Cognitive and cerebrovascular improvements following kinin B₁ receptor blockade in Alzheimer’s disease mice

For selective blockade of the B₁R, the non-peptide, brain penetrant B₁R antagonist SSR240612 [(2R)-2-[((3R)-3-(1,3-benzodioxol-5-yl)-3-{[(6-methoxy-2-naphthyl)sulfonyl]amino}propanoyl)amino]-3-(4-{[2R,6S)-2,6-dimethylpiperidinyl]methyl}phenyl)-N-isopropyl-N-methylpropanamide hydrochloride] was kindly provided by Sanofi-Aventis (Montpellier, France) [23]. SSR240612 is a stable and highly selective blocker of des-Arg⁹BK binding at B₁R with a K _i of 0.48 nM, which was previously tested in man for neuropathic pain [24].

Detection of BK receptor proteins was performed by western blot and immunohistochemistry using selective anti-B₁R and -B₂R antibodies raised in rabbits (Biotechnology Research Institute, Montréal, QC, Canada) against a conserved amino acid sequence from B₁R and B₂R proteins of mouse and rat [25]. The epitopes used contained 15 amino acids localized in the C-terminal region of B₁R (VFAGRLLKTRVLGTL) and 15 amino acids localized in the second extracellular domain of B₂R (TIANNFDWVFGEVLC). Care was taken to avoid sequences with similarity to related mammalian proteins, including the opposite receptor. Two negative controls were run for each antibody: the pre-immune serum and the receptor-specific immunogenic peptide; both completely prevented any immunostaining. Specificity of our B₁R antibody was further determined using mouse kidney extracts, an organ particularly rich in constitutive B₁R [26], from wild-type (WT) and B₁R knockout (KO) mice (kindly provided by Dr. Jean-Pierre Girolami, INSERM U1048, Université Paul Sabatier, Toulouse, France). Western blotting confirmed that the anti-B₁R antibody recognized a single band at 37 kDa in the kidney of WT mice, which was absent from B₁R KO mice kidney extracts (Figure 1A).

The other reagents were as follows: serotonin (5-HT) and acetylcholine (ACh) from Sigma-Aldrich (St Louis, MO, USA), BK from American Peptide (Sunnyvale, CA, USA). Antibodies were rabbit anti-beta-secretase-1 (BACE, M-83, Santa Cruz Biotechnology, Santa Cruz, CA, USA), -glial fibrillary acidic protein (GFAP, Dako, Glostrup, Denmark), -ionized calcium-binding adaptor molecule 1 (Iba1, Wako Chemicals USA, Inc., Richmond, VA, USA), -early growth response protein 1 (Egr-1 or Zif268, C-19, Santa Cruz Biotechnology,), and -matrix metallopeptidase 9 (MMP9, Millipore, Chemicon, Billerica, MA, USA), guinea pig anti-GFAP (Synaptic Systems, Goettingen, Germany), mouse anti-endothelial nitric oxide synthase (eNOS, BD Transduction Laboratories, Mississauga, ON, Canada), -amyloid-beta (Aβ)_1–16 (6E10, Covance, Princeton, NJ, USA), -β-actin (Sigma-Aldrich), -dynein (Santa Cruz Biotechnology), rat anti-cluster of differentiation molecule 11b (CD11b, Serotec, Raleigh, NC, Canada), and goat anti-lipoprotein receptor-related protein 1 (LRP-1, N-20, Santa Cruz Biotechnology). Biotinylated, cyanin Cy2-, Cy3- or Cy5- and horseradish peroxidase-conjugated secondary antibodies were from Vector Laboratories (Burlingame, CA, USA) and Jackson Laboratories (West Grove, PA, USA), respectively. Avidin-biotin complex (ABC), 3,3′-diaminobenzidine (DAB) and slate gray (SG) reagent kits were from Vector Laboratories. ECL Plus kit for enhanced chemiluminescence was from Amersham (Mississauga, ON, Canada).

