To investigate meningeal afferent responses to CSD, we performed two-photon calcium imaging of GCaMP6s-expressing trigeminal afferent fibers innervating the meninges above the visual cortex (n=325 fibers from 9 fields of view [FOVs] from 7 mice, Figure 1A). We triggered a single CSD episode in the frontal cortex with a cortical pinprick. In every experiment (9 CSDs in 7 mice), we detected a slow, CSD-like wave of calcium activity in numerous meningeal afferent fibers within 1 min following the pinprick (Figure 1B, Figure 1—video 1) as well as in background regions (likely reflecting signal from small, out-of-focus afferent branches). These calcium waves proceeded from the pinprick site in an anterior-to-posterior direction across the FOV (Figure 1C). We also observed progressive activation of portions of individual afferent fibers aligned to the wave’s movement direction. To characterize this phenomenon, we focused on sets of regions of interest (ROIs) belonging to the same long afferent fiber oriented along the direction of the calcium wave (Figure 1D, Figure 1—video 1). Compared to baseline afferent calcium signals observed during periods of locomotion, during which all ROIs belonging to an afferent were activated near-simultaneously, as previously reported (Blaeser et al., 2022d), the sequential recruitment of ROIs along an afferent fiber during the CSD-like wave was much slower (Figure 1E–G). The proportion of afferents activated during this period exceeded the proportion activated during locomotion bouts (Figure 1I). The magnitude of activation was also larger (Figure 1J).

CSD gives rise to acute neuronal and glial swelling and shrinkage of the cortical extracellular space (Mazel et al., 2002; Takano et al., 2007; Rosic et al., 2019). Such cortical mechanical perturbations could lead to acute meningeal deformation, which we have shown previously can activate mechanosensitive meningeal afferents (Blaeser et al., 2022d). We postulated that if CSD leads to meningeal deformations, these deformations could drive the acute afferent response during the CSD wave. Assessment of meningeal deformation parameters (see Materials and methods) revealed severe meningeal scaling and shearing during the CSD-evoked afferent calcium wave (6 CSDs in 6 mice, Figure 1—figure supplement 2A–D). In some experiments, the temporal pattern of meningeal shearing (Figure 1—figure supplement 2C) somewhat resembled that of the acute afferent response. However, the pattern of meningeal scaling (Figure 1—figure supplement 2B) was incongruent with the acute afferent response. Surprisingly, the direction of Z-shifts in the meninges during this epoch was inconsistent across mice, resulting in no significant Z-shift on average relative to the pre-CSD epoch (Figure 1—figure supplement 2D). To estimate the relative contribution of the CSD-driven meningeal deformation to the acute afferent responses we observed, we used a general linear model (GLM; see Blaeser et al., 2022d, and Materials and methods). We focused on afferents whose pre-CSD (baseline) activity could be predicted by the GLM based on deformation predictors (n=145 afferents from 6 mice). We then plugged the peri-CSD deformation data into these GLMs to generate predictions of peri-CSD afferent activity and compared them to the observed activity. Overall, we observed a poor match between the real peri-CSD calcium signals and those predicted by the GLMs trained using pre-CSD deformation data (see example in Figure 1—figure supplement 2E and summary GLM fits in Figure 1—figure supplement 2F), suggesting that the model poorly predicted the magnitude of the activity and/or its temporal pattern in response to CSD. Because ~95% of the afferents were acutely activated by CSD (Figure 1I), we propose that this response is mostly driven by other non-mechanical factors, such as the local depolarizing effects of diffusible excitatory molecules.

In anesthetized rats with exposed meninges, CSD drives sustained increases in ongoing activity lasting tens of minutes in ~50% of meningeal afferents (Zhao and Levy, 2015). To directly assess CSD-related changes in afferent ongoing activity in awake mice with intact meninges, we focused on afferent responses during epochs of immobility between locomotion bouts (8 CSDs in 7 mice). We observed low levels of ongoing activity at baseline before CSD (fluorescent events occurring 6.9 ± 0.3% of the time), consistent with our previous study in naïve mice (Blaeser et al., 2022d). Unexpectedly, most afferents (~70%, 201/288) did not display any change in ongoing activity during the 2 hr following CSD (termed ‘post-CSD’). However, we identified sustained increases in ongoing activity in ~10% (30/288) of the afferents during this period. Surprisingly, we also observed a larger afferent population (~20%; 57/288) whose activity was suppressed (Figure 2A–C). Afferents with sustained activation showed increased ongoing activity that emerged at an ~25 min delay on average (Figure 2D). In contrast, afferents with sustained suppression showed decreases in ongoing activity beginning shortly after the passage of the acute calcium wave (Figure 2D). The durations of the afferent activation or suppression were similar, lasting ~25 min on average (Figure 2E).

