Introduction

Resident macrophages are key myeloid immunocytes that play an important role in innate immune surveillance and defense across various peripheral tissues and organs [1–3]. The central nervous system also harbors a large subset of parenchymal macrophages, known as microglia, and several distinct subsets of macrophages localized to the brain’s border tissues, including the choroid plexus, perivascular spaces, and the meningeal compartments that cover, protect, and support the brain [4–8]. Macrophages are the predominant immune cell type within the brain meninges, and recent studies have demonstrated the diverse ontogeny, transcriptomic profiles, and immune functions of meningeal macrophages at steady state [5, 8–12] and in several neuropathological conditions [8, 13–16].

Cytoplasmic calcium (Ca²⁺) signaling underlies a wide variety of cellular homeostatic and inflammatory processes in macrophages [17–23]. In addition to intracellular Ca²⁺ elevation, distinct spatiotemporal dynamics - including oscillation patterns, intracellular propagations, and intercellular synchronization of Ca²⁺ signals - may regulate different macrophage functions during steady state and pathophysiology [20–25]. Despite our increased understanding of the diverse molecular signatures and contributions of meningeal macrophages to homeostasis and neuroinflammation, virtually nothing is known about their Ca²⁺ signaling heterogeneity in both healthy and diseased states.

To comprehensively characterize the Ca²⁺ dynamics of individual macrophages localized to the brain meninges, we combined intravital two-photon Ca²⁺ imaging in a novel reporter mouse line, in which the Ca²⁺ reporter GCaMP6s is expressed in platelet factor 4 (Pf4⁺) meningeal macrophages, with an event-based signaling analysis pipeline. Our data reveal several distinct spatiotemporal Ca²⁺ dynamic features in perivascular versus non-perivascular meningeal macrophages, including a unique coupling between the Ca²⁺ signals of dural perivascular macrophages and behaviorally-driven vasomotion of their associated dural vessels at steady state. Furthermore, our data uncover both increases and decreases in Ca²⁺ activity in distinct subsets of meningeal macrophages in response to cortical spreading depolarization (CSD), a pathophysiological brain hyperexcitability event linked to neuroinflammation [26] in migraine, traumatic brain injury, and stroke [27]. Mechanistically, our data suggest the calcitonin gene-related peptide/receptor activity-modifying protein 1 (CGRP/RAMP1) axis uniquely mediates CSD-evoked macrophage Ca²⁺ elevation, indicating a brain-to-meninges neuroimmune signaling pathway involving the activation of meningeal peptidergic sensory neurons.

Results

Characterizing macrophage Ca²⁺ signaling features in homeostatic brain meninges

Previous intravital imaging studies exploring the spatiotemporal dynamics of tissue-resident macrophages have primarily used CX3C motif chemokine receptor 1 (CX3CR1)-based mouse reporter strains [13, 22, 28, 29]. However, other resident monocyte-derived cells are labeled in these reporter mice [30, 31]. Moreover, brain microglia also express CX3CR1 [13, 31–33], limiting the use of these reporter mice for resolving the spatiotemporal subcellular Ca²⁺ dynamics of macrophages localized to the relatively thin meningeal layers covering the brain parenchyma. To systematically characterize meningeal macrophage Ca²⁺ dynamics in vivo and avoid contamination from Ca²⁺ signals arising from other resident monocyte-derived meningeal cells and superficial parenchymal microglia, we leveraged recent findings showing the expression of Pf4 in meningeal macrophages but not in parenchymal microglia, [10, 16, 34], and generated transgenic reporter mice expressing the highly sensitive Ca²⁺ indicator GCaMP6s (using TIGRE2.0-based Ai162D mice [35]) in Pf4⁺ macrophages (using Pf4^Cre mice [36]). Notably, nearly 100% of meningeal macrophages are labelled in Pf4^Cre-based reporter mice [34]. Although Pf4 is also expressed by megakaryocytes [36], these cells are absent in the brain meninges [5].

