Atp1a2 +/ − is a model mouse of familial hemiplegic migraine type 2. The numbers of astrocytes with elevated Ca2+ concentrations in Atp1a2 +/ − (11.0 ± 2.8 astrocytes) were larger than those in wild‐type mice (5.7 ± 2.5 astrocytes) after induction of cortical spreading depression (which is closely associated with migraine) by application of KCl.
Migraine is a common disorder characterized by recurrent debilitating headache attacks  that are preceded by aura in a third of patients . Cortical spreading depression (CSD) is a propagating depolarizing wave in neurons and glial cells that spreads across the cerebral cortex and that is followed by a subsequent sustained suppression of spontaneous neuronal activity [3, 4]. Migraine patients show multiphasic cerebrovascular changes that are often observed in CSD during visual aura. Multiphasic cerebrovascular changes appear to be directly linked to the aura percept in both space (retinotopy) and time . In addition, migraine headache depends on the activation of the trigeminovascular system [4, 6]. Animal experiments have indicated that CSD activates meningeal nociceptors in the trigeminovascular system . These results indicate that CSD is closely associated with migraine aura and headache .
Findings from the monogenic form of diseases provide clear insights into the molecular pathways of diseases. Familial hemiplegic migraine (FHM) is a rare monogenic form of migraine with aura including motor weakness , and three FHM causative genes have been identified: CACNA1A (coding for the α1a subunit of the CaV 2.1 calcium channel) for the FHM1 ; ATP1A2 (coding for the α2 subunit of Na,K‐ATPase) for the FHM2 ; and SCN1A (coding for the α subunit of the NaV 1.1 sodium channel) for the FHM3 . Gain‐of‐function mutations in the CACNA1A gene causing FHM1 lead to increased Ca2+ influx in presynaptic terminals, which, in turn, results in increased glutamate release in the synaptic cleft [11, 12]. Loss‐of‐function mutations in ATP1A2 causing FHM2 lead to the reduced clearance of glutamate and K+ by astrocytes . Loss‐of‐function mutations in SCN1A causing FHM3 lead to the reduced inhibitory activity of inhibitory interneurons, which results in increased excitatory neuronal activity . Increased glutamate in the synaptic cleft results in increased cortical excitatory neurotransmission [10, 15]. Consistently, a decreased threshold for CSD induction has been observed in FHM1 and FHM2 model animals [16, 17].
Na,K‐ATPase consists of α and β subunits and maintains the electrochemical gradient of Na+ and K+ across the cell membrane using the energy of ATP hydrolysis [18, 19]. There are four isoforms of the α subunit (α1–α4) and three isoforms of the β subunit (β1–β3). The α2 subunit, which is coded by Atp1a2, is predominantly localized in astrocytes in adult rats . Heterozygous Atp1a2 knockout mice, Atp1a2tmCKwk/+ , and heterozygous knock‐in mice carrying the human W887R mutation in Atp1a2  have been developed for FHM2 model animals. CSD was induced by electric stimulation or KCl application into the cortex of these animals, and both types of mice showed a decreased threshold for CSD induction than wild‐type (wt) mice, suggesting increased CSD susceptibility [17, 22]. In addition, Atp1a2tmCKwk/+ mice showed enhanced fear/anxiety behaviors  and obesity by hyperphagia . These observations are consistent with the characteristics of FHM2 patients showing anxiety [24, 25] and obesity , indicating that the Atp1a2tmCKwk/+ mouse reproduces FHM2 symptoms and is a useful model animal for clarifying the pathophysiology of FHM2. However, the properties of astrocytes and their possible influence on neuronal activity during CSD have not been examined in FHM2 model mice.
To this end, we analyzed changes in Ca2+ concentrations using the fluorescent calcium indicator G‐CaMP7  in astrocytes and neurons as a proxy for their excitability [28, 29, 30]. Atp1a2tmCKwk/+ mice crossed with transgenic mice expressing G‐CaMP7 in the cortex and hippocampus  were used to monitor changes in Ca2+ concentration in cortical neurons and astrocytes.
