Abstract
Salient aspects of the anatomy and function of the blood-barrier barrier (BBB) are reviewed in relation to migraine pathophysiology and treatment. The main function of the BBB is to limit the access of circulating substances to the neuropile. Smaller lipophilic substances have some access to the central nervous system by diffusion, whereas other substances can cross the BBB by carrier-mediated influx transport, receptor-mediated transcytosis and absorptive-mediated transcytosis. Studies of drugs relevant to migraine pathophysiology and treatment have been examined with the pressurized arteriography method. The drugs, given both luminally and abluminally, provide important notions regarding antimigraine site of action, probably abluminal to the BBB. The problems with the BBB in animal models designed to study the pathophysiology, acute treatment models and preventive treatments are discussed with special emphasize on the triptans and calcitonin gene-related peptide (CGRP). The human experimental headache model, especially the use of glycerol trinitrate (the nitric oxide model), and experiences with CGRP administrations utilize the systemic administration of the agonists with effects on other vascular beds also. We discuss how this can be related to genuine migraine attacks. Our view is that there exists no clear proof of breakdown or leakage of the BBB during migraine attacks, and that antimigraine drugs need to pass the BBB for efficacy.
Introduction
Migraine is a chronic episodic neurovascular disorder. The hypothesis that the migraine attack starts in the central nervous system (CNS) is supported by the migraine aura and the fact that 82% of patients have premonitory symptoms (1). The complex relationship between the CNS and peripheral nociception, including activation of sensory fibres and dilation of cranial vessels, remains to be elucidated. Migraine pain has been suggested to involve intracranial vessels, trigeminal sensory innervation and the reflex connection of the trigeminal system, and the cranial parasympathetic autonomic nerves (2–4).
Few frustrations in medicine can match that felt by the neurologist who holds a potent drug within inches of an infection or a brain tumour it cannot reach. The brain alone among the organs of the body remains inaccessible to chemotherapy. Although the walls of blood vessels elsewhere in the body are fenestrated, they offer free passage, whereas those of the brain are ‘sealed off’. Basically, the so-called blood–brain barrier (BBB) offers free passage to glucose and a few other selected substances, but fends of most drugs. Circulating catecholamines, indoleamines, peptides, inter alia, can, when present in the CNS, act via specific receptors; their free access to the neurons would have destroyed the specificity of the CNS. In support, all animals with a CNS have a BBB, illustrating the importance of the BBB for higher function.
Research in headache disorders has for many decades touched the interface between the CNS and peripheral nociception, but many questions remain. In this paper we will first provide a brief general overview of some physiological aspects of the BBB, then discuss existing data on the BBB in relation to migraine attacks, and finally relate this to different aspects of migraine treatment. Since our awareness of migraine pathophysiology is nowadays more focused on a process initiated within the CNS, we felt it topical to discuss current data on available drugs in relation to the BBB, with a secondary aim to shed light on their possible therapeutic target.
Blood–brain barrier in physiological conditions
The BBB is a unique feature of the brain circulation; it is a tightly tiled lining, a lipophilic membrane located at the endothelial cell layer in which the endothelial cells are connected with ‘closed’ or ‘fused’ tight junctions not allowing any bulk flow of water or solutes to pass, excluding most substances. The intracerebral capillaries and venules consist mainly of endothelium and pericytes, all surrounded by a basement membrane (5). In addition, there is often a dense contribution of astrocytes with end-foot processes around the intracerebral microvessels; these are suggested to play a coupling role between the synaptic activity and cerebral blood-flow and metabolism (6, 7). The basal lamina is an acellular membrane composed of type IV collagen, fibronectin and laminin interposed between endothelial cells and astrocytic end feet. The BBB is formed by the endothelial cells and the epithelial-like, high-resistance, tight junctions that join the endothelial cells of the capillaries of the brain microvasculature. The resistance across the brain capillary endothelial barrier is about 8000 Ω/cm2(8), which is as high as any membrane barrier has in biology. The electrical resistance across the pial vessels on the cortical surface (small arteries and arterioles) has a slightly different BBB, in which the resistance is somewhat lower, 1000–2000 Ω/cm2(9). Similarly, lanthanum, an electron-dense tracer, is able to penetrate into cerebral arteriolar walls and into extracellular compartments of the surrounding brain (10), indicating a less tight BBB in the cerebral arterioles than in brain capillaries. If a large molecule such as the classical Evans blue albumin is given, it does not pass the pial vessel wall. Due to the presence of the fused tight junctions in the CNS, no paracellular pathway has been found to occur through the inter-endothelial cleft that normally exists in capillaries in peripheral tissues. Pinocytosis, which is a transcellular pathway of free solute movement that takes place in the capillaries of peripheral tissues, is minimal in capillaries of the CNS. Since both the transcellular and paracellular pathways of solute movement across the capillaries are minimal or non-existent in brain microvessels, molecules in the blood gain access to the CNS across the capillaries only minimally; it is non-existent in brain microvessels.
