Spreading depression as a preclinical model of migraine

Migraine with aura occurs in 30–40% of patients diagnosed with migraine and is most commonly a visual disturbance. The visual disturbance can be variable and include fortification spectra, sparkling or shimmering colored dots and blobs, and scotoma [15‐17]. While visual symptoms are the most commonly described aura event of migraine, other auras including sensory and speech disturbance have been described. In one study visual aura occurred in 98% of those with migraine with aura, while sensory symptoms including paresthesias and hypoesthesia occurred in 36% and dysphagic symptoms in 10% [18]. In those with more than one aura symptom, the onset of the second or third aura symptom seem to follow the first or second aura symptom in succession, i.e. the additional aura symptom starts subsequent to the start of the preceding aura symptom. In those with two aura symptoms, the second symptom started after the onset of the first 66% of the time. In those with three aura symptoms, the third symptom started after the onset of the second 82% of the time [19].

There are several clinical studies supporting SD as the likely mechanism involved in the aura event of migraine which has been the topic of multiple well written review articles. In early depictions of migraine aura, Lashley postulated that the positive symptom resulted from a region of cortical hyperexcitability while the scotoma likely related to an area of diminished cortical activity spreading across visual cortex. It was further surmised based on the rate of spread that the velocity of this electrical event was about 3 mm/minute. The event of cortical SD (CSD) recorded by Leao, having congruent temporal pattern and spread, raised the possibility that SD was the underlying electrophysiological event of migraine aura [17, 20]. Several clinical studies have since supported this relationship between migraine aura and SD. Both SD and the migraine aura phase are associated with pronounced oligemia as noted in multiple Xenon based and single photon emission computed tomography imaging studies [17, 21]. In one study examining functional magnetic resonance blood oxygen level dependent (BOLD) signaling during migraine aura, increased BOLD signal propagated across visual cortex retinotopically coincident with movement of the aura a rate resembling SD. This further buttressed the causal relationship of Leao’s SD to migraine aura [13, 22]. Symptoms other than the visual disturbance suggest that brain regions outside of striate cortex may be involved in migraine with aura and possibly affected by the spread of CSD [23] though this has not been confirmed in human studies. While direct clinical evidence that SDs are causally associated with sensory and other non-visual aura symptoms is limited, experimental SDs can be generated from various anterior and posterior cortical brain regions highly suggestive of SD as a neurobiological phenomenon responsible for these aura symptoms. Some migraine with aura sufferers experience sensory and visual symptoms simultaneously (i.e. without succession) raising the possibility that in addition to spread, SD could be generated in multifocal regions simultaneously [18, 19].

There are arguments both for and against a temporal relationship between the migraine aura and headache. While some will experience aura without headache, most migraine attacks with aura are accompanied by headache (91%) [18]. While the headache can occur before or simultaneous with the aura event, the headache in most cases (78%) occurred after the onset of aura either during the aura phase (28.7%), at cessation of the aura (12.1%) or a period after aura cessation (37.6%) [18].