Meningeal deformation associated with locomotion bouts can lead to the activation of mechanosensitive meningeal afferents (Blaeser et al., 2022d). We wondered whether, following CSD, afferent responses to a given level of mechanical deformation would be enhanced (i.e. mechanical sensitization). If so, this could explain the exacerbation of migraine headaches during physical activity. CSD suppresses cortical activity, leading to decreased motor function (Houben et al., 2017), including reduced locomotion in head-fixed mice (Enger et al., 2017). CSD-related vascular changes and reduced extracellular space (Mazel et al., 2002; Takano et al., 2007) could also affect meningeal deformations and the associated afferent response. Hence, we first analyzed the effect of CSD on wheel running activity and the associated meningeal deformation. In most sessions, mice stopped locomoting following the CSD (8/9 CSDs in 7 mice) but resumed sporadic wheel running activity ~6 min later on average (range 0.5–16.5 min). However, the mice ran less during the 2 hr following CSD than during the baseline period (Figure 3B). Locomotion bout analysis also revealed an overall reduction in bout rate during the post-CSD period (Figure 3C). Remarkably, despite the reduction in locomotion following CSD, we observed similar bout characteristics at baseline and post-CSD, including bout duration (Figure 3D) and peak velocity (Figure 3E). Given that CSD had minimal effect on locomotion bout characteristics, we next examined its effect on meningeal deformations. Surprisingly, CSD did not affect bout-related meningeal deformations: we observed similar scaling, shearing, and Z-shift values during the 2 hr post-CSD (Figure 3F–H).

Having shown that locomotion bout charactersitics and the related meningeal deformations are not altered during the 2 hrs following CSD, we next compared afferent responses during locomotion bouts before and after CSD. Initial observations of afferent activation during locomotion indicated enhanced responsiveness following CSD (Figure 4A). To systematically investigate this augmented afferent response, we used GLMs (see Materials and methods and Blaeser et al., 2022d) to assess whether meningeal afferent activity becomes sensitized to the state of locomotion and/or to various aspects of meningeal deformation following CSD. We modeled each afferent’s activity based on variables that describe (1) the binary state of locomotion, (2) mouse velocity, or (3) aspects of meningeal deformation, including scaling, shearing, and Z-shift.

We first focused on afferents whose activity could be predicted by the same variables both at baseline and following CSD (i.e. afferents that exhibited sensitivities to locomotion and/or deformation signals both before and after CSD, n=67/325 afferents, 9 CSDs in 7 mice). Higher GLM coefficients for a given variable post-CSD indicate greater afferent response during an equal expression level of that variable. Thus, we defined an afferent as sensitized by CSD if its GLM coefficients post-CSD were larger than at baseline (i.e. stronger activation of afferents per unit deformation or locomotion; for example, see Figure 4C). Using these criteria, we identified elevated locomotion and/or deformation-related activity (i.e. sensitization) post-CSD in ~51% of afferents (34/67; Figure 4D). In contrast, only 12% of afferents (8/67) showed reduced locomotion- and deformation-related activity (i.e. desensitization) post-CSD. Sensitivity was unchanged in the remaining 37% of afferents (25/67).

Meningeal afferent sensitization following CSD may reflect increased sensitivity to mechanical deformation and/or to other physiological processes that occur in response to locomotion (Blaeser et al., 2022d). Because locomotion and meningeal deformations are partially correlated (Blaeser et al., 2022d), we next estimated their relative contributions to the augmented afferent responsiveness post-CSD by comparing, for each sensitized afferent, the GLM coefficients generated for baseline epochs and for post-CSD epochs. Surprisingly, we found that only the deformation coefficients were increased post-CSD (Figure 4E), suggesting that meningeal afferent sensitization following CSD reflects primarily an increased sensitivity to local mechanical deformations.

