|Home | About | Journals | Submit | Contact Us | Français|
The migraine attack is characterized by alterations in sensory perception, such as photophobia or allodynia, which have in common an uncomfortable amplification of the percept. It is not known how these changes arise. We evaluated the ability of cortical spreading depression (CSD), the proposed mechanism of the migraine aura, to shape the cortical activity that underlies sensory perception. We measured forepaw- and hindpaw-evoked sensory responses in rat, before and after CSD, using multi-electrode array recordings and 2-dimensional optical spectroscopy. CSD significantly altered cortical sensory processing on a timescale compatible with the duration of the migraine attack. Both electrophysiological and hemodynamic maps had a reduced surface area (were sharpened) after CSD. Electrophysiological responses were potentiated at the receptive field center, but suppressed in surround regions. Finally, the normal adaptation of sensory evoked responses was attenuated at the receptive field center. In summary, we show that CSD induces changes in the evoked cortical response that are consistent with known mechanisms of cortical plasticity. These mechanisms provide a novel neurobiological substrate to explain the sensory alterations of the migraine attack.
Migraine is a multi-system disorder involving transient but severe disruptions of homeostasis and sensory processing. Beyond craniofacial pain, sensation is altered during the course of the migraine attack. The aura that can precede a migraine is a frank sensory hallucination, and the photophobia, phonophobia, and allodynia that accompany the attack are painful amplifications of visual, auditory, and tactile perception.
The mechanisms of sensory modulation during the migraine attack are poorly understood. Changes in sensation take minutes to develop, but can persist for hours or even days, suggesting both cortical and subcortical network alterations. The cerebral cortex is integral to conscious sensory experience, and demonstrates more intrinsic plasticity than lower structures in the sensory network (Ahissar et al., 2000; Foeller and Feldman, 2004; Feldman and Brecht, 2005). However, there has been relatively little examination of cortical sensory networks in migraine models.
Cortical spreading depression (CSD) is thought to be the physiological correlate of the migraine aura (Charles and Brennan, 2009). CSD is a propagating wave of neuronal and glial depolarization, involving extracellular voltage changes larger than during seizures (Somjen, 2001). Hemodynamic changes are equally large, with hemoglobin desaturation to levels encountered during ischemia (Chang et al., 2010). Such profound disruptions could be inferred to have conditioning effects on both cortical and subcortical structures, which in turn might contribute to the migraine phenotype.
We tested the effects of CSD on sensory processing in vivo, using a combination of 2-dimensional optical spectroscopy (2dOS), planar multi-electrode array recordings, and naturalistic stimulation. We observed a sharpening of cortical sensory maps, potentiation or suppression of evoked potentials depending on location relative to the sensory map, and reduced sensory adaptation, all of which could contribute to the altered sensorium of the migraine attack.
Adult male Sprague-Dawley rats (n = 11, weight 481+/−80 g) were used in accordance with the University of California at Los Angeles Animal Research Committee guidelines. Anesthesia was induced with 5% enflurane and maintained at 1–2 % for the duration of the experiment, in a 2:1 nitrogen:oxygen mixture. An 8×5 mm2 cranial window in the right parietal and frontal bones (1 mm lateral to midline) exposed the somatosensory cortex. An 8-channel planar electrode array was positioned into the cranial window after retraction of the dura (Figures 1,,2).2). Trials began at least 1 hour after the electrode array was placed onto the cortical surface.
Temperature was monitored and maintained with a rectal probe and homeothermic blanket. A subset of animals (n=4) underwent tracheotomy and ventilation with humidified air. Systemic hemoglobin saturation and PETCO2 were monitored. All systemic parameters remained within a physiologically normal range throughout the experiments (T 36.9–37.2°C; HR 296–350 bpm; SpO2 88–96%; PETCO2 3.6–5.0%) and there was no significant change during or after CSD (linear mixed-effects model with post-hoc Tukey test; p≈1 for all comparisons).
Stainless steel stimulating electrodes (A-M Systems) were placed in the glabrous skin of the forepaw (FP) and hindpaw (HP); return electrodes were inserted into the fore- and hindlimbs. Trials of alternating FP and HP stimulation were performed for 1–1.5 hr with each trial lasting 35–60 seconds. Either paw was stimulated at 5 Hz for 2 seconds, beginning 5 seconds after the start of each trial. Monopolar current pulses were 1.0 ms in duration and 1.0 mA in intensity. Stimulation occurred prior to and after CSD, which was induced by placing a single potassium chloride crystal at the edge of the cranial window. Residual potassium chloride was immediately washed from the cranial window with saline when spontaneous field potentials were eliminated (within 30–60 seconds after application).