Anesthetic agents impact intracellular Ca²⁺ signaling, including in brain macrophages and other non-excitable glial cells [37–39]. We therefore investigated subcellular Ca²⁺ activity of meningeal macrophages in awake, behaving mice. We imaged meningeal macrophages via a chronic cranial window implanted together with a restraining headpost over the intact dura mater overlying the posterior neocortex. This chronic window approach produces minimal inflammatory responses in the cortex and meninges below the window [40, 41]. After at least 7 days of recovery, mice were gradually habituated across multiple days to head restraint while free to run on a wheel (Fig 1b). We tracked Ca²⁺ transients of GCaMP6s-labeled meningeal macrophages using high-speed two-photon microscopy in 37 fields of view (FOVs) from 7 mice. We first corrected the imaging movies for locomotion-evoked meningeal translational shifts using rigid registration [41]. Movies were then processed using the AQuA2 data analysis platform [42], which implements an unbiased event-based approach to capture spatiotemporal Ca²⁺ event dynamics (Fig 1C). Based on spatial analysis, we assigned events (n=1361), with their corresponding Ca²⁺ features, to a given macrophage, with data gathered from a total of 503 cells.

Figure 1.

We characterized Ca²⁺ dynamics in two anatomically distinct macrophage populations: perivascular and interstitial non-perivascular (n=122; n=381, respectively, Fig 2A and Video S1). The vast majority of perivascular macrophages (93.4%, n=114) displaying ongoing Ca²⁺ activity were associated with vessels in the dura mater. The two meningeal macrophage subpopulations exhibited several distinct Ca²⁺ activity features. While the total area of Ca²⁺ activity in the peri- and non-perivascular macrophages was similar (Fig 2B), the signal perimeter of perivascular macrophages was significantly higher (Fig 2C) and exhibited a more elongated shape (Fig 2A, 2D). The Ca²⁺ event duration in the perivascular macrophages was also longer compared to the non-perivascular subpopulation (Fig 2F). The peak Ca²⁺ activity level (Max DF/F0, Fig 2E) and event rate (Fig 2G), nonetheless, were similar in the two meningeal macrophage subpopulations.

Figure 2.

Distinct Ca²⁺ signal frequency spectra may underlie different biological functions of macrophages [22]. We therefore analyzed meningeal macrophage Ca²⁺ signal waveforms by performing a multi-step signal processing and clustering analysis. We clustered cells based on two distinct patterns of Ca²⁺ activity. Cells in Cluster 1 (n=40) exhibited a more noisy-like activity pattern characterized by multiple frequencies, while cells in Cluster 2 (n=463) displayed a single dominant frequency at 0.01 Hz (Fig 3A). To further explore these cellular signaling differences, we combined the clustering data with AQuA2-derived features and observed that macrophages in Cluster 1 showed a larger event perimeter and lower circularity and peak magnitude compared to those in Cluster 2 (Fig 3D-F). Cluster 1 cells also had a lower signal-to-noise ratio compared to Cluster 2 cells (Fig 3I), but had differences in terms of event area, duration, or rate (Figures 3C, G, H). There was a significant association between cell cluster (1 vs. 2) and cell type, with Cluster 1 predominantly comprised of perivascular macrophages, and Cluster 2 including primarily non-perivascular macrophages (Fig 3J), further indicating that these two meningeal macrophage subpopulations have distinct Ca²⁺ signaling properties.

Figure 3.

Intracellular Ca²⁺ signal propagation underlies diverse cellular functions and has been recently identified in macrophages in vitro [21] and in skin-resident macrophages in vivo [43]. By assessing the Ca²⁺ signal propagation maps of each event within a determined cell, we identified two distinct patterns of activity: propagating events in which Ca²⁺ signals traveled throughout the entire cell, and stationary events (Fig 4A). Propagating events had varied signal source regions and directionality. Most macrophages (perivascular, 94.3%, n=115; non-perivascular, 86.9%, n=331) exhibited only propagating events, while a small minority of cells displayed a mix of propagating and stationary events or exclusively stationary activity events (Fig 4B).

Figure 4.