We next considered the possibility that sensitization is also manifested in the unmasking of responsiveness to locomotion and/or meningeal deformation in previously silent (i.e. insensitive) meningeal afferents (Levy and Strassman, 2002; Levy and Moskowitz, 2023). Indeed, among the 245 afferents that were not well fit before CSD (i.e. insensitive), we detected a substantial population that developed a sensitivity to locomotion and deformation variables following CSD (53/245; i.e. afferents whose activity could be well predicted by locomotion and deformation variables following CSD). By contrast, far fewer neurons (13/245) lost their sensitivity to these variables after CSD (i.e. afferents that were well fit before but not after CSD). Overall, the above findings show that four times more afferents displayed increased sensitivity than decreased sensitivity following CSD (87 vs 21 afferents). This sensitized afferent population also displayed post-CSD increases in GLM coefficients related to deformation but not to locomotion (Figure 4F).

To further quantify the importance of the locomotion and deformation variables to the afferent sensitization, we estimated their relative contributions to the overall ability of the models to predict the afferent activity pre- vs. post-CSD. To this end, we calculated the difference in model fit using the full model with all variables or models lacking either the set of deformation variables or locomotion variables. We found that the impact of deformation variables on the model fit was greater post-CSD than pre-CSD, while the impact of locomotion variables was similar pre- and post-CSD. This was true for afferents with models that were well fit both at baseline and following CSD (Figure 4G; purple subset in Figure 4D) and for those with models that were well fit only post-CSD (Figure 4H; magenta subset in Figure 4D). Further analysis revealed that scale, shear, and Z-shift deformations were, on average, equally important in predicting the activity patterns of sensitized afferents (Figure 4I). Taken together, these data suggest that the afferent sensitization following CSD primarily involves increased afferent responsiveness to a mix of meningeal deformation variables rather than to other locomotion-associated processes.

Previous studies in anesthetized rats suggested that the mechanisms underlying meningeal afferent mechanical sensitization are independent of those responsible for increased ongoing discharge in several migraine models, including CSD (Levy and Strassman, 2002; Zhang et al., 2011; Zhang et al., 2013; Zhao and Levy, 2018; Zhao et al., 2021). Here, using the CSD model in awake mice, we also found no association between sensitization and sustained changes in ongoing activity, as similar proportions of sensitized and non-sensitized afferents were activated, suppressed, or did not display any change in their ongoing activity following CSD (Figure 4J).

All experimental procedures complied with the ARRIVE and were approved by the Beth Israel Deaconess Medical Center Institutional Animal Care and Use Committee (protocols #105-2015, 072-2021). All experiments were conducted on adult (8–16 weeks of age) C57BL/6J mice (5 males, 2 females, Jackson Laboratory). Mice were group-housed with standard mouse chow and water provided ad libitum before viral injection (see below). Mice used for in vivo two-photon imaging were singly housed and provided a running wheel, a hut, and a chew bar.

All surgical procedures were performed in anesthetized mice (isoflurane in O₂; 3.5% for induction, 1.5% for maintenance). Animals were given Meloxicam SR (4 mg/kg s.c.) for post-surgical analgesia. For monitoring calcium activity in meningeal afferents, 2.0 μl of AAV2/5.CAG.GCaMP6s.WPRE.SV40 (titer: 1×10¹³; Addgene, 100844-AAV5; RRID:100844) was injected into the left trigeminal ganglion (TG) using the following stereotaxic coordinates: 1.5 mm lateral and 0.3–0.8 mm anterior to Bregma and 7.0–7.2 mm ventral to the dura at a lateral-to-medial tilt with an angle of 22.5° relative to the dorsal-ventral axis. We previously verified this approach by examining GCaMP6s expression in TG somata and meningeal afferent fibers 8 weeks after injection using immunohistochemistry (Blaeser et al., 2022d). Mice used for in vivo two-photon imaging were instrumented with a titanium headpost and a 3 mm cranial window (Goldey et al., 2014; Blaeser et al., 2022d) covering the posterior cortex (window centered roughly 1.5 mm lateral and 2 mm posterior to Bregma over the left hemisphere) 6–8 weeks after AAV injection. For CSD induction, a burr hole (~0.5 mm diameter) was drilled 1.5 mm anterior to the edge of the window until the brain’s surface was barely visible. The burr hole was then plugged using a silicone elastomer (Kwik-Cast, WPI), and the mouse was allowed to recover. To trigger a single CSD episode, a glass micropipette (50 μm diameter) was briefly inserted through the silicon plug ~1 mm deep into the cortex for 2 s (Zhao and Levy, 2016).

