Increased presynaptic excitability in a migraine with aura mutation

Abstract

Migraine is a common and disabling neurological disorder. The headache and sensory amplifications of migraine are attributed to hyperexcitable sensory circuits, but a detailed understanding remains elusive. A mutation in casein kinase 1 delta (CK1δ) was identified in non-hemiplegic familial migraine with aura and advanced sleep phase syndrome. Mice carrying the CK1δ_T44A mutation were more susceptible to spreading depolarization (the phenomenon that underlies migraine aura), but mechanisms underlying this migraine-relevant phenotype were not known. We used a combination of whole-cell electrophysiology and multiphoton imaging, in vivo and in brain slices, to compare CK1δ_T44A mice (adult males) to their wild-type littermates.

We found that despite comparable synaptic activity at rest, CK1δ_T44A neurons were more excitable upon repetitive stimulation than wild-type, with a reduction in presynaptic adaptation at excitatory but not inhibitory synapses. The mechanism of this adaptation deficit was a calcium-dependent enhancement of the size of the readily releasable pool of synaptic vesicles, and a resultant increase in glutamate release, in CK1δ_T44A compared to wild-type synapses. Consistent with this mechanism, CK1δ_T44A neurons showed an increase in the cumulative amplitude of excitatory post-synaptic currents, and a higher excitation-to-inhibition ratio during sustained activity compared to wild-type.

At a local circuit level, action potential bursts elicited in CK1δ_T44A neurons triggered an increase in recurrent excitation compared to wild-type, and at a network level, CK1δ_T44A mice showed a longer duration of ‘up state’ activity, which is dependent on recurrent excitation. Finally, we demonstrated that the spreading depolarization susceptibility of CK1δ_T44A mice could be returned to wild-type levels with the same intervention (reduced extracellular calcium) that normalized presynaptic adaptation.

Taken together, these findings show a stimulus-dependent presynaptic gain of function at glutamatergic synapses in a genetic model of migraine, that accounts for the increased spreading depolarization susceptibility and may also explain the sensory amplifications that are associated with the disease.

Suryavanshi et al. show that the excitability of glutamatergic synapses is increased in the CK1δ_T44A mouse model of migraine, due to an increase in the size of the readily releasable pool of synaptic vesicles. Reversing the increase in synaptic excitability rescues the migraine phenotype.

Introduction

Migraine affects ∼12% of the world population and causes enormous disability, especially to females and to those in working and childbearing years.^1,2 Yet the disease remains poorly understood at the cellular and circuit level. Although migraine is typically associated with headaches, multiple lines of evidence suggest it is better characterized as a disorder of multi-sensory gain.³ In addition to the headache, migraine attacks involve amplification of sensory percepts including light, sound, smell, touch and interoception (allodynia, photophobia, phonophobia, osmophobia and nausea, respectively).^4,5 Migraine sufferers also exhibit larger amplitude sensory-evoked responses and reduced habituation to repeated sensory stimulation during attacks and interictal periods.^6-8 In a third of migraineurs, the attack is preceded by aura, which is caused by a massive spreading depolarization (SD) of cortical tissue.^3,9

The mechanisms underlying the aberrant sensory network activity in migraine are poorly understood. Animal models of monogenic forms of migraine offer unique mechanistic insight,¹⁰ as mutations in single genes, and resultant dysregulation of downstream molecular pathways, produce migraine phenotypes. Established monogenic mouse models harbour mutations found in familial hemiplegic migraine (FHM types 1 and 2) and exhibit an increased probability of release and impaired clearance at glutamatergic synapses, respectively.^11-14 These convergent lines of evidence suggest that increased synaptic excitability is a potential unifying mechanism of the disease. However, these models represent rare and severe forms of migraine, in which the migraine attacks are accompanied by hemiplegia.¹¹ It is thus of interest to determine whether such mechanisms apply more broadly.

Loss-of-function mutations in casein kinase 1 delta (CK1δ) were recently identified in two families exhibiting familial migraine with aura and advanced sleep phase syndrome.¹⁵ Unlike FHM mutations, the CK1δ mutations segregate with phenotypically common migraine with aura. Mice carrying one of the mutations (CK1δ_T44A) exhibit increased sensitivity to tactile and thermal hyperalgesia elicited by the migraine trigger nitroglycerin (NTG), as well as increased susceptibility to SD,¹⁵ thus confirming that two key phenotypes of the disease are represented in the animal model. However, the CK1δ_T44A mutation affects a ubiquitous serine-threonine kinase with broad roles regulating normal cellular function.^16-19 Thus, although the molecular moiety responsible for the migraine phenotypes in CK1δ_T44A mice is known, the underlying cellular and synaptic mechanisms remain elusive. Here, using whole-cell electrophysiology and imaging in vivo and in brain slices, we uncovered presynaptic mechanisms contributing to SD susceptibility in CK1δ_T44A mice.

Materials and methods

Animal handling and experimental design

All experiments were conducted in accordance with and approved by the Institutional Animal Care and Use Committee at the University of Utah. Experiments were conducted on 2–3-month-old male CK1δ_T44A and wild-type (WT) littermate mice (line 827)^15,20 weighing between 20 and 30 g. Animals were housed in temperature-controlled rooms on a 12-h light-dark cycle. The sample size was determined using data from preliminary experiments (in vivo up-state recordings) as well as previous reports, generally n = 5–9 mice for in vivo and n = 15–20 slices (7–10 mice) for in vitro experiments. For patch-clamp electrophysiology, recordings were obtained from one neuron per mouse (in vivo) or 1–2 neurons per slice (in vitro). Experimenters were blind to the animal genotypes in all experiments. Most comparisons were made across genotypes within the same condition or across conditions within the same animal or slice (paired or repeated measures).

In vitro brain slice preparation

CK1_T44A and wild type (WT) littermate mice were deeply anaesthetized with 4% isoflurane and the brain was removed for slice preparation. Coronal sections were cut in ice-cold dissection buffer (in mM; 220 sucrose, 3 KCl, 10 MgSO₄, 1.25 NaH₂PO₄, 25 NaHCO₃, 25 D-glucose, 0.2 CaCl₂) and sections containing somatosensory cortex²¹ were allowed to recover in a chamber containing normal artificial CSF (in mM; 125 NaCl, 3 KCl, 1.3 MgSO₄, 1.25 NaH₂PO₄, 25 NaHCO₃, 25 D-glucose, 1.3 CaCl₂, and saturated with 95%O₂/5%CO₂) at 35°C. For electrophysiology experiments, the slices were transferred to a submerged chamber constantly supplied with artificial CSF (flow rate: 2.5 ml/min, saturated with 95%O₂/5%CO₂) also maintained at 35°C. In experiments manipulating extracellular calcium concentrations ([Ca²⁺]_e), artificial CSF containing 0.65 mM CaCl₂ and 1.95 mM MgCl₂ was used after slice incubation.

