Arguments against the role of cortical spreading depression in migraine

Abstract

Cortical spreading depression (CSD) is a wave of increased electrocortical activity and vasodilation, followed by sustained decreased activity and prolonged vasoconstriction. Although the discovery of CSD has been ascribed to Leão, rather than vasoconstriction, he only observed a depression of neural activity combined with vasodilation, with much weaker stimulation than used by his followers. There is a longstanding belief that CSD underlies migraine aura, with its positive symptoms such as mosaic patterns and its negative symptoms such as scotoma, and a similar propagation speed and vasoreaction pattern. However, there are many arguments against this theory. CSD is difficult to evoke in man, and electroencephalography (EEG) readings are not flattened during migraine (as opposed to EEG during CSD). Moreover, in contrast to CSD, migraine can occur bilaterally, and is not accompanied by a disrupted blood–brain barrier, increased cerebral metabolism, or cerebral cell swelling. Calcitonin gene-related peptide, which is thought to be characteristic of migraine pain, is increased in the blood from the external jugular vein during migraine in humans, but not during CSD in cats or rats. Moreover, CSD does not explain the appearance of premonitory symptoms or allodynia, long before the actual onset of aura. In addition, there is a variation in the pain mechanisms of migraine and CSD, and in their reaction to transcranial magnetic stimulation and several pharmacologic interventions. Finally, the origin of putative CSD in migraine is currently unknown.

Keywords:

• Migraine
• cortical spreading depression
• Leão
• vasoconstriction
• EEG

Background

Migraine is an important health problem. In Western countries, it affects 12–16% of the population. For several decades, it has been generally assumed that migraine’s pathophysiology is based on a phenomenon called Cortical Spreading Depression (CSD), which is thought to be a wave of initial increase, followed by a sustained decrease in electrocortical activity that is associated with initial vasodilation and prolonged vasoconstriction. However, two recent studies have challenged the role of CSD in migraine [1,2]. The present paper provides additional arguments to re-examine the CSD hypothesis. However, first, it will be shown that CSD as initially described in 1944 by Leão differs from CSD reported in anesthetized and awake animals in later studies. This point is often overlooked, and yet, is crucial for completely understanding the arguments presented here.

Lashley’s aura and Leão’s CSD

The headache in approximately one-third of migraine sufferers is preceded by an aura that denotes the slow march of sensory or motor symptoms, which usually gradually develops over 5–20 minutes and lasts less than 60 minutes. Visual aura positive neurological symptoms, e.g. flashing lights and fortification or mosaic patterns, are usually followed by negative symptoms, e.g. scotoma or hemianopia. Aura signs are accompanied by initial vasodilation and prolonged vasoconstriction [3,4].

In 1941, Lashley [5] analyzed his own aura symptoms, and speculated that a wave-front of intense excitation was followed by a wave of complete activity inhibition. He proposed that this wave spread at a velocity of ~3 mm/min from the center of the visual cortex to temporal parts of the brain.

Subsequently, when Leão was studying the origin of epileptic insults in anesthetized rabbits, he described a phenomenon that showed some similarity to the inhibitory wave reported by Lashley [6]. Light touch of the exposed frontal cortex caused a slow self-propagating inhibition of spontaneous and evoked electrocortical activity (including EEG), with a similar spreading velocity (~5 mm/min) as calculated for the wave of inhibition by Lashley. This CSD was associated with strong dilation of arteries and veins. Leão occasionally also observed tonic-clonic discharges during the suppression of activity, and vasodilation that was sometimes followed by moderate vasoconstriction [7]. He noted the following on the association between migraine aura and CSD:

Migraine with marked dilation of major blood vessels and the slow March of scotomata in the visual or somatic sensory sphere is suggestively similar to the experimental CSD, in spite of the fact that known scotomata are still felt to be vasoconstrictive in nature. [8]

CSD in anesthetized animals: differences between Leão’s CSD and CSD in later animal studies