All animal experiments were carried out in a humane manner. The Institutional Animal Experiment Committees of Jichi Medical University and the Saitama University approved this study. This study was conducted in accordance with the Institutional Regulations for Animal Experiments and Fundamental Guidelines for Proper Conduct of Animal Experiments and Related Activities in Academic Research Institutions under the jurisdiction of the MEXT of Japan.
Atp1a2‐deficient mice (Atp1a2tmCKwk/+ , ) and G7NG817 transgenic mice  expressing the fluorescent calcium indicator G‐CaMP7  were crossed. Mice of the F1 generation (Atp1a2tmCKwk/+; G7NG817+/−) were inbred. Then, Atp1a2tmCKwk/+; G7NG817+/+ and Atp1a2+/+; G7NG817+/+ mice were selected from the F2 generation and crossed. In the F3 generation, female Atp1a2tmCKwk/+; G7NG817+/+ and Atp1a2+/+; G7NG817+/+ mice were called Atp1a2+/− and wt mice, respectively, and were used in this study. Mice were housed under a 12‐h light/dark cycle (lights on from 7:00 AM to 7:00 PM) in a temperature‐controlled room (22 °C ± 2 °C). Food and water were provided ad libitum.
Female postnatal 2‐ to 5‐month‐old mice were used (Fig. 1A). On day 1, mice were anesthetized with isoflurane (3% induction, 1.5% maintenance) and placed in a stereotaxic frame with ear bars . After skull exposure, a stainless steel head plate with a circular opening (7 mm diameter) was placed over the left parietal bone and was attached to the skull with dental acrylic. The mice were returned to their home cages.
On day 2, the mice were anesthetized with isoflurane (3% induction, 1.5% maintenance) supplemented with chlorprothixene (1 mg·kg−1, i.p.). Atropine (0.3 mg·kg−1, s.c.) and dexamethasone (2 mg·kg−1 , s.c.) were administered to reduce respiratory secretions and brain edema, respectively . The mice were placed in a stereotaxic frame via the head plates (Fig. 1B). A piece of skull (~ 2 mm diameter) within the circular opening of the head plate was surgically removed for imaging. Then, a small hole (~ 1 mm in diameter) was opened 4.5 mm anterior to the center of the imaging window for KCl application. Ten microliters of 100 μm sulforhodamine 101 was applied to the exposed cortical surface to label astrocytes. After 5 min of application, the cortical surface was washed 5 times with cortical buffer (123 mm NaCl, 5 mm KCl, 10 mm glucose, 2 mm CaCl2, 2 mm MgCl2, and 10 mm HEPES at pH 7.4) and covered with 1% agarose.
During two‐photon imaging, the mice were anesthetized with 1% isoflurane. Body temperature was maintained with a heating pad throughout the imaging session. To avoid the risk of evoking abnormal cortical neural activity by sulforhodamine 101 , images were acquired using a Nikon A1RMP microscope equipped with an Apo LWD 25 × 1.10 objective (Nikon, Tokyo, Japan) after about 60 min from application of sulforhodamine 101. For imaging spontaneous activity, 512 × 512 pixel images (field size 254 × 254 µm at depths of 200 µm and 250 µm) were acquired at 15 frames per second for 5 min. From each mouse, two fields of view were imaged each at 200 µm and 250 µm deep.
Cortical spreading depression was induced after imaging spontaneous activity. Imaging was started immediately after the application of 2 μL of 1 m KCl to the hole. Images in 512 × 512 pixels (field size 508 × 508 µm) were acquired at one frame/ 3 s at a depth of 200 µm for 30 min.
We defined regions of interest (ROI) on neurons manually from average Z‐stacked images (Fig. 2A), and the mean values (F ) of fluorescence intensity within the ROIs in each frame of the time‐lapse image sequences were calculated (Fig. S1A) by fiji (ImageJ) (National Institutes of Health, Bethesda, MD, USA) software . The obtained data were subsequently analyzed by r software . The baseline value (F0) of each ROI was defined as the first quartile of all the data points. F0 was used to calculate ΔF/F as (F‐F0)/F 0 (Fig. 2B). A simple moving average with a window size of five frames was applied to ΔF/F (Fig. S1B). Neuronal activity was defined as the local maxima of ΔF/F above the threshold (4.5 SD from F 0; Fig. S1C,D). Then, the duration, area, number, and peak of activities were measured (Fig. S1E‐G).