Molecules in the blood may gain access to the interstitial fluid via one of the following two mechanisms: (i) free diffusion or (ii) catalysed transport. A number of proteins with transporter and regulatory functions have been identified (11). Two examples can be given: isoform I of the facilitative glucose transporter family is expressed solely in endothelial (and pericyte) domains, and 75% of the protein is membrane-localized in man. Occludin and claudin-5 (like other tight junction proteins) exhibit a restricted distribution, and are expressed solely within interendothelial clefts of the BBB. Monocarboxylic acid transporter and water channel (Aquaporin-4) expression is enriched at the glial-foot processes. Three basic categories of catalysed transport of solutes across the BBB exist, as follows: carrier-mediated transport, receptor-mediated transport and active efflux transport (Fig. 1).
Schematic illustration of how substances may pass the blood–brain barrier; passive diffusion, carrier-mediated transport influx transport, receptor-mediated transcytosis, absorption-mediated trancytosis, and carrier-mediated transport efflux transport.
Free diffusion of solutes across the BBB occurs if the molecule has lipid solubility and molecular weight under a 400–600-Da threshold (12). This assumes that the solute or the drug has minimal plasma protein binding and that it is not a substrate for numerous active efflux transport systems within the BBB. These systems consist of the superfamily ATP binding cassette (ABC) protein transporters that includes P-glycoprotein (P-gp), multidrug resistance-associated proteins (Mrp) and breast cancer resistance protein (6), as well as others that are present at the BBB. Expression and function of these drug transporters at the BBB have been widely investigated by measuring mRNA, protein and solute transport. However, much remains unknown about their distribution and function in the CNS (13). Expression of P-gp in brain microvessels has been observed, but not of Mrp2.
The main physiological purpose of the BBB is to maintain a constant microenvironment within the cerebral tissue. Due to their nature, hydrophilic substances such as sodium, hydrogen, bicarbonate and other ions cannot cross the BBB. There are, however, a few exceptions such as glucose, lactate and amino acids. These substances pass the BBB via a mechanism of facilitated diffusion (carrier-mediated). Thus, a number of important nutrient molecules are transferred across the BBB by this mechanism. The transported molecules cross the plasma membranes by interacting with intramembrane transporter proteins related to water-filled channels. The two systems with the highest capacity are that for C-glucose and certain other sugars (the gene product is Glut I), and that for large neutral amino acids, the so-called L-system. Glucose binds to the glucose carrier, and this complex can pass the lipophilic endothelial barrier. By this mechanism glucose can pass the otherwise impermeable membrane. This represents a true facilitated diffusion, which takes place in the direction of a concentration gradient and it is not an energy-requiring transport mechanism. This system has high transport capacity; 4 µmol min−1 g−1 in the rat and 1 µmol min−1 g−1 in man (14). The L-transporter has been sequenced (15) and shown to have binding sites for neutral amino acids. The transporter has clinical significance, since it carries the dopamine precursor L-3,4-dihydroxyphenylalanine (L-DOPA) (16), used in the therapy of Parkinson's disease. Basically, the cerebral endothelial cells contain DOPA decarboxylase, which converts L-DOPA to dopamine, supplying the brain with this transmitter and thereby alleviating the disease symptoms.