After at least 1 week of recovery following cranial window implantation, mice were head-fixed on a 3D-printed running wheel for gradual habituation (10 min to 1 hr over 3–4 days). To minimize downward forces on the bone and meninges produced while the mouse pushed upward against the headpost, the running wheel was mounted on a cantilever (Ramesh et al., 2018; Blaeser et al., 2022d). The position of the headpost, anterior to where the mouse’s paws touch the wheel, also made it hard for the mouse to push straight up and apply forces to the skull. The strong cement used to bind all skull plates and headpost together (Goldey et al., 2014) further mitigated any movement-induced strain on the skull that might affect the underlying meninges. Mice displaying signs of stress were immediately removed from head fixation, and additional habituation days were added until mice tolerated head fixation without visible signs of stress. Mice received a high-calorie liquid meal replacement (Ensure) via a syringe as part of the habituation process.

Calcium imaging was performed as recently described (Blaeser et al., 2022d) while mice were head-fixed on the running wheel. We used a Nikon 16X, 0.8 NA water immersion objective on a resonant-scanning two-photon microscope (Neurolabware) and a MaiTai DeepSee laser, set to 920 nm with 25–40 mW power for GCaMP6s visualization. Digital zoom was set at 2.4× (626×423 μm² FOVs). In seven experiments, we imaged across a 60 μm volume (3D) using an electrically tunable lens (Optotune) at 1.03/s. We employed volumetric imaging for three main reasons: (1) To capture the activity of afferents throughout the meningeal volume. In our volumetric imaging approach, including in this work, we observed afferent calcium signals throughout the meningeal thickness (see Figure 5 in Blaeser et al., 2022d). However, the majority of afferents were localized to the most superficial 20 µm (Figure S1E in Blaeser et al., 2022d), suggesting that we mostly recorded the activity of dural afferents; (2) to enable simultaneous quantification of 3D deformations and the activity of afferents throughout the thickness of the meninges. This allowed us to determine whether changes in mechanosensitivity could involve augmented activity in response to intracranial mechanical forces that produced meningeal deformation along the Z-axis of the meninges (e.g. increased intracranial pressure); (3) to provide a direct means to confirm that the afferent GCaMP fluorescent changes we observed were not due to artifacts related to meningeal motion along the Z-axis. In two experiments, only single-plane (2D) data were collected at the cranial dura level at 15.5 Hz due to technical issues. In every experiment, we conducted two imaging runs (30 min each) to collect baseline data, followed by four more 30 min runs post-CSD induction. In a subset of experiments in which the FOV included a visible large pial artery (n=4), we verified the induction of a CSD by visualizing its vascular signature, including a brief vasoconstriction followed by dilation (Rosic et al., 2019; Figure 1—figure supplement 1A).

Wheel position during each imaging run was recorded using an Arduino Uno board at 15.5 Hz. The instantaneous velocity was calculated as the time derivative of this signal and was downsampled to match the sampling rate of volume scans. Locomotion state was determined using a two-state hidden Markov model. Locomotion bouts were defined as periods when the locomotion state was sustained for at least 2 s (Blaeser et al., 2022d).

All image processing and analyses were performed in MATLAB 2020a and ImageJ (Fiji, NIH) as described previously (Shipley et al., 2020; Blaeser et al., 2022d). In brief, imaging movies were subjected to several preprocessing steps. For single-plane experiments, all imaging runs were concatenated into a single movie. A reference image was defined as the mean projection over the middle 50% of frames from the second run pre-CSD. Each frame of the concatenated movie was then affine registered to the reference using the TurboReg plugin for ImageJ. For volumetric imaging, we first corrected the lensing effect due to the electrically tunable lens. We next subjected the movies to rigid registration, using the discrete Fourier transform to correct for within-volume (2D) translations and translations along the z-axis. We next used z-interpolation to correct for z-translations of individual imaged planes within each volume. The last step involved affine registration using a reference volume formed from the mean projection of the middle 50% of frames from the second run.