An 8-channel microfabricated planar platinum electrode array (Theriot, 2011) was used to detect evoked field potentials from somatosensory cortex. A stainless steel reference electrode was placed in the temporalis muscle. Field potentials were amplified 1000×, sampled at 1 kHz, bandpass filtered at 10–300 Hz, and notch filtered at 60 Hz. Summed evoked potentials (sEP) (Figure 1) were generated by summing all evoked discharges during the stimulation period, and quantified by taking the magnitude of the negative peak of the summed waveform (Lauritzen, 2001). Spontaneous peaks (amplitude >5 S.D. from baseline) occurring during the 5-second pre-stimulus period for each trial were also sized and counted.
The cortex was epi-illuminated with white light from a voltage regulated halogen source. Light reflected from the cortical surface was divided into four parallel paths (Figure 2) by an image splitter (QuadView, Photometrics) and captured by a cooled CCD camera (PhotonMax 512-B, Princeton Instruments). Every 512×512 image contained four sub-images, each filtered at 560, 568, 577, or 610 nm (+/− 5 nm). A single trial consisted of 30–50 seconds of imaging with a 5–10 second inter-trial interval. Images were acquired at 2 Hz with an exposure time of 400 ms and 1.0× or 0.5× magnifications. The approximate pixel resolution of an image at 1.0× was 16 μm/pixel. Optical data sets consisted of interlaced FP and HP trials, each 60–100 frames. Hemoglobin changes (Δ[HbO2], Δ [HbR], Δ [HbT]) were determined by analyzing the registered multi-wavelength imaging data at each pixel using a modified Beer-Lambert law:
where is the pre-stimulus light reflectance for wavelength λ and Iλ(t) is the reflectance at any pixel over the timecourse of each trial. In vitro phantom models (Sato et al., 2002; Sheth et al., 2004) were used to determine the wavelength-dependent absorption coefficients, αλ.
To generate hemoglobin moiety maps, the median value of all pixels in each frame was plotted and the time point with the largest deviation from the pre-stimulus baseline was determined. The image corresponding to that time point was chosen to represent the peak level of activity for that dataset. This image was thresholded at 50% maximum response to generate the map. Time series of hemoglobin moiety change were generated from averaged pixels within the map. To compare hemodynamic and electrophysiological data, circular optical regions of interest (ROI) were chosen to coincide with each electrode. The resulting timecourses were typically mono-phasic, with a negative deflection for HbR or positive deflections for HbO2 and HbT (Figure 2).
Data were pooled into pre- and post-CSD time bins, each 10 minutes in length. The pre-CSD time period is bin o. Post-CSD bins i–vi cover 60 minutes of post-CSD trials. The first trial with a summed evoked potential greater than 50% of the maximum post-CSD response marked the beginning of bin i. Trial reductions for electrophysiological and optical datasets were identical.
Results in text are means +/− standard deviations. Comparisons were made using a linear mixed-effects model. Variation between animals was treated as random. Differences between time bins were determined using a post-hoc Tukey test for multiple comparisons. Box plots are used to show data distributions: median, upper and lower quartiles, and range exclusive of outliers (data outside +/−2.7 standard deviations from the group’s mean) are depicted. Data reduction and summary statistics were done using Matlab. Tests for significance were performed using R (R Language and Environment for Statistical Computing, http://www.r-project.org).
Pre-stimulus hemodynamic measures were taken over exposed cortex adjacent to each electrode (Figure 2A). Baseline hemodynamic and field potential measures were homogeneous across the array; though there were large changes in all locations after CSD, there was no significant difference between locations (linear mixed-effects model with post-hoc Tukey test; p≈1 for all comparisons). There was a long-lasting hypoperfusion after CSD that returned to baseline within 60–90 minutes. Consistent with prior reports (Piilgaard and Lauritzen, 2009; Chang et al., 2010) there was no significant difference in HbT from baseline at 80 minutes post-CSD; 3/4 animals with post-CSD recordings ≥70 minutes recovered to baseline. Spontaneous electrical activity was undetectable for 7.6+/−4.2 minutes following the CSD wave. Afterward, both the frequency and amplitude of spontaneous discharges were significantly lower than before CSD, returning to baseline with approximately the same kinetics as HbT (no significant difference in spontaneous discharge count from baseline at 80 minutes post-CSD; 4/4 animals with post-CSD recordings ≥70 minutes recovered to baseline) (Figures 2B,,33).
