The Discovery of a New Drug Class for the Acute Treatment of Migraine


  • Patrick P.A. Humphrey PhD, DSc

  • From the Theravance, Inc., South San Francisco, CA, USA.

  • This manuscript is an edited version of a transcript of a lecture given to the New England Center for Headache in Greenwich, CT, November 2004.

Address all correspondence to Dr. Patrick P.A. Humphrey, Theravance, Inc., 901 Gateway Boulevard, South San Francisco, CA 94080, USA.


The history of the scientific ideas and events that led to the discovery of sumatriptan is outlined with personal reminiscences about individuals who influenced the approach. The development of sumatriptan revolutionized the acute treatment of migraine and led to the availability of a number of other triptans. The anti-migraine effects of all the triptans are mediated via 5-HT1B, and possibly 5-HT1D receptors, which transduce their effects via G; proteins. This suggests that agonists at other Gi protein-coupled receptor types appropriately located (eg, somatostatin sst2, adenosine A1) should be examined for their effects on the trigeminovascular system, Studies on such receptor targets may provide insight into a novel approach towards the design of new anti-migraine drugs.

One of the great neurologists in the field of headache and migraine was Harold Wolff (Fig. 1). He was a pioneer of scientific experimentation on headache in the clinic.1–3 It was his work that helped establish the foundation for the advances in antimigraine drug discovery research made to date. Unfortunately, Wolff's work is somewhat ignored today because modern thinking has moved away from his proposed vascular theory of migraine and more toward a neurogenic theory.

Figure 1.—.

Harold G. Wolff, MD (1898-1962).

As to the discovery of sumatriptan, it started back in 1972 when David Jack, the Research Director of Glaxo at Ware, set me on a mission to find an antimigraine drug. I was naïve to the science of migraine, having previously worked in a Physiology Department in a medical school (St. Mary's, London), where I studied skeletal muscle electrophysiology. At Ware I was provided with an empty lab and a graduate assistant, Eira Apperley. At the outset, we studied the clinical literature and found that Wolff proposed that migraine was a vascular disease because he was able to measure pulsations in the superficial temporal artery that correlated temporally with the headache.4 When migraine patients received ergotamine intravenously, the pulsations subsided, as did the pain. One of the reasons this does not impress many researchers today is because of the involvement of the superficial temporal artery, which is not a typical vascular involvement for many migraineurs. However, James Lance's laboratory in Australia reproduced Wolff's observations and reported that about one-third of migraineurs exhibited this type of profile.5 The key point with these studies was the observation that the ability of ergotamine to abort an attack correlated with reduced temporal artery pulsations.

Furthermore there is a wealth of data to show that vasodilators can induce migraine attacks—for example, we know that alcohol can induce a headache for many migraineurs.6 The concept that vasodilation would induce an attack and vasoconstriction would abort an attack was a reasonably well-accepted concept in the 1970s. We also learned that it was not just ergotamine that aborted a migraine attack, but other vasoconstrictors such as noradrenaline could do the same.7 Of particular interest was the fact that 5-hydroxytryptamine (5-HT; serotonin) could also be added to this list.8 This was not only shown in the United States, but Lance's group in Sydney were also able to demonstrate the action of serotonin in aborting a migraine attack.9

Such findings led to the idea that serotonin might have a potential pathophysiological involvement in migraine and to extensive research on the possible role of serotonin in migraine. Studies showed that the platelet 5-HT content decreased during an attack10 and Sicuteri was the first to show that the metabolite, 5-hydroxyindolacetic acid, was present in urine following an attack.11 However, despite numerous follow-up studies the role of serotonin in the etiology of migraine is still unclear even today.

Regardless, at this point it was clear to me that 5-HT was probably acting as a vasoconstrictor on the cranial vessels outside of the brain because I thought it very unlikely that a low dose of exogenous 5-HT would get into the brain and abort an attack by acting centrally. This was based on the knowledge that peripheral 5-HT underwent rapid removal by uptake and metabolism, and that the blood-brain barrier would readily prevent 5-HT from penetrating to the CNS. So I assumed that 5-HT was acting as a vasoconstrictor of cranial vessels outside of the blood-brain barrier.

