Vascular amyloidosis impairs the gliovascular unit in a mouse model of Alzheimer’s disease

Animals

The hAPPJ20 line overexpresses the hAPP (human APP) minigene with Swedish (K670M/N671L) and Indiana (V717F) mutations under control of the PDGFB promoter (Mucke et al., 2000). Mice were on a congenic C57BL/6 J background (>15 backcrosses). Hemizygote males were bred with C57BL/6 females to produce hAPPJ20 mice used for experiments and littermate non-transgenic controls. Male and female mice were used for all experiments (Supplementary Table 1). Mice were kept on a 12-h light/ 12-h dark cycle and had ad libitum access to food and water. The studies were approved by the Institutional Animal Care and Use Committees of the University of Alabama at Birmingham and were conducted in compliance with the National Institutes of Health’s ‘Guide for the Care and Use of Laboratory Animals.’

Electron microscopy

hAPPJ20 and control animals were transcardially perfused with phosphate-buffered saline (PBS) followed by a fixation solution of 2.5% glutaraldehyde and 4% paraformaldehyde in 0.1 M sodium cacodylate buffer, pH 7.4. After fixation, the brain was removed and sectioned into 500–1000 μm slices. Under a dissection microscope, tissue punches were taken to capture cortical grey matter. Punches were placed in fixation solution overnight at 4°C. Samples were washed twice for 15 min in 0.1 M sodium cacodylate buffer, pH 7.4 and then immersed in 1% osmium diluted in 0.1 M sodium cacodylate buffer, pH 7.4 for 50 min at room temperature on a rotator in the dark. Samples were then washed four times in 0.2 M sodium cacodylate buffer, pH 7.4. Samples were dehydrated in 50%, 80%, and 95% acetone for 5 min each. Four washes in 100% acetone for 15 min were performed. Samples were equilibrated for embedding in a solution containing equal parts of 100% acetone and Epon^® 812 with accelerator on a rotator overnight at room temperature. Samples were embedded in 100% Epon^® 812 with the accelerator at 60–70°C overnight. Ultrathin sections were cut and imaged on a Tecnai T12 120 kV Transmission Electron Microscope (FEI).

For 3,3'-diaminobenzidine (DAB) immunoelectron microscopy, paraformaldehyde-fixed 40-μm thick brain slices were incubated in 0.1% glutaraldehyde/ 4% paraformaldehyde in PBS for 2 h at room temperature followed by wash with PBS. This was followed by incubation in 1% sodium borohydride for 15 min at room temperature. The tissue was prepared, stained, and developed using the Mouse On Mouse kit combined with a DAB stain by Vector Laboratories according to manufacturer’s instructions [primary antibody: mouse amyloid-β, clone W0-2 (Millipore; Cat. # MABN10)]. The tissue was then processed for electron microscopy as described above.

Immunohistochemistry

Tissue of hAPPJ20 and control mice was fixed in 4% paraformaldehyde in PBS. Thirty to fifty micrometre coronal slices were sectioned at the vibratome. Slices were incubated in primary antibody solution diluted in staining solution (10% normal goat serum and 0.5 % Triton^™ X in PBS) overnight at 4°C. Primary antibodies used included rabbit aquaporin-4 (Sigma Aldrich; Cat. # A5971; 1:400), chicken GFAP (Abcam; Cat. # ab4674, 1:1000) or mouse GFAP (Millipore; Cat. # MAB360, 1:1000), rabbit S100 (Dako; Cat. # Z0311, 1:100), mouse amyloid-β, clone W0-2 (Millipore; Cat. # MABN10, 1:250) and rat CD31 (BD Biosciences; Cat. # 550274; 1:150). Slices were washed in PBS for at least 10 min twice and stained with secondary antibodies for 1 h at room temperature. Secondary antibodies used included goat anti-rat Alexa Fluor^® 647, goat anti-rabbit Cy3, goat anti-rabbit Alexa Fluor^® 488, and goat anti-mouse Alexa Fluor^® 488. The nuclei stain DAPI (4',6-diamidino-2-phenylindole) was added to the secondary antibody solution. Slices were washed with PBS and mounted on glass slides with Aqua-Poly/Mount (Polysciences). Images were acquired using an Olympus FluoView FV1000 laser scanning microscope equipped with four diode lasers (405, 473, 559, 635 nm) using 20× air, 40× air, or 40× oil objectives (Olympus). Experiments were successfully repeated in at least three different animals per group; limitations in repeatability of immunohistochemistry or the described results were not encountered.