Activation of the trigeminovascular system (TVS) is critical to migraine pathogenesis [24‐30]. CSD may be a key CNS trigger for TVS activation [31]. CSD can activate perivascular trigeminal afferents and evoke a series of cortico-meningeal and brainstem events consistent with the development of headaches [32‐36]. CSD leads to increased expression of the immediate early gene product c-FOS in the trigeminal nucleus caudalis (TNC), sterile neurogenic meningeal inflammation mediated by trigeminal collateral axons, and dilatation of the middle meningeal artery via the trigemino-parasympathetic reflex [32]. Single unit recording studies showed that CSD can lead to delayed and long-lasting activation of meningeal nociceptors in the trigeminal ganglion [33] and central TVS neurons in the TNC [34]. The precise mechanism that triggers TVS activation has yet to be elucidated, but SD can cause the release of inflammatory and diffusible substances in the cortex including prostanoids, nitric oxide, ATP, and K⁺ [37]. In addition to glutamate release and collapse of ionic gradients; SD can activate purinergic receptors and pannexins, large pore channels whose stimulation can produce brain inflammation [36, 38]. In fact, SD can increase brain cytokine release and astroglial activation. In addition to local cortical responses to SD, diffusible substances may reach the overlying meningeal surface and potentially activate trigeminal neuropeptide containing axons leading to peripheral and central release of calcitonin gene related peptide (CGRP) [32, 39, 40]. This and other mechanisms may be involved in meningeal inflammation and peripherally and sensitization of TNC neurons centrally. The stimulation of these nociceptive pathways may be involved in the pain of migraine. That SD can trigger a series of events likely involved in the headache phase of migraine provides a plausible biological link between SD generation and migraine pain that may not be necessary for the generation of migraine pain but in some cases, it may be sufficient. Hence, SD may be not only the physiological substrate of migraine aura, but also a potential cause of headache. Although it is a matter of debate whether migraine patients without aura have asymptomatic SD, a recent study suggests that the visual percept of aura can be clinically silent [41]. While speculative, it is possible that some migraine patients without a perceived aura might have SD-like activities propagating through ineloquent cortex.

There are limitations to the approach of using SD as a model to study mechanisms that may be associated with migraine. Much like other models, it is one component part of a complex heterogenous disease process involving genetic, sex dependent, hormonal and environmental factors. Therefore, like other models including meningeal application of exogenous inflammatory substances, SD does not encompass all the disease complexity of migraine. However, it does allow for the examination of alterations in cortical and subcortical brain excitability and nociceptor activation. There are several shortcomings of the SD model that are detailed below including the invasive conventional methods previously employed which may resemble a model of injury as opposed to migraine. While there are concepts that challenge the link between SD and the headache of migraine [42], including the variable onset of headache following aura symptoms, aura without headache and several incongruent preclinical observations; the evidence for a plausible causal relationship of SD to trigeminal nociceptor activation and therefore likely pain remain convincing. To the extent that SD can activate dural afferents and second order trigeminovascular neurons [40, 43], increase neuropeptide release and alter pain behavior [44]; it is a reasonable experimental model to investigate SD mechanisms involved in migraine with aura. Moreover, since trigeminal activation is a critical component of migraine pain, SD mediated activation of trigeminal neurons and peripheral release of neuropeptides may link the aura of migraine with the pain experienced during an attack.

In this review, we summarized the currently known experimental models of SD, reviewed the triggers, modulators and consequences of SD, and elaborate their relevance to migraine (Fig. 1).

Experimentally-evoked SD in normally metabolizing brain tissue requires intense depolarizing stimuli. An increase of extracellular K⁺ above a critical threshold concentration (12 mM) in a minimum volume of brain tissue (1 mm³) is estimated to be the minimal requirement to provoke SD in rodents [55, 56]. A variety of stimuli, spanning pharmacological, electrical, and mechanical modalities, have been used to induce SD [2, 57, 58]. Each has its own caveats and possibly differential mechanisms [47, 59].

The conventional methods to induce SDs discussed above have been critical in advancing our current understanding of the role of the phenomenon in human disease. However, inferences about the role of SD in inflammation, for example, could be confounded by the invasive nature of conventional SD induction methods. Therefore, a non-invasive approach for the induction of SDs could be a useful complement to conventional methods.

Optogenetics technology enables non-invasive, real-time stimulation of targeted brain cells, and provides the potential for detailed, precise insight into disease mechanisms in awake animals [79, 80]. Investigators have recently developed such a non-invasive approach by utilizing transgenic optogenetic mouse lines where a light-responsive ion channel called channelrhodopsin-2 (ChR2) is expressed in excitatory cortical neurons under the Thy1 promoter [81‐84]. This new optogenetic approach allows for the controlled induction of SDs through intact skull using 470 nM blue wavelength light stimulation. Optogenetic SDs can be induced as single events, repeated to determine the impact of recurrence and produced in both anesthetized or awake and behaving animals. Importantly, the technique enables longitudinal study of SDs over the course of weeks without brain injury confounders caused by invasive SD induction and detection methods. Optogenetic SD induction can be detected with multiple methods including optical intrinsic signal (OIS) imaging. However, when simultaneously examining SD detection using this method in combination with other techniques (electrode recording, laser speckle imaging, and laser doppler flowmetry); the fidelity and reproducibility of the response detected with OIS was indeed comparable to more invasive methods like electrode recording [82].