Around this time Lance's group published a comparative clinical study showing that methysergide was more effective in migraine prophylaxis than a number of other 5-HT receptor antagonists.12 I questioned why this was the case and what was different about methysergide. In discussion with James Lance he told me that patients given methysergide sometimes benefited acutely with headache abrogation, which was unusual for an antagonist drug. So I became very interested in what methysergide did and its precise pharmacological characteristics.

At the same time, Pramod Saxena reported vascular changes in anesthetized dogs given methysergide.13 He reported that intravenous methysergide produced a very localized cranial vasoconstriction; methysergide had no effect on blood pressure, but it reduced the arterial flow in the external carotid bed. We subsequently repeated these studies in our laboratory. Indeed, we also found that the total carotid flow was reduced following administration of methysergide and there was no change in driving pressure. Therefore, these studies suggested that there must be vasoconstriction somewhere within the carotid circulation. Interestingly, blood flow in the femoral arterial bed increased. We went on to show that the cranial vasoconstriction produced by methysergide was mediated by activation of a then unknown vascular 5-HT receptor type, now known as the 5-HT1B receptor.14 The action by methysergide in the femoral bed is also mediated by 5-HT1B receptors, but for this action the receptor is located on nerve terminals, not blood vessels, turning off sympathetic drive and thus neuronally causing vasodilatation.14 So, it was Pramod Saxena, another migraine research pioneer, who in describing vascular changes in an animal model with methysergide, influenced our subsequent experimentation and drug discovery activities in the field of migraine.

This led us to spend several years of work where we focused on characterizing 5-HT receptors throughout the body in various species. We concluded that there was a new receptor, the 5-HT1 receptor, that was activated (not blocked) by methysergide and located on cranial blood vessels.14–16 We predicted that a selective prototypic agonist for the 5-HT1 receptor would constrict cranial vessels but have little or no effect on peripheral vessels, which contain predominantly 5-HT2 receptors. At this time we also knew about the 5-HT3 receptor. Thus we thought we had discovered the third known 5-HT receptor type when we identified the 5-HT1 type. Little did we know then (in a premolecular biology era) there are more than a dozen types or subtypes, such that the 5-HT1 receptor type comprises 5 subtypes, the 5-HT2 comprises 3 subtypes, not to mention the other 5-HT receptor types now known to exist. In fact, we were aware of the existence of the 5-HT7 receptor well before it was named because some of our potential experimental antimigraine drugs actually caused vasodilation, rather than vasoconstriction, in the cranial circulation because they simultaneously activated the 5-HT7 receptor which mediates marked vasodilation.14

In designing a new antimigraine drug, we knew we needed to retain the desirable actions of 5-HT that aborted the attack experimentally, which we believed were mediated by 5-HT1 receptor activation. We also had to eliminate the undesirable effects of serotonin. We knew that 5-HT2 receptor activation leads to all sorts of undesirable things such as platelet aggregation and bronchial constriction, among others. The 5-HT3 receptors are on afferent nerves and mediate depolarization, so again, we wanted to avoid activating these receptors. So we set out to make an analog of 5-HT that would selectively stimulate the 5-HT1 receptor. To some extent we were lucky that it did not stimulate all these other 5-HT receptor types that were identified later at the molecular level. So, we developed sumatriptan, which is a relatively simple analog of 5-HT, but remarkably selective for the 5-HT1B receptor. As predicted it produced selective vasoconstriction in certain isolated blood vessels. When we exposed isolated extracerebral cranial vessels to sumatriptan, they constricted; however, it had little or no effect on peripheral vessels.16,17 Our next steps were to evaluate sumatriptan fully in vivo. Having confirmed its selective carotid arterial vasoconstrictor action in the dog,17,18 we needed to extend this work and study 5-HT1 receptor agonist effects on blood flow and vascular resistance in important vascular beds simultaneously throughout the body.

