Abstract
While cerebrovascular dysfunction and reactive astrocytosis are extensively characterized hallmarks of Alzheimer's disease (AD) and related dementias, the dynamic relationship between reactive astrocytes and cerebral vessels remains poorly understood. Here, we used jGCaMP8f and two-photon microscopy to investigate calcium signaling in multiple astrocyte subcompartments, concurrent with changes in cerebral arteriole activity, in fully awake 7- to 8-month-old male and female 5xFAD mice, a model for AD-like pathology, and wild-type (WT) littermates. In the absence of movement, spontaneous calcium transients in barrel cortex occurred more frequently in astrocyte somata, processes, and perivascular regions of 5xFAD mice. However, evoked arteriole dilations (in response to air puff stimulation of contralateral whiskers) and concurrent calcium transients across astrocyte compartments were reduced in 5xFAD mice relative to WTs. Synchronous activity within multicell astrocyte networks was also impaired in the 5xFAD group. Using a custom application to assess functional coupling between astrocyte end feet and immediately adjacent arteriole segments, we detected deficits in calcium response probability in 5xFAD mice. Moreover, end feet calcium transients following arteriole dilations exhibited a slower onset, reduced amplitude, and lacked relative proportionality to vasomotive activity compared with WTs. The results reveal nuanced alterations in 5xFAD reactive astrocytes highlighted by impaired signaling fidelity between astrocyte end feet and cerebral arterioles. The results have important implications for the mechanistic underpinnings of brain hypometabolism and the disruption of neurophysiologic communication found in AD and other neurodegenerative conditions.
Significance Statement
Astrocytes are an essential component of the neurovascular unit. Chronically reactive astrocyte phenotypes are mechanistically linked to deleterious features of Alzheimer's disease (AD) including impaired cerebral blood flow, hypometabolism, and synapse dysfunction/loss. Here, we use two-photon imaging to show that reactive astrocytes in a fully awake mouse model of AD-like amyloid pathology are spontaneously hyperactive, exhibit impaired functional connectivity, and respond to dilations in immediately adjacent arterioles with poor fidelity. The results reveal a key point of communication breakdown between the brain and the cerebrovasculature.
Introduction
The neurovascular unit, composed of cerebral blood vessels and multiple neural cell types, is essential for promoting brain health (Iadecola, 2017; Schaeffer and Iadecola, 2021). Chronic neurodegenerative diseases like Alzheimer's disease (AD) are characterized by pathognomonic proteinaceous deposits that occur alongside cerebrovascular dysfunction and hypometabolism (Veitch et al., 2019). Most of the cerebral vasculature is ensheathed by astrocyte end feet specializations, while other astrocyte processes are in direct contact with many, if not most, excitatory synapses (Verkhratsky and Nedergaard, 2014; Pivoriunas and Verkhratsky, 2021). Astrocytes also contain the requisite machinery for taking up, processing, and releasing metabolites and therefore represent a critical cellular liaison between blood and brain. In the course of AD and related dementia disorders (ADRDs), astrocytes become reactive and exhibit complex loss- and gain-of-function phenotypes (Escartin et al., 2021; Price et al., 2021; Liddelow et al., 2024). These clues have led us and others to hypothesize that astrocytes play key roles in the deleterious cerebrovascular phenotypes of chronic neurodegenerative conditions (Beard et al., 2021; Price et al., 2021; Stackhouse and Mishra, 2021; Takahashi, 2022).
Recent studies have investigated calcium signaling as a proxy for the altered function of reactive astrocytes in AD/ADRD mouse models (Takano et al., 2007; Kuchibhotla et al., 2009; Delekate et al., 2014; Lines et al., 2022; Shah et al., 2022; Kelly et al., 2023; Lee et al., 2023; Sompol et al., 2023). Findings have been mixed with some reports showing augmented calcium signaling in AD mouse models, characterized by elevations in spontaneous calcium transient frequencies and amplitudes, while others have shown no change or reductions in these parameters. To date, only a few reports on intact AD/ADRD mouse models have described calcium signaling in the end feet (Delekate et al., 2014; Kelly et al., 2023; Lee et al., 2023), which are the most proximal astrocyte subcompartments relative to the cerebral blood vessels. In these studies, end feet calcium was assessed either in the absence of concurrent vascular measures (Kelly et al., 2023) or during the administration of general anesthesia (Delekate et al., 2014; Lee et al., 2023), which can greatly suppress astrocyte signaling and/or alter vascular tone during brain stimulation (Thrane et al., 2012; Gao et al., 2017).
Here, we used the genetically encoded calcium indicator jGCaMP8f and intravital two-photon microscopy to investigate astrocyte calcium transients in multiple cellular compartments of fully awake 5xFAD mice (characterized by robust parenchymal brain amyloid pathology) and age-matched wild-type littermates. Astrocyte calcium signaling in individual subcompartments and astrocyte networks, concurrent with dynamic responses in penetrating arterioles, was assessed in the barrel cortex at rest and during air puff whisker stimulation of the contralateral vibrissae. Using a novel custom vascular modeling application, Localized Analysis of Vascular Astrocytes (LAVA), we also directly assessed temporal and proportional relationships between astrocyte end feet calcium transients and immediately adjacent arteriole segments. Our findings describe novel (dys)functional phenotypes of reactive astrocytes and may offer new insights into how energy coupling breaks down during AD and related dementias.
Materials and Methods
Mice
Male and female 5XFAD and wild-type (B6-SJL; Oakley et al., 2006) mice were purchased from The Jackson Laboratory (#034840-JAX) and bred to produce hemizygous and littermate control F2 mice. Mice were housed in standard laboratory cages (IVC) under 12 h light/dark cycles in a pathogen-free environment in accordance with University of Kentucky guidelines. Mice had access to food and water ad libitum. All animal procedures were conducted in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals and were approved by University of Kentucky Institutional Animal Care and Use Committees.