Quantification of GFAP and AQ4 (encoded by Aqp4) was performed as follows: three or four confocal images were taken per genotype and time point in the cortical grey matter of three different tissue slices using a 20× objective. Images were imported into ImageJ (NIH). Background was subtracted using a rolling ball radius of 50 pixels. Binary images were made and the area covered by AQ4 or GFAP signal was determined using the ‘create selection’ and ‘make measurement’ functions in ImageJ. Values for area measurements are plotted in per cent area. To determine the ratio of polarized (end-foot) to cytoplasmic AQ4 we followed the method described in (Kress et al., 2014). In short, plot profiles of regions of interest (5 μm × 50 μm) were placed such that the long side was oriented perpendicularly to the measured vessel. Signal intensities of the signal at the end-foot and of the surrounding cytoplasm were determined and a ratio between these values was calculated for every vessel.

Drugs and solutions

All chemicals were obtained from Sigma Aldrich unless otherwise specified. 9,11-dideoxy-9α,11α-methanoepoxy prostaglandin F_2α (U-46619) was obtained from Cayman Chemical and diluted with dimethyl sulphoxide. (±)-1-Aminocyclopentane-trans-1,3-dicarboxylic acid (t-ACPD) was obtained from Tocris Bioscience.

Acute brain slice drug application experiments

Animals were sacrificed and brains were placed in ice-cold artificial CSF (cutting artificial CSF) bath solution (135 mM N-methyl-D-glucamine, 1.5 mM KCl, 3.5 mM MgSO₄, 0.5 mM CaCl₂, 1.5 mM KH₂PO₄, 23 mM choline bicarbonate, 25 mM D-glucose and 0.4 mM ascorbic acid, bubbled with 95% O₂/5% CO₂ to maintain pH at 7.4). Two hundred-micrometre coronal slices were sectioned and allowed to recover for 45 min to 1 h at 28°C in recording artificial CSF (125 mM NaCl, 3.0 mM KCl, 1.25 mM NaH₂PO₄, 25 mM NaHCO₃, 25 mM D-glucose, 2.0 mM CaCl₂, 2.0 mM MgSO₄) bubbled with 95% O₂/5% CO₂ to maintain pH at 7.4_. For t-ACPD experiments, vessels were preconstricted in 125 nM U-46619 for 30 min in recording artificial CSF bubbled with 20% O₂/5% CO₂ at 37°C. Slices remained in 20% O₂/5% CO₂ for the duration of the experiment. All other pharmacological experiments were performed without preconstriction in 95% O₂/5% CO_{2 .} Acute brain slices were transferred to a Leica DM LFSA upright DIC-IR fluorescence microscope using a Zeiss Hm digital camera at 1000× magnification (Leica Microscopy). Slices were continually perfused with recording artificial CSF heated to 32°C and bubbled with 20% O₂/5% CO₂. Time lapse images of vessels were acquired using AxioVision 4.7 software (Carl Zeiss Microscopy). Vessel diameter was measured as the cross-section of the vessel using FIJI (ImageJ) software.

Acute brain slice calcium uncaging experiments

After obtaining slices as described above, slices were allowed to recover for 1 h in artificial CSF bubbled with 95% O₂/5% CO₂ at room temperature to maintain pH at 7.4. Slices were loaded with 20 μM Alexa Fluor^® 633 Hydrazide (Invitrogen), 10 μM Rhod-2 AM cell permeant Ca²⁺ indicator (Invitrogen), and 10 μM DMNPE-4 caged Ca²⁺ (Ellis-Davies and Barsotti, 2006) in artificial CSF containing 0.02% pluronic acid for 60 min at room temperature. Slices were then transferred to regular artificial CSF for 30 min to allow for stabilization before recording. Alexa Fluor^® 633 Hydrazide and Rhod-2 AM fluorescence were detected by laser scanning confocal microscopy using an Olympus FV1000 and a 60× /1.1 NA water immersion objective (Olympus). Single plane confocal images of 800 × 800 pixels were obtained about every 3 s. Ca²⁺ uncaging was achieved using a 405 nm laser in region of interest tornado scanning mode at 40% power for 800 ms that was directed at astrocytic end-feet associated with arterioles identified by Alexa Fluor^® 633 Hydrazide dye. The effect of laser scanning on caged Ca²⁺ and arteriole response during confocal image capture was assessed by recording vessel diameter and Ca²⁺ changes in the presence and absence of the higher powered uncaging pulse (Supplementary Fig. 1). Vessel diameter was measured as the cross-section of the vessel using FIJI (ImageJ) software and validated with a motion tracking system (After Effects, Adobe Systems Inc.) that automated vessel wall tracking and calculated the changes in lumen diameter (Supplementary Video 1).