Macrophage intercellular communication, including synchronized activity, potentially across connected macrophage networks, has been implicated in maintaining tissue homeostasis and immune function [18, 21, 44, 45]. To investigate meningeal macrophage intercellular interactions, we characterized spatiotemporal relationships between Ca²⁺ events in distinct cells within each FOV. We compared temporal factors, including the relative onset latency (t0_A – t0_B) between different macrophages and the duration of the first occurring event (t1_A – t0_A) (Fig 4C). We also compared the distances between macrophage pairs exhibiting concurrent events and the number of synchronous events (Fig 4D). Finally, we calculated the proportion of perivascular and non-perivascular macrophages exhibiting synchronous Ca²⁺ events. Across all FOVs, nearly half of the macrophages (49.3%) exhibited synchronous Ca²⁺ activity, involving cells at a wide range of distances (Fig 4E), suggesting that spatial proximity does not influence event synchronicity. Interestingly, there was poor correlation between the distance of cell pairs exhibiting synchronous events and their timing (Fig 4F), suggesting that an external factor, rather than intercellular coordinated networked activity, mediates this intercellular signaling. When comparing synchronous activity among peri- and non-perivascular macrophages, most cells exhibited mixed interactions between the same and different subtypes (Fig 4G).

Dural perivascular macrophage Ca²⁺ activity is tuned to behaviorally-driven dural vasomotion

Brain border macrophages in the leptomeninges and related parenchymal perivascular spaces regulate pial arterial vasomotion indirectly by affecting vessel stiffness [6]. Yet, interaction between vascular-associated macrophages residing in the dura mater, the outermost meningeal layer, and dural vasomotion remains unknown. We employed a Generalized Linear Model (GLM) approach to investigate functional interaction between dural perivascular macrophage Ca²⁺ signals and behaviorally-driven dural vessel dynamics (Fig 5A). Dural arteries constrict during locomotion, while pial arteries dilate [46]. We analyzed the locomotion-associated responses of 86 meningeal vessels (32 FOVs from 5 mice) and identified a subset (22%, n=19), whereby the diameter changes were well fit by a GLM using the locomotion state as a predictor. Of these, we identified 74% (n=14) as dural vessels based on their GLM’s negative coefficients consistent with constriction (Fig 5B and 5C). Next, we fitted the Ca²⁺ signal observed in perivascular macrophages associated with these dural vessels (n=35) to a GLM using the diameter changes as a predictor variable. Overall, the Ca²⁺ activity of 83% (n=29) of these dural macrophages was well predicted by the model (average deviance explained across all well-fit macrophages: 0.43±0.15, mean±SD). Analysis of the macrophage-vascular models’ beta coefficients revealed two distinct interactions. About half of the macrophages (55%, n=16) exhibited negative coefficients (i.e., increase and decrease in Ca²⁺ activity associated with dural vasoconstriction and recovery, respectively; Fig 5B and 5D). The remaining macrophages (45%, n=13) exhibited positive coefficients (i.e., decrease and increase in Ca²⁺ activity in response to dural constriction and recovery, respectively; Fig 5B and 5 E). The coefficients for increased and decreased macrophage Ca²⁺ activity peaked near zero delay relative to the vasoconstriction and were not statistically different (Fig 5D-E and 5H). These data provide evidence that dural perivascular macrophages are functionally coupled to dural vasomotion, either responding to or mediating such behaviorally driven vascular response.

Figure 5.

An acute aberrant pro-inflammatory brain hyperexcitability event drives diverse Ca²⁺ dynamics in meningeal macrophages