Adeno-associated virus (AAV) reagents
cDNAs for jGCaMP8f (Tran et al., 2023) were encoded in vectors downstream of the human GFAP promoter (GFA104 ABCD9). Plasmids were sent to Charles River Laboratories for packaging into AAV2/5 capsids. When injected into adult mice, AAV-GFA104 vectors drive expression of target transgenes selectively in astrocytes. Mice were injected with AAV2/5-GFA104-jGCaMP8f into barrel cortex (see below) at ∼6 months of age to detect and quantify astrocytic calcium transients.
AAV injection and cranial window implantation
Surgical techniques were performed as described in our previous studies (Case et al., 2023; Sompol et al., 2023) with some modifications. Briefly, mice were anaesthetized by 3% isoflurane (SomnoSuite, Kent Scientific) in an induction box and head-fixed to a stereotaxic frame. Anesthesia was maintained by continuous isoflurane inhalation (1.5–3%, via nose cone) during surgery. The injection was marked over the barrel cortex at 1.5 mm AP, ±3 mm ML. The injection site was used as the center point of a 3-mm-diameter craniotomy for cranial window implantation. The freed circular skull fragment was carefully removed with tweezers. AAV (1013 IFU/ml) was delivered at a rate of 0.2 µl/min (total 2 µl) into the brain using a Hamilton 701 10 µl syringe and microinjector (Stoelting QSI controller) at 200 µm depth. After injection, the needle remained in place for an additional 2 min prior to retraction. Then a glass cranial window, made from 3 and 4 mm #1 glass coverslips (0.13 mm) combined with optical glue, was carefully placed over the craniotomy with the smaller coverslip down, to cap the exposed brain in the skull opening. The window was stabilized and affixed to the skull along with a stainless-steel head mount (Proto Science Solutions) using dental acrylic. Analgesic was injected subcutaneously, and animals recovered from anesthesia before return to their home cages.
Two-photon microscopy
Mice were anesthetized and affixed to the imaging platform, via a custom-made fixture (Proto Science Solutions), on a Scientifica Hyperscope utilizing a tunable InSight X3 laser (Spectra-Physics), tuned to 920 nm at 80 mW. Laser power was controlled by Pockels cells (ConOptics Model 302RM) and calibrated before the start of each imaging session. Pockels cells were allowed to warm up for at least 30 min prior to calibration. Bias voltage was set to attenuate laser transmission to its minimum, near zero value at 0% control level. A laser power curve was then generated by measuring power at increasing control levels within the driver software (0–100% in 10% increments). The control level was then set at an output with the desired laser power. The microscope was equipped with two GaAsP photomultipliers (Hamamatsu Photonics), typically operated at a supply voltage of 700 (configured in ScanImage), and a water immersion objective (16×, 0.8 NA) from Nikon Instruments (N16XLWD-PF) with theoretical lateral and axial resolution of 0.53 and 2.68 µm. Fluorescence filters included a primary ZT405/473-488/NIRtpc, emission dichroic T565 LPXR, and fluorescence filters ET525/50 for PMT #1 and ET620/60 for PMT #2. ScanImage acquisition software (MBF Bioscience) was used to drive microscope components and data acquisition (512 × 512 pixels at 30 frames per second) within MATLAB (MathWorks). Mice were aligned under the objective lens and received single retroorbital injections of 500 kDa (5% w/v in saline) TRITC Dextran (#52194-1G, Sigma-Aldrich) to visualize cerebral vasculature. A small air pump, for delivering whisker air puffs, was gated by a solenoid control valve and connected to the vDAQ controlled through ScanImage (MBF Bioscience). Anesthesia was removed, and imaging commenced once mice were fully alert. Mice were considered fully alert when they responded to light stimulus (from a flashlight) and exhibited motor adjustments to the imaging stage (i.e., standing up).
Intravital imaging of vascular dynamics and astrocyte calcium signaling
Imaging sessions began by surveying the cranial surface for expression of jGCaMP8f indicator in barrel cortex. Fields of view (FOV) were selected that contained desirable features for analysis such as fields of fluorescent astrocytes, penetrating arterioles, and capillaries. Arterioles were identified as vessels originating from a penetrating vessel that branched or forked into smaller vessels in the direction of blood flow. For analyses, we chose arteriole segments (vessels of interest) that ran parallel to the imaging plane, which was then focused to the z point of the arteriole segment diameter. Data collection was balanced to ensure that consistent numbers and sizes of FOVs were collected and arterioles of consistent size were used for analysis. Once a FOV was selected, a baseline 5 min recording of vascular and astrocyte activity was acquired followed by a series of timed whisker stimulation trials. Stimulation consisted of 10 s air puff trains (9 Hz) to the contralateral vibrissae, controlled by a custom software trigger integrated into the ScanImage software's frame clock. Videos were captured that contained several astrocytes, cerebral capillaries, and arterioles within the FOV. A minimum 3 min interval separated stimulation trials. At the conclusion of the experiment, mice were then lightly anesthetized for removal from the imaging platform and placed back into their home cages.
