.E., n ϭ 4 mice/group. *, p Ͻ 0.003. E, in stressed APP Sw mice, M 1 receptors are localized in the hippocampal CA1-CA4 and dentate gyrus regions (left), whereas AT 2 receptors are predominantly localized in CA1-CA3 (right), as determined by immunohistochemistry applying anti-M 1 and anti-AT 2 antibodies (original magnification, ϫ18). Preabsorption controls of the antibodies are shown in the lower panels.

TABLE 1 Stress enhances Tau phosphorylation and neuronal loss in the hippocampal CA1 area of APP Sw mice
Tau phosphorylation in CA1 of stressed and nonstressed APP Sw mice was determined by direct binding assay with 125 I-labeled AT8 antibodies, and neuronal cell bodies of CA1 were quantified by 125 I-labeled NeuN antibodies, respectively (1.0 Ci/point; final concentration 5 ϫ 10 Ϫ8 M). Binding assays were performed in triplicates with crude homogenates (0.5 mg of protein/ml) prepared from dissected CA1 regions of stressed (long term stress for 3 months) and nonstressed 15-monthold APP Sw mice. Bound radioactivity of 125 I-labeled NeuN antibodies to neuronal cell bodies in CA1 of stressed APP Sw mice is expressed as a percentage of control (i.e. the value obtained with nonstressed APP Sw mice, which was set to 100%). Data represent mean Ϯ S.E., n ϭ 5 mice/group. for a causal role of AT 2 receptor oligomers in the AD-related G␣ q/11 dysfunction in vivo. A␤ Induces Oxidative and Transglutaminase-dependent Cross-linking of AT 2 Receptors in Vivo-In view of the pathophysiological importance of AT 2 oligomers, we determined the mechanism of AT 2 oligomerization. A two-step process seemed to underlie the formation of AT 2 oligomers, because cerebral microinjection of A␤ induced the sequential dimerization-oligomerization of AT 2 receptors in a dose-dependent manner (cf. Fig. 2A). Accordingly, nontransgenic mice devoid of aggregated A␤ showed predominantly monomeric AT 2 receptors, whereas APP Sw mice with moderate levels of (aggregated) A␤ displayed only monomers and cross-linked AT 2 receptor dimers (Fig. 7A, lanes 1-4 versus 5-8; cf. Fig. 2, A and  F). The initial cross-linking of AT 2 receptors leading to dimers was due to reactive oxygen species (ROS) generated as a consequence of aggregated A␤ (36), because reduction of A␤ plaques or ROS by immunization with an A␤ peptide or by antioxidant treatment with diferuloylmethane, respectively, led to significantly decreased levels of nondissociable (cross-linked) AT 2 receptor dimers (Fig. 7, A (lanes 9 -12 and lanes 13 and 14) and B).

Stressed APP
We thought that the A␤-induced oxidative dimerization could assemble AT 2 receptors for cross-linking by a transglutaminase, because transglutaminases are capable of cross-linking preassembled G-protein-coupled receptors (15), and high transglutaminase activity is a characteristic feature of clinical AD with neurodegeneration (37,38). In agreement with this concept, the formation of AT 2 oligomers was paralleled by the activation/induction of transglutaminase in APP Sw mice upon stress-enhanced A␤ aggregation (Fig. 7C, left) and in nontransgenic mice upon cerebral microinjection of more than 10 pg of (aggregated) A␤ (Fig. 7C, right). Vice versa, expression of transglutaminase in nonstressed APP Sw transgenic mice (lacking AT 2 oligomers and elevated transglutaminase activity) by stereotactic injection of a lentivirus encoding HA-transglutaminase-2 into the CA3 area of the hippocampus (Fig. 7, D and E) was not only followed by DNA strand breaks as a marker of neuronal degeneration but also induced G␣ q/11 -sequestering AT 2 oligomers and G␣ q/11 dysfunction (Fig. 7, F-H). Thus, transglutaminase accounts for the formation of G␣ q/11 -sequestering AT 2 oligomers in vivo in APP Sw mice.
Altogether, the experiments are compatible with a two-step process leading to AT 2 oligomers in vivo. In a first step, AT 2 receptors are oxidized by ROS. In a second step, the assembled AT 2 receptors (dimers) are targeted by activated transglutaminase, resulting in nondissociable tetramers (oligomers). Both steps are related to the pathogenesis of AD by the sequential induction of ROS and transglutaminase activity with increasing doses of (aggregated) A␤.