In vitro whole-cell electrophysiology

All whole-cell patch-clamp recordings were obtained from regular spiking pyramidal neurons²¹ in layer 2/3 (L2/3) somatosensory cortex visualized using differential interference contrast (DIC) microscopy. Whole-cell patch-clamp recordings were obtained using glass microelectrodes (4–6 MΩ resistance, tip size of 3–4 μm). To record intrinsic membrane properties, patch electrodes were filled with intracellular solution containing 130 K-gluconate, 5.5 EGTA, 10 HEPES, 2 NaCl, 2 KCl, 0.5 CaCl₂, 0.3 Na₂GTP, 2 MgATP, 7 phosphocreatine (concentrations in mM, pH 7.2, 289–292 mOsm/kg-H₂O). Baseline membrane voltage (resting membrane potential, V_m), as well as membrane voltage responses to 20 pA current injection steps (from −100 to +400 pA, 500 ms duration), were recorded. Spontaneous excitatory and inhibitory postsynaptic currents (sE/IPSCs) were recorded using patch pipettes containing 130 CsMeSO₄, 3 CsCl, 10 HEPES, 2 MgATP, 0.3 Na₂GTP, 5 EGTA, 10 phosphocreatine, 5 QX-314, 8 biocytin (concentrations in mM; pH 7.2, 291–294 mOsm/kg-H₂O). Excitatory and inhibitory currents were isolated by clamping neuronal membrane potential to −70 mV (near inhibitory reversal potential) and 10 mV (near excitotoxic reversal potential, 20mV in vivo), respectively. Excitatory / inhibitory postsynaptic currents (EPSCs / IPSCs respectively), were pharmacologically blocked by DNQX (50 µM) and picrotoxin (PTX, 20 µM), respectively, confirming the AMPA and GABA_A receptor-mediated nature of the respective currents.

Tonic inhibitory current recording

To isolate tonic inhibitory currents, GABA_A receptor-mediated phasic IPSCs were suppressed using 20 µM picrotoxin (PTX) for at least 5 min.^22,23 A histogram of 10 000 data-points (5 s at 2KHz) was generated at baseline and after PTX treatment. A Gaussian filter was fitted to the part of the distribution not skewed by phasic IPSCs to obtain mean holding currents.²² The PTX-sensitive tonic current was measured by subtracting the mean holding currents following PTX treatment from the baseline.

Stimulus train evoked E/IPSCs

For stimulus train experiments, bipolar stimulation electrodes were fabricated from pulled theta glass pipettes (0.5 to 1GΩ resistance, 4–7 µm tip diameter) and filled with saline. Chlorinated silver wires were inserted into each half of the theta glass and connected to the two poles of the stimulus isolator. Distinct barrels in the L4 somatosensory cortex were identified using IR-DIC microscopy.²⁴ Theta glass stimulation electrodes were placed at the bottom edge of the barrels. Postsynaptic L2/3 neurons from the same cortical column were patched for voltage-clamp recordings. Postsynaptic responses to 100 µs stimuli ranging from 1 to 10 mV intensity were recorded to establish input-output relationships. The range of stimulus intensity was restricted to evoke only column-specific monosynaptic EPSC (at −70 mV) and di-synaptic IPSC (at 10 mV) responses. Paired-pulse ratios (PPRs) were calculated as the ratio of amplitudes of the second and the first evoked response (EPSC2/1). Stimulus intensity sufficient to evoke ∼50% of the maximum responses was selected for the rest of the experiment. Trains of 10 or 30 stimuli were introduced at different frequencies (10/20/50 Hz) and resulting monosynaptic EPSC and di-synaptic IPSC responses were measured by clamping post-synaptic cell voltage at −70 and 10 mV, respectively.²⁵

Recording AMPA/NMDA ratio

For AMPA/NMDA ratio measurements, a concentric bipolar metal stimulating electrode (FHC #30214, tip diameter: 125/25 µm, length: 75 mm) connected to a stimulus isolator (World Precision Instruments) was placed on L4 and L5a (∼400 µm distance from recording site). Post-synaptic evoked AMPA and NMDA receptor-mediated currents were recorded by stimulating L4 and L5a (1 ms, 1 to 10 mV stimuli). Stimulation intensity providing ∼50% of maximum response with no failure was selected for further experiments. Evoked AMPA currents were isolated at −70 mV. Evoked NMDA-mediated currents were isolated at 40 mV in the presence of DNQX (50 µM) and picrotoxin (20 µM).

In vivo whole-cell electrophysiology

Mice (males; 2–3 months old) were anaesthetized using urethane (0.75 g/kg; i.p.) supplemented with isoflurane (∼0.5%). Body temperature was monitored and maintained at 35–37°C using a heating pad. Vital signs (heart rate: 470–540 bpm, SpO₂: 92–98%, respiration: 120–140/min) were monitored (MouseStat, Kent Scientific) throughout the experiment and maintained within a physiologically normal range. An Omega-shaped head bar was mounted on the skull, using glue and dental cement, and affixed with screws to a holding rod attached to the stage. A craniotomy 2 mm in diameter was made above the hind paw region of the somatosensory cortex (1 mm caudal to the bregma and 2 mm lateral to the midline), and filled with 1.5% agarose (in artificial CSF) to keep the cortical surface moist and dampen the movement associated with breathing.²¹ We used in vivo whole-cell techniques to record the membrane potential from layer 2/3 pyramidal neurons in the primary somatosensory cortex (S1 L2/3 neurons) and analysed spontaneous up states in current-clamp mode at resting membrane potential (i.e. −65 to −70 mV). Patch electrodes with 4–6 MΩ resistance and longer taper were used (tip size of 3–4 μm).²¹

Two-photon laser scanning microscopy

Acute slices containing the somatosensory cortex were prepared and incubated similarly to whole-cell electrophysiology experiments. After incubation, slices were placed in a submerged chamber and perfused with artificial CSF (2 to 2.5 ml/min), under a two-photon microscope [Neurolabware with Hamamatsu H11901 and H10770B photomultiplier tubes; Cambridge Technology CRS8 8KHz resonant scanning mirror and 6215H galvanometer scanning mirror; Nikon 16×/0.8 NA objective; Coherent Cameleon Discovery dual laser system (tunable laser = 920 nm; fixed laser = 1040 nm; pulse width ≍100 fs); Semrock emission filters (510/84 nm green and 607/70 nm red); power ≤ 25 mW; 250 µm—1.5 mm field of view (FOV) on its long axis with ∼1.15–1.73 µm/pixel resolution at 2 × digital zoom, Scanbox acquisition software]. Glutamate transients were evoked using a stimulation paradigm similar to the whole-cell electrophysiology experiments, in the presence of glutamate receptor antagonists (CNQX 20 µM, AP5 50 µM).²⁶ Maps of glutamate transients across layers were acquired by scanning the entire field of view (512 lines/frame) at 15.49 Hz. After confirming a measurable stimulus-evoked response [>mean + (3 × standard deviation) of baseline], L2/3 glutamate transients were temporally resolved using limited area scans (100 lines/frame at 79.80 Hz).

Viral injections

iGluSnFR expression was targeted to S1 L2/3 neurons using injections of adeno-associated virus (AAV1.hSyn.iGluSnFr.WPRE.SV40, Penn Vector Core catalogue # AV-1-PV2723) 2–3 weeks before imaging.²⁷ Animals were anaesthetized using isoflurane (∼5% induction, <1.5% for surgery) and fixed in a stereotaxic frame (David Kopf Instruments, Models 962, 923B). Bupivacaine (30μl, subcutaneous) was administered as a local anaesthetic and Rimadyl (5 mg/kg, subcutaneous) was used for postoperative analgesia, if necessary. Using standard sterile practices, bilateral injections were performed through small burr holes placed at coordinates relative to bregma (2 mm posterior, 3.2 mm lateral, 200 µm depth). AAV was injected (750–1000 nl, measured using a Hamilton Gastight syringe) through a pulled glass pipette (0.5 to 1 MΩ resistance) connected to a sterile, Luer-lock injection syringe via polyethylene tubing. Post-surgery, mice were typically housed with 1–2 littermates, though occasionally they were singly housed.

Spreading depolarization induction thresholds

Acute coronal slices containing the barrel cortex prepared from WT and CK1δ_T44A mice (2 to 3 months) were perfused with artificial CSF at a flowing rate of 3 ml/min. Pressure-ejection pulses of 3 M KCl (0.5 bar) of increasing duration (at 8 min intervals in 20 ms steps) were applied through a glass micropipette (R = 0.5–0.72 MΩ, placed over layer1 and/or 2/3) onto the slice surface, using a pressure ejection system (MPPI-2, Applied Scientific Instruments), until an SD was elicited.^13,21 SD was detected by monitoring the associated change in intrinsic optical signal (IOS). SD propagation velocity was calculated as the rate of change of IOS, at a distance of 500 µm from the pressure-ejection pipette tip.