Later studies borrowed Leão’s term ‘CSD’ to describe a similar phenomenon in animals like mouse, rats, rabbits, cats, and monkeys. The phenomenon included an initial increase of neuro-glial activity and vasodilation, followed by a sustained decrease of activity and prolonged vasoconstriction. The CSD demonstrated in these studies (see e.g. [13]) better resembled the excitation and inhibition in migraine aura as theorized by Lashley. A point of difference between CSD described by Leão and by later investigators was the presence of prolonged vasoconstriction in the latter. This constriction could most likely be attributed to the higher stimulus strengths used in later studies. Although Leão either used subliminal electrical stimulation (0.1–1.0 mA), a single tap with a glass rod, or a small piece of filter paper soaked in 135 mM (i.e. 1%) KCl [8], later studies mostly used stronger electric stimulation (e.g. 3–5 mA, [9]), a needle-stick (e.g. [10], which could cause potential injuries), or injection of 0.5–3 M KCl (e.g. [11,12]). All these stimuli involve the release and diffusion of chemical mediators, most likely K⁺ and glutamate. Extracellular K⁺ ions cause vasodilation and are buffered into the cortical capillaries by astrocytes. When this buffering becomes insufficient, due to a high stimulus strength, the extracellular K⁺ concentration may exceed ~20 mM, consequently constricting resistance vessels [13]. A second mechanism for the flow reduction might be the compression of the microvasculature due to the shrinking of the extracellular space, caused by the influx of water and swelling of the hyperosmolar astrocytes [14].

Subsequent studies demonstrated that CSD may not only be invoked by electrical, mechanical, and chemical stimuli (like KCl), but also by a number of other stimuli, like the topical application of N-methyl-D-aspartate (NMDA), which is a specific receptor agonist for the glutaminergic system [15], noninvasively by hypoxia [16], or by optogenetic laser light stimulation of cortical cells in transgenic mice [17]. All these involve the release and diffusion of K⁺ and glutamate. CSD is probably spread by slow gap-junctional calcium waves in the astrocytes at a rate reminiscent of aura propagation. Heptanol, which uncouples gap junctions, blocked both calcium waves and CSD [18].

CSD in awake animals

CSD has also been studied in awake animals since it may differ from CSD in anesthetized animals and because migraine occurs in conscious humans. A decrease in the amplitude of EEG was observed after a subdural injection of 2% KCl solution in conscious rats [19] and rabbits [20], and after an injection of 25% KCl into the cortex of freely moving rats [21]. Further, direct current stimulation with 3–5 mA in conscious rats depressed EEG and evoked a long-lasting reduction in cortical blood flow. Initial vasodilation, however, was only observed when cortical blood flow was diminished beforehand using anesthesia [9]. Conversely, stimulation with 50 mA increased cerebral blood flow in conscious rats to approximately 190% of the control value for 1–3 minutes and decreased flow to approximately 80% of the control value thereafter [22]. Eliciting CSD in awake rats with a topical application of NMDA resulted in a decrease in ECoG amplitude for 5–14 minutes, and episodes of freezing behavior and wet dog shakes. NMDA also induced significant c-Fos expression in the ipsilateral cerebral cortex and amygdala [23]. Similarly, application of KCl decreased locomotor activity and induced freezing behavior, wet dog shakes, and grooming in freely moving rats. It stimulated c-Fos expression in the cortex, trigeminal nucleus caudalis, and amygdala [24,25].

Therefore, the effects of CSD on electrocortical and vasomotor activity in conscious animals do not seem to greatly differ from those observed in anesthetized animals. The only difference was the freezing behavior, wet dog shakes, and grooming observed in awake animals.

Difficulty in evoking CSD in humans

Although it is easy to evoke CSD in lissencephalic animals like mouse, rats, rabbits, opossum, and pigeons, it is difficult to evoke in gyrencephalic animals like cats and monkeys [6,26]. CSD has occurred with an extra-strong stimulation (3.3 M KCl) in monkeys, but it was less frequent than in non-primate models. Further, its propagation was limited to areas near the site of stimulation. In addition, focal hyperemia was not followed by spreading or persistent hypoperfusion as measured with positron emission tomography (PET). The authors concluded that given the features of CSD in primates, reappraisal of the hypothesis that CSD contributes to the pathogenesis of human migraine was required [27]. It has been suggested that inhibitory boundaries in the gyrencephalic brain limit the spread of CSD activity from one sulcus to another [28,29].