Wasyl Feniuk and I went to Pramod Saxena's laboratory to learn the radiolabeled microsphere technique, which was a rather novel technique in those days. There we injected a prototype of sumatriptan into a cat animal model, and we looked at its effects on all the important vascular beds. We later performed similar experiments with sumatriptan back in our own laboratory, also testing the effects of ergotamine (30 μg/kg), which was found to increase vascular resistance in all the vascular beds, as would be expected of a nonselective vasoconstrictor.19 In marked contrast sumatriptan only constricted the arteriovenous anastomoses, which in the anesthetized cat model are predominantly in the carotid circulation.19 Excitingly for us, sumatriptan, unlike ergotamine, had no effect on vascular beds supplying the kidneys, adrenals, the heart, or the brain.

So why should a drug which selectively constricts carotid arteriovenous anastomoses work in migraine and how could we persuade clinicians to evaluate sumatriptan as a potential, acute treatment for a migrainous headache? With regard to this step, we owe gratitude to Hartwig Heyck, another clinical experimentalist, because he proposed the “carotid shunt” theory of migraine.17,20 Although the carotid shunt theory was never well accepted, it was fundamental to our advancement of testing sumatriptan in humans. Heyck collected jugular venous blood during a migraine attack and demonstrated that, on the side of the headache, blood had a higher oxygen content in the external venous jugular drainage than normal.21 This implied a unilateral opening of arterial carotid shunts. In fact, there are some data to suggest that there is significant shunting in the meningeal circulation, and maybe the meningeal circulation is important pathophysiologically in migraine. It was these ideas that led us to working with the German clinicians, Professor Doenicke, and Dr. Brand, who first examined the clinical effects of an earlier prototype of sumatriptan.20,22 These studies were followed by additional larger studies in the United States that clearly showed clinical effects of sumatriptan, which benchmarked this drug as a breakthrough medicine for the acute treatment of migraine. The first 2 large clinical studies were published in 1991 in the New England Journal of Medicine by the Subcutaneous Sumatriptan International Study Group, led by Michel Ferrari, and in JAMA by Roger Cady and his colleagues.23,24

Historically, to reflect back on how we discovered sumatriptan, it was not a complete and uneventful success story from the start. We began working on this project in late 1972. We first discovered a compound (AH24167) that had been predicted to be a vasoconstrictor in the carotid circulation, but it caused vasodilation instead. At this time, some within the team thought the working hypothesis was wrong, and we had to spend about 18 months in research and discovery to identify the 5-HT7 receptor and then get the chemists realigned to find AH25086, a more selective agonist.14,20 AH25086 was unsuitable for oral administration, so we pressed on to get sumatriptan. So that is the history of the drug. It became available for clinical use in 1991 and was initially launched in Holland. Sumatriptan became available in the United States in 1993. Fig. 2 is a photograph of key opinion leaders with the late Princess Margaret during a meeting at Leeds Castle in Kent, in 1995, one of a number of international meetings to discuss the new drug's mechanism and what we had learned.

Figure 2.—.

Migraine Experts and the Late Princess Margaret. Leeds Castle, Kent, England in June, 1995.

So how did sumatriptan work? Many did not accept the idea of vasoconstriction being central to migraine, so Mike Moskowitz proposed a cranial “neuronal extravasation” theory.25 This was the first indication that sumatriptan might have a direct neuronal effect in aborting an attack. In several studies he stimulated the meninges in rats and guinea pigs, so activating neuronal release of neuropeptides which cause extravasation.25 He showed that following sumatriptan administration, there is an inhibition of the effects of meningeal stimulation. These studies led to the concept that perhaps sumatriptan was blocking neuronally mediated protein extravasation, an indication of a possible neuronal antimigraine effect.26

Mike Moskowitz allowed us to come to his laboratory and learn how to set up the model and evaluate it for ourselves. The problem I pointed out from the outset was that 5-HT does not work in the model and yet clinically 5-HT can abort an attack. This illustrates the problem with many animal models of disease. With animal models, it is important that a specific mechanism is studied and that one does not just measure an “all or none” efficacy parameter in a “black box” disease model. Predictably many compounds identified in this model by various companies went on to fail in the clinic. It was later found that the key neuropeptide mediator involved in this animal model was not CGRP but it was the neurokinin, substance P.27 This provides a basis for the disconnect between the animal model and the known clinical pathophysiology of migraine where CGRP, not substance P, is implicated.