Evoked activity was also suppressed during and immediately after the passage of CSD, taking 5.1 +/−2.5 minutes to reach amplitude greater than the half maximum of all post-CSD responses. In contrast to the global suppression of spontaneous electrical activity, evoked responses either increased or decreased in amplitude after CSD, depending on the electrode location and which stimulus (FP or HP) was presented. Figure 4A shows a histogram of the ratio of post-CSD (bin i) to pre-CSD (bin o) amplitude for spontaneous and evoked activity over all electrodes. Unlike spontaneous activity, whose post/pre-CSD amplitude ratio never exceeded unity, summed evoked potential (sEP) activity showed a post-CSD amplitude increase in 18.8% of electrodes during FP stimulation and 34.1% of electrodes during HP stimulation. Six of 11 animals during FP stimulation and 10/11 of animals during HP stimulation had at least one electrode that recorded higher post-CSD i amplitudes.
The locations of potentiated electrodes varied by stimulus (Figure 4B, Figure 5A). When FP was stimulated, potentiated electrodes corresponded to the anatomical location of FP cortex; when HP was stimulated, potentiated electrodes were located in HP cortex (Franklin and Paxinos, 1997). In other words, the same electrode could show either increased or decreased sEP amplitudes, depending on which stimulus was presented.
Hemoglobin maps allow an independent measure of the boundaries of evoked cortical activity (Figure 5B). We used post-CSD i Δ[HbT] maps to test whether an electrode would show potentiation. There was a significantly greater likelihood that potentiated electrodes were located within the boundaries of hemodynamic maps, and conversely, that the suppressed electrodes were located outside map boundaries. Δ [HbT] maps predicted potentiated electrodes correctly 89% of the time for FP (6% false positive, 5% false negative) and 86% of the time for HP (6% false positive, 8% false negative; Fisher’s exact test, 2 tailed: p=3.9×10-7 and p=3.0×10-7 for FP and HP, respectively).
We used Δ [HbT] map classification to further examine electrode behavior over time. Electrodes within hemodynamic maps showed a significant increase in sEP amplitude, and those outside a significant decrease, which lasted at least 60 minutes after passage of CSD. The change in electrode behavior (potentiation or suppression to either FP or HP stimulation) was preserved in the majority of electrodes for the whole duration of even the longest experiments (Figure 5C,D). For experiments (n=4) with 70 minutes or more of post-CSD recording, 2/11 potentiated electrodes and 0/47 suppressed electrodes recovered to baseline values. Moreover, amplitude of potentiated and suppressed electrodes remained significantly different from baseline at 80 minutes post-CSD (p<0.001 for each, linear mixed-effects model with post-hoc Tukey test).
Somatosensory responses typically depress on repetitive train stimulation, and this sensory adaptation can be modulated by changes in network activity (Moore, 2004). In contrast to the adaptation seen before CSD, there was a relative maintenance of evoked potential amplitude throughout the train after CSD (Figure 6). Though there was no difference in the amplitude of the first evoked potential before and after CSD, evoked potentials 2–10 were significantly increased in amplitude compared to the pre-CSD period, and most maintained the increase in amplitude for the duration of even the longest experiments (11/14 electrodes in 4 experiments with post-CSD recordings ≥70 minutes).
Cortical sensory map sharpening (defined as a reduction in the area of an evoked response) (Feldman and Brecht, 2005), has been observed in sensory plasticity paradigms (Prakash et al., 1996; Polley et al., 2004). Sharpening was seen in both electrophysiological and hemodynamic maps after CSD. To quantify electrophysiological sharpening, electrodes that responded within 50% of the maximum response for a particular stimulus were counted. Between pre-CSD o and post-CSD i time periods, there was a 42.9% and 29.3% reduction in the number of electrodes that were within half-height of the maximum response, for FP (10/11 animals) and HP (11/11 animals) respectively (Figure 7). The area at half-height of post-CSD hemodynamic maps was also reduced. The reduction in sensory map area occurred for both stimulus modalities and all hemodynamic signals, and was approximately concentric to maps of pre-CSD vascular activity (Figure 8). Both hemodynamic and electrophysiological sharpening persisted in long-duration experiments (maps continued to have a reduced area after >80 minutes, 4/4 experiments).
We measured the hemoglobin response amplitude within FP and HP maps. In contrast to evoked potential data Δ [HbO2], Δ [HbR], and Δ [HbT] all exhibited a global suppression of evoked response amplitude after CSD (Figure 8). In contrast to baseline (pre-stimulus) hemodynamic signals (Figures 2B,,3D),3D), evoked hemoglobin responses remained suppressed for the duration of even the longest experiments. Comparison of evoked potentials with hemoglobin responses from regions of interest immediately adjacent to electrodes revealed a reduced slope in the evoked neurovascular coupling relationship (Figure 9).