The BBB is so tight that filtration of water due to hydrostatic or osmotic pressure gradients is no larger than the diffusion of water (17). In contrast, the filtration is 50 times higher in skeletal muscle with fenestrated capillaries. The low filtration of water across the BBB adds to the stability of the chemical microenvironment in the brain. Lipophilic substances can more easily cross the BBB, e.g. gases such as carbon dioxide. Since the BBB is impermeable to bicarbonate but highly permeable to carbon dioxide, the perivascular pH in the brain will be determined by the bicarbonate concentration in the brain and the carbon dioxide tension in the blood. This is important, since the perivascular pH is considered a key factor in regulation of local cerebral blood flow (18).
A few active transport mechanisms may here be briefly mentioned. The potassium ion concentration in the brain is lower than in plasma because of the presence of an active transport mechanism for potassium. Movement into or through the cerebral endothelium of molecules of the size of peptides and proteins (e.g. insulin, transferrin, lipoproteins) may depend on receptor-mediated endocytosis. In addition, there is a potent transporter that restricts brain entry of a wide range of drugs. The P-gp is located in the apical (blood-facing) plasma membrane of the endothelium and is able to utilize ATP in pumping certain drugs from endothelial cells into the blood circulation. It also transports waste products and toxic substances out of the brain in order to avoid damage to the CNS by otherwise neurotoxic substances. In view of the important role of P-gp and other drug efflux transporters for drug distribution, the identification of compounds as substrates of P-gp-mediated transport represents a key issue in drug discovery and development, particularly for compounds presumably acting on the CNS. Overexpression of P-gp or members of the Mrp family at the BBB has been implicated in the mechanisms underlying resistance to antiepileptic drugs and in acute as well as prophylactic drugs in migraine. Therefore, it is important to know which drugs are substrates for P-gp or Mrp. In addition, there are data which indicate that substrate recognition or transport efficacy by P-gp differs between man and rodents. Such species differences, which are certainly not restricted to man and rodent, may explain, at least in part, controversial data that have been reported for transport by P-gp or Mrp2 for transport efficacy of efflux transporters.
Eletriptan is an example of an antimigraine drug that is transported out of the CNS by P-gp (19), whereas there are no reports of other triptans as substrates for this transport efflux system (20). Valproic acid was not found to be a substrate for P-gp or Mrp in vitro or in vivo (21), and the molecular identity of the putative transporter mediating the active efflux of valproic acid from the brain remains to be elucidated. There are no reports on candesartan being a substrate for the ABC transporter system. A study in knockout mice has suggested that topiramate may be a substrate for P-gp-mediated transport (22), whereas this is not the case for lamotrigin or gabapentin (22). Verapamil is a substrate for P-gp (23) and is an inhibitor of the protein (24). The β-adrenoceptor antagonists used in migraine treatment, propranolol and metoprolol, are not substrates for P-gp, and propranolol may even be an inhibitor (25).
A further aspect of cerebrovascular endothelial cells is their cytoplasm content of a comparatively high number of mitochondria, which gives an indication of their high metabolic and thus functional activity (26).
The endothelial cell is more than the anatomical interface between cerebrovascular smooth muscle cells and the blood; it is also involved in the expression of a variety of physiological and pharmacological effects on cerebrovascular smooth muscle. As is the case of endothelin-1 in the peripheral vasculature, the cerebrovascular endothelium produces potent vasoactive agents such as those of the endothelin family of peptides, angiotensin II, purines, and relaxing factors such as nitric oxide (NO), prostacyclin and endothelial-dependent hyperpolarizing factor (18, 27–29). Hence, the endothelial cells may participate in haemodynamic regulation by responding to mechanical forces generated by local blood flow alteration in shear stress. It has been proposed that the endothelium acts as a transducer of haemodynamic forces that controls the release of vasoactive agents such as endothelin or NO (30, 31). This would provide a partial explanation for the alterations in cerebrovascular tone that follow a change in arterial pressure (autoregulation).
The blood–brain barrier function in migraine attacks
The BBB may be disturbed in many conditions, such as following neoplastic lesions, hypertensive encephalopathy, inflammation, status epilepticus and stroke. However, we have never come across persuasive data describing transient opening of the BBB in physiological conditions or in migraine attacks with or without aura. Induced cortical spreading depression in rodents reveals a process that shows similarity to the aura (32). In animal studies, activation of matrix metalloproteinases (MMPs) following repeated cortical spreading depressions may result in opening of the BBB (33). Thus, theoretically BBB leakage could be present in migraine with aura. However, the degree of activation necessary for this resembles that seen in stroke (34, 35). In migraine with aura there is oligaemia but no ischaemia during the aura phase and early headache phase (36). Furthermore, diffusion-weighted magnetic resonance images (MRI) examined during the aura phase were normal in one study (37), indicating that there was no observable ischaemia present during the aura (37, 38).