However, Moskowitz's early studies provided the first indication that we should be thinking about the possibility of neuronal inhibitory mechanisms. This notion was supported by Peter Goadsby's studies that showed an increase in the neuropeptide, CGRP, in external jugular venous blood during a migraine attack, a finding not mirrored in cubital fossa blood.28 So it appeared that something was happening in the cranium that causes neuronal release of CGRP during a migraine attack. If sumatriptan was administered it could abolish both the release and the headache but this in itself did not definitively confirm a neuronal inhibitory mechanism as the primary clinical mode of action.29

Another debate at the time revolved around the fact that distended blood vessels do not necessarily cause a migraine attack. For example, sitting in a hot bath causes vasodilation, but does not cause a migraine attack. However, in the study published by Nichols and colleagues, they showed that inflating an intravascular balloon catheter in different regions of the middle cerebral artery could produce a headache in specific regions of the head.30 Upon removal of the balloon catheter, the headache subsided; this could be repeated 1 week later. If cranial vascular distension really can cause headache, the question remains as to why some people get migraine and others not and today we may have some clues as to why that should be so.

In 1991, Jes Olesen and colleagues measured blood flow in the middle cerebral artery and reported that there was no change in blood flow when sumatriptan was administered to a migraineur during an attack, which was very reassuring in those early days.31 However, there was an obvious change in local blood velocity during a migraine attack, which indicated that the conducting vessel had increased in size; sumatriptan decreased it and restored the vessel diameter to normal.31 This provided rather convincing evidence that during migraine, there is distention of the blood vessel walls and this could be reversed by sumatriptan. The question remains as to whether or not it is the vasodilatation per se that is causing the pain.

However, others strongly believed that the clinical mode of action of sumatriptan involved a neuronal inhibition. Goadsby and colleagues mechanically stretched the sagittal sinus in a cat model and measured c-fos expression in the dorsal horn at the level of the trigeminal nucleus caudalis.32 I suggested this mechanical distension approach to get away from concepts of vasodilatation and vasoconstriction and directly attempt to measure a neuronal inhibitory action. This study showed an increase in c-fos expression in the trigeminal nucleus caudalis following mechanical distention of the sagittal sinus. They also reported c-fos expression in C1 and C2, and none in C3, demonstrating the very specific neuronal innervation. Importantly, following sumatriptan administration, there was a significant decrease in c-fos expression, although it was not possible to decide whether the effects of sumatriptan were central or peripheral. However, it seems likely that the predominant action of sumatriptan was on the peripheral terminal endings. Goadsby and colleagues also provided some evidence that supports this, where following electrical stimulation of the sagittal sinus there was no inhibitory effect of sumatriptan on evoked potentials at the C2 level.33 However, after infusion with high concentrations of mannitol, which disrupts the blood-brain barrier, sumatriptan was effective in inhibiting evoked potentials. So this evidence from a cat model suggested that sumatriptan does not normally act centrally.

Subsequent studies showed the newer triptans were active centrally. Thus, naratriptan is more lipophilic and has a central inhibitory effect in the cat model without disrupting the blood-brain barrier.34 This seems true of other triptans too which can be shown to have a central neuronal inhibitory effect, but this does not mean that this is how they are exerting their antimigraine action clinically.

The question still remains as to how all the triptans are working to abort the migraine. Once we knew a lot about serotonin receptors and we could study them, it became clear that all of the triptans activate 5-HT1B and 5-HD1D receptor subtypes. So sumatriptan is a 5-HT1 receptor-selective compound as we showed originally because it does not activate all the other 5-HT receptor types. Most of the triptans, but not all, also stimulate 5-HT1F receptors. Certainly 5-HT1B and 5-HT1D receptors appear to be important. We know that the 5-HT1B receptor is virtually the only 5-HT1 receptor subtype in blood vessels. So, the question is could a 5-HT1D or a 5-HT1F receptor agonist be used to avoid the vasoconstrictor profile mediated via the 5-HT1B receptor? The studies are inconclusive to date. There has been 1 published study to my knowledge that evaluated the effects of a selective 5-HT1D agonist in migraineurs and that drug proved to have no efficacy.35 The potential role of 5-HT1F receptor agonism is also unclear and whether agonists have been tested that are selective enough remains to be seen.36

There were still many unanswered questions about triptans and their mechanism(s) of action.