We used heterozygous transgenic adult C57BL/6 mice (10 months old) that express the human amyloid precursor protein (hAPP) carrying the Swedish (K670N, M671L) and Indiana (V717F) familial AD mutations directed by the platelet-derived growth factor (PDGF) β-chain promoter (APP mice, J20 line) [22] and age-matched WT littermates, with approximately equal numbers of females and males. None of the parameters measured in the present study were affected by gender. We selected the J20 mice because at 10 months of age they display the full spectrum of cerebrovascular, cognitive, neuroinflammation and amyloid pathologies [22, 27, 28], hence allowing therapeutic rescue rather than preventive intervention, which we believe has high relevance for AD patients. Two cohorts (n ≥48 each) were divided in four groups (WT-vehicle, WT-treated, APP-vehicle, APP-treated) and treated with SSR240612 for periods of 5 or 10 weeks. Due to compound solubility and minipump limitations, the maximal administrable dose was 10 mg/kg/day. As promising, yet incomplete beneficial effects were observed after 5 weeks of treatment, a second cohort received SSR240612 during 10 weeks. SSR240612 was diluted at 215 mM in 40% DMSO and 60% sterile saline (NaCl) added in this sequence. SSR240612 (10 mg/kg/day) and vehicle (40% DMSO, 60% NaCl) were delivered (0.11 μl/hr, 5 weeks) through osmotic minipumps (model 1004, Alzet™, Durect Corporation, Cupertino, CA, USA), which were implanted subcutaneously under isoflurane anesthesia. For the treatment of 10 weeks, minipumps were removed after 5 weeks and replaced by new ones for an additional 5-week period. Treatment did not affect APP transgene expression in brain or vascular tissues (Figure 1B).

The ability of mice to learn and remember the location of a platform in a predefined (target) quadrant in a circular pool filled with opaque water (17°C) using visuospatial cues was tested during eight consecutive days (three days of visible platform training followed by five days of hidden platform trials), as described elsewhere [29]. Platform location and visual cues distribution were altered between the visible and hidden platform testings. Twenty-four hours after the last hidden platform trial, on day 9, mice were submitted to a 60-sec probe trial (platform removed). Visual acuity and locomotor ability were comparable between all groups, as assessed during the visible platform testing. All experiments were started at the same time every day. Daily escape latencies to the platform, percentage of time spent and distance traveled in the target quadrant during the probe trial, along with swim speed, were collected with the 2020 Plus tracking system and analyzed with the Water 2020 software (HVS Image, Buckingham, UK).

Laser Doppler flowmetry measurements (Transonic Systems Inc., Ithica, NY, USA) of evoked cerebral blood flow (CBF) in response to sensory stimulation were carried out one week following the Morris water maze in anesthetized mice (n = 4 to 6, ketamine 80 mg/kg intraperitoneally; Wyeth, St-Laurent, QC, Canada) fixed in a stereotaxic frame [28]. CBF was recorded over the contralateral somatosensory cortex before, during and after unilateral stimulation of the right whiskers (20 sec at 8 to 10 Hz). Six recordings were acquired every 30 to 40 sec and averaged for each mouse. The entire procedure lasted less than 30 min, a time window when all physiological parameters remain stable [30]. Cortical CBF changes were expressed as percentage increase relative to baseline.

Mice were killed by cervical dislocation and middle cerebral artery (MCA) segments immediately tested in vascular reactivity studies. For immunohistochemistry (IHC), western blot (WB) and ELISA studies, mice were exsanguinated by intracardiac perfusion of sterile 0.9% NaCl under deep sodium pentobarbital anesthesia. Vessels of the circle of Willis and their branches free of pial membrane along with cortex and hippocampus of one hemibrain were collected, snap-frozen on dry ice and stored (−80°C). The other hemibrain was fixed by overnight immersion in 4% paraformaldehyde (PFA) in 0.1 M phosphate-buffered saline (pH 7.4), cryoprotected, frozen in isopentane (−45°C) and stored (−80°C) until cutting into 25 μm-thick free-floating coronal sections using a freezing microtome.