Nevertheless human CSD has been observed with electrocorticography or scalp EEG in patients with aneurismal subarachnoid hemorrhage and hemispheric stroke [30], as well as in patients suffering from brain trauma [31,32]. These conditions are, however, absent in healthy migraineurs, and the CSD was not accompanied by aura [33].

Attempts to evoke CSD in conscious humans have been performed in awake epileptic patients whose cortices were widely exposed, prior to the surgical excision of their epileptogenic focus. In nearly 1000 patients, mechanical deformation or electrical stimulation of the cortex did not elicit a response remotely resembling aura or CSD [34]. Similarly, mechanical, electrical, chemical (15% KCl), or thermal stimulation was not sufficient to evoke a CSD in 23 conscious patients, although authors were able to induce CSD in rats using the same techniques [35]. Only Sramka et al. [36] recorded negative DC potential shifts (5–10 mV) following injection of 5% KCl into the caudate nucleus or hippocampus, in 6 of the 9 patients who were part of their study. Although there was a suggestion of spread of some of these DC potential changes, they were not associated with the expected EEG suppression. Further, Penfield [37] extensively stimulated exposed brain fields without noticing an aura or pain. Moreover, there have been no reports of migraine aura following concussion, or transcutaneous electric or magnetic stimulation.^{Table 1}

Table 1. Differences between the characteristics of cortical spreading depression (CSD) and migraine aura.

Cortical spreading depression	Migraine aura
Water diffusion leading to cell swelling	No change in water diffusion
Blood–brain barrier disrupted	Blood–brain barrier intact
Somatosensory neuronal silencing	Only positive symptoms in many aura cases
Cerebral metabolism is enhanced	Cerebral metabolism remains unchanged
Regional cerebral blood flow is only moderately reduced to non-ischemic levels	Regional cerebral blood flow may fall to ischemic levels
Symptoms immediately after initiation	Premonitory symptoms up to 72 hours before the onset of aura
Little to no cerebral damage after CSD	Localized deficits in the visual field and subcortical white matter
EEG flattened	No EEG flattening
MEG: positive or negative shifts immediately after CSD initiation	MEG: positive or negative shifts at approximately 10 minutes after the initiation of the aura

Arguments against a role for CSD in migraine aura (see Table 1)

No change in water diffusion. A dramatic change in the distribution of ions between extra- and intracellular spaces underlies the occurrence of CSD in animals. K⁺ and H⁺ ions are released from cells, while Na⁺, Ca⁺⁺, and Cl⁻ ions enter the cells, together with water [38], causing them to swell. This consequently reduces the volume of the extracellular compartment to approximately half of the control value [39]. In contrast, a human fMRI study did not demonstrate significant changes in water diffusion, during or after spontaneous migraine aura, despite reductions in regional cerebral blood flow of up to 53% [40].

The blood–brain barrier is disrupted by CSD [41,42]; however, it remains intact during attacks of migraine with aura [43] and migraine without aura [44].

Cerebral metabolism in CSD experiments on rats is distinctly enhanced [10,11]. However, it remained unchanged in four patients experiencing migraine with aura [45].

Regional cerebral blood flow in experimental CSD is only moderately reduced [46], while it can drop by approximately 50% [40] or may even fall to ischemic levels during migraine aura attacks [47].

Only positive symptoms like flashing lights, castellations, scintillations, and pins and needles are described by a number of patients with aura, without negative symptoms, like scotoma, hemianopsia, or numbness and loss of sensibility as expected from neurosilencing by CSD [29]. In other patients, visual aura symptoms are superimposed on a normal visual image, or numbness and tingling exists without a loss of touch [2]. The preservation of function seems incompatible with a process such as CSD, during which electric activity in the somatosensory cortex is strongly depressed [48].

Premonitory symptoms. The occurrence of premonitory symptoms, such as difficulty with speech and reading, increased emotionality, sensory hypersensitivity, and allodynia, up to 72 hours before the onset of the aura, in more than 70% of patients with migraine [49,50], cannot be attributed to an aura-generating CSD. In addition, premonitory symptoms occur simultaneously, as opposed to progressing consecutively, which would be expected since CSD progresses from the occipital to parietal to frontal cortices.