Data quantification and analysis

Whole-cell electrophysiology

All recordings were acquired at 20 kHz and filtered at 2 kHz (low pass) using a Multiclamp 700B amplifier. Analogue data were digitized using Digidata 1330 digitizer and Clampex 9 software (Axon Instruments). Access resistance was monitored throughout recordings (5 mV pulses at 50 Hz). Recordings with access resistance >25 MΩ or with >20% change in the access resistance were discarded from the analysis. Series resistance compensation (70%) was applied to recorded currents in voltage-clamp setting. Offline data processing was done with Clampfit 10 (Axon Instruments).

Stimulus-evoked glutamate transients

Image stacks were converted to maximum intensity projections and regions of interest (ROIs) were determined using a binary map of all pixels with a greyscale value >2× standard deviation of maximum intensity projection (avoiding the slice anchor and stimulation electrode if these present within the FOV). The mean fluorescence intensities of the initial 50 frames of each trial served as F₀. Fluorescence intensity traces were collected as the average intensity of all pixels within the ROI and normalized as ΔF/F₀. Peak amplitude, area under the curve (AUC), and full-width at the half-maximum response (FWHM) of the glutamate transients were calculated using MATLAB (findpeaks function).

Analysis and statistics

Data analysis was performed using GraphPad Prism 8 (GraphPad Software, San Diego, California), MATLAB R2020a (Mathworks), Stata (StataCorp) and Microsoft Excel (Microsoft Corp). The normality of distributions was determined using the D'Agostino-Pearson K2 test and Shapiro-Wilk test. Outliers were identified based on data distribution (Grubbs’ test for parametric data and 1.5 × interquartile range for non-parametric data). However, no outliers were identified and excluded from the data presented in the manuscript. Experimenters were blind to the animal genotypes during the analysis. Average values for individual cells were compared across genotypes using a two-tailed, unpaired t-test (parametric data) or Mann-Whitney U-test (non-parametric data). Paired comparisons were done using a two-tailed, paired t-test. Normalized (%) distributions of individual events were compared across genotypes using a two-sample Kolmogorov–Smirnov (KS) test. Data representing multiple time points within the same slice were compared across groups using two-way ANOVA (Friedman’s test for non-parametric data) or within groups using one-way ANOVA (Kruskal Wallis test for non-parametric data). For multiple comparisons, Bonferroni’s correction or Tukey’s test were used post hoc for equal sample sizes. To compare the kinetics of input current-voltage (I/V) and action potential frequency-input current (F/I) relationships, linear regression was fitted across datasets and slopes were compared between genotypes. To estimate the size and replenishment rate of the readily releasable pool (RRP), we fitted linear regressions to steady-state cumulative amplitudes of E/IPSCs. Y-intercepts (RRP size) and slopes (RRP replenishment rates) of the linear regression fit were measured for grand means for genotypes or individual neurons.^28,29 For pharmacology experiments, comparisons between cells before and after drug treatment within a genotype were done by paired t-test; comparisons between control and drug treatment groups across genotypes were done by two-way ANOVA with Tukey’s post hoc test. Statistical significance was set at P < 0.05, *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

Results

CK1δ_T44A mice have hyperexcitable sensory cortical circuits despite hyperpolarized membrane potentials in vivo

Hyperexcitable cortical sensory circuits can render the cortex susceptible to SD ^12,13 as well as confer tactile and thermal hypersensitivity,^30,31 similar to that observed previously in CK1δ_T44A mice.¹⁵ Under anaesthesia or in quiet wakefulness, cortical networks can oscillate between depolarized ‘up states’ and quiescent ‘down states’.^32,33 Up/down states are regulated by the balance of excitatory and inhibitory synaptic activity. As such, these states dynamically modulate cortical circuit gain, as neurons are more likely to fire action potentials (APs) during up states and less likely during down states.^33,34 We recorded L2/3 pyramidal neuronal membrane potential (V_m) oscillations during up and down states in vivo, to characterize network-driven excitability in WT and CK1δ_T44A mice (Fig. 1A).

Figure 1.

Increased up-state duration and V_m variance in CK1δ_T44A mice, despite hyperpolarized neuronal resting V_m. (A) Schematic of in vivo whole-cell patch-clamp experiment along with representative current-clamp traces (L2/3 pyramidal neurons) of up and down states from wild-type (WT) and CK1δ_T44A mice. (B) CK1δ_T44A neurons were significantly hyperpolarized compared to WT. (C) Frequency histograms of up-state half-width show a positive shift in the up-state durations in CK1δ_T44A mice. (D) CK1δ_T44A mice show an increase in the mean up-state durations of individual animals and (E) an increase in membrane potential variance (log standard deviation) during up-states. Statistical analyses: unpaired t-test (B and D), two-sample Kolmogorov–Smirnov (KS) test (C and E). Pooled data are means ± standard deviation. *P < 0.05, **P < 0.01 and ****P < 0.0001. Exact P-values can be found in Supplementary Table 1.

To our surprise, neuronal resting V_m during down states was significantly hyperpolarized (two-tailed t-test, P < 0.01), in CK1δ_T44A mice—which is commonly associated with reduced excitability (Fig. 1B). However, the duration of individual up states in CK1δ_T44A mice was significantly increased (two-sample Kolmogorov–Smirnov test, P < 0.0001, Fig. 1C), as was the mean duration of all up states per animal (two-tailed t-test, P < 0.05, Fig. 1D). V_m variance, which represents the magnitude of persistent synaptic barrages during up states,³³ was also significantly higher in CK1δ_T44A mice (two-sample Kolmogorov–Smirnov test, P < 0.0001, Fig. 1E), confirming hyperexcitable cortical circuits in CK1δ_T44A mice, despite the hyperpolarized resting V_m.

Hyperpolarized membrane potential in CK1δ_T44A neurons is due to increased tonic inhibition

For further dissection of the circuit phenotypes in CK1δ_T44A mice, we used in vitro whole-cell electrophysiology in acute cortical slices (Fig. 2A). Consistent with our in vivo finding, CK1δ_T44A neurons were hyperpolarized compared to WT littermates (two-tailed t-test, P < 0.01, Fig. 2B), confirming the viability of brain slice preparation as a platform for more detailed examination.

Figure 2.

Increased tonic inhibition in CK1δ_T44A neurons primarily contributes to hyperpolarized resting V_m. (A) Schematic showing in vitro whole-cell patch-clamp electrophysiology in acute coronal brain slices and a representative image of a typical biocytin-labelled excitatory L2/3 neuron. (B) Significantly hyperpolarized resting membrane potentials in CK1δ_T44A neurons compared to WT (WT). (C) Representative trace from a WT neuron showing a reduction in tonic inhibitory current, after application of 20 µM picrotoxin (PTX) (GABA_A antagonist). Right: % distribution histograms along with Gaussian fits (10⁴ data-points) showing PTX-sensitive holding current (tonic inhibitory current) as the difference between the % distribution means. (D) Tonic inhibitory currents were significantly larger in CK1δ_T44A neurons compared to WT. (E) Representative traces from a WT neuron showing resting membrane potential as well as voltage response to depolarizing current pulse before and after PTX application. (F) The difference in the resting V_m between individual WT neurons as well as CK1δ_T44A neurons before and after PTX treatment. (G) Pharmacological blockade of tonic inhibitory current lead to the ‘rescue’ of hyperpolarized membrane potentials in CK1δ_T44Aneurons. Statistical analyses: unpaired t-test (B), Mann-Whitney U-test (D), paired t-test (F) and two-way ANOVA (G). Pooled data are means ± standard error of the mean (SEM). *P < 0.05, **P < 0.01 and ***P < 0.001. Exact P-values can be found in Supplementary Table 1.