In order to assess the impact of B₁R blockade on the reactivity of cerebral vessels, isolated, pressurized and submaximally precontracted (5-HT, 2.10^-7 M) MCA segments (diameter 40 to 70 μm) from WT and APP mice were tested for dilatation to ACh (10^-10 to 10^-5 M) and BK (10^-10 to 10^-5 M) using online videomicroscopy [31]. Percentage changes in vessel diameter from pre-constricted tone were plotted as a function of agonist concentration. The maximal response (EAmax) and the concentration eliciting half of EAmax (EC₅₀ value, or pD₂ = −[log EC₅₀]), were generated by the GraphPad Prism software (version 4, San Diego, CA, USA) and used to evaluate agonist efficacy and potency, respectively.

Sections were pretreated with 3% H₂O₂ (20 min) and incubated overnight at room temperature (RT) with either rabbit anti-B₁R (1:1500) or -Egr-1 (1:250) antibodies diluted in a blocking buffer, followed by biotinylated anti-rabbit IgGs and the ABC kit; labeling was revealed with 0.05% DAB (B₁R) or SG (Egr-1). To study basal Egr-1 expression levels and avoid task-induced changes [32], Egr-1 immunohistochemistry was done on animals sacrificed three days post-water maze. For detection of dense core amyloid plaques, sections from APP mice were stained with 1% thioflavin-S (8 min). Sections from all groups were incubated with rabbit anti-GFAP (1:400), -Iba1 (1:300), mouse anti-Aβ_1–16 (1:1000), or goat anti-LRP-1 (1:250), followed by species-specific Cy2- (GFAP and Iba1) or Cy3- (LRP-1) conjugated secondary antibodies (1:300) for the detection of activated astrocytes, microglia, diffuse and dense-core Aβ plaques, or LRP-1, respectively. Sections were observed under a Leitz Aristoplan microscope using bright field or an FITC filter and epifluorescence (Leica, Montréal, QC, Canada) and digital pictures were acquired with a digital camera (Coolpix 4500; Nikon, Tokyo, Japan). For double immunofluorescence, sections were simultaneously incubated with a rabbit anti-B₁R antibody and either a guinea pig anti-GFAP, rat anti-CD11b, or mouse anti-Aβ_1–16 antibody, followed by donkey anti-rabbit Cy3- and species-specific Cy2-conjugated IgGs. Triple immunolabeling was also performed by co-incubating sections with rabbit anti-B₁R (with Cy5 secondary), rat anti-CD11b (Cy2) and mouse anti-Aβ_1–16 (Cy3) antibodies. Sections were observed and images acquired under a Zeiss LSM 510 laser scanning confocal microscope (Carl Zeiss Ltd., Toronto, ON, Canada) equipped with appropriate filters.

Digital images (two to three sections/mouse, n = 4 to 6 mice) taken under the same conditions were analyzed with MetaMorph (6.1r3, Universal Imaging, Downingtown, PA, USA). The areas of interest (somatosensory/cingulate cortex, dorsal hippocampus) containing thioflavin S-, 6E10-, GFAP-, Iba1- and LRP-1-positive elements were manually outlined in MetaMorph on low-power digital pictures (as exemplified in Figure 1A by dashed lines), whereas high-power images of the hippocampus were used for quantification of Egr-1 immunostaining. For microglial proliferation, the numbers of Iba-1-positive cell bodies were manually counted on digital pictures (two pictures from cortex, one from hippocampus including the dentate gyrus (DG), from three immunostained sections per mouse). The area occupied by Iba1- and GFAP-positive cells was also quantified and expressed as a percentage of surface occupied by labeling within the delineated areas of interest. Since clusters of activated microglial emit a stronger immunofluorescence signal, the maximal intensity of Iba-1 labeling was measured by dividing the maximal gray value of staining by the total gray value of the delineated area. The area occupied by thioflavin S- and 6E10-positive plaques, as well as by LRP-1-positive cells was also measured, together with the intensity (total gray value) of LRP-1 immunofluorescence. For better LRP-1 staining illustration, colors on micrographs were inverted and desaturated. Quantification of Egr-1 immunoreactivity in the hippocampus included both the total gray value in CA1-CA2 areas and the number of positive nuclei in the DG.

Levels of soluble Aβ_1–40 and Aβ_1–42 were measured in homogenized hemibrains from APP mice (n = 4 to 5) using ELISA kits (BioSource International, Camarillo, CA, USA), as described before [28]. Data were collected as optical density values in the tissue supernatants and expressed as a percentage of untreated APP mice.

