Abstract
For the past 25 years, it has been hotly debated whether cortical spreading depression (CSD), the underlying mechanism of aura, can trigger migraine headache through activation of the trigeminovascular pain pathway. According to one school of thought, CSD leads to a cascade of molecular events (1) that activate pain fibers in meningeal areas overlying the affected cortices (2). According to another school of thought, the aura (and therefore CSD) and the headache are coincidental events with no causal relationship (3). In the past 2 years, this debate has been put to rest by a wealth of scientific evidence that CSD is a noxious event that activates the trigeminovascular pathway (4,5).
The recent publication by Lambert and colleagues opens yet a new debate on whether the activation of the central trigeminovascular neurons by CSD depends on peripheral input from meningeal nociceptors or central input from the cortex (6). In their study, the authors suggest that activation of brainstem trigeminovascular neurons by CSD can occur independently of peripheral input (6). This stands in sharp contrast to a large body of evidence for the critical role of peripheral nociceptors in the activation of central trigeminovascular neurons (4,5,7–11). Settling this issue is therefore critical, as it is relevant to the very biological underpinnings of migraine in general, and to the relationship between the aura and headache in particular. Unfortunately, as experts in the experimental techniques used in this study, we believe that the veracity of the findings are questionable, for a number of reasons.
First, the recordings are not single units, as claimed, or even multi-units, but instead appear to reflect background electrical noise (e.g. Figure 5 lower trace; multi-unit discharge appears only after the second arrow, but the rate meter registers high levels of activity before that point, even though no discharge is visible). Even the top panel in Figure 1, which is vastly better than the one shown in Figure 5, shows that firing of several neurons are registered into the data analysis system as coming from only one neuron (6).
Second, the blockade of peripheral input to the spinal trigeminal nucleus is incomplete. For the study to be valid, a complete blockade of transmission of all action potentials from the dura to the spinal trigeminal nucleus should have been achieved. Only then could the authors conclude with any kind of certainty that the activation of brainstem trigeminovascular neurons by CSD was partially independent of peripheral input (1). As shown in Figure 2 (6), lignocaine injections in the trigeminal ganglion did not eliminate all neuronal responses to stimulation of the dura and skin. At that point, the study should have been stopped or redesigned as it could no longer refute the possibility that the activation seen in the spinal trigeminal nucleus after the induction of CSD was mediated solely by the peripheral input from meningeal nociceptors, whose activity, by the way, was increased by CSD in a similar manner (5). (Scientifically, there are two ways to explain the incomplete blockade of peripheral input from skin and dura to the spinal trigeminal nucleus: (a) incomplete inhibition of the trigeminal ganglion by the lignocaine injection, especially the difficult-to-reach neurons in the ophthalmic division of the ganglion, and (b) peripheral input from intact sensory neurons in the 2nd and 3rd dorsal root ganglia.)
Third, the blockade of peripheral input to the spinal trigeminal nucleus is assessed inadequately. The inhibition of sensory signal transmission in the trigeminal ganglion, as presented in Figure 2, was assessed by skin and dural stimulation that induced responses that lasted 10–40 hundredths of a single second (Figure 1 bottom, histograms) (6). These responses are too short (e.g. shorter than an eye blink reflex) and thus cannot reflect the true effects of any drug intervention. An adequate way to assess the lignocaine effects in the trigeminal ganglion would have been to show a continuous (120 min) recording of single-unit activity in response to quantitative stimulation that induced firing that lasted many seconds or even minutes (e.g. noxious heating of the skin or capsaicin on the dura).
Fourth, the exclusive sampling of wide-dynamic range (WDR) neurons introduces sampling bias that leads to type II error. For their conclusions to be meaningful and interpretable, the authors (6) should have studied high-threshold (HT), low-threshold (LT) and WDR neurons in different laminae of the spinal trigeminal nucleus, as well as neurons whose activity is driven by Aδ and, even more importantly, C-fiber input (4). As pointed out by the authors, they only studied WDR neurons in deep laminae (we do not really know which laminae as no histology is provided) whose activity was driven by Aδ fibers (6). For this reason, they cannot refute the possibility that other classes of neurons (HT, LT), especially those that receive input from C-fibers and those located in laminae I–II, are activated solely by the peripheral input from meningeal nociceptors (simply because they did not study them).
Fifth, the lack of a time-controlled protocol by which changes in firing of all studied neurons are measured over one time period (e.g. 5 or 10 min) and at the same time points after CSD or lignocaine injection (e.g. 15, 30, 60, 120 min) renders the statistical group analysis (Figure 4) invalid because the comparison is based on the experimenter’s subjective judgment rather than objective predetermined parameters (6). (As stated in the paper (6), ‘CSD was initiated with a cortical KCl micro-injection and recording continued until no further changes in discharge rate of the neuron occurred. At this point, 10 µg lignocaine was injected slowly into the trigeminal ganglion and recording continued until no further changes in discharge rate of the neuron occurred.)
Based on these observations, it is difficult to reconcile the results reported in this study (6) with those in the authors’ recent report that CSD does not activate trigeminovascular neurons in the spinal trigeminal nucleus (12). It is likely that sampling error led them to find no activation in the first study and activation that is partially independent of peripheral input in the current study.
Finally, the authors (6) cite a study that shows that CSD activates meningeal nociceptors in the trigeminal ganglion (5). Given that the central axons of these neurons terminate in several laminae of the spinal trigeminal nucleus, the rationale for proposing that this functional anatomical circuit does not drive activity in central neurons is unclear. This issue is especially bewildering because the latency for activation by CSD in the spinal trigeminal nucleus was similar to the latency for activation of meningeal nociceptors in the trigeminal ganglion. If the activation of central trigeminovascular neurons by CSD is in fact induced by direct (or even indirect) cortical projections to the spinal trigeminal nucleus, one would have expected it to occur within a second, rather than the 10–20 min shown in this and the Zhang et al. (5) studies.
Regarding the last point, we are in full agreement with the attention given by Lambert and colleagues to the possibility that cortical projections to the medullary dorsal horn have an important role in trigeminal nociception (6). In support of this concept, Noseda and colleagues (13) showed recently that laminae I–II neurons in the spinal trigeminal nucleus receive direct facilitatory input from the insular cortex, whereas laminae III–V neurons receive direct inhibitory input from the primary somatosensory cortex. Synthesizing these data with all previous studies that showed activation of central trigeminovascular neurons by CSD, it seems reasonable to propose that the initial activation of these neurons is achieved through incoming signals from (peripheral) meningeal nociceptors, whereas the magnitude and duration are determined by the direct and/or indirect inputs they receive from different cortical and brainstem areas.