Tonic inhibitory currents can contribute to neuronal V_m hyperpolarization.²³ We recorded tonic holding currents in voltage clamp (holding potential 10 mV) before and after blocking GABA_A receptors with 20 µM PTX (Fig. 2C). We found that CK1δ_T44A neurons had significantly larger PTX-sensitive tonic inhibitory currents compared to WT (Mann Whitney test, P < 0.05, Fig. 2D). To determine if increased tonic inhibitory currents in CK1δ_T44A neurons substantially contribute to hyperpolarization, we recorded neuronal resting V_m before and after PTX treatment. As expected, PTX treatment depolarized neuronal resting V_m and reduced rheobase in both WT and CK1δ_T44A neurons (Fig. 2E and Supplementary Fig. 1). However, PTX-induced depolarization was significantly larger in CK1δ_T44A neurons compared to WT (Fig. 2F). PTX also rescued neuronal V_m in CK1δ_T44A neurons (two-way ANOVA, Fig. 2G), suggesting that increased tonic inhibition was the primary cause of resting V_m hyperpolarization.

Increased frequency of evoked action potentials due to higher synaptic feedback excitation in CK1δ_T44A neurons

Thus far, our results show that CK1δ_T44A neurons were hyperpolarized due to increased tonic inhibition, yet there was an increase in cortical circuit excitability in vivo. Therefore, we examined whether CK1δ_T44A neurons were intrinsically hyperexcitable, by measuring membrane voltage (V_m) responses to a range of input currents (Fig. 3A). There was no difference in the neuronal membrane resistance (R_m) measured below the AP threshold (Supplementary Fig. 2) or the rheobase between the genotypes (Fig. 3D). However, in response to larger input currents, CK1δ_T44A neurons fired APs at a significantly higher frequency compared to WT neurons (two-way ANOVA, Fig. 3B), thus increasing the frequency-input current (FI) slopes in CK1δ_T44A neurons (Mann Whitney test, P < 0.05, Fig. 3C). Consistent with the F/I slope data, the inter-spike interval between APs was significantly shorter in CK1δ_T44A neurons at larger input currents (two-way ANOVA, Fig. 3H–J). Interestingly, there was no difference in AP half-width and after-hyperpolarization amplitude between the genotypes (Fig. 3E–G), suggesting that altered AP kinetics did not contribute to the increased firing in CK1δ_T44A neurons. To determine a possible effect of increased tonic inhibition on suprathreshold activity in CK1δ_T44A neurons, we recorded evoked APs before and after PTX treatment (Fig. 2E) and found no significant difference in F/I slopes after PTX treatment in either genotype (Supplementary Fig. 1D).

Figure 3.

Increased evoked action potential frequency in CK1δ_T44A neurons is due to increased synaptic feedback excitation. (A) Representative traces of L2/3 excitatory neurons firing action potentials (APs) to suprathreshold current injections. (B) AP frequency in CK1δ_T44A neurons was significantly increased, especially at higher input currents along with the slopes of linear regression lines fitted to the frequency-input current (F/I) plots. (C) Comparison of F/I slopes quantified for individual neurons revealed that CK1δ_T44A had significantly higher F/I slopes compared to WT (WT). (D) No significant difference in rheobase between WT and CK1δ_T44A neurons. (E) Representative AP traces for WT and CK1δ_T44A neurons, as well as phase plots showing the rate of change of V_m during APs as a function of V_m (dV_m/dt versus V_m). (F) No difference was observed in AP half-width as well as (G) the after-hyperpolarization (AHP) amplitude. (H) Representative traces of WT and CK1δ_T44A neurons firing APs in response to 400 pA current injection. (I) Comparison showed inter-spike intervals between WT and CK1δ_T44A neurons were significantly reduced following 400 pA current injection, suggesting reduced spike frequency adaptation. (J) However, the inter-spike interval was not different in CK1δ_T44A neurons at 300 pA. (K) Schematic showing pharmacological inhibition of postsynaptic glutamate receptors to block synaptic excitation during intracellular current injections. (L) AP frequency/input current (F/I) plots demonstrating that inhibition of postsynaptic glutamate receptors normalized the input-current-dependent increase in AP frequency seen in CK1δ_T44A neurons to WT levels. (M) No difference in the slopes of the linear regression fitted to the F/I plots for individual neurons, between genotypes. Statistical analyses: two-way ANOVA with Dunnett’s test for multiple comparisons (B, I and J), unpaired t-test (C, F, G and M), ANCOVA (B) and Mann Whitney U-test (D). Pooled data are means ± SEM. *P < 0.05 and **P < 0.01. Exact P-values can be found in Supplementary Table 1.

Neuronal AP bursts elicited by >500 ms long input currents, typically used in slice electrophysiology paradigms,^35,36 can generate synaptic feedback excitatory currents within neocortical microcircuits.³⁷ To test whether recurrent feedback excitation contributed to increased firing frequency in CK1δ_T44A neurons, we recorded evoked APs in the presence of ionotropic glutamate receptor antagonists (AMPARs and NMDARs) to abolish postsynaptic excitatory currents (Fig. 3K). Glutamate receptor antagonists normalized the elevated AP frequency (two-way ANOVA, Fig. 3L) and larger F/I curve slopes in CK1δ_T44A neurons (two-tailed t-test, Fig. 3M). Therefore, increased feedback synaptic excitation facilitated a higher frequency of evoked APs at larger input currents, suggesting that elevated excitatory synaptic activity may underlie the increased cortical excitability in CK1δ_T44A neurons.

Baseline phasic excitatory and inhibitory neurotransmission is unaffected in CK1δ_T44A neurons

Balanced synaptic excitation and inhibition mediate ‘normal’ cortical circuit excitability.^33,38 We next examined whether baseline excitatory and inhibitory synaptic transmission contributed to hyperexcitability in CK1δ_T44A circuits. We recorded AP-independent miniature excitatory and inhibitory postsynaptic currents (mE/IPSCs) in the presence of 1 µM tetrodotoxin (Supplementary Fig. 3A and B). We found that neither mEPSC amplitude nor frequency was significantly different between WT and CK1δ_T44A neurons (two-tailed t-test, Supplementary Fig. 3C and E). Similarly, mIPSC amplitude and frequency were not different between the genotypes (two-tailed t-test, Supplementary Fig. 3D and F). We found no difference in AP-dependent spontaneous excitatory and inhibitory neurotransmission (Supplementary Fig. 4). Finally, we recorded evoked E/IPSCs in L2/3 neurons in response to a single stimulus in layer 4 (L4). Again, we found no difference in the E/IPSC amplitude ratio between the genotypes (two-tailed t-test, Supplementary Fig. 4H).

NMDA receptor-mediated currents are a key component of excitatory neurotransmission and are modulated through phosphorylation by CK1 family proteins.³⁹ To test if NMDA receptor-mediated currents were altered in CK1δ_T44A mice, we calculated the ratio of NMDA and AMPA components of responses (AMPA/NMDA ratio) evoked by a single stimulus (L4 to L2/3). We observed no significant difference in the AMPA/NMDA ratio in CK1δ_T44A compared to WT neurons (two-tailed t-test, Supplementary Fig. 4I). Taken together, our results suggest that neither phasic excitatory nor inhibitory synaptic transmission was altered in CK1δ_T44A neurons.