We identified ROIs on 20 neurons and 20 sulforhodamine 101‐labeled astrocytes selected manually from images after propagating waves of transiently increased G‐CaMP7 fluorescence (Fig. 3E,M and H,P). The mean values (F) of fluorescence intensity within each ROI in each frame of the time‐lapse image sequences were obtained by fiji software (Fig. S2A). To determine the baseline value, Ca2+ waves were defined as the periods during which the mean value of F across 40 ROIs was above 2500 (Fig. S2B,C). Then, the baseline value (F0) was defined as the median value of fluorescence intensity before the first Ca2+ waves (Fig. S2D). F0 was used to calculate ∆F/F as (F‐F0)/F 0 (Fig. S2E). The number of Ca2+ waves in each image sequence was counted visually as the number of discontinuous peaks of Ca2+ waves (Fig. S2C).
Minimum values of fluorescence intensity were obtained from ΔF/F time series from which increased intensity during Ca2+ waves was omitted (Fig. S2F,G). To analyze changes in fluorescence intensity after Ca2+ waves, Ca2+ waves and 30 frames before and after the Ca2+ waves were excluded from the ΔF/F data (Fig. S2H). The rate of change in fluorescence intensity during the periods after (or between) Ca2+ waves was quantified as the slope obtained by linear regression (Fig. S2I). The slopes obtained from multiple periods were then averaged for each ROI.
The ΔF/F time series after or between Ca2+ waves generally appeared as upward curves. To evaluate the transient elevation of fluorescence intensity in neurons and astrocytes, the baselines were normalized by calculating ΔF/F − (slope × frame + intercept). Slope and intercept were calculated by linear regression using the mean values of ΔF/F from 40 ROIs. Peaks of activity were defined as the local maxima of normalized ΔF/F values above the threshold (5, 7, and 10 times the SD of fluorescence intensity before the first Ca2+ waves; Fig. S2J,K). The number, duration, and peak of activity events for each ROI and fractions of active ROIs among a total of 20 ROIs were calculated.
In each mouse, pairwise correlation of the changes in fluorescence intensity between two different ROIs was calculated using the normalized ΔF/F data (Fig. 6A,B). The data were binarized according to the presence and absence of activity (1 and 0, respectively) after thresholding (five times the SD of fluorescence intensity before the first Ca2+ waves; Fig. 6C). Then, the frames in which all of the ROIs were not active (Fig. 6D) and the ROIs that did not show activity were omitted (Fig. 6E). We then calculated the correlation coefficients between ROIs (Fig. 6F). Correlation coefficients for neuron–neuron, neuron–astrocyte, and astrocyte–astrocyte pairs were separately averaged. To estimate the correlation between changes in fluorescence intensity within single astrocytes, ROIs were manually defined on their cell bodies and processes (Fig. S5A). The mean values (F) of fluorescence intensity within each ROI in each frame were then calculated. To determine the baseline value, Ca2+ waves were defined as the periods during which the mean value of F across all ROIs was above 3700. Then, the baseline value (F0) was defined as the median value of fluorescence intensity before the first Ca2+ waves. F0 was used to calculate ΔF/F as (F‐F0)/F0. To analyze changes in fluorescence intensity after Ca2+ waves, the Ca2+ waves and 30 frames before and after the Ca2+ waves were excluded from the ΔF/F data. The baselines were normalized by calculating ΔF/F − (slope × frame + intercept). The slope and intercept were calculated by linear regression using 10 percentile values from all ROIs. Correlation coefficients of fluorescence changes within single astrocytes were calculated as described above for those between cells.
The speed of propagating waves of transiently increased G‐CaMP7 fluorescence was measured by the distance between the wavefronts of increased fluorescence imaged at nth and (n + 1)th frames divided by the time interval between the two frames. The position of the wavefront at the (n + 1)th frame was defined as the intersectional position of the perpendicular line from the position at the nth frame to the wavefront at the (n + 1)th frame. In a wave, we selected three different positions where we could easily draw the perpendicular line, and measured the lengths of three perpendicular lines. Then, the mean values were used for statistical analysis.
