Abstract
Recent progress in the genetics of migraine has refocused attention on cortical dysfunction as an important component of the pathophysiology of this disorder. In previous work, we have demonstrated functional changes in the visual cortex of migraine patients, using an objective transcranial magnetic stimulation technique, termed magnetic suppression of perceptual accuracy (MSPA). This study aimed to replicate previous findings in migraine with aura (MA) and to use the technique to examine migraine without aura (MoA). Eight MA patients, 14 MoA patients and 13 migraine-free controls participated. MSPA assessments were undertaken using a standardized protocol in which computer-presented letter targets were followed at a variable delay interval by a single magnetic pulse delivered over the occiput. MSPA performance is expressed as a profile of response accuracy across target-pulse delay intervals. The profiles of migraine-free controls exhibited a normal U-shape. MA patients had significantly shallower profiles, showing little or no suppression at intermediate delay intervals. MoA patients had profiles that were similar to controls. Recent animal evidence strongly indicates that the U-shape of the normal MSPA function is caused by preferential activation of inhibitory neurons. Shallower MPSA profiles in MA patients are therefore likely to indicate a functional hyperexcitability caused by impaired inhibition. The finding of normal MPSA profiles in MoA patients is novel and will require further investigation.
Introduction
A substantial focus in recent migraine research has been on the genetic basis of the disorder. For reasons that are well documented (e.g. (1)), studies on familial hemiplegic migraine (FHM) have been at the forefront of scientific progress in this area. There is a continuing debate, however, about the likely clinical and pathophysiological relationship between FHM and the more common migraine phenotypes, migraine with aura (MA) and migraine without aura (MoA). Recently, Moskowitz et al. (2) have suggested how the two ion channel deficits responsible for FHM may also allow insights into the mechanisms responsible for MA. In brief, Moskowitz and colleagues argue that the spread of cortical spreading depression (CSD) may be facilitated by excessive glutamate accumulation in the extracellular space. This, in turn, is caused by the combination of increased calcium influx to the presynaptic terminal (the FHM-1 calcium channelopathy), resulting in increased glutamate release and decreased clearance of glutamate from the synaptic cleft by adjacent astrocytes (the FHM-2 sodium channelopathy).
These extraordinary advances in migraine genetics give new and additional impetus to the discussion about the role of the cortex in migraine pathophysiology. Increasingly, events in the cortex are seen as important: Pietrobon and Striessnig (3), for example, in a recent paper in Nature Reviews Neuroscience, comment ‘… cortical spreading depression (CSD) is the most probably primary event in trigeminovascular system activation in migraine with aura and perhaps also migraine without aura’ (p. 390). This re-orientation is consistent with a view that has been expounded over the past decade or so by a small community of neuroscientists who have developed a well-articulated account of interictal cortical hyperexcitability predisposing the brain to develop episodes of CSD (4–7). Animal models confirm that susceptibility to CSD is conferred by the FHM-1 mutations: knockin mice expressing this mutation had a lowered charge threshold for CSD induction by electrical stimulation than their wild-type littermates (8). The mutation also resulted in a 150% increase in CSD velocity from approximately 3 mm/min to approximately 4.5 mm/min.
Evidently, invasive procedures to determine susceptibility to CSD are impossible in human subjects. A surrogate marker of cortical hyperexcitability has, for some years, been an important goal, sought by us (9–12) and others (13–15). Transcranial magnetic stimulation (TMS) techniques have been an important tool in progress towards this goal. TMS is a non-invasive technique that allows investigation of the functional status of superficial cortical areas in patients and controls. The early use of TMS to investigate migraine focused on the method of phosphene induction: stimulation of the primary visual cortex (Brodmann's area 17) with eyes shut can result in the subject perceiving flashes, sparkles or bright zig-zags in the visual field: these are termed phosphenes. Some investigators have demonstrated that the magnetic field intensity required for phosphene induction is lower in MA than in matched controls (11, 16, 17); others, however, have demonstrated the reverse (18, 19). This heterogeneity of findings has led to considerable debate (20–22) and a general feeling that the phosphene threshold is insufficiently objective.