Frequency-dependent adaptation deficit at CK1δ_T44A excitatory synapses

Thus far, we found that CK1δ_T44A mice exhibit an increase in local circuit excitability, yet there are no changes in phasic excitatory or inhibitory neurotransmission. We reasoned that this might be due to an altered postsynaptic response of CK1δ_T44A neurons to sustained synaptic stimulation.^33,40 To test this, we recorded evoked E/IPSCs following short stimulus trains (L4-L2/3, 10 or 30 stimuli) at different frequencies (10, 20, 50 Hz) (Fig. 4A–B). We found no difference in the input-output relationship of stimulus-evoked EPSCs between the genotypes (Supplementary Fig. 5). To avoid polysynaptic responses, we verified the response latency and column-specificity of stimuli by confirming a lack of evoked responses outside the cortical column (Supplementary Fig. 6).

Figure 4.

Stimulus-frequency-dependent adaptation deficits due to increased RRP size at CK1δ_T44A excitatory synapses. (A) Schematic representation of L2/3 cortical microcircuit showing the placement of stimulation as well as recording electrodes. (B) Representative traces showing evoked EPSC and IPSC responses to a train of 30 stimuli, at 20 Hz. Traces of EPSCs evoked to the first two stimuli (B’), used for the quantification of paired-pulse ratio (PPR). (C) No significant difference between genotypes in the normalized EPSC amplitude as well as PPR at 10 Hz and (D) 20 Hz stimulus trains. (E) Normalized evoked EPSC response to 50 Hz stimulus trains was significantly higher at CK1δ_T44A synapse. Inset: Significant difference in the PPR at 50 Hz between genotypes. (F) Schematic of a model synapse showing dynamic regulation of readily releasable pool (RRP) with simultaneous release and replenishment of synaptic vesicles upon repeated stimulation. (G) Schematic showing the analytical approach employed to explore the components of the presynaptic release machinery (RRP size and replenishment rate). (H–J) Cumulative amplitude plot for mean EPSC amplitudes: there was no significant difference in the slope of the linear regression fitted to the steady state between WT (WT) and CK1δT44A at any stimulus frequency. However, y-intercepts of the linear regression were significantly different between genotypes at both 20 Hz (I) and 50 Hz (J). (K–M) Slopes and y-intercepts for individual neurons: no significant difference in slopes, but significant difference in y-intercept at 50 Hz (M). Statistical analyses: repeated measures two-way ANOVA with Dunnett’s test for multiple comparisons and unpaired t-test (C, D and E), ANCOVA (H, I and J), unpaired t-test or Mann Whitney U-test (K, L and M). Pooled data are means ± standard error of the mean (SEM). *P < 0.05 and ****P < 0.0001. Exact P-values are in Supplementary Table 1. E/IPSCs = excitatory/inhibitory postsynaptic currents; n.s. = not significant.

Wild-type synapses showed the expected adaptation of postsynaptic responses to trains of 30 stimuli, as well as a paired-pulse ratio (PPR) consistent with that described in the literature for adult L4-L2/3 primary sensory cortex synapses.⁴¹ CK1δ_T44A neurons showed a similar response to WT at 10 Hz stimulation. However, as stimulus frequency increased, the evoked EPSC adaptation showed a weak attenuation at 20 Hz, and a robust and significant reduction at 50 Hz (two-way ANOVA, Fig. 4C–E), with a significant increase in PPR (two-tailed t-test, Fig. 4E, inset). We observed a similar frequency-dependent adaptation deficit in CK1δ_T44A EPSCs following shorter duration (10 stimuli) trains (Supplementary Fig. 7). Interestingly, the impaired adaptation in CK1δ_T44A neurons was specific to excitatory currents, with evoked inhibitory currents exhibiting adaptation and PPR similar to WT at all stimulation frequencies (Supplementary Fig. 8). Taken together, these results suggest a reduced adaptation of excitatory currents in CK1δ_T44A neurons following repeated high-frequency stimuli.

Increased size of readily releasable pool impairs CK1δ_T44A excitatory synaptic adaptation

The rapid adaptation of excitatory postsynaptic currents is due to a presynaptic phenomenon involving Ca²⁺-mediated depletion and replenishment of the readily releasable pool (RRP) of synaptic vesicles.⁴² The RRP size and its rate of replenishment influence the adaption of postsynaptic responses.^28,43,44 We used the approach initially described by Schneggenburger et al.^45-47 in the Calyx of Held, which was later adapted to cortical synapses by Abrahamsson et al.,⁴⁸ to explore the components of the presynaptic release machinery that contribute to the reduced adaptation at CK1δ_T44A synapses. This model uses cumulative postsynaptic current amplitude as a function of stimulus number, with the slope of linear extrapolation of the steady-state indicating the RRP replenishment rate and the y-intercept corresponding to the RRP size (Fig. 4F and G).⁴⁸

For EPSCs, there was no difference in the linear regression slope (replenishment rate) between genotypes at any of the frequencies tested. At 10 Hz there was no difference in y-intercept (RRP size); however, at both 20 and 50 Hz, there was a significant increase in the y-intercepts (more robust at 50 Hz), consistent with a frequency-dependent increase in RRP size in CK1δ_T44A compared to WT animals (ANCOVA, P < 0.0001, unpaired t-test, P < 0.05, Fig. 4H–M).

In contrast, neither replenishment rate nor RRP size was different for IPSC amplitudes between WT and CK1δ_T44A at any stimulus frequency (Supplementary Fig. 8D–I), confirming no difference in presynaptic adaptation at inhibitory synapses. These data predict that a larger presynaptic RRP during high-frequency stimulation at CK1δ_T44A glutamatergic synapses is responsible for the observed adaptation deficit.

The adaptation deficit at CK1δ_T44A excitatory synapses is [Ca²⁺]_e-dependent

Recent evidence suggests that the RRP is a subset of fusion-competent and primed vesicles that may or may not be docked.²⁸ Vesicle docking, priming and release are Ca²⁺-dependent mechanisms, often modulated by extracellular [Ca²⁺],^45,48 that ultimately determine the RRP size.^49,50 Following high-frequency stimulation, residual Ca²⁺ and subsequent activation of biochemical cascades enhance vesicle priming and fusion, increasing the RRP size.⁴³ We next tested whether increased RRP size at CK1δ_T44A glutamatergic synapses following high-frequency stimulation was dependent on [Ca²⁺]_e. We evoked EPSC responses to stimulus trains (10, 20, and 50 Hz) in the presence of low Ca²⁺ artificial CSF (0.65 mM versus 1.3 mM in normal artificial CSF; Fig. 5A and B). We found that the adaptation deficit at CK1δ_T44A excitatory synapses following 50 Hz stimulation was ‘normalized’ with low Ca²⁺ artificial CSF (two-way ANOVA, Fig. 5E), with RRP size (unpaired t-test, Fig. 5I) and replenishment rate (unpaired t-test, Fig. 5J) similar to WT levels. Therefore, the adaptation deficit at CK1δ_T44A excitatory synapses is dependent on [Ca²⁺]_e, consistent with a presynaptic RRP-dependent mechanism.

Figure 5.

Impaired adaptation of evoked EPSCs at high stimulation frequencies is dependent on [Ca²⁺]_e. (A) Schematic showing cortical microcircuits and experimental design to record evoked EPSCs in the presence of low [Ca²⁺]_e. (B) Representative trace showing evoked EPSCs (at 10, 20 and 50 Hz) recorded in low Ca²⁺ (0.65 mM) artificial CSF, in response to a train of 30 stimuli. (C–E) In low [Ca²⁺]_e conditions, we found no difference in normalized EPSC amplitude and paired-pulse ratio (PPR) between WT (WT) and CK1δ_T44A neurons at any frequency. (F–H) Similarly, we found no difference in the y-intercepts of linear regressions fitted to the steady-state cumulative EPSCs, in low [Ca²⁺]_e conditions between WT and CK1δ_T44A neurons. (I and J) In low [Ca²⁺]_e conditions, linear regression y-intercepts and the slopes of the cumulative EPSCs quantified for individual neurons were not different between genotypes for any stimulation frequency. Statistical analyses: repeated measures two-way ANOVA (C, D and E), ANCOVA (F, G and H), and unpaired t-test (C, D, E, I and J). Pooled data are means ± SEM. Exact P-values in Supplementary Table 1. E/IPSCs = excitatory/inhibitory postsynaptic currents; n.s.  = not significant.