All statistical analyses were performed using r . Significance was set at P < 0.05. Analysis of variance (ANOVA) was performed to examine the effect of genotype and observation depth. Differences between Atp1a2+/− and wt mice were tested by t‐test or Wilcoxon rank‐sum test.
To evaluate whether spontaneous neural activity in FHM2 model mice Atp1a2tmCKwk/+  is different from that in wt mice, we observed spontaneous changes in Ca2+ concentrations in the cortex of anesthetized mice by monitoring the fluorescence intensity of the fluorescent calcium indicator G‐CaMP7. Atp1a2tmCKwk/+ and G7NG817 transgenic mice expressing G‐CaMP7  were crossed. Images were obtained from the cortex at depths of 200 µm and 250 µm in Atp1a2tmCKwk/+; G7NG817+/+ (Atp1a2+/−) and Atp1a2+/+; G7NG817+/+ (wt) mice. Sums of the duration, area, and number of activities in each ROI did not show significant effects of genotype and observation depth in two‐way ANOVA (genotype × depth) (Fig. S1H‐J). Similarly, sums of the duration, sum area, and number of activities in each field of view showed no significant effects in two‐way ANOVA (genotype × depth) (Fig. S1K‐M). Peak ∆F/F of each activity also showed no significant effects in two‐way ANOVA (genotype × depth) (Fig. S1N). Because there is no effect of depth on spontaneous activity, we summed the duration, area, number, and peak ∆F/F of activities in each mouse (Fig. 2C‐F). Again, sums of the duration, area, number, and peak ∆F/F of activities in each mouse were not significantly different between Atp1a2+/− and wt mice (Fig. 2C‐F). The duration and area divided by the number of activities were similar between Atp1a2+/− and wt mice (Fig. 2G,H). The number of active ROIs among all ROIs (40 ROIs) in each mouse (%) was not different between Atp1a2+/− and wt mice (Fig. 2I). These observations indicate that spontaneous neural activity did not show any significant differences between Atp1a2+/− and wt mice. Spontaneous activity was not observed in astrocytes labeled by sulforhodamine 101 (data not shown) in our experiment, although it has been reported that the G7NG817 transgenic mice show spontaneous activities in astrocytes . The discrepancy may be due to the differences of method of anesthesia and/or the degree of recovery of mice from surgery.
To examine whether changes in astrocyte properties influence neuronal activity during CSD, we investigated Ca2+ concentration changes in astrocytes and neurons during CSD. CSD can be induced by KCl directly applied to the cortex . After the application of KCl, propagating waves of transiently increased G‐CaMP7 fluorescence intensity were observed across the cortex in Atp1a2+/− (Fig. 3I‐O, Video S2) and wt mice (Fig. 3A‐G, Video S1). Some mice showed multiple propagating waves after a single application of KCl. The number of propagating waves was not different between Atp1a2+/− and wt mice (Fig. 4A). However, the speed of the second propagating waves in Atp1a2+/− mice (33.72 ± 4.33 µm/s) was significantly faster than that in wt mice (27.56 ± 1.38 µm/s; Fig. 4B). This result was consistent with the results showing the faster propagation of CSD in electrophysiological recordings in FHM2 model mice [17, 22]. In electrophysiological recordings, propagation speeds of Atp1a2tmCKwk/+ and wt mice were about 4.5 and 3.9 mm/min, respectively . Interestingly, some astrocytes and neurons showed increased fluorescence intensity after the propagation of the waves (Fig. 3N,O). We then quantified the fluorescence intensity in neurons and astrocytes after the propagation of the wave by establishing ROIs on them. Changes in fluorescence intensity in each ROI were quantified as minimum values and slopes calculated by regression analysis. The minimum values of fluorescence intensity (Fig. 4C) and the slopes of fluorescence intensity (Fig. 4D) were similar between Atp1a2+/− and wt mice in astrocytes and neurons. In each ROI, minimum values and slopes of fluorescence intensity showed significant differences between Atp1a2+/− and wt mice neither in astrocytes nor in neurons (Fig. S3A–D). The results show that the minimum values and slopes of Ca2+ concentration changes after the propagation of the wave were similar between Atp1a2+/− and wt mice.