For protein quantification (n = 5 to 6 mice/group), vessels were homogenized in Laemmli buffer and cortex and hippocampus in a lysis buffer [31]. In brief, extracts were protein assayed, loaded (5 to 50 μg) in 10% SDS-polyacrylamide gels, separated by electrophoresis and transferred to nitrocellulose membranes. Membranes were incubated (1 h, RT) in a blocking buffer containing 5% skim milk and then (overnight, 4°C) with either rabbit anti-B₁R (1:500), -B₂R (1:500), -BACE (1:1000), -MMP9 (1:1000), or mouse anti-Aβ_1-16 (1:1000), -eNOS (1:1000), -β-actin (1:10000), or -dynein (1:25000). Membranes were further incubated (1 h, RT) with horseradish peroxidase-conjugated secondary antibodies (1:2000) and proteins visualized by chemiluminescence (ECL Plus kit) using a phosphorImager (Scanner STORM 860; GE Healthcare, Piscataway, NJ, USA), followed by densitometric quantification (ImageQuant 5.0, Molecular Dynamics, Sunnyvale, CA, USA).

All data are means ± SEM and were analyzed with GraphPad 4 (San Diego, CA, USA) or Statistica 10 (StatSoft, Tulsa, OK, USA) software. Two-group comparisons (effects of treatment on amyloidosis) were analyzed by unpaired Student’s t tests. Four-group comparisons were analyzed using two-way analysis of variance (ANOVA, genotype and treatment as factors) followed by a Newman-Keuls post hoc test when the interaction or at least one factor was significant. P <0.05 was considered significant. Except for vascular reactivity, anatomical and western blot studies, experiments were conducted blind to the mouse identity. The relationships between amyloid burden and cognitive performance or LRP-1 immunoreactivity were assessed by plotting the averages of Aβ plaque load (percentage area) from thioflavin-S and 6E10 stainings with those of the cognitive score (percentage of time and percentage of distance spent in the target quadrant) or LRP-1 immunostaining (percentage of intensity or percentage of occupied area) in the cortex, hippocampus and DG in five non-treated and five treated APP mice from the 10-week treatment cohort.

Since localization and expression of BK receptors in the rodent brain has been described by autoradiography of radioligand binding sites and by western blot [9, 17, 18], we first investigated the immunohistochemical distribution of the B₁R protein in WT and APP mice using highly selective primary antibodies. B₁R immunoreactivity was very low to absent in the cortex and hippocampus of WT mice (Figure 2A), with only a faint, barely detectable staining in the DG (Figure 2A, inset). In APP mice, B₁R immunostaining was increased in the hippocampus (Figure 2B, arrows), primarily in the DG that displayed abundant immunopositive cells in the hilus region (Figure 2B, inset), whereas only few B₁R-positive cells were found in cortex (Figure 2B, arrowhead). In APP hippocampus (Figure 2B,C), B₁R-positive clusters contained glia-shaped cells with hypertrophied processes surrounding empty parenchymal zones characteristic of amyloid deposition, as confirmed in high-power micrographs of double immunofluorescence for B₁R and Aβ_1-16 (Figure 2D; see also 2F). The upregulated B₁R immunoreactivity did not co-localize with the CD11b marker for activated microglia (Figure 2E,F), but with the typical astroglial GFAP marker (Figure 2, bottom). B₁R-positive astrocytes in APP mice displayed enhanced GFAP immunostaining, hypertrophic processes and soma (Figure 2, bottom, white arrows). No B₁R-positive blood vessels were detected in pia or parenchyma of the mouse brain, in accordance with B₁R not mediating BK-mediated vasomotor responses [33].

In order to investigate whether B₁R blockade could alter behavioral outcome in APP mice, cognitive performances were assessed in the Morris water maze paradigm in our two cohorts of mice. Although APP mice were slightly slower than WT mice in reaching the visible platform at specific time points, they had no visual or motor deficit and all groups performed similarly on day 3 (Figure 3A,B). In contrast, both learning and memory were significantly impaired in APP mice compared with WT littermates. APP mice displayed increased latencies to reach the hidden platforms (Figure 3A,B, left) and reduced time spent and distance traveled in the target quadrant (−50%; P <0.01) during the probe trial (Figure 3A,B, right) although swim speeds were similar among all groups (data not shown). Both 5 and 10 weeks of SSR240612 treatment greatly improved learning and memory performances in APP mice, without any effect on WT mice. Recovery was comparable after 5 or 10 weeks of drug delivery.