Larger glutamate transients following high-frequency stimulation in CK1δ_T44A slices

Our data thus far indicate that a larger RRP size at glutamatergic synapses during repeated high-frequency stimulation results in reduced adaptation. A larger RRP size would likely entail an increase in glutamate release, which can be measured with fluorescent indicators.²⁷ Employing an approach similar to that described by Armbruster et al.,²⁶ we imaged glutamate transients in L2/3 upon stimulation of L4 afferents, using two-photon laser-scanning microscopy (Fig. 6A and B).

Figure 6.

Larger glutamate transients in CK1δ_T44A slices evoked by high-frequency stimulation. (A) Schematic showing experimental design for two-photon fluorescent glutamate imaging in acute slices with a representative image of a limited area scan (100 lines/frame at 79.80 Hz) showing viral expression of iGluSnFR in neurons (left). Image panels showing stimulus-evoked glutamate transient response over time (50 Hz train of 30 stimuli). Scale = 200 µm (Supplementary Video 2). (B) Representative traces of evoked glutamate transient responses at 10, 20 and 50 Hz, in WT (WT) and CK1δ_T44A slices. (C) Traces of glutamate transient responses for individual slices (grey) and grand means (WT: cyan, CK1δ_T44A: purple) evoked after 10 Hz stimulation. (D) At 10 Hz, the amplitude and area under the curve (AUC) of evoked glutamate transients were not significantly higher in CK1δ_T44A slices. (E) Traces of glutamate transient responses for individual slices and grand means evoked after 20 Hz stimulation. (F) No difference in the amplitude or AUC of evoked glutamate transients evoked to 20 Hz stimuli. (G) Traces of glutamate transient responses for individual slices and grand means evoked after 50 Hz stimulation. (H) Both amplitude and AUC of glutamate transients evoked by 50 Hz stimuli were significantly increased in CK1δ_T44A slices. Statistical analyses: Mann-Whitney U-test (F and H) and unpaired t-test (D). Pooled data are means ± SEM. **P < 0.01. Exact P-values are in Supplementary Table 1. n.s.  = not significant.

Glutamate signals were evoked using a stimulation paradigm similar to that used for electrophysiology experiments. Glutamate maps acquired by scanning the entire field of view (at 15.49 Hz) showed a column-specific distribution of evoked responses, similar to evoked EPSC recordings (Supplementary Fig. 6 and Supplementary Video 1). L2/3 glutamate transients in the appropriate column were then temporally resolved by faster limited area scans (79.80 Hz at 100 lines/frame, Supplementary Video 2, ‘Materials and methods’ section). Glutamate transients evoked in L2/3 in response to 10 and 20 Hz stimulus trains were not significantly different between genotypes (Mann Whitney U-test, Fig. 6C–F). However, there was a significant increase in peak amplitude and AUC of glutamate transients in CK1δ_T44A slices compared to WT during 50 Hz stimulations (Mann Whitney U-test, Fig. 6G and H), suggesting increased glutamate release^51,52 following high-frequency stimulation of CK1δ_T44A cortical synapses. This finding offers convergent evidence of a presynaptic gain of function as a key underlying mechanism and a potential bridge between the cellular and circuit CK1δ_T44A phenotypes.

An excitatory shift in the cortical circuits of CK1δ_T44A mice, in vitro and in vivo

Reduced adaptation at excitatory, but not inhibitory synapses, can result in a net excitatory shift in cortical microcircuits following stimulation, especially given the finding that reduced synaptic adaptation is associated with increased glutamate release. To test this hypothesis, we generated E/I ratios by measuring the amplitude of evoked E/IPSCs during stimulus trains (10 stimuli) recorded from the same neurons (Fig. 7A). As expected, we observed a significant excitatory shift in CK1δ_T44A neurons after 50 Hz stimulation (two-way ANOVA, P < 0.05), but not at 20 or 10 Hz (Fig. 7B–D). We then tested whether the net excitatory shift upon intense synaptic stimulation is preserved in intact neural networks in vivo (Fig. 7E). During up states, cortical neurons exhibit both excitatory and inhibitory synaptic barrages when recorded using voltage-clamp electrophysiology (refer to the ‘Materials and methods’ section). We observed a robust increase in the duration of up-state evoked excitatory currents (Fig. 7G), without a significant difference in the duration of inhibitory currents (Fig. 7F), consistent with reduced adaptation observed exclusively at glutamatergic synapses in brain slices. Therefore, intense activity results in an excitatory shift in the output of cortical synapses in vitro and in vivo, in CK1δ_T44A mice.

Figure 7.

An excitatory shift in the cortical circuits of CK1δ_T44A mice. (A) Schematic showing cortical microcircuit and experimental design along with representative traces of evoked E/IPSCs from a WT (WT) neuron in response to a train of 10 stimuli, recorded from the same neuron. (B–D) At 50 Hz (D), E/IPSC amplitude ratios (E/I ratio) recorded from individual neurons show a significant shift towards excitation in CK1δ_T44A neurons. No difference in the E/I ratio between genotypes at 20 and 10 Hz (B and C). (E) Representative traces showing inhibitory (holding potential  10 mV for in vitro; 20 mV for in vivo recordings) and excitatory (holding potential −70 mV) synaptic barrages recorded in voltage clamp mode during an upstate in vivo. (F) Comparison of IPSC half-width (animal means, left) shows no difference between WT and CK1δ_T44A mice. Frequency histograms (right) show only a modest shift in half-width of the inhibitory events in CK1δ_T44A mice. (G) Comparison of EPSC half-width (animal means, left) shows a significant increase in CK1δ_T44A mice along with a frequency histogram (right) showing a positive shift in the half-width distribution of the excitatory events in CK1δ_T44A mice. (H) Schematic showing experimental design as well as an example image sequence showing the progression of a focally induced SD wave across space and time under control conditions. Scale bar = 200 µm. (Supplementary Video 3). Representative traces showing an increase in transmittance (ΔT/T₀), obtained from a region of interest 500 µm from the induction site, as well as intracellular voltage rise recorded from different WT slices (arrows indicating the time of SD ignition). (I) CK1δ_T44A slices exhibit significantly reduced SD thresholds that were partially rescued in low [Ca²⁺]_e conditions. Statistical analyses: repeated measures two-way ANOVA (B, C and D), two-way ANOVA with Tukey’s test for multiple comparisons (I), Mann Whitney U-test and two-sample Kolmogorov–Smirnov (KS) test (F and G). Pooled data are means ± SEM. *P < 0.05. Exact P-values are in Supplementary Table 1. E/IPSCs = excitatory/inhibitory postsynaptic currents; n.s.  = not significant.

Reduced [Ca²⁺]_e normalizes spreading depolarization susceptibility in CK1δ_T44A slices

CK1δ_T44A mice exhibit increased susceptibility to experimental SD induction in vivo.¹⁵ Similar observations of SD susceptibility in other migraine models are attributed to the increased release probability,¹³ or impaired clearance,¹⁴ of glutamate. To test whether the glutamatergic gain-of-function due to impaired adaptation at CK1δ_T44A excitatory synapses results in the facilitation of SD , we measured the threshold for SD induction (Materials and methods) in acute cortical slices (Fig. 7H and Supplementary Video 3). As expected, the threshold for SD initiation was significantly reduced in CK1δ_T44A slices compared to WT littermates (two-way ANOVA, Fig. 7I).