Risk for brain lesions. No cerebral damage has been detected after CSD in adult anesthetized rats [51]. In contrast, the prevalence of localized deficits and hyperexcitability in the visual field [52], and the presence of small hyperintensity lesions on MRI in subcortical white matter, deep white matter, periventricular white matter, and basal ganglia of migraine patients [53], strongly suggests that migraine is associated with a significantly increased risk of brain lesions. Further, it is well known that neurotransmitters, like glutamate, which are excessively present in migraineurs, could lead to excitotoxicity [54].

No EEG depression. EEG activity is depressed in CSD [6,9,19,20,23]. Although EEG has been seldom recorded during the aura phase of migraine, EEG rhythms were slowed down within 24 hours of the onset of the attack in 29 of 40 children with migraine [55]. Further, slow or sharp EEG waves appeared in 8 of 20 patients with aura during the headache phase [56]. In another study, although EEGs acquired more than 3 hours after the attack showed slow-wave abnormalities, those acquired within 2–3 hours of the beginning of aura did not reveal any abnormalities [57]. In two other aura studies, one demonstrated an increase in the θ activity [58], while the other did not show EEG changes [59]. These findings suggested that the aura was not initiated by a CSD. Recent studies on patients with brain injury, comparing changes in scalp EEG with changes in electrocorticography of the exposed cortex (ECoG), demonstrated that scalp EEG does have the capacity to identify spreading depression as a flattening of the signal amplitude [31]. The absence of these signs during migraine aura is evidence against a role for CSD in migraine.

Magnetoencephalography (MEG) is more sensitive than EEG. Using MEG, Bowyer et al. [60] demonstrated DC-shifts in multiple cortical areas in patients with spontaneous and visually induced migraine with aura, but not in controls. Although these DC-shifts were comparable to those induced by CSD in gyrencephalic animals [61], they were initiated after the beginning of aura and could have been caused by factors other than CSD, e.g. short periods of local anemia [62]. Further, using AC-coupled MEG, Barkley [63] demonstrated the suppression of neuronal activity in 5 of 8 patients with aura, and in one patient without aura, for periods ranging from 2 to 10 minutes. The suppression started at more than 20 minutes from the beginning of the recordings (see Figure 2 in [63]), and could therefore not been caused by CSD.

Arguments against the role of CSD in migraine headache

Many defenders of the CSD theory propose that CSD is not only the underlying mechanism of the migraine aura, but also of the subsequent headache. Migraine attacks without aura, which occur in more than two-third of patients with migraine, are thus assumed to be preceded by a ‘silent’ CSD, i.e. one involving areas of the brain that would not generate a perceived aura [49]. However, Hauge et al. [64] demonstrated that although tonabersat reduced the frequency of aura attacks, it was ineffective for migraine without aura, which challenged the concept of the silent CSD.

No protein leakage. It is assumed that CSD evokes pain via activation of trigeminal afferents from meningeal nociceptors. This could cause reflex-like neurogenic inflammation with protein leakage and edema, particularly in the postcapillary venules of the dura [65–68]. However, there was no evidence of protein extravasation around dural and brain blood vessels during attacks of migraine with [69] and without aura [70].

No neurogenic inflammation. In patients, migraine headache was not prevented or reduced by inhibitors of substance P [65,68,71], neurokinin-1 (lanepitant) [72], or endothelin (bosentan) [73]. However, all these antagonists prevented animal CSD-induced neurogenic dura mater inflammation. During migraine attacks in another study, a short-lasting dilation of the pain-sensitive cranial veins was induced by applying pressure on the internal jugular veins (Queckenstedt’s maneuver), but there were no significant differences in the headache intensity ratings from the control (i.e. when an equal pressure was applied onto the lateral aspect of the neck) [74]. These findings make it unlikely that a neurogenic inflammation in the cephalic venous system is of major importance in migraine pain mechanisms.