Behavioral animal models are vital in translational studies of human diseases. While migraine can be defined clinically, preclinical methods used to study mechanisms of migraine model component features thought to be critical for the pathological generation of a migraine attack. However, awake animal models of SD are scarce. In awake and freely moving rats, SD-evoked blood flow changes are consistent with those identified in anesthetized animals [141], suggesting that SD models in awake animals may be useful for modeling migraine aura. Whether awake SD models could recapitulate migraine headache-like behaviors is not yet known. One study showed that KCl injection but not pinprick over the cortex in freely moving rats induced tactile allodynia of the face and hindpaws, and increased Fos expression within the TNC [142]. However, application of KCl onto the dura, without eliciting SD events, could also elicit cutaneous allodynia and increase TNC Fos staining [142]. Therefore, it seems that sustained activation of trigeminal afferents required to establish cutaneous allodynia may be independent of SD. In free-moving rats, induction of a single SD with topical NMDA evoked freezing behavior and wet dog shakes but not ultrasonic vocalization consistent with pain calls (22–27 kHz), suggesting that SD induces anxiety and fear (possibly via amygdala activation) rather than severe pain [61]. Nevertheless, while cutaneous allodynia and ultrasonic vocalization are not completely synonymous with headache; these studies did not refute the proposed link between SD and trigemonovascular activation observed in anesthetized rats [143]. The behavioral responses to “repetitive SDs” evoked by topical KCl have also been evaluated in studies of awake free-moving rats, which demonstrated that SD could propagate into thalamic reticular nucleus and significantly decrease locomotor activity and induce freezing behaviors [144]. It remains uncertain to what extent these behaviors represent pain. However, taken together these neurobiological disturbances are consistent with the migraine condition in humans. While animals cannot be queried about whether or not they have a migraine, these SD associated pathological consequences would suggest that SD is functionally important for the symptomatology of a migraine attack in those that have migraine with aura.

Using the mouse grimace scale [145] it was shown that topical 1 M KCl induced painful craniofacial expression in mice [36]. Although 1 M KCl would readily induce SD in mice, it might also cause significant chemical irritation to the dura and cortex. The newly developed non-invasive optogenetic methods (see above) might circumvent this shortcoming and better address the link between SD and headache. Awake FHM1 R192Q and S218 L mutant mice, exhibit behavioral changes suggestive of spontaneous unilateral head pain, including increased amount of head grooming with unilateral oculotemporal strokes and increased blink rates with one eye closed, induced by novelty and/or restraint stress. In addition to potential signs of head pain, FHM1 mice displayed signs of photophobia [122].

Migraine is a repetitive neurological attack of disabling headache accompanied by sensory and gastrointestinal disturbances. The classification criteria for migraine takes into account its recurrent nature [146]. Chronic migraine is an implacable incapacitating form of migraine characterized by very frequent attacks. However, the ability to model the recurrent nature of episodic migraine and the very frequent attacks of chronic migraine is a challenge [147]. Despite SD being one of the most widely used models of migraine; using SD to model recurrent episodic or chronic migraine has been hampered by the invasive nature of prior SD models, which often resulted in a barrage of SDs. Injurious methods involving pinprick trauma or direct continuous topical KCl application require placement of a burrhole and likely produce meningeal damage and irritation as part of the surgical preparation. Furthermore, the barrage of SDs occurring at a frequency of 9~12 per hour does not align well with the experience of migraine aura, which would likely be the result of a single SD event. These represent just a few of the challenges involved in using SD to model recurrent or chronic migraine.