Image analysis
Experimenters were blinded from genotype during all analyses, which were conducted using in-house Matlab algorithms described here. First, files containing time-lapse videos (i.e., .tifs) were motion-corrected by the widely used NoRMCoRRE algorithm (Pnevmatikakis and Giovannucci, 2017). Metadata containing scan speed, spatial dimensions, and stimulation timing were extracted for common use by analyses applications. Event-based detection methods were used to identify astrocytes in the FOV. Further segmentation of astrocyte compartments, such as soma and end feet, was accomplished through convolutional filtering and masking techniques that accentuate astrocyte features. Convolutional filtering was performed using a custom-developed tool. A custom filter was applied in Fourier space to accentuate astrocyte features, eliminate Poisson noise, and equalize unevenly illuminated FOVs. The technique is used nondestructively for masking purposes. Masking for soma ROI determination was performed by inputting time series vectors and calculated signal parameters into a principal component analysis and the k-means unsupervised learning clustering algorithm. Three clusters were determined to be both the minimum and optimal number of clusters for differentiating active cells from background pixels and fine processes. For astrocyte end feet, the perivascular space was masked prior to event-based ROI detection. In-depth analyses of vascular motion and spatially paired astrocyte calcium signaling were characterized using our custom MATLAB-based application called LAVA. Vascular motion was assessed by linearizing vessels of interest to better define spatial alignment between arteriole segments and end feet compartment activity. Cross-sectional vectors placed over arterioles of interest were used to determine cross-sectional area and to sample fluorescence values of adjacent end feet. The resulting “dataset” provides spatial-temporal information that can be used to quantify dynamic relationships between vascular tone, motion, and calcium fluorescence, which, in turn, can offer insights into the directionality of vascular tone fluctuations and potential calcium mediated changes. LAVA can also be used to correlate this activity in relation to other ROI segmentations (e.g., somata and capillary end feet signals mentioned previously and not shown in this report), for predefined trial phases (i.e., before, during, and poststimulation). The workflow for the application is as follows: In each FOV, dilating and contracting vessels of interest are identified and vessel boundaries are outlined. Vessels are then binarized and archived. The resulting data structure contains a time series vector of calculated vessel segment volumes and fully cross-sectioned and linearized vascular and calcium channel image stacks. User-defined perivascular ROIs are then identified from the linearized and scaled pseudo-images of fluorescence changes along the vessel of interest. Results from vascular motion measurements are optionally used to update ROI positions for arteriole end feet where vascular movement contaminates extracted calcium signals. The application then extracts and saves the resulting time series vectors from these ROIs, along with calculated maximum relative fluorescence changes. Correlations between end feet calcium signals and vascular activity were quantified by Pearson's r. Once ROI-based end feet analyses were completed, LAVA conducted within vessel comparisons of fluorescence time series data and local vascular tone, by correlations within cross-sectional vectors. Video outputs of processed data are available from within LAVA and are demonstrated in Movie 3. For astrocytic processes, the segmentation algorithm is repeated and performed last, with soma- and end feet-identified time series vectors excluded from clustering.
Postmortem measures and correlations to physiologic measures
After imaging, a small subset of mice was transcardially perfused with PBS, and brain tissue was harvested and placed in 4% paraformaldehyde. Twenty-four hours later, the tissue was transferred to 30% sucrose solution until brains sank to the bottom of the tube. Free-floating coronal mouse brain sections of 40 µm thickness were prepared and stored in antifreeze cryoprotectant solution (catalog #006799, Bioenno Lifesciences) until use. Prior to staining, sections were washed three times for 10 min in 0.1% PBS-T (PBS with 0.1% Triton X-100) at room temperature. For blocking, sections were incubated for 1 h at room temperature in 3% normal goat serum (GS) diluted in PBS with 0.1% Triton X-100. Primary antibodies were diluted in blocking solution and added to 1.5 ml tubes and then sections added. The following primary antibodies were used: rabbit anti-Aβ (Abcam ab201060) and mouse anti-GFAP (Cell Signaling GA5), each at a 1:1,000 dilution. Sections were incubated in primary antibodies overnight at 4°C on a shaker. Negative controls were incubated with blocking solution only. The following day, sections were washed in PBS-T and incubated with goat anti-rabbit Alexa Fluor 594 (A11072, RRID: AB_2534116) and goat anti-mouse Alexa Fluor 488 (A11029, RRID: AB_2534088) secondary antibodies, each at 1:200 dilution in PBS-T containing goat serum. Incubations were done for 2 h at room temperature with shaking. Sections were then washed three times for 10 min each in PBS-T. Finally, sections were mounted on slides and coverslipped using EverBrite TrueBlack Hardset Mounting Medium (catalog #23017) to reduce autofluorescence. Slides were imaged using the Nikon BioPipeline Slide Scanning System with a Plan APO 10× objective, and images were stitched together to visualize whole-brain slice photomicrographs. Analyses were performed on confocal micrographs acquired from user-defined regions over the barrel cortex using a Nikon confocal Tie2 microscope (Nikon Instruments). Two independent Z stacks were compiled and values averaged per mouse. Images were thresholded and denoised prior to assessment of fluorophore labeling area using NIS-Elements. Labeling measures were then compared within-animal to calcium and vasomotion parameters using Spearman's correlation and data fit using linear regression.
Statistics
Most calcium transient and vessel property parameters were compared across groups using unpaired Student’s t tests. Arteriole diameter changes before and during air puff stimulation were evaluated with repeated-measures ANOVA, followed by simple effects tests. Correlations between time series data were determined by Pearson's r tests. Specifically, for network synchronicity analysis, events were cropped to only the rise phase of calcium event transients before statistical analysis. Roughly equal numbers of males and females were used per genotype condition (WT: 5M and 6F; 5xFAD: 5M and 9F). The total group numbers for animals were 11 WT and 14 5xFAD. At least two FOVs per mouse were used and three replicate trials each for stimulation experiments. Not all parameters could be cleanly measured from each mouse (e.g., interruption in imaging session, lack of identifiable arterioles, poor expression of calcium indicator), resulting in exclusion from statistical analyses. Exclusion was performed before decoding of transgene groups. Unless, otherwise noted, statistical samples shown are average values obtained within each mouse. Significance for all statistical tests was set at p < 0.05. For most parameters evaluated, there were no significant sex differences in either transgene group and so males and females were combined together.
Software accessibility
The software described herein and associated documentation are available for download at https://github.com/BlaineWeiss/STONE-LAVA/ and the Norris Laboratory webpage https://norris.createuky.net/. Localized Analysis of Vascular Astrocytes (LAVA) and Spatial & Temporal Observation of Network Events (STONE) © 2024 University of Kentucky were developed by Blaine Weiss and are distributed for noncommercial academic use only (see repository for full terms). Documentation is shared under the Creative Commons Attribution-Non-Commercial-Share-Alike 4.0 License (CC BY-NC-SA 4.0).