In vivo multi-photon imaging through a cranial window

Before surgery, animals were anaesthetized with 2–5% isoflurane and given preoperative analgesics and antibiotics. After surgical plane anaesthesia was achieved, the hair and skin of the skull was removed. A metal bar was secured to the skull behind lambda by dental cement. Removal of the skull anterior to lambda and posterior to bregma was performed on one hemisphere and a round cover glass was affixed and sealed with dental cement. After a recovery period, animals were injected retro-orbitally with 100 µl of 5 mg/ml 70 000 MW tetramethylrhodamine dextran (Invitrogen) with a 28-gauge, 0.5-inch long needle (BD Biosciences) before commencing visualization. During multi-photon imaging, animals were kept under light anaesthesia [∼100 beats per minute (bpm)] and their vitals were constantly monitored (MouseOx, Starr Life Sciences Corp.). To visualize amyloid, methoxy-XO4 (2 mg/kg) was administered to hAPPJ20 and non-transgenic animals 12 h before the imaging experiment via intraperitoneal injection. Animals were placed on a custom built apparatus with a heating pad, and the metal bar was secured to immobilize the animal’s skull. Arterioles were identified by retro-orbitally administering 100 µl of 200 µM Alexa Fluor^® 633 Hydrazide (Invitrogen). Optical sections were acquired using a four-channel multi-photon laser scanning fluorescence microscope (Olympus FV1000MPE) equipped with a XLPLN25X/1.05 NA water-immersion objective (Olympus). Z projections were created using FIJI (ImageJ) software and 3D reconstructions were created using Imaris 7.5.2 software (Bitplane Scientific Software). When necessary, channel unmixing was done using FIJI (ImageJ) software. Vessel diameter was measured as the cross-section of the vessel lumen at the area of maximum response. Two micrometre cross-section slices at the area of maximum response were converted to binary images and then analysed and measured using particle analysis and Feret’s diameter (FIJI).

In vivo vascular smooth muscle cell activation

Arterioles, identified by retro-orbitally administered Alexa Fluor^® 633 Hydrazide (Invitrogen), with lumen diameters of 15–35 µm were selected only if they ran parallel to the cortical surface. This allowed for comprehensive imaging of vessel dynamics across larger parts of the vessel. Imaging took place at subsurface depths of up to 200 µM. The VSMC-containing media layer of the arterial wall was targeted for laser irradiation by measuring 4 µm from the luminal surface (labelled by retro-orbital injection of 70 000 MW tetramethylrhodamine dextran). VSMC stimulation was achieved using a 5 µm² region of interest in tornado mode for 100 ms at 7–10× imaging power levels. The optical path and laser source for VSMC activation were the same as those used in the two-photon fluorescence imaging (modified from a technique previously described; Choi et al., 2010).

Statistical analysis

Statistical tests were performed using GraphPad Instat or GraphPad Prism 6.0. All measurements are reported with standard error of mean. Statistical data are reported in the figure legends or results. The Kolmogorov–Smirnov normality test was performed for each data group. If normality was not achieved, the appropriate non-parametric test was performed. The specific test is stated in the results or figure legends. Outliers were excluded from the analysis where noted. The number of replicates is reported in the results or figure legends.

Patterns of vascular amyloid in hAPPJ20 mice

In patients with Alzheimer’s disease, amyloid-β accumulates around the cerebrovasculature (Mandybur, 1975; Vinters, 1987). This vascular amyloid deposition is replicated in mouse models of the disease (Yang et al., 2011; Park et al., 2014). To visualize vascular amyloid in the brains of aged hAPPJ20 mice, we injected methoxy-XO4 (Klunk et al., 2002) intraperitoneally before in vivo imaging by multi-photon microscopy. Multi-photon optical slice data were captured through a chronic cranial window in live hAPPJ20 mice using TRITC-Dextran to label the lumen of the cerebrovasculature. Similarly to previous studies in other models, we found a close relationship between amyloid-β and the cerebral vasculature in hAPPJ20 mice (Fig. 1 and Supplementary Video 2).

