Abstract
We would like to address the points raised in the Viewpoint article entitled ‘Can cortical spreading depression activate central trigeminovascular neurons without peripheral input? Pitfalls of a new concept’ by Burstein et al. (1).
We take issue with each of the six points raised in the Viewpoint and give detailed rebuttals to each. Firstly, however, it must be pointed out that it appears to us that Burstein et al.’s (1) overriding assertion is that we claimed in our publication that migraine pain can occur ‘without peripheral input’. We have never believed this and have spelled out why in previous articles (2). There certainly cannot be migraine headache in the absence of any trigeminovascular traffic. Rather what is at issue here is whether the pain of migraine is always the result of the generation of increased traffic in the periphery or whether it could also be the result of the modulation of existing traffic by a descending central mechanism. We believe that the experiments which we have carried out over several years strongly favour the latter interpretation. The core of our idea is that changes in cortical excitability (which in migraine with aura seems to be associated with cortical spreading depression – CSD) act to disinhibit trigeminovascular neural transmission so that existing traffic from the periphery is perceived to be painful. Cortical excitability changes and the resulting disinhibition would occur in all forms of migraine and not just those associated with CSD. Perhaps all we migraine researchers have placed too much emphasis on CSD as a tool and this emphasis may have obscured the fact that other ‘triggers’ of migraine, such as bright flashing light produce similar loss of descending inhibition, as we have previously shown (3–5). CSD could certainly produce its effects on trigeminovascular sensation by peripheral means, but it is hard to invoke these mechanisms to explain the effects of light flash. This idea is not new in its base concept, being first articulated by the gate control theory of Wall (6). What is new in our ideas is that this particular ‘top-down’ influence of the cortex is selective for dural sensation and hence selective for the pain of migraine only.
‘First’ point (≡‘signals are merely noise’). We reject this suggestion. We have more than 30 years experience in recording from single neural elements in monkeys, cats and rats. We were, in fact, the first laboratory ever to record trigeminovascular neuronal activity in the CNS, as far back as 1979 (7,8). The publications just cited described and illustrated the existence of a second-order trigeminovascular neuron with mechanoreceptor input from the superficial temporal artery. The activity of such neurons was linked to the pulsations of the artery and to the ECG; this activity was enhanced by an intracarotid injection of bradykinin and abolished by local application of lignocaine.
These signals are most certainly neural action potentials from perikarya. The signals which form the basis of our paper were fed through a window discriminator of our own design, which has both vertical (voltage) and horizontal (time 0.5–1.2 ms) windows. This device therefore filters out signals which are too-short, too-long, too-small and too-large, eliminating all forms of artefact, including many forms of electrical noise, signals arising from fibres of passage and stimulus-induced artefact. Only those signals which fulfilled the predetermined criteria were accepted into the study. In addition, in creating the consequent post-stimulus histograms (PSTHs), we used an automated process, using Visual Basic software to strip out any artefacts which have made their way through the filtering process. It should also be noted that the counts of ‘action potentials’ in the original paper were counts only of the window discriminator output and not of all the underlying spikes that can be seen in the graphics produced by the software. These window discriminator ‘spikes’ generally arose from only a small number of well-discriminated perikarya, often only that of a single neuron.
Our confidence that these are, in fact, neural signals from perikarya can be tested by the following ‘Q&A’, derived from our experiments in cats and rats, as far back as the 1979 work cited above. These are criteria we have applied, and do apply, to our work since the first experiments in 1979.
Do the signals disappear in animals which die unexpectedly during the course of the experiment? YES. Do the signals occur anywhere and everywhere? NO. Our original work in cats showed that field potentials associated with single neuronal signals were restricted to the dorsal horn of the trigemino cervical complex (9). Are the signals time-locked to stimuli? YES. This is obvious in the figures of the original paper and of all of our previous work. Do the signals occur after mechanical stimulation (where there is no electrical noise or stimulus artefact) as well as after electrical stimulation? YES. This was clear in the original paper. Are the signals accelerated by the iontophoretic application of glutamate (i.e. do they arise from neuronal perikarya)? YES (10). Can the signals be abolished by trigeminal nerve section and by application of local anaesthetic to the dura mater? YES. The work cited above (7,11) showed that application of lignocaine to a cranial artery blocked both stimulus-induced and vascular-linked responses of the second-order neuron which was being examined. These tests were carried out in the early years of our research and even included experiments in decerebrate animals where all parts of the sensory path from the dura (which was left intact) to the trigeminal ganglion could be manipulated surgically or with drugs. Such experiments were a necessary control to ensure that the ‘responses’ to dural stimulation were ‘real’ and not the consequence of some other mechanism, such as by direct cortical activation. Are the signals elicited only from restricted areas of the skin and dura mater? YES. The work which attracted the Viewpoint article gave summary details of the cutaneous receptive fields involved. In experiments in cats, where mechanical stimulation of the dura mater was used extensively (12), dural receptive fields were often quite restricted.