Results
Spontaneous calcium transients in 5xFAD mice are smaller but occur more frequently in astrocyte somata and processes
Studies were performed on male and female 5xFAD mice and age-matched WT littermates (7–8 months of age). The 5xFAD strain is an aggressive rapidly developing model of parenchymal amyloid pathology (Oakley et al., 2006) but exhibits comparatively less vascular amyloid (Szu and Obenaus, 2021). Nonetheless, 5xFAD mice do show some cerebral vessel abnormalities along with deficits in cerebrovascular function including capillary stalls (Cruz Hernandez et al., 2019), impaired neurovascular coupling (Mughal et al., 2021), and reduced cerebral blood flow (Igarashi et al., 2020). Like most other amyloid mouse models, 5xFAD mice also exhibit key phenotypes associated with AD/ADRDs including synaptic deficits, neuronal hyperexcitability, and cognitive loss (Kimura and Ohno, 2009; Crouzin et al., 2013; MacPherson et al., 2017; Sompol et al., 2017). By 8 months, 5xFAD mice exhibit profound amyloid pathology and extensive astrocyte reactivity (Oakley et al., 2006), making them an attractive model to assess functional phenotypes of reactive astrocytes.
Two-photon imaging was used to assess astrocyte calcium transients in barrel cortex of fully awake mice at a depth of 150–300 μm. Spontaneous calcium transients were first assessed in the absence of movement across a 5 min time period (Movie 1), in which no external stimulation was administered. Regions of interest (ROIs) representing astrocyte somata and processes were segmented across the field of view (FOV; Fig. 1A–D) and a convolution-based event detection technique was used to detect significant events over baseline calcium levels. Calcium transients (ΔF/F) in representative ROIs from a single FOV in a WT and 5xFAD mouse are shown in Figure 1E. Calcium transients in both cellular compartments occurred more frequently in 5xFAD mice, especially in astrocyte processes (Fig. 1F, somata, p = 0.06; Fig. 1K, processes, p = 0.02). However, transient amplitudes in 5xFAD astrocytes were also smaller (Fig. 1G, somata, p = 0.03; Fig. 1L, processes, p = 0.03) with trending reductions in area under the curve (AUC) measures (Fig. 1I, somata, p = 0.05; Fig. 1N, processes, p = 0.07). No differences in rise time properties were observed (Fig. 1H,M), but transient decay was accelerated in the 5xFAD group (Fig. 1J, somata, p = 0.02; Fig. 1O, processes, p = 0.03).
Video demonstration of spontaneous astrocyte calcium signaling and vasoactivity in the barrel cortex of an awake WT mouse. Representative 10 min video (at 10× speed) of typical vascular and astrocyte network activity in the absence of air puff stimulation. FOV is 573 µM × 573 µM. Astrocyte calcium activity of varying intensity occurs across much of the FOV (e.g., at 1:45 and 3:20) and also in highly localized areas: e.g., at 4:20, right side of FOV (arrow) and at 5:00, bottom left of FOV (arrow). Varying degrees of “spontaneous” vasoactivity is also observed. [View online]
Segmentation of astrocyte sub compartments and spontaneous signal analyses in barrel cortex of awake mice. A, Two-photon micrograph of a cerebral penetrating arteriole and surrounding capillary bed (red) and summated activity map of astrocyte calcium signals (green). B, Activity feature map of signal properties (amplitude, red; variability, green; and shape, blue) represents the strength of each feature. Signal features for each pixel were inputted into an unsupervised learning algorithm and clustered to identify pixels composing active astrocyte compartments. C, D, Somata (C) and fine process (D) segmentation masks used for ROI signal extraction. E, Representative spontaneous calcium traces extracted from individual soma compartments in FOVs from a WT and a 5xFAD mouse. Regions in red denote where individual calcium transients were detected and measured. F–O, Scatterplots show calcium transient parameters extracted from astrocyte somata (F–J) and processes (K–O) and averaged across FOVs for WT and 5xFAD mice. Parameters included the following: average number of calcium transients per ROI (F, K), calcium transient amplitude (G, L), rise time (H, M), total AUC (I, N), and decay (J, O). Each plot symbol represents an individual mouse. p values derived from two-tailed t tests.
Evoked dilations in penetrating arterioles are reduced in 5xFAD mice
Functional hyperemia, as assessed by stimulus-evoked cerebral vessel dilations or increased blood flow, is impaired in multiple AD/ADRD rodent models including 5xFAD mice, when assessed under general anesthesia (Mughal et al., 2021). To determine if similar deficits occur in fully awake 5xFAD mice, we delivered air puffs (9 Hz, 10 s) to contralateral vibrissae and measured hyperemic responses in penetrating arterioles of the barrel cortex (Fig. 2A–D). Vessels of interest were identified and cross-sectioned to generate a linearized vessel model (Figs. 2E, 5A–C). Representative two-photon images show penetrating arterioles before (Fig. 2B) and during (Fig. 2C,D) the air puff, while representative time plots illustrate differences in arteriole dilation across WT and 5xFAD mice (Fig. 2F). Before-and-after (air puff) scatterplots are shown in Figure 2G and differences determined by repeated-measures two-way ANOVA. No genotype differences were observed for resting (prestimulation) arteriole tone (Fig. 2G, round plot symbols). But, while air puff stimulation led to increased arteriole cross-sectional area in all animals (baseline vs stimulation: F(1,24) = 69.3, p < 0.001), this increase was significantly smaller in 5xFAD mice (interaction: F(1,24) = 5.097, p = 0.033; Fig. 2H, p = 0.002), indicative of a neurovascular coupling deficit.
Air puff-evoked arteriole dilations in barrel cortex of awake mice. A, Mice received a 10 s duration 9 Hz air puff train to the contralateral vibrissae. B, C, Pseudo-colored two-photon micrographs showing penetrating arterioles, before (B) and during (C) air puff stimulation. D, Subtracted image of panels B and C showing the extent of vessel dilation (red arrows, green vessel segments) during air puff stimulation. E, Arteriole modeling and analysis protocol using LAVA application. F, Time plots showing representative arteriole diameter changes for WT and 5xFAD mice before, during, and after air puff stimulation (shaded rectangle). G, Raw arteriole cross-sectional areas measured within each mouse at rest (circles) and during (squares) air puff stimulation. Basal arteriole diameter was comparable across genotypes, and both groups showed significant dilatory responses during air puff delivery (baseline cross-sectional area vs stimulation cross-sectional area: F(1,24) = 69.3, p < 0.001), though dilations were greater in the WT group (interaction: F(1,24) = 5.097, p = 0.033). H, Scatterplot shows percent arteriole dilation during air puff stimulation for both transgene groups. Each plot symbol represents an individual mouse. p values derived from two-tailed t tests.