  • • It is still unclear whether selectively targeting 5-HT1D or 5-HT1F receptors would provide superior efficacy and tolerability over the current triptans.
  • • We still do not know why sumatriptan, when given subcutaneously or even intravenously, is much more effective than any triptan orally. It would seem to be something to do with the rate of rise in plasma concentration. I am sure if we looked at that in more depth we would get a better understanding and it might teach us something.
  • • We do not know whether the vasoconstrictor action is just important or whether it is essential for terminating a migraine attack.37 It certainly seems to be a feature of all the triptans.
  • • We do not know the role central 5-HT receptor activation plays in terminating migraine. Most of the triptans seem to be able to get into the brain, but whether that is how they act is another matter; it is not obvious to me that sumatriptan works in this way. Although there is the possibility that during a migraine attack there may be localized changes in the permeability of the blood-brain barrier.

What we do know about the mechanism of action of triptans is that:

  • 1They cause selective extracerebral intracranial vasoconstriction.
  • 2They inhibit trigeminal nerve terminals innervating extracerebral vessels.
  • 3They potentially produce central neuronal inhibition in the nucleus caudalis.

The 5-HT1B receptor, which is what we identified from our dog work years ago, mediates all 3 of these actions. 5-HT1B receptors not only mediate vasoconstriction but they also mediate neuronal inhibitory effects as we showed on sympathetic neurons years ago. So in order to specifically address what mechanisms are important and what receptors are responsible for specific actions, we will need to develop a selective agonist each for the 5-HT1B, the 5-HT1D, and the 5-HT1F receptor and find out.

So, where are we today in trying to integrate our studies on blood vessels and the early focus on the pharmacology of cranial blood vessels with what we know now about pain pathways and the idea of a “migraine generator” in the brain? I think Rami Burstein's current concepts of sensitization of meningeal sensory afferent pathways provide an explanation that is most attractive.38,39 Thus we can assess individual migraineurs and in some cases identify cutaneous allodynia, toothache, jaw ache, and other related symptoms. Maybe in such cases there is second-order neuronal sensitization, although we still have to work out how this occurs and there are many questions not answered including what is abnormal about these pathways and what might be happening pathophysiologically? My favorite notion is that central sensitization may arise from a disorder of descending inhibition, which would actually fit with the idea of a hypothetical migraine generator or something that is abnormal in the brain. Also, if there is altered withdrawal of descending inhibition unilaterally, it would explain why there is hemicrania. Clearly there is something different about a migraineur's brain and its pathways, we just have not yet been able to ascribe defects in specific neuronal pathways to account for the abnormal neuronal activation that leads to headache.

There is some experimental evidence linking neuronal sensitization to the pathophysiology of migraine and the action of the triptans. One such study was done by Richard Hargreaves and colleagues,40 where they measured the middle meningeal artery diameter and recorded neuronal firing in the region of the trigeminal nucleus caudalis. Following an injection of CGRP they recorded an increase in vessel diameter (vasodilatation) as we might have expected. They also stimulated the whiskers in an anesthetized rat model and recording centrally. When CGRP was administered and vasodilatation occurred, they noted sensitization of the inputs from the whiskers. This sensitization actually extended beyond the period of the vasodilatation. This group also very nicely showed that the triptans demonstrated inhibitory effects on the afferents from the blood vessels, not on the afferents from the whiskers, demonstrating remarkable selectivity of the action of the triptans. Afferent impulses from blood vessels are likely to be important in headache, whether they are the “windup” inputs or whether they are just the neuronal input that through sensitization becomes the nociceptive “painful” input remains to be seen. Perhaps ongoing studies with the new CGRP antagonists will help elucidate these different mechanisms. One scenario may be that when perivascular afferent nerves become sensitized, they release CGRP, which causes vasodilatation. Through sensitization this vasodilatation mediates painful afferent firing, and blocking the vasodilatation may be enough to reduce afferent neuronal inputs and abort the headache and associated symptoms.

Another approach might be to evaluate