We then studied Egr-1 (Zif268) protein levels in hippocampus, a transcription factor related to synaptic activity [34], and required for memory induction and consolidation [35]. Basal expression of Egr-1 is reduced in APP mice [30, 36], and upregulated in adult APP mice with pharmacologically restored memory [30]. We confirmed a drastic reduction (50 to 70%, P <0.001) of Egr-1 staining intensity in the CA1-CA2 region (Figure 4A,B) of the hippocampus in APP mice. However, despite significant improvement in memory after SSR240612, no beneficial effect of treatment was evidenced on Egr-1 expression in this region. In contrast, in addition to a higher Egr-1 staining intensity readily detectable in the hilus of the DG of treated compared to untreated APP mice (Figure 4C), a partial but significant recovery (+25%; P <0.05) in the reduced number (−60%; P <0.001) of Egr-1-positive nuclei was also observed following 10, but not 5 weeks of SSR240612 treatment (Figure 4C).

As expected [22], transgenic APP mice featured high brain levels of soluble Aβ peptide, as measured by ELISA (Aβ_1-40 and Aβ_1-42, Figure 5A,B), and of aggregated and deposited Aβ species as measured by thioflavin-S staining (dense-core Aβ plaques, Figure 5C) and 6E10 immunohistochemistry (total amyloid, Figure 5D). Following short- and long-term SSR240612 administration, levels of soluble Aβ_1-42 species were selectively decreased in hemibrains (Figure 5A,B), and mature and diffuse Aβ plaque loads were significantly reduced in the somatosensory/cingulate cortex and dorsal hippocampus (Figure 5C,D). Both the 5- and 10-week treatments exerted reducing effects on amyloidosis, but the total Aβ plaque load in neocortex was significantly decreased only after the longer treatment (Figure 5D).

Since brain Aβ homeostasis is highly regulated by Aβ clearance at the BBB by LRP-1 [37, 38], we investigated whether SSR240612 treatment could impact on this Aβ efflux system. WT and APP mice showed few LRP-1-immunoreactive neurons in cortex and hippocampus, with immunoreactive vessels being present along the cortical surface and throughout the hippocampus (Figure 6A). SSR240612 treatment decreased LRP-1 immunoreactivity in brain parenchyma and microvessels of treated WT mice. In contrast, 5 and 10 weeks of SSR240612 treatment significantly increased (approximately 30 to 40%) LRP-1-labeled area and intensity in APP mice depending on the brain region (Figure 6A). Neuronal LRP-1, in contrast, remained unchanged after treatment, compatible with neuronal LRP-1 expression being independent of amyloidosis in AD mice [39]. As MMP9 is involved in both BBB integrity and Aβ clearance [40‐43], its protein levels were also measured. Following 10 but not 5 weeks of drug administration, a 50 to 60% increase (P <0.01) in MMP9 protein levels was evidenced in the cortex and hippocampus of WT and APP mice (Figure 6B), whereas those of the Aβ-generating enzyme BACE-1 were not altered (Figure 6B), the latter supporting that SSR240612 acted beyond Aβ synthesis.

Knowing that neuroinflammation mediated by astrocytes and microglia is induced in AD [44], as replicated in the brain of APP mice by the enhanced GFAP (astrocytes, Figure 7A) and Iba1 (microglia, Figure 7B) immunofluorescence, we sought to investigate whether the SSR240612-induced decrease in amyloidosis would be accompanied by a reduction in activated glial cells. Upregulated GFAP immunostaining in APP mice occurred primarily as patchy islands (Figure 7A, insets) reminiscent of Aβ deposits in neocortex and hippocampus, but SSR240612 treatment (5 and 10 weeks) did not lessen the extent of astroglial activation (Figure 7A). In contrast, the increase in microglial Iba1-positive area and, mainly, Iba1 staining intensity in APP brain were significantly reduced following SSR240612 treatment (Figure 7B). The patchy areas associated with plaques (Figure 7B, insets) were much less apparent in all regions after 10 weeks of drug delivery. Hippocampus, including the DG, was more responsive to treatment than cortical areas. In contrast, microglial proliferation was not affected by genotype or by treatment, as evidenced by the unchanged Iba1-positive cells number in all conditions (Table 1).