In familial hemiplegic migraine that is severe and long-standing (39, 40) or in attacks of persistent symptoms (41), a gadolinium enhancement (indirect sign of enhanced passage) has been observed with MRI. In four patients with persistent migrainous visual disturbances gadolinium was administered and no enhancement was reported (42). Migraine with aura is seldom investigated with MRI during attacks, first, because there is no diagnostic uncertainty with typical aura, and second, because the aura phase is normally < 1 h (International Headache Society classification) and hence too brief to catch. There exist, however, a few series of MRI examinations during migraine aura (37, 43, 44). The emphasis has been on perfusion-weighted imaging, but a possible gadolinium enhancement during migraine aura was in one study investigated by inspection of the post contrast T1-weighted images (37); no enhancement was reported in four patients. Regional cerebral blood flow (rCBF) in the relevant occipital lobes decreased from 16 to 53% (37). In 20 migraine patients both with and without aura (44) no gadolinium enhancement was observed (Sanchez del Rio, personal communication). The mean decrease in occipital rCBF was 27% (range 14–35%) (44). In line with the results in these last two series showing no enhancement with gadolinium, our view is that the BBB is most likely intact during migraine attacks.
Consequences of preserved BBB for migraine treatment
Investigations of the blood–brain barrier with pressurized arteriography
The standard method for in vitro pharmacological investigation of vessels has for three decades been to hook up a vessel segment on two prongs, one fixed and the other connected to a strain gauge for recording of vasomotor activity (45). All components of the vessel wall are in this preparation exposed to the drug. An alternative method has been developed, namely the pressurized arteriography method (46–48). In this in vitro method large cerebral arteries such as the rat middle cerebral artery (MCA) or branches thereof are perfused via luminal glass micropipettes, allowing selective luminal and abluminal administration of substances or drugs. If a drug has an effect both luminal and abluminal it will usually be because the drug can cross the BBB of the artery provided the endothelium is intact (checked at the start of the experiments). This is the case for ergotamine and dihydroergotamine (DHE), which both result in quite similar contractile responses when given luminally or abluminally (49). In agreement, Goadsby found tritium-labelled DHE in the brainstem of cats following systemic administration (50). However, if a drug has an endothelium-dependent effect it can cause an effect without crossing the BBB. This is the case for sumatriptan, which in the rat results in relaxations (sic!) when given abluminally or luminally (51). It has been shown that luminal sumatriptan results in relaxation by an endothelium-dependent mechanism involving the release of NO. This shows that rat cerebral vessels are unsuitable for investigating the contractile effect of triptans because rats have mainly endothelial 5-hydroxytryptamine 1B (5-HT1B) (a weak contractile 5-HT1B receptor has been found in the smooth muscle cells) (52). In man, contractile smooth muscle 5-HT1B receptors dominate (53). In contrast, if a drug has an abluminal effect but shows no luminal response, it is because it cannot cross the BBB. Thus, calcitonin gene-related peptide (CGRP) [molecular weight (MW) 3700] relaxed the MCA after abluminal administration, but there was no effect when CGRP was given luminally (54). The CGRP blocker olcegepant (BIBN4096BS; MW 870) effectively inhibited CGRP-induced relaxation when given abluminally (CGRP was also given abluminally), but was mainly ineffective when given luminally (54). These two large molecules, CGRP and olcegepant, apparently do not cross the BBB of the rat MCA. However, when increasing the dose we observed that high doses of olcegepant can attenuate ablumial CGRP dilation. These results with migraine-relevant drugs and substances demonstrate the suitability of pressurized arteriography in migraine pharmacology.