There have been only a few preclinical studies on migraine chronification and associated phenotypic behaviors. In one method, the epidural surface or cortex is exposed after scalp reflection and burrhole drilling through the skull. Two methods of chronic daily SD lasting 1–2 weeks were employed. In the first method, a cotton ball soaked in 1 M KCl is placed on the epidural surface for 1 min, followed by a saline wash, to induce a single CSD. In the second method, Tungsten stimulation electrodes are implanted 1-mm below the cortical surface. One second square pulse direct bipolar cathodal stimulation (100–8000 μC) is delivered until a single SD is elicited. Between stimulations, the animals are re-sutured to mark the sites where epidural application of KCl and electrical stimulations were made. An increase in astrocyte staining and decrease in SD susceptibility were observed with these techniques [148]. In a variation of this technique, a 2-mm burrhole was drilled through the skull, taking care to leave the dura intact. A plastic tube (2.5 mm inner diameter) was then affixed to the skull surrounding the burrhole with dental acrylic. The tube was capped to keep the dura moist. Through this tubing, 10–100 mM NMDA or 1–3 M KCl solution (10–20 μl) was allowed to diffuse to the cortical surface below and produce an SD [61]. Using these methods, it is possible to examine the effects of repeated SD on freezing behavior, periorbital mechanical allodynia, and anxiety behaviors [149]. One potential limitation of these techniques is it still involves the potential direct stimulation of the meninges with burrhole drilling and direct application of supraphysiological concentrations of NMDA and KCl directly onto the meningeal surface. Although SD is produced, it is unclear whether the observed changes are due to the SD itself or the disturbance of meningeal nerve terminals.

An optogenetic approach offers the opportunity not only to produce SD noninvasively but also to do so repeatedly [83]. In our laboratory, we have constructed two methods for repeated single event SD induction using optogenetics. In the first approach, a glass coverslip is fixed to the intact skull after a single scalp incision [81]. The durability of the glass coverslip allows daily blue light stimulation (470 nM) for up to 2 weeks. In the second approach, two 10-μL plastic pipette tips cut to a 5-mm length are glued to the intact skull overlying the stimulation site, through which an optical fiber can contact the skull, and the recording site, through which a laser doppler fiber can be placed. SDs are then detected by a characteristic change in the laser doppler flow signal following light stimulation. After SD induction, the fibers are removed, and the animals can be returned to their cages until the next stimulation. This procedure can be done repeatedly in both line 9 and line 18 Thy1-ChR2 YFP transgenic animals. However, we observed an increase in SD threshold with repeated stimulation in this latter method, which may become prohibitive in the line 9 animals as compared to the line 18 animals, as the line 18 animals tend to have lower thresholds (data not published).

These methods can be used to examine changes in pain behavior, anxiety, and cognition as well as changes in light sensitivity and social interactions free of the confounding factors of the invasive induction paradigms previously employed. The use of repeated, non-invasive, optogenetically-induced SD may be able to help answer important questions about the sensory, psychiatric, and cognitive dysfunctions that can accompany chronic migraine. Given the differences in life span of rodents as compared to humans, it is unclear if a direct correlation can be made between the frequency of attacks in humans and those experimentally produced in mice. In that sense, the model is used to examine the nature of change that occurs with repetitive single event less invasive SD but does not (and likely cannot) perfectly recapitulate the human condition of migraine in timing and frequency.

In vivo models can be challenging and time consuming due to microsurgical preparation and maintenance of stable systemic physiological conditions under anesthesia. Nevertheless, they are essential for preclinical therapeutic testing. In vitro models in brain slices or chicken retina are also critical in SD research [150‐152]. The key advantage of the brain slice over a whole animal preparation is that parameters such as temperature, oxygenation, pH, ionic, and pharmacological environment can be precisely controlled. Cellular resolution imaging and high-quality electrophysiological recordings can be better performed in a slice than in vivo. Slice preparations also allow access to brain regions that are difficult to access in in vivo studies, especially in human brains. In vitro SD studies in chicken retina, which has characteristics similar to brain slices, also allowed systemic evaluation for SD pharmacology [151]. However, these in vitro models are not networked nervous system. To understand the complex brain circuitry involved in migraines, the information obtained from in vitro models is limited.