To better understand the physical relationship between vascular amyloid and its associated vessels we created high-resolution 3D reconstructions and volumetric optical section projections from in vivo and ex vivo multi-photon and confocal data. Three dimensional reconstructions illustrate how the bands of amyloid can span the entire circumference of vessels in some areas (Fig. 1B) but not others (Fig. 1C). In other Alzheimer’s disease models, vascular amyloid forms a banding pattern between established VSMCs (Domnitz et al., 2005), and we observed a similar pattern in hAPPJ20 mice (Fig. 1D–F). As the disease progresses, the gaps shrink and vascular amyloid starts to form contiguous structures (Domnitz et al., 2005). In high-resolution confocal z-projections, vascular amyloid had a banding pattern resembling an exoskeleton around blood vessels (Fig. 1D). Three dimensional reconstructions illustrated this exoskeleton framework at even higher detail, revealing the formation of bridges between the amyloid bands (Fig. 1E). Quantitative assessment of multi-photon 3D reconstruction data showed that vascular amyloid volume was ∼20% of total vessel volume (Fig. 1G and Supplementary Fig. 2). Taken together, vascular amyloid appears to form a vascular exoskeleton covering a significant surface area of cerebral blood vessels in aged hAPPJ20 mice.

Vascular amyloid displaces astrocyte end-feet in hAPPJ20 mice

As astrocyte end-feet cover over 99% of the cerebrovascular surface (Mathiisen et al., 2010) (Fig. 2A), we next asked if vascular amyloid deposits interfered with astrocyte end-foot placement. To answer this question, we first imaged acute brain slices from hAPPJ20 mice using confocal imaging. To label vascular amyloid, mice were injected with methoxy-XO4 12 h before the experiment. Bath application of the Alexa Fluor^® 633 Hydrazide dye was used to stain the elastin layer of arterial vessels ≥10 μm. As there is extensive literature reporting preferential loading of astrocytes by fluorescent acetoxymethyl ester indicators such as Rhod-2 AM (Ellis et al., 1979; Harder et al., 1995; Niwa et al., 2000; Hirase et al., 2004; Tian et al., 2005; Takano et al., 2006), we used this label to visualize astrocyte end-feet. We found that amyloid deposits mostly displaced astrocytes and the end-feet no longer abutted the arteriole elastin layer in areas where vascular amyloid was present (Fig. 2B). However, there were a few locations where the end-foot was still in its normal vessel juxtaposition (Fig. 2C) as seen in non-transgenic animals (Fig. 2D). We cannot exclude that Rhod-2 AM loads into endothelial cells or pericytes as previously reported for Fluo-4, albeit this has been reported to occur only with poor efficiency and attenuated strength of the fluorescent label (Hirase et al., 2004). Furthermore, endothelial cells should localize inside of the Alexa Fluor^® 633 Hydrazide label, which stains the elastin layer of arterial vessels (Fig. 1F), but this was not where we observed Rhod-2 AM staining. The contractile cells of vessels with this diameter are vascular smooth muscle cells, not pericytes.

As Rhod-2 AM does not label all astrocytes, we complemented the above studies by labelling astrocyte end-feet with antibodies to AQ4, a water channel (encoded by Aqp4) exclusively expressed in astrocytes and highly enriched in their end-feet (Bragg et al., 2006) (Fig. 2A and Supplementary Fig. 3B). We observed a decrease of AQ4 expression in the cortical grey matter as early as 3 months in hAPPJ20 mice, even before the appearance of amyloid-β plaques, vascular amyloid (Esposito et al., 2006), or pronounced astrogliosis (Fig. 2E and F). Similar to previous reports (Kress et al., 2014), AQ4 expression tended to decrease in aged control animals (Fig. 2E). As the brain tissue of aged animals is characterized by severely increased autofluorescence, in part due to lipid granule formation (Eichhoff et al., 2008) (also see Supplementary Fig. 4), the values shown are likely an overestimation of true AQ4 levels and differences between controls and hAPPJ20 mice were difficult to establish (Fig. 2H). To avoid autofluorescence and assess if AQ4 expression might also be mislocalized, as previously described in a different Alzheimer’s disease mouse model (Yang et al., 2011), we determined the ratio between end-foot labelling and cytoplasmic AQ4 expression. Brains from hAPPJ20 mice had a significantly decreased ratio of end-foot to cytoplasmic AQ4 staining (Supplementary Fig. 3A, B and E), indicating that AQ4 was indeed mislocalized in hAPPJ20 mice. This was corroborated by a significantincrease in intensity of cytoplasmic AQ4 immunolabeling in hAPPJ20 brains (Supplementary Fig. 3C, D and F).