Experimental animal models
Pathophysiological models
Intravenous (i.v.) dopamine has no effect on trigeminovascular neurons, whereas when given locally (iontophorized) an inhibitory effect on these neurons was seen, indicating a tight BBB to dopamine in this animal model (55). This fact is well known from the work of MacKenzie and McCulloch: only when the BBB was opened by osmotic shock could noradrenaline or dopamine pass the BBB and induce changes in cerebral blood flow (CBF) and metabolism (56–59). In studies of peripheral (60) and central sensitization (61) with chemical stimulation of the intracranial dura mater, the stimuli are administered outside the BBB because the meningeal arteries, and the dural microvessels as part of the dura, have no BBB. This fact has been discussed repeatedly during recent decades (62). It may of course be an advantage when designing experiments to examine the peripheral ramifications of the trigeminal nerve.
Another model is to stimulate meningeal structures and then examine various part of the trigeminovascular system. Thus, cells in the trigeminal nucleus caudalis (TNC) in the brain stem respond to stimulation of the superior sagittal sinus (SSS), the dura mater or sometimes even specifically the middle meningeal artery (MMA) with ramifications via the sensory C- and A-fibres (63); the TNC cells also respond to local l-glutamate and α-CGRP administrations (injected into the TNC). Microiontophoresis of olcegepant was found to inhibit the SSS-evoked stimulation activity in the TNC. In addition, i.v. administration of olcegepant in a fairly high dose inhibited SSS-evoked neural responses (63). Furthermore, activation of meningeal nociceptors resulted in activation of trigeminal neurons; only olcegepant in a very high (0.9 mg/kg) i.v. dose could inhibit neuronal activity, whereas local meningeal application of the CGRP antagonist had no effect (64). This is in excellent agreement with the work of Levy; they pointed out that their data do not support that peripheral CGRP or meningeal vasodilation can result in the generation of migraine headache because meningeal vasodilation could not activate mechanosensitive A- or C-units in the trigeminal ganglion or TNC (65). Furthermore, Lennerz et al. (66) found no evidence of CGRP, receptors in cranial dura mater on sensory axons. This is in excellent agreement with the first series of data on the trigeminovascular reflex (67–69). In these experiments we observed that cortex cerebral pial artery dilation did not activate the trigeminovascular reflex; however, the reflex was activated only by vasoconstriction induced by a number of vasoconstrictors tested and it involved primarily CGRP as the neuronal messenger (2, 4).
Acute treatment in animal models
Sumatriptan was originally developed as an intracranial vasoconstrictor, since this was the supposed mechanism of the antimigraine effect. However, it was subsequently observed that it could inhibit the release of CGRP (70–72). Consistent with this view, triptans were found to inhibit the central effects of meningeal activation (73). This may be due to the presynaptic effect, since 5-HT1B receptors have been observed only on primary afferent terminals and not on cell bodies in the TNC (74). Interestingly, when sumatriptan was applied on the dura mater, no effect could be detected on the mechanically evoked discharge of dural nociceptors, either under baseline conditions or after sensitization with inflammatory mediators (75). In contrast, sumatriptan excited these neurons. In agreement, systemic sumatriptan evoked discharge of the dural nociceptors, suggesting another site of the therapeutic effect of triptans (76). Zolmitriptan was developed to act on peripheral 5-HT1B/1D receptors and to have access to central inhibitory receptors on the trigeminal neurons (77, 78). In the studies zolmitriptan inhibited trigeminovascular neuronal activity after stimulation of the SSS (77, 78), whereas sumatriptan was unable to do so (78, 79). It is, however, uncertain whether zolmitriptan blocked the ability of meningeal primary afferents to conduct the signals to the TNC or whether it directly inhibited the TNC. Naratriptan is more lipophilic than sumatriptan and inhibits trigeminal neuronal activity in animal studies (80, 81). Interestingly, there was no effect on neurons in the spinal dorsal horn (80). This is consistent with the clinical observations that the triptans are not general analgesics (82). Rizatriptan inhibited the activity of neurons in the TNC in response to electrical stimulation of the dura mater (83). For both naratriptan and rizatriptan the reservations mentioned above hold true. The 5-HT1B/1D agonist L-741,604 was developed to have the ability to pass the BBB and seemingly blocked the development of central sensitization (82), but the receptive field of neurons did not expand and their responsiveness to stimuli did not increase, making it uncertain whether central sensitization occurred in that study. Thus, there is no evidence for a central action of this drug. Binding sites for tritium-labelled DHE have been localized to several places in the CNS, such as the dorsal horn, the cervical cord and in the medulla oblongata (50), demonstrating that DHE has access to parts of the brain that are relevant for an action in migraine. Similar studies with ergotamine are lacking. Sumatriptan appears to act not only via intracranial vasoconstriction (53), but also through presynaptic 5-HT1B/1D receptors on the trigeminovascular neurons (84). This study has thus demonstrated a CNS effect of sumatriptan. In another study, sumatriptan inhibited the mast cell-induced increase in c-fos expression in the TNC (85). This could be due to a 5-HT1D action on the peripheral neurons or to an action in the brainstem. To summarize, it is likely that triptans act therapeutically on different parts of the trigeminovascular system, mainly to inhibit the release of CGRP.