DISCUSSION
The current study identified and characterized AT 2 receptor oligomers in brains of AD patients and mice. AT 2 receptor oligomers assemble in brain by a two-step process of oxidative and transglutaminase-dependent cross-linking that is triggered consecutively by (aggregated) A␤ in a dose-dependent manner. An initial oxidative cross-linking step accounts for the formation of AT 2 receptor dimers from monomers. The oxidation of AT 2 is linked to the pathogenesis of AD, because oxidation is due to high levels of ROS generated as a consequence of (aggregated) A␤ (i.e. reduction of A␤ plaque burden and antioxidant treatment of APP Sw transgenic mice reduced the amount of oxidized AT 2 receptors). By assembling AT 2 receptors, oxidative cross-linking generates the target for the second transglutaminase-mediated cross-linking step leading finally to AT 2 tetramers/ oligomers. Relying on activated transglutaminase, the second cross-linking step is also linked to aggregated A␤ and the pathogenesis of AD, because transglutaminase activity is elevated in the brains of AD patients (37,38), and cerebral microinjection of (aggregated) A␤ or stress-enhanced generation of (aggregated) A␤ in APP Sw transgenic mice was accompanied by the activation/induction of transglutaminase.
A␤-induced AT 2 oligomers seem to contribute to the well known G␣ q/11 dysfunction in AD: (i) impaired activation of G␣ q/11 in prefrontal cortex specimens of AD patients and hip- pocampal tissue of mice with AD-like pathology correlated with the appearance of AT 2 oligomers; (ii) the formation of G␣ q/11 -sequestering AT 2 oligomers was causally linked to the pathogenesis of AD, because (aggregated) A␤-induced ROS and transglutaminase were required for the formation of AT 2 oligomers; and (iii) stereotactic inhibition of cross-linked AT 2 oligomers by RNA interference (this work) or a mutated AT 2 receptor (see accompanying article (41)) prevented the development of G␣ q/11 dysfunction. By sequestering G␣ q/11 proteins, AT 2 oligomers of AD patients and mice seem to act as dominant negative receptors, because G␣ q/11 -sequestering AT 2 oligomers induced the virtual arrest of a constitutively activated G␣ 11 -protein in vitro (cf. accompanying article (41)).
Although the G␣ q/11 -protein defect per se is a well established hallmark of clinical AD in patients (4 -8), its pathophysiological role is barely understood. A direct relationship between G␣ q/11 dysfunction and the pathogenesis of AD leading to neurodegeneration and dementia is suggested by the important role of G␣ q/11 proteins in neuronal survival and memory (9,10,39,40). In addition, many of the cognitionenhancing effects of acetylcholine in mice are mediated by G␣ q/11 -coupled muscarinic receptors (39). The M 1 receptor is a major target of G␣ q/11 dysfunction of AD patients, and impaired G␣ q/11 coupling of M 1 receptors correlates with disease severity (8).
The current study supports a pathophysiological role of the G␣ q/11 signaling defect in AD, because G␣ q/11 -sequestering AT 2 receptor oligomers seem to be part of the A␤-induced neurodegenerative process during the progression of AD. (i) A threshold dose of aggregated A␤ triggers the formation of AT 2 oligomers from dimers. (ii) The induced AT 2 oligomers mediate G-protein sequestration and dysfunction. (iii) The resulting FIGURE 6. Down-regulation of AT 2 oligomers by RNA interference delayed the development of G␣ q/11 dysfunction and Tau phosphorylation. A, AT 2 protein expression (anti-AT2) was determined by immunohistochemistry in the CA1 area of stressed APP Sw mice injected with a control lentivirus targeting ␤-galactosidase (left; Control RNAi) or with a lentivirus targeting AT 2 by RNA interference (right; AT2-RNAi; original magnification, ϫ600). B, M 1 receptor protein expression was determined by immunohistochemistry (anti-M1) in the CA1 area of stressed APP Sw mice injected with a control lentivirus (left, control RNAi) or with a lentivirus targeting AT 2 by RNA interference (right, AT2 RNAi; original magnification, ϫ600). C, immunoblot detection of AT 2 receptors with anti-AT 2 antibodies (IB: anti-AT2) in the hippocampus of stressed APP Sw mice (n ϭ 4 mice/group) injected with a control lentivirus targeting ␤-galactosidase (Control-RNAi) or with a lentivirus targeting AT 2 by RNA interference (AT2-RNAi). The control immunoblot (bottom) shows similar protein levels of G␣ q/11 of the two groups (IB: anti-G␣ q/11 ). D, basal and carbachol (Cch)-stimulated G␣ q/11 activation in the hippocampus of stressed APP Sw mice injected with a control lentivirus or with a lentivirus targeting AT 2 by RNA interference. Data represent mean Ϯ S.E., n ϭ 4 mice/group. *, p Ͻ 0.05; **, p Ͻ 0.001). E, Tau phosphorylation (top) was determined by immunoblot (IB: AT8) in the hippocampus of stressed APP Sw mice (n ϭ 4 mice/group) injected with a control lentivirus or with a lentivirus targeting AT 2 by RNA interference. A control immunoblot (IB: anti-Tau; bottom) shows equal protein levels of (dephosphorylated) Tau of the two groups. F, down-regulation of AT 2 expression by RNA interference prevented the decrease in the activity of ␣-secretase of stressed APP Sw mice (AT2-RNAi versus Control-RNAi). Data are given as percentage of control (i.e. the ␣-secretase activity of nonstressed mice (Non-str.) injected with a control lentivirus (i.e. 100%)) and represent mean Ϯ S.E., n ϭ 4. *, p Ͻ 0.01). G and H, levels of SDS-soluble (G) and SDS-insoluble A␤ (H) in the hippocampus of stressed APP Sw mice injected with a control lentivirus targeting ␤-galactosidase or with a lentivirus targeting AT 2 by RNA interference. Data represent mean Ϯ S.E., n ϭ 4. *, p Ͻ 0.05; **, p Ͻ 0.002. impairment of neuroprotective G␣ q/11 -dependent signaling (10, 40) may enhance the neurodegenerative process, as revealed by a decreased Tau phosphorylation and A␤ aggregation upon down-regulation of AT 2 (oligomers). And indeed, additional experiments confirmed a role of G␣ q/11 -inhibitory AT 2 receptor oligomers in neurodegeneration (i.e. specific inhibition of AT 2 oligomerization by an AT 2 mutant delayed the development of neurodegenerative symptoms in "AD mice") (cf. accompanying article (41)). A␤-induced AT 2 oligomers may thus constitute a previously unrecognized signature of ongoing neurodegeneration in AD.