As a result of this, a number of objective TMS indices of cortical function have been developed over the past 3 years. Battelli et al. (23) used single-pulse TMS over visual area V5 to induce phosphenes in migraineurs and controls. Brighina et al. (24) showed that 1-Hz repetitive TMS had an effect in MA patients that was consistent with increased excitability. Our group has concentrated on the use of a technique pioneered in visual neuroscience by Amassian et al. (25). We term the technique magnetic suppression of perceptual accuracy (MSPA). This technique examines the extent to which perception of simple letter stimuli is suppressed by a magnetic pulse over the occipital skull, which overlies the primary visual areas of the brain.
In 2001, Mulleners et al. (12) demonstrated that the MSPA profiles of MA patients were significantly shallower than those of controls. The MSPA technique is advantageous in the following respects. First, it is straightforward for patients to understand—their sole task is to report letters that are flashed onto a computer screen. Second, both the presentation of letters on screen and the assignment of stimulus onset asynchrony (SOA; the delay between the onset of the visual stimulus and the start of the magnetic pulse) are done randomly by computer, eliminating any potential for experimenter effects. Third, patients are not aware of the variation in SOA, making it very unlikely that their reports of the stimulus letters could be contaminated by any demand characteristics of the technique or wish to impress upon the experimenter the severity of their condition. Fourth, patients should demonstrate improved performance over controls, so it is unlikely that patient–control differences could be ascribed simply to a general deterioration of function in the patient group.
The aim of this study therefore was twofold. We expected to replicate the MSPA findings of Mulleners et al. (12), in that the MSPA profiles of MA patients should be shallower than those of controls. We also wished to examine MSPA profiles in MoA patients: relatively little is known about cortical hyperexcitability in this group and it is important to know whether the same degree of excitability is associated with this condition.
Methods
Subjects
Thirty-five subjects participated in this study, eight MA, 14 MoA and 13 migraine-free controls. Clinical characteristics of the patient groups are shown in Table 1. Although case-by-case matching of patients was not possible, there were no statistically significant differences in gender ratio, age, migraine frequency or disease duration across MA and MoA groups. All subjects were recruited from the local population using media advertisements and posters in the surgeries of local general practitioners. All subjects were sent a study information sheet accompanied by an expression of interest form. Subjects were asked to return the completed expression of interest form if they wished to take part in the study. This form asked for details regarding migraine history, current medication and general medical history. The forms were screened on return to exclude subjects for whom TMS could be considered unsafe, i.e. those with a history of epilepsy, presence of a cardiac pacemaker or other implanted medical devices. Additional exclusion criteria were used for this study in order to ensure safety and standardize the study population. These exclusions were: the presence of an ophthalmological condition other than refractive error, baseline overall accuracy at the MSPA task of <50% or >95%, a history of significant psychiatric illness, a history of unstable cardiac disease and the routine use of certain concomitant medications, which included antidepressants, tranquillizers, lithium, antiepileptics, antiparkinsonians, muscle relaxants (i.e. baclofen, dantrolene, tinazidine), systemic anticholinergics, calcium entry blockers, antiemetics, betahistamine, cinnarizine, piracetam and hormone replacement therapy. The use of migraine prophylaxis was also not permitted in this study.
Clinical characeristics of the patient groups
Following screening and the exclusion of unsuitable subjects, the remaining subjects were invited to attend for testing. Prior to testing the migraine diagnosis for each subject was confirmed by the second author using the International Headache Society (IHS) diagnostic criteria (26). Any diagnostic uncertainties were resolved by correspondence between the second author and the third author, who is a neurologist.
The visual acuity of each volunteer was assessed using a Snellen chart prior to commencing TMS. Local Research Ethics Committee approval was obtained from the Morecambe Bay LREC for this study and all subjects gave written informed consent prior to testing.