Intravenous aspirin in a high dose (30 mg/kg) inhibited central trigeminal neuronal activity following the stimulation of the SSS in cats (86); this was suggested to demonstrate that the drug can pass the BBB. However, if aspirin can act on other parts of the trigeminovascular system, outside the brain stem, a similar end result would occur. In fact, we may reason in the same way for the triptans, suggesting activity on different parts of the trigeminovascular system. The cyclooxygenase (COX) 1/COX2 inhibitors ketorolac and indomethacin blocked sensitization of meningeal (dura mater) nociceptors and suppressed ongoing sensitization in spinal trigeminovascular neurons (87). Similarly, i.v. naproxen was able to exert direct inhibition of central trigeminovascular neurons in the dorsal horn (88). These results suggest that the above three COX1/COX2 inhibitors can to some extent cross the BBB in the rat. Similarly, ketorolac was effective in allodynic patients in an open study (87).
The CGRP antagonist olcegepant, which is effective in the treatment of acute migraine attacks (89), can inhibit trigeminocervical neuronal firing in cats (67, 69). This indicates that olcegepant can pass the BBB. However, the CGRP antagonist olcegepant could not inhibit the effect of CGRP on rCBF and MCA diameter in man (90) or in rat (91), indicating that the drug has limited possibility to cross the BBB in vivo. It is clear that receptors for CGRP are situated abluminally on the vessels (on the smooth muscle cells) and in the trigeminal-cervical complex (63), since systemic α-CGRP did not penetrate the pressurized MCA (54). The CGRP antagonist given intravenously inhibited SSS-evoked activity (63) or activation of meningeal afferents. The ED50 dose was 31 µg/kg. This suggests that olcegepant in a clinically relevant dose can to some extent pass the BBB. In concert, it was not effective as an inhibitor of the CGRP-induced increase in pial artery diameter, but inhibited CGRP-induced vasodilation of the MMA (91), which does not possess a BBB. However, if olcegepant is given by local microiontophoresis, it is a potent inhibitor of activated trigeminocervical neurons in vivo (a way to circumvent the BBB) (63) or given systemically in very high doses (1 mg/kg) (64). These studies demonstrate the presence of functional CGRP receptors on the second-order trigeminal neurons or presynaptically on the central aspects of the trigeminal sensory fibres.
The NO synthesis inhibitor NG-monomethyl-
Preventive treatment in animal and human models
A wide range of drugs have been found to have prophylactic effects in migraine. However, the mode of interlinking them has been elusive. Chronic administration of the effective preventive drugs topiramate, valproate, propranolol, amitrityline and methysergide inhibited evoked cortical spreading depression (CSD) in rats (97), seemingly demonstrating a relevant antimigraine CNS effect in this animal model (at least on the aura). Whether this is a specific antimigraine effect would depend on the effect of, for example, antiepileptic drugs such as carbamazepine and phenytoin, which are not of proven efficacy in the prevention of migraine. One would expect metoprolol to work also in this model since metoprolol is the only β-adrenoceptor antagonist with proven efficacy in migraine with aura (98). However, metoprolol was without any effect on CSD given acutely in rats (99). In one study topiramate dose-dependently inhibited trigeminal neurons (100), and it was suggested that this effect was responsible for the prophylactic effect of the drug. In addition, there have been no studies on the effect of angiotensin converting enzyme inhibitors or of the angiotensin AT1 blocker candesartan on CSD. Nevertheless, it is an attractive model, but it does not explain why the preventive drugs work in migraines both with aura and without aura.