To determine whether impaired adaptation at glutamatergic synapses had a causal role in the threshold difference, we measured SD thresholds in WT and CK1δ_T44A slices perfused with low Ca²⁺ artificial CSF, with the rationale that the reduced adaptation in CK1δ_T44A slices was [Ca²⁺]_e-dependent (Fig. 5). As SD induction itself depends on [Ca²⁺]_e,⁵³ low [Ca²⁺]_e significantly increased the SD threshold in slices from both genotypes, along with reduced maximum SD propagation (Supplementary Fig. 9). However, a comparison between genotypes in low [Ca²⁺]_e conditions revealed a ‘rescue’ of reduced SD thresholds in CK1δ_T44A slices (two-way ANOVA, Fig. 7I) suggesting a [Ca²⁺]_e-mediated mechanism underlies the reduced threshold in CK1δ_T44A compared to WT. Taken together, our results demonstrate that [Ca²⁺] _e-mediated presynaptic adaptation deficit at excitatory synapses likely contributes to cortical network hyperexcitability and SD susceptibility in CK1δ_T44A mice (Fig. 8).

Figure 8.

Model of cortical network excitability in CK1δ_T44A mice. (A) CK1δ_T44A excitatory (not inhibitory) synapses demonstrated reduced presynaptic adaptation to repeated high-frequency stimuli, mediated by [Ca²⁺]_e-dependent enhancement of readily releasable pool (RRP) size, despite normal synaptic transmission at rest. (B) This led to an increase in the glutamate release and a resultant increase in cumulative amplitude of evoked EPSCs along with a higher excitation-to-inhibition ratio during sustained activity, both in vivo and in brain slices. (C) At the local microcircuit level, trains of action potentials (APs) elicited from single neurons increased glutamatergic feedback excitation in CK1δ_T44A compared to WT (WT) brain slices, further increasing firing frequencies at more intense input currents. (D) At a network level in vivo, CK1δ_T44A mice show an increase in the duration of up-state activity, which is dependent on local circuit synaptic feedback excitation. (E) Finally, SD susceptibility in CK1δ_T44A brain slices was returned to WT levels with reduced [Ca²⁺]_e, consistent with presynaptic contributions to the phenotype. Together, these findings show that a presynaptic stimulus-dependent glutamatergic gain-of-function mediates a cortical network-level hyperexcitability phenotype in CK1δ_T44A mice. E/IPSCs = excitatory/inhibitory postsynaptic currents.

Discussion

Despite the pervasive and debilitating nature of migraine, the underlying mechanisms remain poorly understood.^1,2 In CK1δ_T44A mice expressing a mutation found in humans with migraine with aura, we found a calcium-dependent reduction in presynaptic adaptation (Fig. 4) and increased release (Fig. 6) exclusively at glutamatergic synapses following intense circuit stimulation. This occurred despite increased tonic inhibition (Fig. 3) as well as normal subthreshold and synaptic activity (Supplementary Figs 1–4). The stimulus-dependent increase in excitability bears an intriguing resemblance to migraine symptom descriptions and electrophysiological phenotypes in humans,^7,54,55 and it provides a synaptic and circuit mechanism for the increased cortical excitability noted in CK1δ_T44A mice (Fig. 1).¹⁵ Finally, we show a [Ca²⁺]_e-dependent increase in SD susceptibility CK1δ_T44A slices (Fig. 7), directly implicating [Ca²⁺]_e-dependent presynaptic mechanisms in migraine-relevant circuit excitability (Fig. 8).^13,53

Although migraine is more commonly polygenic,³ animal models of monogenic forms of the disorder offer a unique opportunity for mechanistic dissection.^12-14,56,57 Recently, a loss-of-function mutation in the CK1δ_T44A gene was identified in two families with a combination of familial migraine with aura and advanced sleep phase syndrome.^15,20 Mice harbouring this mutation showed migraine-relevant network excitability phenotypes including increased SD susceptibility and tactile and thermal hyperalgesia. These phenotypes were associated with increased c-fos expression in the trigeminal nucleus caudalis after administration of the migraine-triggering drug nitroglycerin.¹⁵ SD-associated vascular responses were also heightened in CK1δ_T44A mice. As SD likely triggers craniofacial pain at least partly through activation of vascular afferents,^9,58 this phenotype might be evidence of a larger perturbation of pain-sensitive structures. The combined cortical excitability, vascular and pain phenotypes in CK1δ_T44A mice¹⁵ likely have broad disease relevance, despite the rarity of the mutation.

Sensory amplifications, during and between attacks, are a key feature of migraine.^59-61 Sensory amplifications correlate with headache frequency and disease severity^62-64 and are associated with poor clinical outcomes.⁶⁵ Interestingly, migraineurs also report a lack of habituation to repeated sensory stimuli.^6-8,66 Presynaptic adaptation is one of the principal mechanisms regulating sensory habituation and gain modulation in sensory circuits.^67,68 In the cortex, a precise balance between synaptic excitation and inhibition is important for establishing the timing and signal-to-noise ratio necessary for signal processing.⁶⁹ Stimulus-evoked and spontaneous activity in the sensory cortex involves a stereotypical pattern of excitation followed by inhibition.^70-72 Even under normal conditions, synaptic excitation adapts more slowly than inhibition following repeated stimulation in the sensory cortex,^73,74 generating a temporal window for the sensory signal integration.^24,75 Reduced adaptation at excitatory—but not inhibitory—synapses upon high-frequency stimulation in CK1δ_T44A cortical neurons can thus significantly enhance signal transmission, resulting in sensory amplification.

Our findings are quite consistent with a scenario of enhanced cortical excitation due to the selective deficit in the adaptation of CK1δ_T44A excitatory neurons. The adaptation deficit is caused and accompanied by increased glutamate release. Increased glutamate release is a likely reason for the glutamate-dependent increase in feedback excitation we observed upon high-frequency stimulation (Fig. 3). It also likely accounts for the increased up-state duration and membrane potential variance (Fig. 1), which represents local synaptic barrages during depolarized up-states.

Interestingly, impaired adaptation is also present in other genetic models of migraine (FHM1)^76,77 and it also occurs after SD,⁷⁸ potentially providing a common presynaptic mechanism for the habituation failure to repeated stimuli reported in migraineurs.⁷ However, it is also important to note differences between these gain-of-function phenotypes in different models, as these may also have clinical relevance. The presynaptic gain-of-function phenotype in FHM1 differs from that of the CK1δ_T44A model, as a single stimulus is sufficient to elicit larger amplitude postsynaptic excitatory responses in FHM1.¹³ In the case of the CK1δ_T44A mouse model, repetitive high-frequency stimuli are necessary to elicit a glutamatergic adaptation deficit. Similarly, astrocytic transporter currents evoked by single stimuli, that represent the rate of glutamate clearance, are slower in FHM2 mice.¹⁴ These phenotyic differences appear to correlate with the differences in the severity of disease between FHM syndromes (associated with hemiplegic or paralytic auras) and CK1δ_T44A mutation carriers, who report only the non-hemiplegic auras that are also more common in the general migraine population.^3,11 The response to repetitive rather than single stimulation is also consistent with patient reports of triggering migraines with high-intensity sensory inputs.^7,8

Sensory perception is a multi-synaptic, multi-circuit process and the mechanisms underlying altered perception extend well beyond the single-synapse level. Therefore, understanding migraine as a disorder of sensory gain requires an examination of circuit-wide mechanisms in vivo. Cortical slow oscillations are a well characterized form of dynamic gain modulation in sensory circuits:⁷⁹ depolarized ‘up states’ increase the likelihood of AP generation^34,80 and have been proposed as coincidence detectors for feedforward signal transmission⁸¹ during quiet wakefulness. Interestingly, increased up-state frequency and duration have also been reported as a marker for network excitability in epilepsy models.^82,83 Both migraine and epilepsy patients show alterations in high and low-frequency cortical and thalamocortical oscillations, and the low-frequency oscillations are proposed as correlates of up states.^33,54,84 Up state phenotypes thus provide a cortical circuit mechanism that can potentially bind the phenotypes observed in both humans and animal models of migraine.