CSD is an aberrant brain hyperexcitability event linked to parenchymal inflammation and pain in migraine, traumatic brain injury, and stroke [27, 47, 48], and could also affect meningeal macrophages [49]. In anesthetized mice subjected to an acute CSD episode, a small subset of meningeal macrophages undergoes morphological changes resembling an inflammatory state [26]. Given the direct anatomical and functional connection between the brain and meninges [11, 50] and the involvement of intracellular Ca²⁺ influx in macrophage inflammatory activation [51], we asked whether brain-to-meninges signaling drives intracellular Ca²⁺ elevations in meningeal macrophages in the wake of CSD. We therefore triggered a single CSD episode in the frontal cortex [52, 53] and characterized the related changes in meningeal macrophage Ca²⁺ dynamics. In each experiment, we verified CSD induction based on the associated acute meningeal deformation and/or pial vasoconstriction observed in awake animals ([53] and Video S2). We studied CSD-related Ca²⁺ dynamics in 249 macrophages (perivascular, n=64; non-perivascular, n=185; 13 FOVs from 10 mice). For each cell, we compared Ca²⁺ event rates during the passage of the CSD wave (1 min) and the PostCSD period (30 min) to baseline (PreCSD, 30 min) to assess acute and persistent Ca²⁺ responses, respectively. We characterized cells as exhibiting persistently increased (event rate > 2× PreCSD), decreased (event rate < 0.5× PreCSD), or unchanged responses. Given the low Ca²⁺ activity rate observed under steady state and the likelihood that no spontaneous Ca²⁺ elevations occur during the brief period of the CSD event, we considered cells to be either acutely activated or exhibiting an unchanged Ca²⁺ response. CSD evoked both acute and persistent Ca²⁺ activity changes in (Fig 6 A-E and Video S2). While smaller subsets of meningeal macrophages exhibited acute (21.3%, n=53) and/or persistent increase (22.1%, n=55) in their Ca²⁺ activity, the majority of cells (58.6%, n=146) exhibited a persistent decrease. An acute increase was observed more often in peri-vs non-perivascular macrophages (perivascular, 32.8%, n=21; non-perivascular, 17.3%, n=32, Fig 6F). Persistent changes in Ca²⁺ activity were observed similarly in peri- and non-perivascular macrophages (increases; perivascular, 28.2%, n=18; non-perivascular, 18.4%, n=37; decreases; perivascular, 50.0%, n=32; non-perivascular, 61.6%, n=114, Fig 6G). The macrophages’ propensity to develop a persistent Ca²⁺ increase was unrelated to their acute response (Fig 6H), suggesting that the mechanisms underlying these two temporal responses are distinct. However, cells showing no acute activation were more likely to display decreased Ca²⁺ activity post-CSD (Fig 6H). Finally, we observed that macrophages exhibiting a persistent increase in Ca²⁺ activity had a lower baseline activity when compared to cells showing a persistent (Fig 6I), suggesting that this post-CSD response is influenced by the macrophages’ basal response activity.

Figure 6.

CGRP/RAMP1 signaling mediates CSD-evoked persistent increase in meningeal macrophage Ca²⁺ activity

Many meningeal macrophages are localized near peptidergic sensory axons [16]. In the wake of CSD, cortex-to-meninges signaling enhances the responsiveness of meningeal sensory neurons and can drive the release of the neuropeptide CGRP from their peripheral nerve endings [52–54]. CGRP-expressing sensory neurons regulate tissue immunity, including meningeal macrophage function, via the CGRP/RAMP1 neuroimmune axis [16, 55]. We therefore asked whether the CSD-related changes in meningeal macrophage Ca²⁺ dynamics we observed involve CGRP/RAMP1 signaling. We pretreated mice with the RAMP1 antagonist BIBN4096 and then imaged meningeal macrophage Ca²⁺ activity (42 cells; perivascular, n=14; non-perivascular, n=28; 3 FOVs from 3 mice) before and after CSD. As expected, RAMP1 blockade did not affect CSD triggering [56]. When compared to control saline treatment, RAMP1 blockade also did not reduce basal macrophage Ca²⁺ activity (Fig 7B). Blocking CGRP/RAMP1 signaling neither affected the incidence of acute increases in Ca²⁺ activity (Fig 7A and 7C) nor the magnitude of that response (Fig 7D). RAMP1 antagonism, however, inhibited the CSD-evoked persistent increase in the macrophage’s Ca²⁺ activity, without affecting the incidence of the persistent decrease (Fig 7E). The data suggest that in the wake of CSD, the CGRP/RAMP1 axis, likely involving a peripheral sensory neuroimmune signaling, is responsible for the prolonged enhancement of Ca²⁺ signaling in a subset of meningeal macrophages, which could potentially mediate their pro-inflammatory response.

Figure 7.