The effects of SSR240612 on brain and vascular BK receptors were measured using western blot. B₁R protein levels were significantly increased only in the hippocampus from APP mice, as shown in the two mouse cohorts treated for 5 or 10 weeks (+41%, P <0.01; and +32%, P <0.05; respectively) (Figure 9A,B). SSR240612 treatment slightly reduced this upregulation in hippocampus, bringing B₁R protein levels in treated APP mice indistinguishable from those of WT controls after 10 weeks of drug delivery (Figure 9B, right). Neither genotype nor treatment altered B₂R protein levels in cortex and hippocampus. Knowing that B₂R and endothelial NOS (eNOS) activation mediate the effects of BK on brain arteries [33], their protein levels were measured in pial vessels. B₁R protein was not detected in vascular extracts (Figure 9C), confirming our immunohistochemical observations. However, in APP mice, B₂R and eNOS protein levels were increased in cerebral arteries (+40 to 50%; P <0.05), and both were normalized by SSR240612 administration independent of treatment duration (Figure 9C).

Previous in vivo studies in rats [17] or mice [18] showed increases in brain B₁R binding sites or protein levels after a single i.c.v. infusion of human Aβ_1-40. Here, with western blot analysis we found an upregulation of B₁R protein levels in the hippocampus, but not cerebral cortex of approximately 11- to 12-month-old APP mice. Furthermore, using the cellular resolution of immunocytochemistry, we confirmed that B₁R were upregulated in the hippocampus of APP mice, particularly in the hilus of the DG, a key segment of the entorhinal-hippocampal network for learning and memory that is impaired early by APP/Aβ overproduction [47]. Interestingly, recent evidence suggests that upregulation of B₁R might occur in early stages of AD-associated neuroinflammation [2], particularly in the hippocampus where components of the KKS are activated in response to inflammatory stimuli [48].

Astrocytes and microglia are instrumental in the neuroinflammatory processes associated with AD pathogenesis [49], and activation of astrocytes and microglia can be evidenced in the brain of APP mice with their respective typical association outside and within Aβ plaques [50]. Yet, the prominent finding from double-immunofluorescence experiments was that upregulation of B₁R was limited to astrocytes, mainly those located in the vicinity of Aβ deposits, that displayed characteristics of reactive astrocytes such as hypertrophic processes and soma, and upregulated GFAP immunostaining [51]. Our findings add to the previously reported upregulation of B₁R on glial or neuronal cells in various pathologies, including in the brain of epileptic patients [52] and spinal cord of diabetic rats [53].

In contrast to a recent report of B₁R upregulation in brain microvessels of Tg-SwDI mice, published online while our work was under review [54], we did not detect cerebrovascular B₁R expression in our 12-month-old APP mice. Beyond the use of different anti-B₁R primary antibodies, this apparent discrepancy is likely due to the robust deposition of cerebrovascular amyloid (CAA) in Tg-SwDI mice [55]. Indeed, J20 APP mice at the age used in our study are literally free of CAA [30]. Together, these findings reinforce our conclusion that astroglial B₁R upregulation is associated primarily, if not exclusively, with aggregated Aβ species.

We found that chronic B₁R antagonism improved spatial learning and memory in APP mice, abilities that depend largely on hippocampal integrity [56]. Moreover, prolonged B₁R blockade enhanced the baseline levels of the memory-related Egr-1 protein in the DG, a brain region previously associated in APP mice with cognitive deficits, reduced immediate-early gene expression and altered synaptic activity [57, 58]. Although further experiments would be required to confirm this hypothesis, it is conceivable that under chronic B₁R blockade with SSR240612 in APP mice, BK can then act on its normal target and preferentially activate constitutive B₂R that are considered neuroprotective and able to prevent memory loss [59].