Our in vivo investigations of cortical slow oscillations showed increased up state duration and membrane voltage variance in CK1δ_T44A mice, despite hyperpolarized resting membrane potentials at baseline. During up-states, we found that CK1δ_T44A neurons received excitatory post-synaptic currents for a significantly longer duration than WT neurons, tipping the excitatory-inhibitory balance towards excitation. These in vivo findings bear an interesting resemblance to the increased frequency of action potentials evoked using intense stimuli in vitro in CK1δ_T44A neurons as both phenomena are dependent on glutamatergic input and rely on local feedback excitation. Since up states are spontaneous rather than evoked events that can occur during quiet wakefulness,³³ CK1δ_T44A circuits may be hyperexcitable at baseline, without exposure to repetitive high-intensity stimuli.

At the synaptic level, our results showed a Ca²⁺-dependent increase in RRP size in CK1δ_T44A mice following high-frequency stimulation at excitatory synapses. The RRP is functionally defined as a subset of fusion-competent and primed vesicles in a presynaptic bouton that are more easily released than the remaining vesicle population.^28,49,50 The RRP is thus dynamic, and its size changes as it is depleted and replenished simultaneously during repeated stimulation.²⁸ This dynamic modulation of RRP size is dependent on [Ca²⁺]_e^45,47 as well as stimulus frequency.^85,86 An increase in RRP size, as we observe in CK1δ_T44A on intense stimulation, has been shown to increase stimulus-evoked release probability,⁸⁵ which can enhance Ca²⁺ -dependent glutamate release.⁸⁷

The high-frequency stimulus-dependent increase we observed in the RRP of CK1δ_T44A synapses prompts a comparison with post-tetanic potentiation (PTP), which is attributed to elevated residual Ca²⁺, or changes to release machinery, including vesicle priming.^43,88 Residual Ca²⁺ following high-frequency stimulation has been shown to activate biochemical targets like protein kinase C, Munc13, synapsin and calmodulin/CaM kinase II, all of which regulate the size of the RRP.^89-91 PTP is dependent on enhanced Ca²⁺ sensitivity of the vesicle fusion process, mediated by protein kinase C and Munc13.^89,90 Moreover, genetic manipulations that impair RIM- and Munc13-mediated vesicle priming affect the RRP size.^28,92,93 Interestingly, CK1 kinases co-localize with synaptic vesicular markers and phosphorylate a subset of synaptic vesicle-associated proteins (SV2).¹⁸ Phosphorylation of SV2_A protein by CK1 family kinases controls the retrieval of synaptotagmin-1, a calcium sensor for the SNARE complex mediating the release of synaptic vesicles.^29,94

Development-dependent morphological changes in synapses can also affect vesicle release probability. Recent evidence from CK1δ loss-of-function mutants in Caenorhabditis elegans suggests a potential role in axon maturation and stabilization.⁹⁵ Multiple studies have demonstrated the variations in the presynaptic RRP size and release efficacy at different stages of axonal maturation.^96,97 Therefore, a developmental role of the kinase in CK1δ_T44A mice could contribute to the presynaptic phenotypes observed in this study.¹⁸ Taken together, these findings support a possible role of CK1δ in the regulation of presynaptic function and suggest avenues for further investigation to uncover the molecular mechanisms responsible for the CK1δ_T44A adaptation phenotype.

Intriguingly, we observed an increase in the tonic inhibitory currents of CK1δ_T44A neurons, leading to a hyperpolarized resting V_m, which is generally consistent with a reduction in excitability. At first approximation, this result appears to counter the other cellular and circuit phenotypes and suggests a possible compensatory mechanism. However, there is an equally tenable alternative explanation. Tonic inhibitory currents are mediated by extra and peri-synaptic GABA_A receptors²³ and can be elicited by low levels of ambient GABA, due to their high GABA affinity.^22,98 Tonic inhibitory currents are modulated by vesicular GABA release as well as re-uptake and therefore their activity correlates with the intensity of local network activity.^22,99 Neuronal and astrocytic GABA transporters responsible for the GABA re-uptake play an important role in regulating the ambient levels of GABA, which in turn can influence inhibition at tripartite synapses.^100,101

Under baseline conditions, GABA transporters remain close to equilibrium (reversal) potential.¹⁰² As a result, even moderate circuit activity (brief neuronal depolarization or bursts of APs) is sufficient for GABA transporter reversal, facilitating both vesicular as well as non-vesicular GABA release,^103,104 both of which can contribute to tonic inhibitory currents.¹⁰⁵ Thus, a plausible explanation of the increased tonic inhibitory currents in CK1δ_T44A neurons is that they are a direct consequence of the hyperexcitable cortical circuits mediated by the presynaptic gain-of-function, rather than a compensatory response. It should also be considered that tonic inhibition may even contribute to network excitability, as the resulting increase in [Cl⁻]_i can result in shunting inhibition or render GABA excitatory, due to depolarized GABA reversal potential.^106-108 That said, tonic inhibition as a more regulated compensatory response to circuit excitability (e.g. via reductions in GABA transporter expression)¹⁰¹ cannot be ruled out. Ultimately, despite the increased tonic inhibition, CK1δ_T44A cortical circuits remain hyperexcitable.

Apart from [Ca²⁺]_e-dependent glutamate release, other mechanisms such as insufficient glutamate or extracellular K⁺ re-uptake may also contribute to the presynaptic hyperexcitability in CK1δ_T44A mice. Like GABA transport, glutamate re-uptake through neuronal (EAAT3-4) and astrocytic (EAAT1-2) transporters is dependent on ionic gradients and thus, can be susceptible to prolonged synaptic activity and membrane depolarization.^109,110 Moreover, astrocytic gap-junction protein connexin-43 is a substrate for CK1δ.¹¹¹ Reduced phosphorylation by CK1δ_T44A can adversely affect astrocytic gap junction formation and consequently, syncytial K⁺ buffering.^15,112 Reduced glutamate transporter function or altered K⁺ buffering in CK1δ_T44A, especially after repeated high-frequency synaptic stimulation, could contribute to the reduced glutamatergic adaptation. These putative non-neuronal mechanisms underlying adaptation deficit at CK1δ_T44A excitatory synapses need further investigation.

In conclusion, we uncovered evidence of a presynaptic gain-of-function at glutamatergic synapses in CK1δ_T44A mice, due to a calcium-dependent increase in RRP size that results in increased synaptic glutamate release. This gain-of-function caused a stimulus-dependent adaptation deficit of glutamatergic synaptic activity and was accompanied by enhanced network activity both in vitro and in vivo that contributed to SD susceptibility. Taken together, our findings in CK1δ_T44A mice show presynaptic origins of cortical circuit excitability associated with phenotypically common forms of migraine (Fig. 8).

Data availability

The authors confirm that the data supporting the findings of this study are available within the article and its Supplementary material. Raw data were generated at The University of Utah. Raw and derived data supporting the findings of this study are available on request from the corresponding author.

Funding

This work was supported by the United States National Institutes of Health: R01 NS102978 and NS104742; and the United States Department of Defense: PR200891 (K.C.B.).

Competing interests

The authors report no competing interests.

Supplementary material

Supplementary material is available at Brain online.