Discussion

Resident macrophages in the brain meninges are essential for maintaining brain homeostasis, regulating central nervous system immune surveillance, and mediating neuroimmune responses under pathological conditions [12, 14, 16, 57–59]. Macrophages rely on Ca²⁺ signaling to mediate many of their functions [17–23]. Here, using two-photon microscopy in awake, behaving Pf4^Cre:GCaMP6s reporter mice, we describe the heterogeneity of meningeal macrophage Ca²⁺ dynamics at steady state and in response to CSD, an aberrant cortical hyperexcitability event associated with migraine, traumatic brain injury, and stroke. Our data suggest the presence of two subsets of macrophages in discrete meningeal niches, perivascular and non-perivascular, that exhibit several distinct Ca²⁺ signal properties at steady state. We further demonstrate Ca²⁺ activity in dural perivascular macrophages tuned to behaviorally-driven dural vasomotion. Finally, we describe opposing Ca²⁺ responses of meningeal macrophages following CSD, and demonstrate the contribution of the CGRP/RAMP1 axis in mediating CSD-evoked persistent Ca²⁺ elevations.

The exact link between the distinct Ca²⁺ signal properties of meningeal macrophage subsets observed herein and their homeostatic function remains to be established. The lower event magnitude and noisier signal observed in dural perivascular macrophages may reflect functional interactions with the pulsation dynamics of dural vessels [46]. Indeed, by combining vascular and macrophage Ca²⁺ imaging, we demonstrate a tight temporal association between the diameter fluctuations of the dural vessels and the Ca²⁺ signal of their associated macrophages in awake, locomoting mice. Direct vascular-macrophage coupling may underlie this interaction, involving macrophages sensing vascular-related mechanical changes via Piezo1 signaling [60]. Functional vascular-macrophage interaction may also involve mural cells as inter-mediators [58, 61]. Studying whether macrophage Ca²⁺ signaling regulates dural vasomotion will require an experimental approach that has yet to be developed, allowing for their selective manipulation. The paucity of ongoing Ca²⁺ activity in perivascular macrophages situated in the leptomeninges we observed supports recent findings that subdural perivascular macrophages indirectly affect pial and parenchymal vasomotion via extracellular matrix remodeling [6].

Intracellular Ca²⁺ signal propagation has been observed in various non-excitable cells, such as astrocytes [62]. We show that the majority of meningeal macrophages, including peri- and non-perivascular, exhibit intracellular Ca²⁺ signals that propagate throughout the entire cell, suggesting microdomain intracellular Ca²⁺ signal elevation, following release from internal stores. By characterizing the spatiotemporal relationships between Ca²⁺ signals in distinct cells, we also demonstrate synchronous events that are independent of spatial proximity, suggesting that synchronous Ca²⁺ activity is not driven by intercellular communication but rather by an external signal. Interestingly, our data indicate that synchronized events involve both peri- and non-perivascular macrophages, despite having distinct Ca²⁺ elevation features, suggesting that these meningeal macrophage subtypes similarly sense external signals.

Cortex-to-meninges signaling involves a relatively slow flow of soluble molecules in the cerebrospinal fluid reaching the subarachnoid space and advancing via arachnoid cuff exit points into the dura mater [11, 50]. Our finding of an acute Ca²⁺ elevation in a subset of extrasinusoidal perivascular dural macrophages coinciding with the CSD event suggests a rapid transfer of soluble signaling factors from the cortex across all meningeal layers [47]. Nevertheless, we cannot exclude a mechanically driven macrophage response to the acute meningeal deformation produced by the neuronal and glial swelling and shrinkage of the cortical extracellular space during CSD [53, 63, 64]. Our data also indicate a delayed and prolonged increase in Ca²⁺ signal in a relatively small subset of macrophages post-CSD, which could underlie their proinflammatory-like morphological change [49, 51]. Our findings also support the view that meningeal neuroimmune CGRP/RAMP1 axis serves as a mechanism responsible for this macrophage Ca²⁺ response, likely via the activation of their CGRP receptors [16]. Whether the relatively small subset of meningeal macrophages featuring increased Ca²⁺ signaling serves a protective role [65, 66] or a proinflammatory, destructive function [67] remains to be elucidated. Intriguingly, our data points to a persistent decrease in macrophage Ca²⁺ activity post-CSD, not involving CGRP/RAMP1 signaling, as the most prevalent response, raising the possibility that a hyperexcitable cortex dampens meningeal immunity.