Stimuli and apparatus
Visual stimuli consisted of low-contrast letter trigrams presented centrally within a frame. The letters were presented in upper case Helvetica, font size 48. The purpose of the frame was to help equalize any crowding effect on the letters and improve their legibility (27). The letters used were chosen from a subset of letters of approximately equal legibility (28). Visual stimuli were presented on a Dell computer monitor with a measured refresh rate of 12 ms, driven by a Dell desktop computer running SuperLab software (Cedrus Corp., Phoenix, AZ, USA). For all volunteers the contrast and brightness of the monitor were set at 50 cd/m2.
TMS was administered using the MagStim SuperRapid (The Magstim Company, Cardiff, UK). A 90-mm circular coil was used in this study, which gives a maximal output of 2 T. All volunteers were asked to wear a tightly fitting EEG skull- cap, which had been stripped of its metal-containing electrodes and wiring and had an orientation grid marked on the back. The grid was marked symmetrically over the occipital area of the scalp. The inion, nasion and periauricular points were used as reference points in order to ensure correct fitting of the cap. The orientation grid covered a rectangular area ranging from 4 to 10 cm superior to the inion and 4 cm from the midline on either side; stimulation points were marked on the EEG cap at 2-cm intervals along the X and Y axes.
Procedure
All subjects were tested interictally at least 24 h after their last migraine attack. The letter trigrams were presented on screen for 24 ms and the subjects were asked to report verbally the letters in their correct order; the experimenter then recorded this. If subjects were unable to report any letters they were asked to say ‘blank’ or ‘don't know’. A magnetic pulse was given following presentation of the letters at varying intervals. Testing consisted of three phases.
Phase one—familiarization
All subjects were asked to complete a number of practice trials before magnetic stimulation was commenced. The purpose of these practice trials was to ensure that the volunteers were able to see the letter trigrams clearly. Subjects were first shown three trigrams, which were each presented for 100 ms. Following this, subjects were shown seven trigrams at a presentation speed of 24 ms and finally 10 trigrams at 24 ms. Subjects were allowed to continue to the next phase of the study provided they reported at least 85% of the trigrams correctly.
Phase two—threshold determination
Having ascertained that subjects were able to see the trigrams when presented for 24 ms, subjects were then shown trigrams with a TMS pulse administered 100 ms after presentation of each trigram (SOA). The coil was positioned 6 cm above the inion. As a safety precaution, the SuperLab program was set to ensure an interval of 5 s between the administrations of each TMS pulse. Previous work in our laboratory has found that the 100-ms interval provides peak suppression in volunteers (12). The magnetic stimulation was commenced at 60% stimulus intensity and increased in 5% steps until the subject was unable to identify two out of the three letters presented correctly and in the right order. Once this level of suppression was achieved the stimulus intensity was fine tuned by adjusting the level in 1% steps. Once the optimal level of stimulus intensity was achieved, the coil was moved around the grid to exclude a coordinate with a lower intensity at criterion. A previous study of phosphene levels using a similar method found interindividual variability required the scanning of the occipital area to determine the optimal stimulation point (11).
Phase three—time course of suppression
In phase three subjects were presented with 54 trials in which the letter trigrams were followed by the TMS pulse at a variable interval. Five SOAs were examined—40, 70, 100, 130 and 160 ms—with volunteers completing nine trials at each SOA. The presentation of trials was randomized for each volunteer within the SuperLab program. Throughout phase three the level of intensity used was set at 85% and the point of stimulation used was as determined in phase two. As in the previous phase, an interval of 5 s was observed between each pulse and the subject attempted to report verbally the letters as they were presented, with the experimenter recording their responses.
All subjects were followed up approximately 48 h after testing to record incidence of migraine and other adverse events.