Loss of AQ4 has been reported in other CNS pathologies presenting with reactive astrocytes (Wolburg-Buchholz et al., 2009; Noell et al., 2012; Iliff et al., 2014). Thus, we used the astrocyte-specific marker GFAP to compare the extent of astrogliosis at different ages in hAPPJ20 mice. Interestingly, at 3 and 9 months of age, when the area covered by AQ4 was already reduced, GFAP expression in hAPPJ20 mice was comparable to age-matched littermate controls. At 12–16 months of age, we observed an increase in reactive astrocytes around amyloid plaques as previously reported in a different Alzheimer’s disease mouse model (Sirko et al., 2013). At 24 and 29 months of age, astrogliosis was widespread throughout the cortical grey matter in hAPPJ20 mice (Fig. 2E–G). To a lesser degree, we also found GFAP-positive hypertrophied astrocytes in 24-month-old control animals, consistent with literature on age-related changes of the brain (Rodriguez-Arellano et al., 2015). However, the area covered by GFAP-positive astrocytes was not significantly increased in aged control animals (Fig. 2G). Overall astrocyte numbers were assessed by immunohistochemistry for S100. Astrocyte density was comparable in 29-month-old control and hAPPJ20 mice (Supplementary Fig. 5) pointing toward chronic astrogliosis in the absence of an increase in astrocyte numbers in hAPPJ20 mice. As AQ4 expression was already reduced prior to visible vascular amyloid deposits, this is unlikely to reflect a physical displacement of the astrocyte end-feet but instead a decrease in overall AQ4 expression and loss of polarized expression with preservation of the end-feet along the vessel.

Because AQ4 immunohistochemistry was not a good indicator for end-foot displacement, we next used confocal imaging and high-resolution electron microscopy (Fig. 2I–K and Supplementary Fig. 4) to assess the structural relationship of astrocyte end-feet and blood vessels in control and hAPPJ20 mice. We identified astrocyte end-feet, vascular amyloid, endothelial cells, and the lumen of vessels by differences in electron density and morphology, and confirmed this by immunolabelling of vascular amyloid in electron microscopy (Supplementary Fig. 6) and immunohistochemistry (Supplementary Fig. 3C and 7A). Astrocyte end-feet are characterized by a less electron-dense (lighter) cytoplasm with substructures of filamentous networks and bundles of glial filaments in addition to glycogen granules (Nakazawa and Ishikawa, 1998), whereas amyloid deposits presented as relatively darker structures close to the vessel wall. In control mice, we did not observe these darker structures. Instead, endothelial cells were completely covered by astrocytic end-feet (Fig. 2I and Supplementary Fig. 4 and 6). In hAPPJ20 mice, electron-dense amyloid-β deposits were found between the astrocytic end-feet and the vessel walls (Fig. 2J), although some vessels maintained astrocyte-vessel contact (Fig. 2K). Labelling vascular amyloid with an antibody against amyloid-β and astrocyte end-feet with an antibody to GFAP corroborated this finding (Supplementary Fig. 7A). While some larger vessels did show coverage of GFAP⁺ end-feet in controls and hAPPJ20 mice (Supplementary Fig. 7A and B), most end-feet along the cortical cerebral vasculature expressed GFAP levels below the detection threshold (Supplementary Fig. 7C and D) excluding GFAP as a suitable marker for the presence or absence of an end-foot on these vessels. Together these data suggest that astrocyte polarity is altered in hAPPJ20 mice and that vascular amyloid can disrupt the intimate end-foot–vessel relationship.