In man there are three studies on the effect of proven or possible prophylactic drugs using the NO-induced headache model (101–103). The effect of propranolol was not superior to placebo (103); there was no effect of tonerbasat, a drug that inhibits CSD (104), in a placebo-controlled investigation (101). After valproate (1500 mg for 2 weeks) there was a tendency for less migraine compared with placebo (P = 0.125) (102). The conclusion of this work may be: (i) insufficient dose was used. However, the doses for propranolol and valproate work in migraine prophylaxis; (ii) the usefulness of the NO model should be investigated more for studies of preventive drugs in migraine attacks.
Experimental human models
Pathophysiological models
The development of a headache model in man has helped in understanding essential components of migraine pathophysiology (105). The i.v. infusion of glyceryl trinitrate (GTN), a NO donor, caused an immediate mild to moderate headache with a ceiling effect at 0.5 µg kg−1 min−1(106). In migraine patients it caused a delayed migraine-like attack in about 80% of cases (107, 108), and this has recently been confirmed (109–111). NO is a freely diffusible agent, and there is no BBB restriction with this compound. It can pass the endothelium and act on the vascular smooth muscle cells to result in an increase in the local level of cyclic guanosine monophosphate (112) and with subsequent vasodilation. Effects on other parts of the trigeminovascular system are less clear. The immediate headache is most likely a direct vascular headache because it is accompanied by dilation of the middle cerebral and the temporal arteries (106, 108). This immediate headache is seen in all subjects, and well known to our cardiac patients treated with GTN preparations in conjunction with attacks of angina pectoris. The NO-induced delayed migraine-like attack (see below) could in some cases involve the CNS because it is preceded by premonitory symptoms in 37% of cases (110). The delayed headache is usually mild to moderate, and occurs 1–3 h after GTN administration (106).
Histamine is another agent tested in experimental human headache; it is hydrophilic and cannot pass the BBB. In one study i.v. histamine was found to cause both an immediate headache and a delayed migraine-like headache attack in migraine patients (113). It was suggested that histamine induced NO formation in the endothelium of cerebral arteries. However, it should be noted that histamine activates histamine H1 and H2 receptors located on both smooth muscle cells and endothelium in the human cranial arteries (114). This histamine response could not be blunted by a histamine H1 blocker, whereas a H2 antagonist had a minor effect on the headache (115). In addition, a NO synthase inhibitor was ineffective in preventing histamine-induced headache (116). It is our view that more basic research should be carried out to understand the details of the various types of drug actions in this human headache or migraine-like mode, particularly since there is renewed interest in the role of mast cells located in the dura mater (73, 85).
CGRP has been infused in migraine patients and observed to result in delayed migraine-like headache (117). Ever since the discovery of CGRP in the trigeminovascular system (118), we have argued for a prominent role of this neuronal messenger in primary headaches (for review see (119)). The work in collaboration with Goadsby has stressed its importance for understanding the pathophysiology of this group of disorders. In further studies, CGRP has been found to cause dilation of the MCA in vivo (117), which is in excellent agreement with the observations that human cerebral, middle meningeal and temporal arteries dilate potently to CGRP in vitro (120–123) and in vivo (124). In healthy volunteers a small increase (12%) was observed in rCBF (90), but this was not seen in migraine subjects outside the attacks (only a small 5% effect) (125). However, in migraine patients and in healthy volunteers there was an increase in the MCA diameter to the same degree. This supports studies on the peripheral circulation showing that there is no general difference between migraine and control subjects in responsiveness to CGRP (126, 127). In guinea pigs CGRP caused an increase in CBF (13%) after the BBB was disrupted following osmotic shock with urea (128). In the clinical study with olcegepant (89) an i.v. dose of 2.5–10 mg was required to show a significant effect. In man 2.5 mg intravenously resulted in a peak blood concentration of 170 ng/ml olcegepant (90) or a C max > 200 nM. Since the pEC50 of olcegepant is 0.1 nM in man (129), it corresponds to the use of a dose about 1000 times higher than that necessary to inhibit the CGRP receptor if olcegepant has complete access to the abluminal receptors without having to pass the BBB. Such a difference in concentration would support an effect abluminal of the BBB. This need to use a high dose for an antimigraine effect in the clinic strongly suggests an action of olcegepant at a site located inside the BBB and thus effectively rules out the neurogenic inflammation theory in the dural meningeal circulation as the prime target for the antimigraine effect of the CGRP receptor blocker (130).