Results
Suppression threshold
Stimulation thresholds for perceptual suppression took the form of percentages of maximal stimulator output. Mean thresholds were 88% for the MA group, 84% for the MoA group and 84% for the C group. These percentages were entered into a univariate analysis of variance (
Time course of suppression
Raw data from each subject took the form of the percentage of letters correctly reported at each SOA. Mean percentages by diagnosis and SOA are shown in Table 2. In our previous study using the MSPA technique (12) we analysed these raw percentages; however, this approach is potentially problematic in that it does not take account of baseline accuracy at the letter perception task. A suppression ratio was therefore calculated at each SOA by dividing the percentage correct at that SOA by the maximum percentage correct at any SOA. Profiles of suppression ratios across SOA are shown for each group in Fig. 1. Suppression ratios were then entered into a univariate analysis of variance (
Profiles of supression ratios (expressed as percentages) across stimulus onset asynchrony (SOA), by diagnostic group. ▪, Migraine with aura; □, migraine without aura; •, control.
Uncorrected mean percentages of letters correctly identified, by diagnostic group and stimulus onset asynchrony (SOA)
Two follow-up
Adverse events
Four patients reported the onset of a migraine attack within the 48-h follow-up period. All attacks were reported as being normal in nature and resolved as expected with or without treatment. Three other volunteers had minor symptoms of neck stiffness, disorientation or nausea immediately following testing; these symptoms also resolved quickly without treatment.
Discussion
Although the design of the two studies differed somewhat, the results of this study replicate those of Mulleners et al. (12), in that MA patients exhibited a flattened MSPA profile that was significantly different from the profiles of control subjects. The differences between the two studies are threefold. First, in contrast to our previous study where we used a MagStim 200, a MagStim Super Rapid was employed to perform the experiments for the current paper. The biphasic pulse produced by the MagStim Super Rapid may have a different effect on the visual cortex (29). Second, as we noticed that in some individuals, even in the absence of stimulation, target recognition never reached 100% accuracy, we decided to express the task performance as suppression ratios to correct for individual differences (such as variation in contrast sensitivity) that are not affected by the magnetic stimulation. Third, MoA patients were investigated for the first time using the MSPA technique and found to have profiles that were similar to those of controls.
Several features of the experiment are noteworthy. The MSPA technique is non-invasive and generally well tolerated by both patients and control subjects. The technique also allows an assessment of the status of visual cortex in migraine patients that is objective: patients simply have to report the presence or absence of target letters. The timing and randomization of stimulus materials is computerized, making it hard if not impossible for responses to be deliberately manipulated. Finally, the performance of MA is superior to that of control subjects: this would again be difficult to explain in terms of hypochondriacal responding on the part of patients. In general, therefore, the MSPA technique offers an assessment tool for migraine that appears to be both objective and reliable.
Despite several efforts to maximize recruitment, the number of subjects in each group was rather limited, notably in the MA group. However, as in our pilot study (where we tested seven MA subjects), the effect in even the smallest (MA) group was robust and statistically significant. Since this paper is the first to report similar MSPA profiles in MoA and controls, independent replication in a substantial number of subjects is warranted.
It has been argued recently that changes in cortical evoked potentials may occur several days before an attack and may thus hamper reliable TMS assessments unless timing-to-attack is controlled for (30). Two TMS studies designed to address this issue failed to confirm both the influence of attack timing and menstruation on TMS-derived excitability parameters and therefore we decided not to exclude the data of the four subjects that reported headache on follow-up (31, 32).
In line with our previous paper, we found no significant differences in the threshold TMS pulse intensity required for perceptual suppression between diagnostic categories. This may be due to a lack of statistical power: as the parameter was not of primary interest for this study, the number of data at each stimulus intensity was limited, resulting in a high variation around the mean. It is notable, however, that the somewhat higher threshold in MA patients is at least consistent with their MSPA profiles. We intend to explore the relationship between threshold and profile measures in detail in future studies.