Beyond the aura, the migraine attack involves alterations in sensory perception. CSD, the proposed mechanism of the migraine aura, is brief but leads to long-lasting changes in cortical perfusion, metabolism, and electrical response (Piilgaard and Lauritzen, 2009; Chang et al., 2010). We reasoned that CSD might alter sensory processing, in ways that might explain the sensory symptoms of a migraine attack.
We observed a long-lasting increase in receptive field center and decrease in surround amplitude in response to paw stimulation. These sensory-map-specific changes in evoked potential amplitude persisted longer than whole-field hemodynamic and electrophysiological changes (Piilgaard and Lauritzen, 2009; Chang et al., 2010) (Figures 2B, ,3D),3D), suggesting their mechanisms were distinct. Such long-lasting alterations resemble synaptic plasticity phenomena such as long-term potentiation and long-term depression (LTP, LTD)(Bliss and Lomo, 1973; Malenka and Bear, 2004).
Changes consistent with both LTP and LTD have been induced by CSD in slice models (Footitt and Newberry, 1998; Berger et al., 2008); both potentiation and suppression of evoked responses have been observed in different in vivo studies (Guiou et al., 2005; Piilgaard and Lauritzen, 2009; Faraguna et al., 2010; de Souza et al., 2011). Our results reconcile these apparently contradictory findings by showing that potentiation and suppression depend on location relative to stimulated receptive field. Classical LTP and LTD can also be induced in somatosensory neocortex in vivo (Foeller and Feldman, 2004). The simultaneous potentiation of center responses and suppression of surround responses might be explained by heterosynaptic long-term depression (LTD), which is induced in inactive neurons whose neighbors undergo LTP (Buonomano and Merzenich, 1998).
Purely neural mechanisms may not be the only explanation for changes in evoked potential amplitude after CSD. Astrocytes also undergo depolarization and calcium activity during CSD (Sugaya et al., 1975; Chuquet et al., 2007), and they contribute to synaptic plasticity in vitro and in vivo (Halassa and Haydon, 2010). Astrocytic release of glutamate, K+, ATP, and neuromodulators could favor excitatory transmission on arrival of sensory impulses. Indeed, astrocytic release of d-serine is essential to hippocampal LTP (Henneberger et al., 2010). Alternatively, the metabolic challenge of CSD and its aftermath might impair astrocytic reuptake of glutamate and K+ from the extracellular space, once again favoring excitatory transmission.
We observed a post-CSD reduction in the area of evoked electrophysiological and hemodynamic FP and HP maps, which (like evoked potential amplitude changes) persisted beyond the recovery of previously characterized post-CSD changes. To our knowledge, sensory map sharpening has not been reported after CSD. However, it is frequently seen in sensory plasticity paradigms such as environmental enrichment (Polley et al., 2004) and sensory deprivation (Foeller et al., 2005). Importantly, sharpening is seen on short time scales: a prominent example of this is the sharpening of sensory maps on arousal (Ferezou et al., 2006).
The decreases in surround amplitude that cause sharpening could be caused by synaptic plasticity mechanisms resembling LTD. They could also be explained by an increase in surround inhibition in proportion to center potentiation after CSD. Inhibitory sharpening is a known mechanism of plasticity after sensory deprivation (Foeller et al., 2005). Moreover, such plasticity is possible on the short time scales we observed: significant changes in the excitatory/inhibitory postsynaptic potential ratio have been seen in lateral interactions less than one hour after a peripheral sensory lesion (Hickmott and Merzenich, 2002).
Hemodynamic sensory maps also show sharpening on both acute and chronic time scales (Prakash et al., 1996; Polley et al., 2004), and we observed a dramatic sharpening of hemoglobin maps after CSD. It could be argued that this sharpening could be due to the attenuation in evoked hemodynamic response: a reduced amplitude response would be expected to have a smaller footprint. While we cannot rule out this possibility, we suspect that the reduction in area of activation in our hemodynamic maps is a reflection of electrophysiological sharpening.
After CSD, we observed a sustained attenuation of the normal decremental response of somatosensory cortex to repetitive stimulation. As with sensory map sharpening, sensory adaptation changes have not previously been shown after CSD, but they make sense in the context of known mechanisms of sensory plasticity.
Adaptation to repetitive stimuli is a fundamental response of sensory cortices; its role appears to be the modulation of gain (Chung et al., 2002; Castro-Alamancos, 2004; Moore, 2004). Cellular mechanisms of sensory adaptation are incompletely understood, but for rodent somatosensory cortex short-term synaptic depression is involved (Chung et al., 2002; Heiss et al., 2008). We recorded from the cortical surface, which makes it most likely that we sampled synaptic activity generated in layers I–III, and our observations are consistent with the adaptation shown in layer II/III neurons to somatosensory stimulation (Ahissar et al., 2001; Chung et al., 2002).