Registered 2D movies were analyzed using a PCA/ICA package (Mukamel et al., 2009) to extract masks of pixels with correlated activity. Users screened each prospective ROI for morphology and fluorescence signal quality. Afferent ROIs with <50 pixels were rejected. For each ROI included in subsequent analyses, we generated a dilated mask extending 8–21 pixels from the outer edge of the ROI, excluding any pixels that belonged to another ROI. This ‘neuropil’ mask was used for the subtraction of background signals. In volumetric imaging, we generated a mean projection over all planes containing afferents for each volume. Then, we ran the resulting 2D movie through the PCA/ICA procedure, yielding an initial set of 2D masks representing putative 3D ROIs. An initial fluorescence trace extracted from each 2D mask was calculated by averaging fluorescence across all pixels in the mask. To identify which voxels in the original volumetric dataset contributed most strongly to each fluorescence trace, we calculated the Pearson correlation of this trace with the fluorescence time course of each voxel, resulting in a 3D volume of correlation values. These 3D correlation volumes were then screened manually for quality of morphology and signal. The surviving volumes were thresholded at the 75th percentile of correlation across all voxels to form putative 3D ROI masks.

We calculated raw fluorescence signals at each time point for the ith ROI (Fⁱ_ROI) and its corresponding neuropil mask (Fⁱ_np), as the simple arithmetic means of all pixels/voxels within each ROI mask. Next, we calculated Fⁱ = Fⁱ_ROI – Fⁱ_np +<Fⁱ_np > , where brackets denote the mean across the entire recording. We then calculated the corresponding baseline signal Fⁱ₀ as the 10th percentile of a moving window for the last 32 s (Sugden et al., 2020). We then calculated the normalized, baseline-subtracted time series ΔF/F₀ = (Fⁱ –Fⁱ₀)/Fⁱ₀. This signal was standardized akin to a Z-score operation by subtracting the median value and dividing by the standard deviation (calculated during quiet wakefulness, an epoch with low levels of evoked activity). Fluorescence events were defined as periods where the signal consistently exceeded a value of 1 for at least 1 s and where peak fractional change in fluorescence (∆F/F₀) was at least 5%.

To analyze calcium activity related to an afferent fiber, sets of ROIs putatively belonging to the same axon were initially identified using a previously described method (Liang et al., 2018; Blaeser et al., 2022d). We calculated the pairwise fluorescence event correlation between ROIs during quiet wakefulness, thresholding at 0.7 correlation. We then calculated the cosine dissimilarity between the full set of correlation coefficients for each pair of ROIs, which was used to calculate the linkage between each pair. Finally, hierarchical clustering was performed using a cutoff value of 2. This procedure generated sets of ROIs that were mutually highly correlated. We then visually inspected each cluster and manually identified the subsets of ROIs that unambiguously covered the same specific afferent fiber without any branching. We used the mean activity of these ROI subsets to analyze each afferent’s calcium activity.

To detect the CSD-associated meningeal calcium wave (see Figure 1—figure supplement 1B), we first calculated the mean fluorescence signal over all voxels in the FOV, F_FOV(t), and its time derivative dF_FOV/dt, in the first 5 min after cortical pinprick. To identify the timing of the CSD, we first defined an initial period, t_inital, between the derivative’s maximum and minimum. We then calculated baseline fluorescence, F_pre, as the 10th percentile value of F_FOV in the 30 s before the initial period and a normalized fluorescence signal, ΔF/F_pre = (F_FOV – F_pre)/F_pre. Next, we defined a threshold value as 0.1*max(ΔF(t_initial)/F_pre). The wave’s final onset and offset times were defined as the points around the peak where ΔF_FOV/F_pre crossed the threshold value.