What can we learn about the BBB from treatment of migraine?
Triptans are effective in migraine treatment (131). In the meta-analysis of 53 trials, there were more (7%) CNS-related adverse events (AEs) with sumatriptan 100 mg than with placebo, apparently suggesting that sumatriptan can to some extent pass the BBB. However, subsequent analysis of the eletriptan trial programme indicated that treatment-emergent CNS symptoms following triptan therapy are part of the migraine attack (132). Naratriptan (10 mg subcutaneously) is a very high dose with 71% AEs. In a double-blind, placebo-controlled trial the pain-free response to subcutaneous naratriptan (10 mg) was 88%, whereas the response was 55% after subcutaneous sumatriptan (6 mg) (133). The difference in response could be due to the fact that naratriptan is two-to-threefold more potent than sumatriptan in some animal models relevant to migraine (134). Alternatively, the superiority of naratriptan could be due to the fact that it is more lipophilic than sumatriptan. Naratriptan thus inhibited trigeminal neuronal firing after SSS stimulation in cats (81). Finally, a dual action, in the periphery and centrally, could be a possibility. Against this is the fact that zolmitriptan, which has the possibility for a dual action (78), is not more effective than sumatriptan (131). Minor sedative CNS effects of sumatriptan as studied in non-migraine subjects (135, 136) again suggest that sumatriptan can to some extent cross the BBB.
The migraine preventive drug propranolol, a non-selective β-adrenoceptor antagonist, readily passes the BBB (137, 138), as does metoprolol, which causes CNS AEs such as vivid dreams and depression (138). Flunarizine, a CNS active agent, has been used as add-on treatment in epilepsy (139, 140). Valproate and topiramate are antiepileptic drugs with proven efficacy in migraine (141). Amitriptyline is a well-established antidepressant (142). Flunarizine is not a classical calcium entry blocker (143), but has proven efficacy in migraine (144). Nimodipine, a dihydropyridine type of calcium entry blocker, can penetrate the BBB slowly, but is without proven efficacy in migraine. Verapamil has no proven efficacy in migraine (143). Thus, calcium entry blockade per se is therefore most likely not effective in migraine prevention, and when developing new prophylactic drugs for migraine one should aim at drugs which can to some extent cross the BBB.
It has been suggested that the BBB in migraine patients may be defective, perhaps only transiently or locally, either in the structural or the enzymatic component. In cerebral ischaemia, however, it takes several hours for the BBB to be broken down (145). This occurs via MMP-2 and -9, which digest the proteins clatrin and arrestin that are key elements in forming the tight junctions between endothelial cells (35, 146). In an experimental model of CSD, significant up-regulation of the MMP-9 levels was seen at 6 h after CSD (33). Since MMP-9 cleaves basement membrane proteins between the endothelial cells, it was observed that CSD could decrease the MMP substrate protein laminin and the zonula occludens-1 tight junction protein, and could be prevented by using a MMP-9 blocker, GM6001 (33). In situations such as those in stroke (cerebral ischaemia) with repetitive intense CSDs, these contribute to the evolution of the infarcts and are associated with BBB disruption (147, 148). In a recent report by Kruit and colleagues, deep white matter lesions were found to occur in the posterior circulation (‘infarct like lesions’) in migraine subjects with frequent attacks of migraine with aura (149). The nature of the lesions has so far not been determined. Importantly, the lesions were not seen in cortical layers as would be expected after repetitive CSDs. Therefore, it is our opinion that a fundamental structure such as the BBB remains intact during a functional episodic disorder such as migraine attacks and only in prolonged and severe situations may it be altered (such as after cerebral ischaemia).
Acknowledgement
Supported by the Lundbeck Foundation via the Lundbeck Foundation Centre for Neurovascular Signaling (LUCENS), Denmark, and the Swedish Research Council, Sweden.