To evaluate the increase in fluorescence intensity after the propagation of the wave, thresholds were set to 5, 7, and 10 times the SDs of fluorescence intensity before the first Ca2+ waves. Activity was defined as fluorescence intensity above these thresholds. For the 5×, 7×, and 10× SD thresholds, Atp1a2+/− mice showed significantly higher percentages of active astrocyte ROIs than wt mice (Fig. 5A‐C). The sum of the activity duration of Atp1a2+/− in astrocytes was longer than that of wt mice at 7× SD threshold (Fig. 5E). In neurons, Atp1a2+/− mice showed significantly higher percentages of active astrocyte ROIs than wt mice at only 7× SD threshold (Fig. 5B). Except for Fig. 5E, there was no difference in the sum of the duration of activity, the duration of activity per the total number of activities, and peak ∆F/F between Atp1a2+/− and wt mice at all thresholds (Fig. 5D–L). In each ROI, Atp1a2+/− mice showed significantly higher percentages of active ROIs than wt mice at all thresholds (Fig. S4A–C). From these observations, we conclude that more astrocytes in Atp1a2+/− mice had higher percentages of cells with increased fluorescence intensity compared with that of astrocytes in wt mice after the propagation of the wave.
Finally, to investigate the interaction between astrocytes and neurons, we evaluated synchrony in the changes in fluorescence intensity after Ca2+ wave peaks (Fig. 6A,B). In this analysis, Ca2+ wave peaks were excluded from the data, and the fluorescence intensity after the peaks in the frames that contained at least one active ROI was analyzed (Fig. 6A–E). We calculated the correlation coefficient for activity between active ROIs, and the correlation coefficients were separately averaged for neuron–neuron, neuron–astrocyte, and astrocyte–astrocyte pairs (Fig. 6F). The mean correlation coefficients were not significantly different between Atp1a2+/− and wt mice in neuron–neuron, neuron–astrocyte, and astrocyte–astrocyte pairs in the 5× SD threshold condition (Fig. 6G). We also evaluated the correlation in changes in fluorescence intensity within an astrocyte (Fig. S5A). Although astrocyte processes were not always active when the astrocyte soma was activated, the mean correlation within astrocytes was not different between Atp1a2+/− and wt mice (Fig. S5B,C). General anesthesia decreases spontaneous activities of neurons and astrocytes and astrocyte–astrocyte activity correlation compared with those of awake mice . We performed all experiments under isoflurane anesthesia. Therefore, it is not excluded that the activities and the correlation were underestimated.
In this study, CSD was induced by KCl application to the cortex, and then, propagating waves of Ca2+ were monitored by G‐CaMP7 fluorescence using GLT‐1‐G‐CaMP7 mice. The second Ca2+ waves in Atp1a2+/− mice had a faster propagation speed than those of wt mice (Fig. 4B), and the percentages of astrocytes that showed elevated Ca2+ concentrations after the propagation of Ca2+ waves were significantly higher in Atp1a2+/− mice than in wt mice (Fig. 5A–C). In CSD induction, the propagation of Ca2+ waves is shown to occur with CSD propagation, and the propagation speed of the Ca2+ wave is similar to that of CSD . Therefore, the difference in the propagation speed of Ca2+ waves likely reflects the difference in the propagation speed of CSD. CSD propagation has been proposed to be mediated by the interstitial diffusion of K+ or glutamate [40, 41, 42]. Interstitial K+ diffusion initiates positive feedback cycles that induce CSD in contiguous dendritic regions, and the clearance of K+ and glutamate by astrocytes limits the rate and spatial extent of CSD propagation [39, 42]. Reuptake of glutamate from the synaptic cleft into astrocytes is driven by glutamate transporters GLAST and GLT‐1, utilizing the electrochemical gradient of Na+ generated by Na,K‐ATPase [19, 43, 44]. In Atp1a2+/− mice, when CSD causes a transient increase of glutamate levels in the synaptic cleft, capacity of the glutamate transporters to reuptake the glutamate would become limiting due to the decreased electrochemical gradient of Na+. As the result, glutamate levels in the synaptic cleft in Atp1a2+/− mice are expected to remain higher than those in wt mice. Indeed, mice harboring loss‐of‐function Atp1a2 mutations show defective glutamate and K+ clearance by cortical astrocytes . Taken together, the faster Ca2+ wave propagation in Atp1a2+/− FHM2 model mice is caused by the increased glutamate levels in the interstitial space due to the decreased reuptake of glutamate. It is also known that glutamate promotes increased Ca2+ concentrations in astrocytes . Increased Ca2+ concentrations after the propagation of Ca2+ waves may represent astrocytic reactions to interleukin‐1β (IL‐1β). Studies demonstrate that IL‐1β is released from neurons during CSD  and induces transient Ca2+ elevations in cultured mouse cortical astrocytes .