Conclusions

We provide a detailed characterization of macrophage Ca²⁺ dynamics in homeostatic meninges, thereby expanding our understanding of their biological diversity. The coupling of dural perivascular macrophage Ca²⁺ signals and dural vasomotion may represent a unique homeostatic functional dural macrophage-vascular unit that controls dural perfusion. The diversity of meningeal macrophage Ca²⁺ responses to CSD further highlights the complexity of brain-to-meninges neuroimmune signaling and meningeal macrophage function in neurological disorders such as migraine, traumatic brain injury, and stroke. Our study also provides essential genetic and data analysis tools to further understand the molecular signaling underlying macrophage function at steady state and neuropathological conditions.

Methods

Animals

All experimental procedures were approved by the Beth Israel Deaconess Medical Center Institutional Animal Care and Use Committee. Experiments were conducted on adult Pf4^Cre/wt:TIGRE2.0^GCaMP6s/wt Ca²⁺ reporter mice (8-17 weeks) (9 males, 5 females). Mice were generated by crossing Pf4^Cre (C57BL/6-Tg (Pf4-icre) Q3Rsko/J, Jackson laboratory, Strain #008535) mice with GCaMP6s^fl/fl (Ai162D; B6.Cg-Igs7^{tm162.1(tetO-GCaMP6s,CAG-tTA2)Hze}/J, Jackson laboratory, Strain #031562) mice. Animals were genotyped by Transnetyx Inc.

Surgical procedures

Animals were anesthetized using isoflurane in 100% O₂ (induction: 3%; maintenance: 1.5-2%) and placed on a heating pad with a rectal probe attached to a stereotaxic frame to monitor animal body temperature during surgery. Animals received dexamethasone (8 mg/kg, i.p.) and Meloxicam SR (4 mg/kg, s.c.) to reduce inflammation and improve surgical outcomes. An eye ointment was used to prevent ocular drying. Mice were implanted with a titanium headpost and a 3 mm glass cranial window (1.5 mm lateral and 2 mm posterior to Bregma) over an intact dura covering the left posterior neocortex [53]. Immediately after surgery, the mouse cage was placed on a water-circulating heating pad for faster recovery. Animals were then single-housed with access to a running wheel and a hut and allowed to recover for at least one week.

Wheel running acclimation

After the cranial window surgery, mice were allowed to recover for at least a week. To reduce stress associated with head-fixation during imaging and habituate to wheel running, the mice received multiple training sessions (10 min to 1 h over 3–4 days). In each session, the mouse was placed on a 3D printed running wheel, with its headpost attached to two clamps, and allowed to locomote freely.

Two-photon Ca²⁺ imaging

Awake-behaving mice were head-fixed to the running wheel by its headpost (Figure 1). We used a two-photon microscope (Neurolabware) with a Nikon 16X, 0.8 NA objective to acquire images at 15.5 Hz with digital zoom set at 4X (312 x 212 µm² FOV). A MaiTai laser set to 920 nm with 25–40 mW power was used to excite fluorescence. The Scanbox package for MATLAB (Neurolabware) was used to control the microscope and acquire images and wheel running data. To image the meningeal vasculature, mice were administered 70kDa TRITC-Dextran tracer (50 mg/kg, i.v.; Sigma-Aldrich).

Behavioral tracking during two-photon imaging

We recorded running speed in MATLAB using a custom-made encoder (Arduino) coupled to the 3D-printed running wheel. See below for details of analyses of behavioral variables.