The physiological basis for the difference in MSPA profiles between MA patients and controls has been the subject of some discussion. Randomization of the SOA across trials ensures that the profile is not merely an artefact of anticipation of a potentially bothersome TMS pulse. Further, it is unlikely that the MSPA profiles are the result of transient after-images for several reasons. The letter trigrams are presented at low contrast for a short duration, without colour contrast, so retinal after-images are minimized. Although after-images after fundoscopy are more frequently observed in migraineurs and thought to represent increased sensitivity of the visual system, no difference are reported between MA and MoA (33). We (5, 12) have previously suggested that the decrease in suppression seen in MA patients is a result of an impairment of GABA-ergic inhibitory neural networks in the visual cortex of these patients, resulting in functional hyperexcitability. Others (e.g. (19)) argue that the visual cortex is hypoexcitable in migraine. A recent animal study helps to resolve this debate. Moliazde et al. (34) stimulated cat primary visual cortex (Brodmann's area 17) with a magnetic stimulator, while simultaneously undertaking microelectrode recordings from single cells within the same area. Although technically challenging because of the large artefact caused in the single-cell recordings by the magnetic stimulation, their results clearly indicated that both spontaneous neural activity and activity evoked by visual stimuli are suppressed immediately after TMS stimulation (with intensities above 50% of maximal output). Moliazde et al. conclude from their impressive study: ‘It is likely that suppression of activity is the result of activated local inhibitory circuits’ (p. 677). As GABA is the predominant inhibitory neurotransmitter in the visual cortex (35), we are increasingly confident that the MSPA technique reflects the functioning of GABA-ergic inhibitory networks in this brain area. As inhibitory function declines, it becomes harder to recruit inhibitory cells and the net result is that visual perception is more difficult to disrupt.
The interpretation of flattened MSPA profiles in MA as reflecting an impairment of GABA-ergic intracortical inhibition appears at first glance to conflict with the recent work implicating glutamate in the pathophysiology of the aura (2). It is clear, however, that a variety of neurotransmitter systems may be implicated in the aura and may operate at different phases of the initiation, spread and resolution of an episode of CSD. For the initiation of CSD, disruption of GABA-ergic systems may be of crucial importance. A little-cited early paper by van Harreveld and Stamm (36) shows that CSD may be induced in rabbit visual cortex by directing light flashes into the animal's eyes, so long as the cortex is rendered hyperexcitable. This was achieved simply by intravenous administration of pentylenetetrazol, a GABA-A receptor blocker; no other chemical or mechanical stimulation of the cortex occurred. More recently, in the rat, minimal interference with inhibition by application of low-dose bicucculine, a GABA-A receptor blocker, has been found to produce spontaneous CSD in cortical slices (37). This finding is consistent with various studies showing that tissue may be preconditioned for CSD by reducing chloride concentrations in the extracellular space (38). It is therefore reasonable to suggest that a GABA-ergic deficit may be important for the initiation of an episode of CSD, while decreased clearance of glutamate from the extracellular space may provide the conditions under which a CSD wave may propagate across the cortex.
It is clearly the case from this study that the MSPA profiles of MoA patients are similar to those of controls. This finding is consistent with our previous work, in which a psychophysical index of cortical inhibitory function revealed differences in MA patients (compared with matched controls) but not in MoA patients (10). We next speculate on possible reasons for this finding. If there is truly a pathophysiological difference between MA and MoA, it is possible that CSD is simply not a part of the sequence of events constituting an attack of MoA. Alternatively, it is possible that CSD does occur in MoA, but in a cortical region or regions that remain clinically silent with the passage of the CSD wave (39). One interesting, albeit preliminary, way of examining this hypothesis would be to use the MSPA technique to assess MA patients with somatosensory aura (but not visual aura). In this group of patients, it might be expected that MSPA profiles would be normal, but that appropriate tests of the somatosensory cortex would reveal functional hyperexcitability. By analogy, therefore, the MSPA technique may be examining a portion of cortex that is irrelevant to the attack sequence in MoA patients. Evidently, a considerable continuing research enterprise will be necessary to clarify the ‘whether’, ‘where’ and ‘when’ of CSD in both MA and MoA. We feel that focused functional neuroscientific methods such as the MSPA technique will have an important part to play in this enterprise.
Acknowlegements
This work was partialy supported by a grant to E.P.C. from The Dr Hadwen Trust for Humane Research. We thank Dr Kentaro Hayashi for statistical advice.