Evoked calcium activity in astrocytes is impaired in individual cell compartments and across multicell networks in 5xFAD mice
Air puff stimulation led to a large population calcium response across astrocytes in the FOV, with highly correlated activity observed in both WT and 5xFAD mice (Movie 2, Fig. 3). Figure 3A shows representative extracted calcium traces from each ROI in a FOV during a whisker stimulation trial. Calcium transient kinetics were very similar and synchronous across some ROIs (Fig. 3B, asterisks in linearized traces), while calcium transients in other ROIs showed dissimilar kinetics (Fig. 3B, pound symbols in linearized traces). Figure 3C shows representative astrocyte network responses in a WT and 5xFAD mouse with overlaid correlograms. Lines connect coactive ROIs. Line color indicates the degree of synchronicity between each ROI pair (red-yellow is high; blue is low, Pearson's r). Assessment of population responses in astrocyte networks showed reduced synchronicity in the 5xFAD group, especially in astrocyte processes (Fig. 3D, somata p = 0.01; Fig. 3I, processes p = 0.007 n = 9, 13).
Video demonstration of astrocyte network synchronicity analyses. A, 80 s video (at 5× speed) of astrocyte network activation from an awake WT mouse. At the 10 s mark, air puff stimulation (10 s duration) was delivered to contralateral vibrissae. B, Grayscale video with network activity overlaid by frame. Calcium events from active cells are correlated by the alignment of transient rise phases using Pearson's r. Red circles mark each active cell, with each size relating to its overall synchronicity to other cells in the FOV. The color of each line between circles represents the strength of each individual connection (blue, low synchronicity; red, high synchronicity/r value). Pairs with negative correlation are not plotted. C, 3D time series plots of calcium events from each ROI in the FOV (shaded box denotes delivery of air puff stimulation). [View online]
Astrocyte network activation and evoked signal analyses in awake mice. A, Individual calcium events (shown in different colors) were extracted from unique ROIs in the FOV after whisker stimulation (gray rectangle). B, Temporal alignment of calcium signaling events from unique ROIs. Each row represents calcium events from a unique ROI (y axis). The * symbol exemplifies well-synchronized events across ROIs, while the # symbol denotes poorly synchronized calcium events across ROIs. C, Comparison of representative network diagrams of field astrocyte signal synchronization in a WT and 5xFAD mouse (also see Movie 2). Line colors represent Pearson's r correlation values. (Negative correlations not shown.) D–M, Scatterplots showing evoked astrocyte calcium signaling parameters in WT and 5xFAD mice. Each plot symbol represents an individual mouse. Network synchronicity (D, I) reflects the average Pearson’s correlation coefficients for all correlated ROI pairs (astrocyte somata, D; astrocyte processes, I) within the FOV. Calcium transient amplitude (E, J), rise time (F, K), AUC (G, L), and decay (H, I) were assessed in individual astrocyte compartments (somata D–H; processes I–M) and averaged within the in the FOV for each mouse. p values derived from two-tailed t tests.
In addition to network-wide calcium responses, we also assessed evoked calcium transient parameters from individual cell compartments (somata, Fig. 3E–H; processes, Fig. 3J–M). Individual calcium transients were reduced in amplitude in 5xFAD mice (somata, p = 0.002, Fig. 3E; processes, p = 0.01, Fig. 3J), along with trends for reduced AUC measures (somata, p = 0.11, Fig. 3G; processes p = 0.08, Fig. 3L). The decay time for evoked calcium responses was accelerated in the 5xFAD group, especially in astrocyte processes (somata, p = 0.08, Fig. 3H; processes, p = 0.03, Fig. 3M). No changes in rise time for evoked transients were observed (Fig. 3F,K). The results show that evoked astrocyte activity, as assessed in individual subcompartments and across astrocyte networks, is impaired in 5xFAD mice.
Changes in perivascular calcium signaling in 5xFAD mice
We next assessed perivascular astrocyte calcium signaling dynamics, specifically. This entailed activity in the immediate vicinity of all cerebral vessels (arterioles, venules, capillaries) and was enriched for astrocyte end feet and distal astrocyte process compartments (Fig. 4). Representative masking and segmentation steps are shown in Figure 4A,B. Representative activity from multiple perivascular ROIs in the FOV is shown in Figure 4C. In the absence of air puff, spontaneous events occurred in 5xFADs at a higher rate, relative to WTs (Fig. 4D, p = 0.02), but tended to be smaller in amplitude (Fig. 4E, p = 0.1) and integrated area (Fig. 4G, p = 0.05), with accelerated decay (Fig. 4H, p = 0.02). During air puff stimulation trials, population responses in perivascular regions showed reduced synchronicity in 5xFAD mice (Fig. 4I, p = 0.004). Evoked calcium transients in perivascular spaces were also smaller in amplitude (Fig. 4J, p = 0.001) and integrated area (Fig. 4L, p = 0.05) and tended to decay more rapidly (Fig. 4M, p = 0.06) in 5xFADs relative to WTs.
Perivascular astrocyte calcium events in awake mice. A, Two-photon micrograph of cerebral vessels (red) and summated activity map of astrocyte calcium signals (green). B, Mask of perivascular regions (white) for calcium event extraction. C, Representative calcium traces extracted from perivascular ROIs. In the left panel, regions in red denote where individual calcium transients were detected and measured. In the right panel, extracted calcium events from unique ROIs were temporally aligned for network analyses, as described in Figure 3B. D–H, Scatterplots for spontaneous perivascular calcium transient parameters including the average number of transients/ROI (D), amplitude (E), rise time (F), AUC (G), and decay (H). I–M, Scatterplots for air puff-evoked perivascular calcium transient parameters including correlated activity across all end feet pairs in the FOV (I, network synchronicity as described in Fig. 3), and average calcium transient amplitude (J), rise time (K), AUC (L), and decay (M) in individual perivascular ROIs within the FOV. Each plot symbol represents an individual mouse. p values derived from two-tailed t tests.