Elevated Ca2+ concentrations in astrocytes can trigger the release of gliotransmitters (D‐serine and glutamate), prostaglandin E2, and K+ , which are capable of modulating the function of neighboring glial, neuronal, and vascular cells [48, 49, 50, 51, 52, 53, 54, 55]. In contrast to higher percentages of astrocytes with elevated Ca2+ concentrations, we observed little differences in changes in Ca2+ concentrations in neurons in Atp1a2+/− mice (Fig. 5). Furthermore, the correlation of Ca2+ concentration changes in neuron–neuron, neuron–astrocyte, and astrocyte–astrocyte pairs (Fig. 6G) was similar between Atp1a2+/− and wt mice. These results suggest that astrocytes in Atp1a2+/− mice with elevated Ca2+ concentrations may exert modulatory actions on vascular cells without altered astrocyte–neuron coupling. The release of gliotransmitters, prostaglandin E2, and K+ via elevated Ca2+ concentrations in astrocytes could induce vasodilation [53, 54, 55] together with the change in blood flow. Although regional cortical blood flow (CBF) changes during CSD do not differ significantly between Atp1a2+/− and wt mice , increased percentages of astrocytes with elevated Ca2+ concentrations may lead to delayed CBF changes, as suggested previously . In addition, cerebral artery vasodilation is frequently associated with prolonged aura in an FHM2 family . Increased percentages of astrocytes with elevated Ca2+ concentrations can affect aura symptoms through vasodilation. Because IL‐1β is known to induce elevated Ca2+ concentrations in astrocytes , the increased percentage of astrocytes with elevated Ca2+ concentrations in Atp1a2+/− mice may represent enhanced inflammatory responses by the IL‐1β‐mediated activation of the trigeminovascular system.
It is noted that Ca2+ concentration in astrocytes during CSD is higher than that of wt mice in FHM1 model mice, Cacna1aR192Q/R192Q . Although CACNA1A is predominantly localized in neurons , Cacna1aR192Q/R192Q shows the increased activity of astrocytes as observed in Atp1a2+/− mice. Therefore, it is likely that the activation of astrocytes causes migraine by a similar mechanism to Atp1a2+/− mice as described above. Furthermore, FHM3 model mice carrying R1407X mutation in Scn1a show the reduced inhibitory activity of inhibitory interneurons, which results in the increased excitatory neuronal activity  leading to the increased glutamate levels in the interstitial space. Therefore, it is plausible that the activation of astrocytes occurs by the increased glutamate levels, although this has not been examined. From these, we suggest that the increased activity of astrocytes is a common basis of pathophysiology of three types of FHM, FHM1, FHM2, and FHM3. Astrocytes with elevated Ca2+ concentrations, our novel findings in Atp1a2+/− mice, could cause the migraine aura and headache through vasodilation and inflammatory responses in FHM, although it remains to be elucidated. In future studies, elucidating the role of astrocytes in vasodilation and inflammatory responses during CSD could help reveal the pathophysiology of FHM.
We thank H. Hirase for providing the G7NG817 transgenic mice and Y. Motegi for advice on fiji. This work was supported by JSPS (Japan Society for the Promotion of Science) KAKENHI (Grant‐in‐Aid for Scientific Research C), Grant Number 18K07373 (HS), and the Subsidies for Private Universities.