Altered inhibition might also contribute to adaptation changes after CSD. Paired-pulse inhibition is attenuated after CSD (Krüger et al., 1996), and the inhibitory component of transcallosal evoked potentials is reduced in amplitude (Piilgaard and Lauritzen, 2009). However, adaptation is not modified by bicuculline (Nelson, 1991) or accompanied by changes in membrane potential or input resistance (Chung et al., 2002), suggesting inhibition may not play a major role.
CSD causes massive disruptions in perfusion, metabolism, neurovascular coupling, and even tissue structure (Piilgaard and Lauritzen, 2009; Chang et al., 2010). It is reasonable to infer that the sensory evoked changes we observe might be artifacts of these underlying alterations. While it is clear that neural activity and the structural/metabolic milieu are intertwined, we suspect our observations must rely (at least in part) on neural plasticity. The most important reason is the specificity of our responses. Post-CSD changes in perfusion, metabolism, tissue swelling, DC potential, and spontaneous electrical activity are global in nature. In striking contrast, the changes in our evoked sensory responses are spatially, temporally, and modality specific. It is difficult to explain these effects based on pan-cortical changes in perfusion, astrocytic function, or tissue structure and resistivity (Makarova et al., 2008). A second reason is the nature of the changes in sensory response. Potentiation, suppression, sharpening, and adaptation are all known mechanisms of plasticity in sensory systems (Malenka and Bear, 2004; Moore, 2004; Feldman and Brecht, 2005), whose substrates are incontrovertibly neural, and which occur independent of CSD. Finally, the kinetics of the evoked changes we observe appear different from our measures of spontaneous activity (which are comparable to previously measured post-CSD changes (Piilgaard and Lauritzen, 2009; Chang et al., 2010)) – potentiation, suppression, sharpening and adaptation persist beyond the recovery of spontaneous hemodynamic and electrophysiological activity.
The migraine attack is defined by noxious sensory amplification (International Headache Society Classification Subcommittee, 2004; Lipton et al., 2008). Previous studies in animal models have shown how alterations in subcortical excitability, particularly in trigeminal ganglion (TG), trigeminal nucleus caudalis (TNC) and thalamus, might contribute to the enhanced sensitivity of the migraine attack (Strassman et al., 1996; Burstein and Jakubowski, 2004; Burstein et al., 2010; Noseda et al., 2010; Zhang et al., 2010, 2011; Lambert et al., 2011). Our work shows how migraine-related sensory amplification might be caused by mechanisms of plasticity in the sensory cortex. It also suggests how post-CSD cortical plasticity might explain alterations in excitability of lower structures, with cortex directly modulating TNC and thalamus activity through corticofugal connections (Noseda et al., 2010). Though speculative, this is an appealing explanation because TG and TNC do not typically demonstrate significant plasticity in adult animals. In contrast, the cortex, and to a lesser extent the thalamus, retains plasticity during adulthood (Foeller and Feldman, 2004; Feldman and Brecht, 2005), and corticothalamic networks can respond to stimuli with state- and gain-changing output within seconds (Nicolelis and Fanselow, 2002).
Given the common anatomical and functional features of primary sensory cortices (Markram, 2010), our observations are likely generalizable beyond limb sensation, to facial sensation, hearing, and vision. The fact that we elicited similar responses in two different limb cortices (one used for exploration, the other much less so) is strong evidence that other sensory regions might show a similar response.
Finally, because our experiments use naturalistic sensory stimulation, they are directly comparable to human work using non-invasive imaging and electrophysiological techniques. Interestingly, a recent magnetoencephalography study shows potentiation of the P100m response in patients with persistent aura – a finding completely consistent with our results (Chen et al., 2011).
We have used naturalistic stimulation and distributed optical and electrical recordings in vivo to demonstrate that CSD, the proposed mechanism of the migraine aura, induces cortical plasticity mechanisms that could explain the sensory alterations of the migraine attack.
This work was supported by NSF IGERT (JJT), NIH R01 MH52083 (AWT), NIH NS059072, NS070084, and the NIH Loan Repayment Program (KCB).
Conflict of Interest:
None of the authors have any conflict of interest to declare
Author contributions:JJT, AWT, NP, and KCB conceived the experiments; JJT, NP, YSJ and KCB designed the experiments; JJT performed the experiments; JJT, YSJ and KCB performed analysis; JJT and KCB wrote the paper.