The vasoactive signalling axis in hAPPJ20 mice

The brain regulates blood flow in response to neuronal activity and energy demands, a process called functional hyperaemia. This regulation is partially governed by astrocytes, which act as liaisons between neurons and the vasculature (Attwell et al., 2010). Specifically, activation of astrocytic metabotropic glutamate receptors (mGluRs) leads to increased intracellular calcium ([Ca²⁺]_i) and a phospholipase A₂-mediated generation of vasoactive molecules that dictate arteriole smooth muscle cell constriction and dilation, dependent on the brain’s metabolism (Zonta et al., 2003; Gordon et al., 2007; Girouard et al., 2010). While one study raised doubts about the presence of mGluR5 (encoded by Grm5) in astrocytes in adult mice (Sun et al., 2013), a more recent publication confirmed the lack of mGluR5 pharmacologically but found astrocytic Ca²⁺ signals in response to group I mGluR agonists using a more sensitive Ca²⁺ indicator (Tang et al., 2015). This suggests that astrocytes respond to mGluR agonists such as trans-ACPD (t-ACPD) with Ca²⁺ increases through mGluR1 (Grm1) activation. We used a well-established slice assay in combination with pharmacology (Gordon et al., 2007; Girouard et al., 2010; Watkins et al., 2014) to assess if this signalling axis remained intact in hAPPJ20 mice.

Astrocytes respond to the vasoactive molecules t-ACPD and norepinephrine with increased [Ca²⁺]_i, which, depending on the oxygen tension in the tissue, cause either vasodilatation or vasoconstriction (Zonta et al., 2003; Mulligan and MacVicar, 2004; Gordon et al., 2007; Tang et al., 2015). Using video-microscopy analysis, we measured changes in arteriole diameter in response to bath application of t-ACPD or norepinephrine in acute slices from control and hAPPJ20 mice (Fig. 3A and B). As expected, most arterioles in control slices dilated in response to t-ACPD and a paired analysis indicated a highly significant effect of t-ACPD (Fig. 3A, C and D). However, in hAPPJ20 vessels, there was no significant effect of t-ACPD in a paired analysis (Fig. 3D), suggesting impaired dilation of vessels in hAPPJ20 mice in response to t-ACPD. A similar effect was observed with norepinephrine, which induced significant vasoconstriction in control vessels (Fig. 3B, F and G) but not in hAPPJ20 vessels (Fig. 3G). However, in direct comparisons of the response magnitude (Fig. 3E and H), we did not observe significant differences, thus these data do not unequivocally support an abnormality of vascular responsiveness in hAPPJ20 mice. Statistical power in these analyses was reduced by high variability in these responses, especially in the hAPPJ20 group, likely because in these experiments we had no means of identifying which vessels were affected by vascular amyloid. Suspecting that the high variability in hAPPJ20 mice was due to inclusion of both vessels with and without vascular amyloid, in subsequent experiments we included vascular amyloid labelling to allow differentiation between affected and unaffected vessels.

Gliovascular coupling is impaired on vessels covered with vascular amyloid

To visualize vascular amyloid, we injected methoxy-XO4 12 h prior to sacrificing control or hAPPJ20 animals and used confocal imaging to identify vessels with or without vascular amyloid. To determine changes in vessel diameter, we used the arteriole-specific dye Alexa Fluor^® 633 Hydrazide. Furthermore, to avoid possible off-target effects of the t-ACPD and norepinephrine that raise [Ca²⁺]_i in astrocytes, we specifically elevated [Ca²⁺]_i in individual astrocytes using the caged Ca²⁺ compound, DMNPE-4 (Ellis-Davies and Barsotti, 2006), in combination with the Ca²⁺-sensor Rhod-2 AM.

Focal laser uncaging of the caged Ca²⁺ DMNPE-4 allowed us to elevate [Ca²⁺]_i in single astrocytes on or close to arterioles (Fig. 4). Cortical grey matter arterioles in acute slices from control mice, and arterioles with and without amyloid in hAPPJ20 mice, were imaged during selective astrocyte stimulation by focal laser Ca²⁺ uncaging. In acute slices from control mice, Ca²⁺ uncaging in astrocytes led to an immediate rise in [Ca²⁺]_i followed by biphasic vessel constriction with an immediate constriction followed by a slower sustained vessel response (Fig. 4A and Supplementary Video 3). In slices from hAPPJ20 mice, uncaging in astrocytes near arterioles resulted in similar increases in [Ca²⁺]_i (Fig. 4B and C). To determine if the impaired vasoconstriction in hAPPJ20 mice was due to the amyloid burden on the vasculature, we compared the vasoconstriction of arterioles with and without vascular amyloid (Fig. 4B–E). In hAPPJ20 mice, the response of arterioles without amyloid was comparable to arterioles from control mice (Fig. 4B, D, E and Supplementary Video 4). In contrast, the presence of vascular amyloid significantly decreased vasoconstriction in both the initial and the maximal response (Fig. 4C–E and Supplementary Video 5). To ensure that this effect was not due to variability in Ca²⁺ release we quantified the mean Ca²⁺ ΔF/F and found no significant difference between groups (Fig. 4F). These data suggest that arterioles burdened by vascular amyloid have a diminished range of contractility.