To measure the propagation of the calcium wave throughout the meninges, we estimated the wave’s onset time at different X and Y positions within the FOV. Briefly, we gathered all the data from the peri-CSD wave from –6 to 14 s relative to the calcium wave onset time determined above. All voxels belonging to or neighboring (within 8 pixels) ROIs were excluded to focus on the overall wave-like advancement in the background ‘neuropil’ fluorescence signal (which integrates background fluorescence across depths and thus reflects a smooth estimate of mean activity in a given region). The resulting series of images was then divided into 40×40 pixel spatial bins, and mean fluorescence signals F_np^bin(t) were calculated. The time of estimated onset of the CSD-associated wave at a given bin was estimated by fitting F_np^bin(t) to a logistic function, A/(1+exp(-(t-t_onset)/τ))+K. Fits with R²<0.5, or τ>2 s, or τ<0 s, were excluded. The wave’s speed was estimated by linear regression of onset times as a function of distance. The direction of the wave was determined by estimating the contour of the wavefront at the median bin onset time, fitting this contour to a line, and then calculating the vector orthogonal to that line.

To specifically examine the propagation of the fluorescence signals along long (>200 μm) individual afferent fibers, we determined the timing of CSD-associated wave onset by fitting each ROI’s peri-wave fluorescence signals to a sigmoidal function and then estimating the speed of propagation by linear regression of onset times as a function of distance. For comparison, this procedure was repeated using the peri-locomotion bout signals. Since locomotion-associated activation occurred essentially simultaneously (i.e. Δt=0 s), even for distant ROIs, we report these results in terms of the pace (the inverse of speed) to avoid dividing by zero.

To determine changes in ongoing afferent activity, we focused on the periods of immobility between locomotion bouts (stillness). We minimized any residual effects of locomotion-related activity by excluding 30 s epochs before and following the locomotion bouts. We then analyzed fluorescence events as above. We estimated levels of ongoing activity rate from the normalized Z-score time courses as the fraction of time the afferents exhibited fluorescent events (defined as above) during each 1 min interval. We defined afferents with post-CSD increases or suppression of activity if changes in ongoing activity lasted >10 consecutive min and began within 30 min following the CSD wave (Zhao et al., 2021).

To estimate the degree of meningeal deformation in locomoting mice, we extracted the values produced from the affine registration procedure used to correct intra-volume (XY) and inter-volume (Z) image displacements (see above). Corrections made to account for XY displacements were used to quantify the amount of scaling (expansion and compression) and shearing. Corrections made along the z-axis measured how much each plane moved up or down relative to a reference volume (‘Z-shift’). Scaling and Z-shifts were converted from pixels to microns. Positive Z-shifts indicate meningeal movement toward the skull.

To classify afferent responses, we fit Gaussian GLMs for locomotion, deformation, and fluorescence using the glmnet package in MATLAB. To allow for the possibility of a delay between the predictor and response, we expanded the set of predictors to include variables at varying delays relative to the calcium activity (Driscoll et al., 2017; Ramesh et al., 2018; Blaeser et al., 2022d). Specifically, we generated a set of temporally shifted versions for each variable spanning a time window from –6 to +6 s. These sets of arrays of temporal shifts for each variable were joined to form an array of temporally shifted predictor signals. The GLM was fit on 75% of the data for each cell with elastic net regularization (α=0.01). We then used the GLM coefficients to measure the deviance explained on the remaining 25% of the data. The relative explanatory value of locomotion or deformation variables was calculated by refitting the GLM after excluding the family of predictors (locomotion state and velocity for evaluating the effect of locomotion and scale, shear, and Z-shift for examining the effect of deformation) and calculating the difference in deviance explained by the full model versus the model lacking a given family of predictor variables. All GLMs underwent 10-fold cross-validation.

Data analyses were performed in MATLAB 2020a and Prism 9. Sample sizes were not predetermined by power analysis but are similar to previous studies (Sugden et al., 2020; Blaeser et al., 2022d). Two-tailed paired t-tests and one-way analysis of variance followed by a post hoc Tukey’s test were used for all parametric data. Data with non-Gaussian distributions were analyzed using a Wilcoxon matched-pairs signed rank sum test or a Mann-Whitney U-test. Corrections for multiple comparisons were adjusted using the false discovery rate approach. Bootstrapped confidence intervals and hypothesis tests were generated using the ‘iboot’ iterated bootstrapping package (Penn, 2020). Unless otherwise noted, data are presented as averages ± standard error of the mean (SEM). P-Values are indicated as follows: p<0.05 (*), p<0.01 (**), p<0.001 (***), p<0.0001 (****).