Calcitonin gene-related peptide (CGRP), which is thought to be characteristic of migraine pain [75], is increased in human blood taken from the external [76] or internal jugular [77] or cubital [78,79] veins during migraine, although this finding could not be reproduced in another study [80]. Although CGRP is also released from rat neocortical slices during CSD [81], it was not elevated in the blood acquired from the external jugular vein, in vivo during CSD, in cats [82] or rats [83].

Single transcranial magnetic stimulation of the occiput in patients with migraine abolishes their pain within 2 hours [84]. Similar stimulation of the occipital cortex in the cat, however, failed to inhibit the majority of CSD [85].

Uni- and bilaterality. CSD does not cross the medial longitudinal fissure. It remains in the hemisphere in which it is evoked [42]. However, migraine pain is unilateral in 60% of patients and bilateral in 40% of patients [86], indicating that it is not caused by putative CSD alone.

Awake rats did not demonstrate avoidance behavior to the dark compartment in which multiple CSD waves were induced by 3 M KCl. This indicated that CSD is not an aversive experience, thus, disputing whether CSD causes pain in animals [87]. In addition, when a single CSD was evoked in freely moving rats with topical NMDA onto the dura mater, this only elicited freezing behavior, but not ultrasonic vocalizations, which would have been consistent with severe pain [23]. In another study with freely moving rats, CSD could be elicited by pinprick alone or by pinprick and KCl injection into the occipital cortex. CSD evoked by pinprick plus KCl, but not pinprick alone, provoked tactile cutaneous allodynia in the face and hind paws. This highlighted the role of KCl in causing cutaneous allodynia, but excluded the role of CSD evoked by pinprick alone in eliciting this phenomenon [88].

Pharmacological arguments

If CSD underlies migraine, it is expected that successful anti-migraine agents will inhibit CSD. In cats anesthetized with α-chloralose, however, dihydroergotamine, acetylsalicylic acid, lignocaine, metoprolol, clonazepam, and valproate, which are all migraine-preventive or -abortive agents, did not reduce CSD in the cat [89]. Further, droperidol [90], riboflavin [91], and sumatriptan [92], which alleviate migraine, did not diminish CSD in rats. In addition, propofol, fentanyl, morphine, and sufentanil, which are all analgesics and sedatives that clearly reduce migraine, did not inhibit CSD, elicited during acute brain injuries in humans [93]. Inhalation of 10% CO₂ also discontinued the visual disturbances of aura and prevented headache [94], probably via vasodilation. Although CO₂-induced vasodilation was also present in anesthetized rats, it was absent after CSD [95,96]. Moreover, CGRP, which is known to cause migraine with and without aura [97], did not evoke CSD in rat brain slices [81]. Further, although Tozzi et al. [81] reported inhibition of CSD by CGRP antagonists in rat brain slices, CGRP antagonists, which are known to depress migraine, did not block KCl-induced CSD in anesthetized cats [98]. Conversely, several compounds that are able to suppress CSD in the rat, like the α₂-receptor agonist, clonidine [99,100], had no effect on migraine aura [101,102].

Nevertheless, a number of migraine preventive therapies reduced the susceptibility to CSD [103]. One study demonstrated that chronic treatment with topiramate, valproate, propranolol, amitriptyline, or methysergide reduced the frequency of repetitive CSD, which were evoked by continuous stimulation. It must be noted that the acute administration of these drugs was ineffective [100]. Similarly, anti-epileptic therapies (i.e. lamotrigine and gabapentin) alleviated migraine and also markedly reduced CSD [41]. Conflicting results have been obtained with the anti-migraine effect of glutamate receptor antagonists (e.g. ketamine) [104], and an inhibitor of gap-junction communication (tonabersat) [105,108], which both are claimed to diminish CSD.

Thus, although several agents did not show a pharmacological relationship between migraine and CSD, this association was confirmed for others. These findings indicated that migraine and CSD are different processes with some shared mechanisms. Examples of such mechanisms might include the increase of glutamate in the cerebrospinal fluid of patients with migraine [106] and CSD in animals [107], or neuronal-glial cell gap-junction communication during migraine [105], and CSD [109]. A partial relationship similar to that of migraine and CSD has also been found for migraine and epilepsy. Despite their separate appearances, migraine can be successfully treated with a number of anticonvulsants [41].