Induction of CSD and pharmacological treatment

For CSD induction, a 1 mm burr hole was drilled at the frontal bone (1.5 mm anterior to the cranial window) to allow access to the brain cortical surface. A small amount of silicone elastomer (Kwik-Cast, WPI) was placed to cover the burr hole opening, and the animal was left to recover for at least one week. CSD was induced using a brief 2-second cortical pinprick [53]. CSD induction was confirmed by the identification of a short-lasting meningeal deformation and/or transient pial constriction [53]. On the experimental day, 30 minutes before baseline recording, mice were pretreated with the CGRP/RAMP1 antagonist BIBN4096 (0.3 ml, 1 mg/kg, i.p., Tocris) [54] or 0.3 ml of saline (Vehicle control).

Quantification and statistical analysis

Two-photon imaging movie processing

We used a discrete Fourier transform to perform rigid registration to correct translation changes due to brain motion during locomotion. Movies were then downsampled to 1.03 Hz. Locomotion signals were detected as described [41]. All image processing and locomotion signal extraction were performed in MATLAB 2021b (Mathworks).

Ca²⁺ signal detection pipeline

For detecting macrophage Ca²⁺ signals, we used the Activity Quantification and Analysis (AQuA2) platform that implements an event-based approach with advanced machine learning techniques for temporal and spatial segmentation of Ca²⁺ fluorescence events³⁸. Importantly, this computational platform captures event dynamics beyond traditional ROI-based approaches. The following user-defined parameters were input: 0.49 µm/pixel spatial resolution, 1.03 Hz temporal resolution, 1-second minimal event duration detection; window of event size between 25% (125 µm²/ 523 pixels) and 150% (750 µm²/ 3140 pixels). Every event and cell identified was followed by a manual visual check. Subsequently, each cell was labelled as perivascular or non-perivascular according to its location relative to vessels. Morphological features (area, perimeter, circularity) and spatiotemporal aspects of the Ca²⁺ signals (i.e., ι1F/F₀ dynamics, frequency, amplitude, duration) were used to analyze cell-specific characteristics and Ca²⁺ activity profiles. We employed the AQuA2 automatically-generated function [42] to characterize intracellular Ca²⁺ propagation. Intercellular Ca²⁺ activity was evaluated by analyzing temporally co-occurring events (synchronized event pairs), as well as their corresponding spatial localization in the FOV. All post-processing of AQuA2-generated data was performed using MATLAB 2021b (Mathworks).

Clustering of Ca²⁺ dynamics

We used a Savitzky-Golay filter to detrend and smooth Ca²⁺ activity traces. Polynomial order and frame size were optimized for each cell by selecting the combination of parameters that yielded a better signal-to-noise ratio. From the filtered signals, we extracted the dominant frequency using Fast Fourier Transform, and peak counts using minimum peak prominence (threshold of 10% signal amplitude range). The optimal number of clusters was determined using the Elbow method, after which k-means clustering was applied to group cells based on their signal characteristics.

Analysis of CSD-related changes in macrophage Ca²⁺ dynamics

For analyzing the effects of CSD on macrophages’ Ca²⁺ dynamics, we divided the activity of each cell into three phases: ‘PreCSD’ (minutes 0 – 30), ‘DuringCSD’ (minutes 30 – 31), and ‘PostCSD’ (minutes 31 – 61). Ca²⁺ event rate during CSD and PostCSD was compared against that of the PreCSD baseline for evaluation of acute and persistent Ca²⁺ responses, respectively. Ca²⁺ responses were categorized as increased (event rate > 2x PreCSD), decreased (event rate < 0.5x PreCSD), or unchanged.

Analysis of locomotion

We extracted running speed (cm/sec) from the wheel encoder. To infer the locomotion state, we first concatenated all velocity signals obtained from a given mouse across all experiments and trained a two-state Hidden Markov Model using the MATLAB function ‘hmmtrain’. Then, the locomotion state was inferred for each individual imaging run by applying the MATLAB function ‘hmmviterbi’ with the model trained on the concatenated data. Locomotion bouts were defined as periods when the locomotion state was sustained for at least two seconds.

Vascular signals

We calculated changes in vascular diameter by first generating a maximum intensity projection of the red channel and drawing polygons (ROIs) around each vessel. Subsequently, for each frame, pixels inside each ROI were extracted, and a Radon transform was applied to get a 1D vessel profile. Radon transform of the first frame was used as a reference to normalize the results of all subsequent frames, resulting in a time series of normalized diameter traces.