The findings of decreased soluble Aβ_1-42 and Aβ plaque load in both cortex and hippocampus after SSR240612 treatment suggested that B₁R blockade interfered with the amyloidogenic cascade and, particularly, with the deposition of the fast aggregating Aβ_1-42 species. Interestingly, long-term B₁R blockade in APP mice resulted in increased production of MMP9 that is known to promote non-amyloidogenic processing of APP [60], and in increased cerebrovascular LRP-1 immunoreactivity, a vascular mediator of Aβ efflux from brain to blood [61]. Our results may thus suggest a stimulated Aβ clearance at the BBB following B₁R blockade. To date, both soluble [62, 63] and insoluble Aβ levels have been incriminated in cerebrovascular and cognitive deficits observed in APP mice [64, 65], but there is also strong evidence for no relationship between the Aβ pathology and cognitive performance in APP mice [30, 66, 67]. We cannot ascertain whether or not the normalized memory performance in APP mice after B₁R blockade is related to the concurrent decreases in soluble Aβ_1-42 species and Aβ plaque load. However, the relationships between high Aβ plaque load and low cognitive scores - particularly striking in the DG - and of high Aβ plaque load and low LRP-1 immunostaining (Figure 10) were shifted toward low plaque load and high cognitive performance, and toward low plaque load and high LRP-1 following SSR240612 treatment, suggesting a negative impact of Aβ on cognition possibly due to its lack of clearance from brain.

Aβ is the main initiating factor of the inflammatory cascade in APP mice, and specific Aβ domains can stimulate the KKS [68], and it has been argued that brain injury in AD is primarily caused by Aβ-induced neuroinflammation [69]. Interesting to the present study was that B₁R antagonism selectively reduced microglial activation, and most strikingly reactive microglia associated with Aβ plaques. Suggestive of microglial migration and aggregation around plaques, we show that despite an unchanged total microglial cells number in APP mice brain, the intensity of anti-microglia staining is increased in APP compared to WT controls, and significantly reduced following B₁R blockade. It thus seems that the reduction in Aβ plaque load contributed to the silencing of the reactive microglial cells. This would be consistent with Aβ-induced migration of phagocytic microglia, and with plaque-associated reactive microglia contributing to the inflammatory response and exhibiting a neurotoxic phenotype [70, 71]. Intense and patchy GFAP-immunostained astrocytes persisted after SSR240612 treatment, suggesting that astrocytes have maintained a reactive phenotype, as supported by the enduring increases in B₁R protein levels in activated astrocytes in APP mouse hippocampus (Figure 2). This might indicate a compensatory astroglial activation in response to the reduction in microgliosis [72], possibly related to the ability of reactive astrocytes to attenuate microglia-derived neurotoxicity [44].

A remarkable outcome from the present study was the normalization of CBF responses and cerebrovascular reactivity in adult APP mice by chronic B₁R antagonism, irrespective of the treatment duration. Growing evidence supports cerebrovascular impairment as an early event of AD pathogenesis [1]. Since B₁R blockade failed to reduce soluble Aβ_1–40 in APP mice, commonly perceived as a seed for Aβ_1–42 deposition in brain vessels leading to CAA [73], recovery of cerebrovascular reactivity in APP mice likely happened through reduction of soluble Aβ_1–42 since they display virtually no CAA at this age [30]. Aβ_1–42 is detrimental to cerebrovascular function [45] through increased oxidative stress [31, 74] and inflammation [75, 76]. Hence, cerebrovascular recovery likely resulted, at least in part, from the ability of SSR240612 to antagonize the B₁R-mediated production of vascular NADPH oxidase-derived reactive oxygen species [10] and inflammatory cytokines [77], as recently documented in aorta from a rat model of insulin resistance. In treated APP mice, SSR240612 normalized the increased levels of B₂R and eNOS, two key proteins in endothelial-mediated vasomotor responses, which are known to be upregulated by oxidative stress and inflammation [8]. Hence, the high eNOS protein levels in APP mice may reflect a compensatory upregulation in response to the reduced NO bioavailability following its trapping by reactive oxygen species [31, 74], as also reported in a model of diabetes-associated vascular disease [78].