An alternative hypothesis

It is now widely accepted that migraine is a complex brain network disorder that involves multiple cortical, subcortical, and brainstem regions, which account for the aura, pain, and wide constellation of symptoms characterizing the attack. [92]. With the knowledge that more than one mechanism may initiate a sequence of events leading to a migraine attack, we proposed that one of these mechanisms might be shear-induced platelet aggregation [110]. Platelets might aggregate by increased shear stress in a narrowed vessel to the brain or in the circle of Willis [111]. Locally released serotonin results in endothelial 5-HT_2B-receptor-induced nitric oxide (NO) formation and release of CGRP [112], causing pain and dilation of extracerebral arteries (migraine without aura) in the respective hemisphere. As known from the hemispheric localization of speech function with aid of intracarotid injection of sodium amytal, carotid and vertebral blood flow are mainly directed to the ipsilateral part of the brain. Stronger platelet aggregation with higher plasma serotonin levels provokes vasoconstriction [113]. In addition to platelet-released glutamate [54,114], a high serotonin concentration may induce gap-junctional slow calcium waves in astrocyte syncytium [115,116], thus being responsible for the slowly proceeding aura signs [117,118]. Platelet aggregation may be initiated by triggers that further enhance an already elevated platelet aggregability [119], such as emotional and physical stress, awakening in the morning, additives in food and beverages, oral contraceptives, and the monthly decrease of estrogen during the menstrual cycle [110]. Shear stress-induced platelet aggregation is less sensitive to aspirin, but may be blocked by adenosine diphosphate (ADP) inhibitors, such as clopidogrel [120]. To our knowledge, this is the first theory that may explain the unilateral nature of most migraines. It may also explain why certain patients may have migraine with or without an aura on different occasions, i.e. this phenomenon may be dependent on the strength of aggregation.

Discussion

Despite the common practice in mentioning Leão as the discoverer of the CSD, it must be noted that he described CSD with vasodilation alone, and does not consider vasoconstriction seen in later studies with stronger stimulation. The latter showed similarity to the positive and negative symptoms of aura, as it starts with an increase and is followed by a silencing of nervous activity, with similar propagation speed and vasoreaction pattern. However, as outlined in Table 1, there are many differences between CSD and migraine aura. Further, there are also differences between CSD and migraine headache (Table 2). The occurrence of spontaneous CSD in patients with migraine has never been demonstrated and thus, its possible triggers and mechanisms of initiation remain unknown. Several CSD stimuli used in animals, i.e. a high concentration KCl and strong electrical, mechanical, or optogenetic stimulation, are absent in patients with migraine. Theoretically, CSD could be induced in humans via hypoxia, ischemia, hypoglycemia [41], microemboli [121], endothelin-1 from irritated endothelium [122], and increased glutamate levels [23]. However, given the many differences found between CSD and migraine (see Tables 1 and 2), these possibilities seem unlikely. Although CSD occurs in humans with subarachnoid hemorrhage, stroke, and brain trauma, it is difficult to generate in healthy people. Nevertheless, a pharmacological relationship between migraine and CSD seems to be confirmed for agents like glutamate receptor antagonists and gap-junction communication inhibitors. This suggests that although migraine and CSD are different processes, they may have some shared mechanisms. We concluded that, despite its beneficial side effects for prophylactic therapeutics, the study of CSD will not solve the enigma of migraine generation.

Table 2. Differences between the characteristics of cortical spreading depression (CSD) and migraine headache.

Cortical spreading depression	Migraine headache
Protein leakage	No protein leakage
Neurogenic inflammation	No neurogenic inflammation
CGRP not increased in jugular vein blood	CGRP increased in jugular vein blood
sTMS fails to inhibit CSD	sTMS abolishes pain
Unilateral	Unilateral and bilateral

Note:

CGRP: calcitonin gene-related peptide; sTMS: single transcranial magnetic stimulation.

Contributors

PB conceived and designed the study, obtained funded and ethics approval, analysed the data, wrote the article in whole/part, revised the article.

Disclosure statement

The author declares that he has no competing interests.