General linear models (GLM)

We investigate the functional interaction between dural perivascular macrophage Ca²⁺ activity and dural vasomotion by fitting a Gaussian GLMs using the GLMnet package (MATLAB 2021b), with elastic net regularization (α=0.01) and ten-fold cross-validation. We employed a two-step modeling. The first model was used to identify dural vessels by evaluating the correlation of vessel diameter changes to locomotion state, classifying dural or pial vessels according to their vasoconstriction and vasodilation dynamics, respectively [46]. The second modeling step evaluated the relative contribution of perivascular dural Ca²⁺ activity signals to dural vessel diameter changes. All signals were downsampled to 1.03 Hz to match the frame rate used when extracting fluorescence Ca²⁺ signals in AQuA2. To allow potential anticipatory or delayed responses of diameter- to-locomotion or Ca²⁺ fluorescence-to-diameter, the time interval used to analyze the response to the predictor was set from -60 s to +60 s. The GLM was trained on 75% of the data, and all predictions and model performances reported are from the remaining 25% testing set. A threshold of 0.1 goodness-of-fit deviance explained was set. For each predictor temporal shift, a response coefficient was generated. Centroid delay between predictor and response was calculated, weighted by the absolute value of the coefficient (i.e., center-of-mass, COM).

Statistical analysis

All statistical analysis was performed using GraphPad Prism 10.4 and MATLAB 2021b. Data were analyzed using a Wilcoxon matched-pairs signed rank sum test or a Mann-Whitney U-test. Distribution of categorical data was analyzed using the Chi-square or Fisher’s exact test. P-values are indicated as follows: *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001.

Additional information

Data availability

All data needed to evaluate the conclusions in the paper are present in the paper. The code used for analyzing the data in this study was deposited in the Levy Lab GitHub account.

Code for movie processing is available at: https://github.com/levylabheadache/MovieProcessing/tree/SCN;

Code for locomotion analysis is available at: https://github.com/levylabheadache/Locomotion/tree/SCN;

Code for post Aqua2-processing is available at: https://github.com/levylabheadache/Aqua2Processing;

Code for vascular segmentation is available at: https://github.com/levylabheadache/Vasculature/tree/SCN;

Code for GLM is available at: https://github.com/levylabheadache/GeneralLinearModel_Macrophages

Any additional information is available from the corresponding author upon request.

Funding

The study was supported by NIH grants: R21NS130561; R01NS115972 and R01NS133625 to D.L.

Authors information

D.L. conceived the project. S.C-N and D.L. wrote the manuscript. S.C-N performed two-photon imaging and data analysis with help from D.L. S.C-N, C.W., and A.G. performed surgeries.

Additional files

Video S1. Macrophage Ca²⁺ dynamics in homeostatic meninges of awake mice. Example two-photon imaging showing Ca²⁺ activity in perivascular and non-perivascular meningeal macrophages.

Video S2. Acute and persistent changes in meningeal macrophage Ca²⁺ activity in response to CSD. Example two-photon imaging of meningeal macrophage Ca²⁺ activity at baseline, during, and following CSD. The arrow indicates a macrophage showing an acute Ca²⁺ elevation, the arrowhead depicts a delayed and persistent Ca²⁺ elevation, and the asterisk, a macrophage showing a persistent decreased Ca²⁺ activity.

Abbreviations

• AQuA2: Activity Quantification and Analysis

• Ca²⁺: Cytoplasmic calcium

• CGRP: Calcitonin gene-related peptide

• CSD: Cortical spreading depolarization (depression)

• CX3CR1: CX3C Motif chemokine receptor 1

• GLM: General Linear Model

• NP: Non-perivascular

• P: Perivascular

• Pf4: Platelet factor 4

• RAMP1: Receptor activity modifying protein 1

• TRITC: Tetramethylrhodamine isothiocyanate

Funding

National Institute of Neurological Disorders and Stroke (R21NS130561)

National Institute of Neurological Disorders and Stroke (R01NS115972)

National Institute of Neurological Disorders and Stroke (R01NS133625)