Reduced signaling fidelity between astrocyte end feet and arterioles in 5xFAD mice
To investigate dynamic interactions specifically between cerebral arterioles (Fig. 2) and astrocyte end feet, we used a custom MATLAB-based application called LAVA (Movie 3, Fig. 5). Analyses were restricted to end feet pairings with cerebral arterioles because of the relative difficulty of observing/measuring dilations in veins/venules and capillaries. Figure 5A shows a diagram illustration of arteriole segments demarcated by blue lines along with adjacent astrocyte end feet. Arrows indicate expansion/constriction of vessel diameter with air puff stimulation. Figure 5B shows a two-photon micrograph of an arteriole (red) and summated astrocyte calcium activity (green) in the FOV. Arterioles of interest are binarized and linearized (Fig. 5C) and immediately adjacent perivascular calcium events (1 and 2, shown in green on either side of linearized vessel) are extracted and mapped back to the raw channels (Fig. 5D, vessel segment shown in white, adjacent end feet compartments 1 and 2 shown in fuchsia). Figure 5E shows the temporal relationships between vasoactive responses observed in the arteriole segment and immediately adjacent astrocyte end feet shown in Figure 5D. In nearly all arteriole–end feet pairings, calcium changes in astrocyte end feet lagged behind arteriole dilations (also see Movie 3). A total of 55 and 74 end feet–arteriole segment pairings were evaluated in the barrel cortex of 10 WT and 13 5xFAD mice, respectively (Fig. 5F,G). For the WT group, calcium transients occurred in 80% of the end feet in response to a stimulation-induced dilation of the immediately adjacent arteriole segment (Fig. 5F). For ∼18% of the end feet–arteriole pairings in the WT group, dilations occurred in the absence of calcium transients, while in roughly 2% of the pairings there was neither a dilation nor a calcium change (Fig. 5F). In contrast to WT mice, an evoked calcium transient was observed in only 58% of the end feet–arteriole pairings in the 5xFAD group (Fig. 5G). In 32% of the pairings, arteriole dilations occurred in the absence of an end foot calcium transient, while ∼4% the pairings showed an end foot calcium response in the absence of an arteriole dilation (Fig. 5F). The remaining pairings (5.41%) in the 5xFAD group showed neither a dilation nor a calcium change. In 5xFAD pairings where there was coactivity, end feet calcium transients emerged more slowly after local arteriole dilations (Fig. 5H, p = 0.027) and were smaller in amplitude (Fig. 5I, p = 0.003). We next investigated whether calcium changes in astrocyte end feet were proportional to the magnitude of the arteriole dilation (Fig. 5J). For WT mice, end feet calcium transient amplitudes were positively correlated to dilation magnitudes observed in immediately adjacent arterioles (r = 0.38, p = 0.0062). In contrast, we observed no correlation between arteriole dilation magnitudes and end feet calcium changes in 5xFAD mice (r = 0.046, p = 0.71; 5xFAD vs WT, z = 1.851, p = 0.032). Together, these results suggest that end feet calcium signaling is relatively uncoupled from local arteriole activity in 5xFAD versus WT mice.
Spatial-temporal analyses of vascular-astrocyte end feet activity by LAVA application. A, Diagram illustration showing analyses of astrocyte end feet pairings within individual arteriole segments (blue lines) during airpuff-induced dilations (white arrows). Created in BioRender. https://BioRender.com/u91p773. B, Two-photon micrograph of a cerebral penetrating arteriole (red) and summated activity map of astrocyte calcium signals (green). C, Modeling and sampling of arteriole cross sections are used to produce linearized channel clips containing the vessel of interest (white) and perivascular spaces (calcium activity in green). Vascular tone is measured from cross-sectional area measurements, as well as close capture of astrocyte end feet (1 and 2). D, Same FOV shown in panel B assessed during whisker stimulation. End foot ROIs (1 and 2 pink) and vascular mask (white) determined from the linearized vessel model (panel C) are mapped onto raw channels. E, Measurements of vascular tone from linearized model in panel C and calcium levels recorded from each end foot (1 and 2) in panel D are plotted to display temporal associations of field activity. F, G, Pie charts showing the functional make-up of end feet–arteriole pairings for WT (F) and 5xFAD (G) mice. Red indicates arteriole dilations with a subsequent end foot calcium response; blue indicates arteriole dilations in the absence of an end foot response; yellow indicates end foot calcium response in the absence of an arteriole dilation; black indicates pairings with neither an arteriole dilation nor an end foot calcium response. H, Scatterplots showing the end foot calcium response latency (i.e., latency between initial stimulation-induced vasodilation and subsequent end foot calcium signals). I, Scatterplot shows evoked calcium transient amplitudes in arteriole-associated end feet. J, Scatterplots show the proportional relationship between calcium transient amplitude in each end foot relative to vessel cross-sectional area in the immediately adjacent arteriole. In H and J, each plot symbol represents an individual end foot-arteriole pairing. In I, each plot symbol represents an individual mouse. p values derived from two-tailed t tests.