Vascular smooth muscle cell/pericyte function is impaired in hAPPJ20 mice

Astrocytes release vasoactive substances that stimulate contractile VSMCs or pericytes on arterioles or capillaries, which then constrict or dilate (Attwell et al., 2010). Hence, the reduction in vessel contraction following astrocyte stimulation could also result from impaired VSMC/pericyte function. To assess VSMC/pericyte function directly, we used drugs that bypassed the astrocyte component of the GVU and instead stimulated VSMCs/pericytes directly (Fig. 5). For this we chose endothelin-1 (ET-1, encoded by Edn1), a potent vasoconstrictor that binds to endothelin receptors found on VSMCs (Edwards and Trizna, 1990), and U46619, a thromboxane A₂ agonist vasoconstrictor that binds to TXA₂ receptors on VSMCs (Dorn and Becker, 1993; Blanco et al., 2008). Using video-microscopy, we measured changes in arteriole diameter in response to bath application of ET-1 (Fig. 5A) and U46619 (Fig. 5B). Arterioles from hAPPJ20 mice had significantly less vasoconstriction in response to ET-1 (Fig. 5 A, C and E) and U46619 (Fig. 5B, D and E) compared to control mice. Taken together, these data suggest that the ability of VSMC/pericytes to properly function in regulating vessel diameter is impaired in hAPPJ20 mice.

To assess if a greater initial vessel tone in hAPPJ20 mice could account for decreases in vessel responses, we compared baseline diameters of both capillaries and arterioles. Neither baseline capillary, nor arteriole diameters were significantly changed in hAPPJ20 mice (Supplementary Fig. 8). These data indicate that the reduced constriction of vessels seen in hAPPJ20 mice cannot be due to a difference in baseline diameter.

Vascular amyloid dampens the vessel response in vivo

An important next step was to examine potential changes in VSMC function in vivo, where blood flow is present and vessels are pressurized. For these experiments we adopted minimally invasive in vivo multi-photon imaging. We visualized the vasculature and amyloid through a chronic cranial window in live mice. The vascular lumen was labelled with TRITC-Dextran, arterioles with Alexa Fluor^® 633 Hydrazide, and vascular amyloid by intraperitoneal injection of methoxy-XO4 12 h before the imaging session. We took advantage of direct focal optical stimulation of VSMC (Choi et al., 2010) to elicit a vascular response in vivo. With this technique, we were able to consistently elicit and visualize vasoconstriction by laser-stimulation of single VSMCs. VSMCs were identified by selecting arterioles positive for Alexa Fluor^® 633 Hydrazide and placing the stimulation focal point at 4 µm distance from the border of the vessel lumen as previously described (Choi et al., 2010). We compared the vasoconstriction of arterioles with and without vascular amyloid in hAPPJ20 mice. Arterioles without vascular amyloid constricted after activation of VSMCs (Fig. 6A, E and Supplementary Video 6), while constriction of arterioles with vascular amyloid was reduced by more than half (Fig. 6B, E and Supplementary Video 7). In vivo, we observed a rapid response immediately after stimulation of the VSMC. The observed change in vessel diameter (Fig. 6E) was comparable to the maximum response observed after pharmacological stimulation of VSMCs in brain slices (Fig. 5E).

To determine if the impaired vasoconstriction was restricted to vessel areas with vascular amyloid only, we compared responses of areas with and without vascular amyloid on the same arteriole (Fig. 6C). Areas with vascular amyloid constricted to a lesser extent than areas just downstream of the amyloid on the same vessel (Fig. 6C and D). Laser-stimulation of a single VSMC caused subsequent vasoconstriction in multiple areas along arterioles without vascular amyloid (Supplementary Video 8) likely due to the activation of neighbouring VSMCs. Interestingly, when this chain reaction arrived at areas with vascular amyloid, there was less vasoconstriction than in adjacent amyloid-free areas (Supplementary Video 8). Like our ex vivo data, these in vivo studies show that arterioles have a reduced range of constriction specifically affecting areas covered by vascular amyloid.