‘Second’ point (≡‘only 100% block of stimulus- induced responses will prove the case’). It is not necessary to achieve complete blockade of stimulus-induced responses to show that there must have been some other mechanism to account for the persistence of CSD-induced acceleration of trigeminovascular neurons. Although the responses to lignocaine were variable, there were neurons in which the ongoing discharge rate was reduced to below 10% of control and, perhaps more significantly, there were neurons in which the evoked responses were abolished entirely, indicating complete blockade in the trigeminal ganglion. Even in these neurons CSD still produced a rise in discharge rate. In many instances, the acceleration induced by CSD was, in fact, enhanced when the induced-responses were completely blocked.
‘Third’ point (≡‘inadequate assessment’). Burstein et al. (1) have a point in criticising the use of only single time-locked stimuli to test trigeminovascular traffic integrity. However, there would seem to be no way in which one could produce a long-lasting (many minutes) stimulus, such as by the application of capsaicin, without producing irreversible effects in the periphery. The use of single shocks to assess responses of central trigeminovascular neurons is a tried and true method and is used very widely. In our original paper, especially with the control experiments with lignocaine alone, we could choose to monitor, measure and display simultaneously either ongoing
Fourth point (≡‘WDR is not the whole story’). This is valid insofar as it goes, but it ignores the fact that these neurons received dural input that was activated by dural mechanical, chemical and inflammatory stimuli. They were also activated by light flash and CSD- two stimuli which are followed by migraine pain in humans. Although C-fibre neurons may behave differently, we cannot ignore the body of evidence (some of it from the Burstein et al. (1) laboratory) that A-δ WDR neurons play a significant role, and one that is affected by ganglion block and has an alternative explanation in a CNS pathway. We emphasise that any effect of the brainstem nuclei on sensory processing in the trigeminal nucleus would be exerted by an effect on the receptors (possibly 5-HT receptors) on presynaptic terminal fibres (10) and not on the second-order neuron itself. If so, then the relative importance of C-fibre versus A-δ fibres in contributing to the changes which occur in migraine would be dependent on the populations of 5-HT receptors on each. On this question, nothing much is clear, save that 5-HT1D receptors seem to predominate on each, but especially on fibres descending from the dura mater (14).
Fifth point (‘≡ subjective judgement’). We carried out a range of statistical tests during data acquisition (e.g. Poisson distribution-fitting) and post-experimentally, e.g. Kolmogorov-Smirnov and critical ratio tests (which can be done ‘on the fly’). Variance tests were applied post hoc on set periods as Burstein et al. (1) suggested, i.e. 5 minutes ‘before’ and ‘after’ the application of a protocol and these all show the same consistent results. Our PETH and PSTH software is particularly useful here because it applies a variance analysis to pick the points of change, rather than relying on an operator’s perception. In any case, the rate changes induced by CSD were sometimes astoundingly high and produced vanishingly small p values. It was often hard to squeeze all the data sets of these changes into the non-logarithmically scaled diagrams.
‘Final’ point (‘≡reference to our previous work which did not show an effect of CSD’). This harks back to an old (15,16), and still-running (17), debate about whether the methods used to induce CSD are in themselves painful and not the consequence of cortical neural activity changes. In their ‘Viewpoint’, Burstein et al. (1) use the fact that we did not find that CSD induced neuronal activity changes in our previous study (18) which took place when this debate first began (15). The reviewers of our paper also raised this question and we were able to satisfy them on this point. We stand by our old results and the conclusions which we drew from them. However, that work was in cats, in which CSD is much more difficult to induce and may well be different from rats. The cerebral cortex of cats is most certainly different from that of rats and this appears to have an effect on the global propagation of CSD. The use of CSD as a migraine model in lissencephalic animals such as the rat is a fraught one, but a convenient one. We should not extrapolate too far from it.
In our laboratories we have moved from being steadfast adherents of the vascular theory of migraine, towards the view that migraine and the consequent pain have mostly neural origins and mechanisms. We came to this change quite reluctantly but we could not see a way forward on the basis of a peripheral vascular theory, a theory which has overwhelming animal-based evidence in its favour, but almost no pathology-based evidence in migraineurs (2). We are greatly encouraged by the recent startling results of Oshinsky et al. (19), which demonstrate the existence of what appears to be a model of primary headache in rats that seems to favour a neural, rather than a neurovascular, generation mechanism. We hope that pursuit of alternative central neural mechanisms may eventually lead to findings in humans which will resolve the issue of a ‘missing pathology’. There is already encouraging evidence that this is starting to happen (20,21).