Video demonstration of LAVA application to assess end feet–arteriole spatiotemporal activity. A, 80 s video (at 5× speed) of operating cerebrovasculature and surrounding astrocyte network activity in an awake WT mouse. At the 10 s mark, air puff stimulation (10 s duration) was delivered to contralateral vibrissae. B, Modeling and sampling of arteriole cross sections are used to produce linearized channel clips containing the vessel of interest and perivascular spaces. Vascular tone is measured from cross-sectional area measurements, as well as close capture of astrocyte end feet (1 and 2). C, End foot ROIs (pink) and vascular mask (white) determined from the linearized vessel model (panel B) are mapped onto raw channels. D, Measurements of vascular tone from panel B and calcium levels recorded from each end foot (1 and 2) in panel C are plotted to display temporal associations of field activity. [View online]
Relationships between physiologic parameters and postmortem measures of Aβ pathology and astrocyte reactivity
Limitations in our imaging station (i.e., only two acquisition channels) precluded the simultaneous assessment of local amyloid deposition with astrocyte calcium signals and vasomotion in live animals. However, a subgroup of mice (5 WTs and 6 5xFAD mice) were processed for postmortem assessment of GFAP labeling intensity (a measure of astrocyte reactivity) and parenchymal Aβ1–42 pathology (Fig. S1A–C). As expected, the cortical labeling density for both GFAP and Aβ (Fig. S1A–D) was dramatically increased in 5xFAD versus WT mice (GFAP, p = 0.0002; Aβ, p < 0.0001, two-tailed t tests). For the 5xFAD group, within-mouse correlations were then established between postmortem Aβ and GFAP measures in barrel cortex to antemortem measures of calcium signaling and vasomotion (Table S1). No significant correlations were observed between Aβ levels and any of the physiologic measures collected, suggesting that Aβ pathology may be at, or near, a ceiling at this time point. In contrast, GFAP levels showed significant correlations to spontaneous calcium transient frequency, particularly in the soma (R2 = 0.79, p = 0.042) and the end feet (R2 = 0.79, p = 0.041; Fig. S1E,F; Table S1) and a correlation with astrocyte network synchrony (R2 = 0.51, p = 0.11) that trended toward significance (Fig. S1G, Table S1), suggesting that increased astrocyte reactivity is associated with calcium hyperactivity and impaired network synchronicity.
Discussion
This study is among the first to investigate astrocyte calcium signaling concurrently with vasomotion in a fully awake mouse model of AD. Reactive astrocytes in 5xFAD mice were spontaneously hyperexcitable but showed deficits in evoked activity. Most strikingly, calcium signaling in astrocyte end feet of 5xFADs was largely uncoupled from local arteriole dilations, which could have important implications for brain metabolism in the context of disease. The results underscore the sensitivity of end feet to developing pathological conditions and suggest these cellular specializations offer modifiable targets for improving blood flow, metabolism, and brain health.
Humans with abundant AD neuropathology can exhibit many comorbid vascular pathologies (other than cerebral amyloid angiopathy, CAA) and other in vivo neuroimaging abnormalities indicative of cerebral small vessel disease (SVD; Hilal et al., 2017; Koncz and Sachdev, 2018; Ali et al., 2023; Swinford et al., 2023; Wu et al., 2023; Edwards et al., 2024; Song et al., 2024; Zhang et al., 2024). Though 5xFAD mice do show some cerebrovascular abnormalities, they show relatively little CAA (Szu and Obenaus, 2021), making vascular stiffness less of an issue for the functional experiments performed here. These observations imply the existence of pathogenic synergies between AD-type neuropathology and SVD that go beyond CAA and which may relate to astrocyte dysfunction. Calcium fluctuations in astrocytes are linked to diverse cellular functions (e.g., potassium regulation, gliotransmitter release, cytoskeletal remodeling/degeneration, and gene expression; Straub and Nelson, 2007; Sompol and Norris, 2018; Verkhratsky, 2019; Lim et al., 2021), many of which are altered in neurodegenerative conditions (Olabarria et al., 2010; Bellot-Saez et al., 2017; Price et al., 2021). Moreover, targeted inhibition of calcium signaling pathways in astrocytes ameliorates pathophysiologic outcomes in intact AD/ADRD mouse models (Furman et al., 2012; Sompol and Norris, 2018), suggesting a central role for astrocytic calcium abnormalities in chronic neurodegeneration.
To address potential changes in astrocyte calcium signaling in the context of AD-like pathology, many studies have turned to imaging approaches in primary cultures (Lim et al., 2013; Ronco et al., 2014; Mitroshina et al., 2022), brain slices (Huffels et al., 2022; Paumier et al., 2022; Abghari et al., 2023), and intact animals (Takano et al., 2007; Kuchibhotla et al., 2009; Delekate et al., 2014; Lines et al., 2022; Shah et al., 2022; Kelly et al., 2023; Lee et al., 2023; Sompol et al., 2023). But, while astrocyte–blood vessel interactions are among the most important ways that the periphery communicates with brain, few studies have investigated whether calcium signaling in reactive astrocytes retains fidelity with vasomotion. The choice of experimental preparation to address this issue is important because astrocyte morphologic features/functions depend on the presence of adjoining blood vessels and neurons (Lange et al., 2012; Takano et al., 2014). Even in intact animals, use of anesthesia may alter vascular tone and reduce astrocyte activity, which could uncouple astrocytes from normal metabolic duties (Thrane et al., 2012; Gao et al., 2017). For these reasons, we chose to investigate astrocyte–blood vessel interactions in awake 5xFAD mice at an age where glial reactivity is extensive and neural function is compromised.
Spontaneous calcium activity is augmented in 5xFAD astrocytes, while evoked activity is impaired
Similar to previous reports (Kuchibhotla et al., 2009; Delekate et al., 2014; Takano et al., 2014; Abjorsbraten et al., 2022; Huffels et al., 2022; Lines et al., 2022), we observed signs of spontaneous hyperexcitability in 5xFAD astrocytes compared with WTs. Increased rates of calcium activity may explain why calcium-dependent pathways, like calcineurin/NFAT, show aberrantly high activity/expression in reactive astrocytes (Lim et al., 2013; Pleiss et al., 2016; Sompol et al., 2017; Kraner et al., 2024). The source of calcium hyperactivity may be intrinsic to astrocytes, as previous studies observed greater transient frequency in APP/PS1 mice even when neuronal activity was independently suppressed (Kuchibhotla et al., 2009). While astrocytes in 5xFADs appeared spontaneously hyperactive, evoked responses were impaired across astrocyte networks and in astrocyte subcompartments. In both transgene groups, whisker stimulation elicited robust calcium responses that were highly correlated across multiple astrocytes in the FOV. However, correlated activity in 5xFADs was lower compared with WTs, perhaps because of gap–junction abnormalities that disrupt intercellular coupling (Semyanov, 2019). Reductions in evoked calcium transient amplitudes in 5xFAD mice were also observed in astrocyte somata, processes, and end feet, similar to what others have reported in other amyloid models (Lines et al., 2022; Abghari et al., 2023).
Loss of signaling fidelity between end feet and arterioles in 5xFAD mice
A novel observation of our work is that end feet in 5xFADs responded to arteriole dilations with poor fidelity. To evaluate concurrent arteriole and end feet activities, we developed a MATLAB-based application called LAVA. The application is unique in its versatility for analyzing challenging video files. Its preprocessing options for motion correction, filtering, dissection-by-trial phases, and adaptive ROI identification allow for precise assessment of vascular tone and small proximal calcium signals, even for narrow, moving end feet. LAVA also enables comparison of its outputs to existing ROI signal analysis. Perhaps the biggest advantage is that LAVA can determine the directional nature of signaling within the neurovascular unit. For instance, we found that evoked end feet calcium transients almost always followed dilation of local arterioles. These observations are similar to earlier work on awake mice during whisker stimulation (Tran et al., 2018; Del Franco et al., 2022; Sompol et al., 2023). While we cannot rule out the possibility that faster events at end feet (e.g., transmembrane potassium fluxes) contribute to rapid vessel dilations, the LAVA analyses conducted here, in combination with the most sensitive GCaMP available (jGCaMP8f), suggests that astrocyte calcium signaling does not initiate local arteriole dilations. Moreover, our work suggests that astrocytes in 5xFAD mice do not respond to arteriole dilations with the same fidelity seen in WTs. Calcium transients in 5xFAD end feet occurred only 58% of the time after local arteriole dilations (vs 80% in WTs), with a lag time of ∼30 s (vs 17 s in WTs). While 5xFAD mice also exhibited smaller evoked arteriole dilations, it is unlikely this deficit was the primary cause of impaired end feet signaling, as we found no correlation between vessel dilation magnitude and end feet calcium transient amplitude in 5xFADs (unlike WTs, which did exhibit proportionality). Together, these results suggest that astrocytes, a major metabolic hub in brain, respond sluggishly and disproportionately in 5xFAD mice to the most important mechanism for delivering energy on demand (i.e., functional hyperemia).
Limitations and future directions
Imaging in a single focal plane, combined with the use of a modestly powered (16×) objective, and ever-present movement artifacts (inherent to fully awake imaging paradigms) were limitations that could diminish spatial resolution of fast fluorescent changes in microdomains. Nonetheless, our work suggests that astrocyte calcium transients are smaller and respond to vasomotion with lower fidelity in the context of AD-like pathology. Future studies should address why astrocytes exhibit these deficits. One possibility is lower neuronal drive. Like many amyloid models, 5xFAD mice exhibit smaller evoked synaptic responses compared with WTs (Sompol et al., 2017; Forner et al., 2021; Chen et al., 2022), leading to a possible reduction in neuron-astrocyte communication. End feet also typically show signs of degeneration and/or loss of polarity in AD/ADRDs including detachment from cerebral vessels, and/or mislocalization of signaling mediators like AQP4 (Okoye and Watanabe, 1982; Wilcock et al., 2009; Hawkes et al., 2013; Kimbrough et al., 2015; Sudduth et al., 2017; Abbrescia et al., 2024). Any of these changes could disrupt astrocyte signaling, brain metabolism, and/or clearance of pathogenic factors into the perivascular space. It is also possible that the evoked calcium deficits we observed in 5xFAD astrocytes reflect a ceiling effect due to elevated resting calcium levels. While assessing absolute calcium concentrations is difficult with GCaMPs, another group recently reported higher resting astrocyte calcium using a ratiometric indicator in a similar amyloid model (Kelly et al., 2023). If basal calcium is elevated, this could lead to augmented calcium-dependent potassium fluxes at end feet, which could affect vessel tone. In addition to end feet communication to the vasculature, cerebral vessels also communicate directly with end feet through calcium-permeable mechanosensitive channels like TRPV4, which can, in turn, regulate vascular tone directly or indirectly through vasculo-neuronal coupling (Kim et al., 2015; Haidey et al., 2021). Dysregulation of TRPV4 could therefore be an additional source for the functional uncoupling we report here.
Finally, our work focused on a single mouse model during a relatively advanced stage of pathology. Future work should evaluate how end feet coupling abnormalities develop as amyloid pathology progresses. And because amyloid is one of many pathologies linked to astrocyte reactivity, it will be important to investigate AD/ADRD models that lack amyloid pathology or exhibit comorbid pathologies, which will help determine whether our results reflect generalized phenotypes of reactive astrocytes or are more sensitive to specific factors like aging, tau pathology, or other disease-related changes.
Conclusion
We showed that reactive astrocytes in awake amyloid-bearing mice are spontaneously hyperactive and exhibit impaired functional connectivity. Changes in calcium signaling observed here are consistent with the calcium hypothesis of aging and AD, which suggests that cellular calcium dysregulation is a driving force in neurodegenerative diseases (Landfield and Pitler, 1984; Thibault et al., 2007; Bezprozvanny and Mattson, 2008; Sama and Norris, 2013; Schrank et al., 2020; Lim et al., 2021). Perhaps most notable, we found that astrocyte end feet respond to dilations in immediately adjacent arterioles with poor fidelity indicating a key point of communication breakdown between brain and the cerebrovasculature.
Footnotes
This work was supported by National Institutes of Health (NIH)–National Institute on Aging Grants AG078116 to C.M.N., Y.K., Y.J., P.T.N., D.M.W., O.T., and P.S. and AG027297 C.M.N.; the Hazel Embry Research Trust; and the Sylvia Mansbach Endowment for Alzheimer’s Disease Research.
The authors declare no competing financial interests.
This paper contains supplemental material available at: https://doi.org/10.1523/JNEUROSCI.0349-25.2025
- Correspondence should be addressed to Christopher M. Norris at cnorr2{at}uky.edu.
