Introduction

The trigeminal autonomic cephalalgias (TACs) are a group of primary headache disorders characterized by unilateral head pain that occurs in association with generally prominent ipsilateral cranial autonomic features [1]. TACs include cluster headache (CH), paroxysmal hemicrania (PH), and short-lasting unilateral neuralgiform headache attacks with conjunctival injection and tearing (SUNCT). Several lines of evidence, particularly functional neuroimaging data, suggest a role for the hypothalamus in the pathogenesis of these disorders. Advances in the management of these disorders have meant that the majority of these patients are effectively treated with medical treatments or occipital nerve stimulation. However, a small group of patients with these disorders is completely intractable to these treatment modalities. Based on the reports of posterior hypothalamic involvement in functional and structural imaging studies, various centers have successfully treated intractable TAC patients by ipsilateral electrode implantation and stimulation of this region. However, it has been suggested that the functional imaging data in CH shows activation in the midbrain tegmentum, in close proximity to the hypothalamus but not the hypothalamus per se [2]. This, consequently, raises the issues about the precise anatomical location of the deep brain stimulation (DBS) target used in these disorders, as the target is based on the co-ordinates derived from the functional imaging data [3].

This article reviews the evidence implicating a role for the hypothalamus in TACs and addresses the issue of the anatomical location of the DBS target utilized in treating these patients.

The Hypothalamus and Trigeminal Autonomic Cephalalgias

Clinical, Neuroendocrine, Genetic, and Experimental Animal Studies

There is a striking circadian and circannual periodicity seen in CH, and to a lesser extent in PH. This led to the suggestion that the suprachiasmatic nucleus of the hypothalamus, which is the area involved in the human biological clock system, plays a role in the pathophysiology of these disorders [4].

Neuroendocrine studies provided indirect evidence supportive of deranged hypothalamic function in CH. It was initially demonstrated that plasma testosterone concentrations were altered during the CH period in men [5]. Subsequently, it has been observed that there are abnormalities in the secretion of melatonin and cortisol, alterations in the secretion of luteinizing hormone and prolactin, and altered responses of luteinizing hormone, follicle-stimulating hormone, prolactin, growth hormone, and thyroid-stimulating hormone to challenge tests in patients with CH [6]. There is preservation of the hypothalamic-pituitary axis, and the abnormalities are more supportive of a primary dysfunction of the hypothalamus. Although these pathological neuroendocrine findings may be partially related to pain, stress, or interrupted sleep, the persistence of some abnormal findings during remission would be consistent with hypothalamic dysfunction in CH.

Further support for a hypothalamic abnormality in CH has emerged from genetic and experimental animal studies. Orexin-A and orexin-B are neuropeptides processed from a common precursor, prepro-orexin. Orexin-containing cells are located exclusively in the posterolateral hypothalamus, with widespread projections to the entire neuroaxis. Two receptors, orexin receptor-1 (OX1) and orexin receptor-2 (OX2), have been identified. The peptides of the orexin system influence a wide range of physiologic processes, including pain transmission, autonomic function, and neuroendocrine function. Two studies have reported that a polymorphism of the OX2 gene (1246 G>A) is associated with CH [7, 8]. An experimental animal study examined the effects of orexin receptor activation in the posterior hypothalamic area on dural nociceptive input. Orexins were microinjected into the posterior hypothalamus, and the effects on responses of neurones in the caudal trigeminal nucleus were studied. Orexin-A decreased the A-fiber and C-fiber responses to dural electrical stimulation as well as spontaneous activity, whereas orexin-B elicited increased responses to dural stimulation in A-fiber and C-fiber responses and resulted in increased spontaneous activity [9]. Another experimental animal study examined the role of orexin-A and orexin-B in the modulation of central and peripheral trigeminovascular pain pathways and found that orexin-A inhibits both central and peripheral pathways, whereas orexin-B facilitates central transmission [10]. These genetic and experimental animal data taken together suggest a possible pathogenetic mechanism for CH.

Functional Neuroimaging Studies

Functional neuroimaging studies have played a pivotal role in providing fundamental insights into the pathophysiology of primary headache disorders. Table 1 provides an overview of the brain regions reported to be active in these functional neuroimaging studies.

Table 1 Pattern of activation of brain structures in functional neuroimaging studies of primary headache syndromes and experimental head and facial pain
Full size table

May et al. [11] performed the landmark neuroimaging study in CH. They employed H 152 O positron emission tomography (PET) to study nine chronic cluster headache (CCH) patients during nitroglycerin-induced attacks. They reported specific activation of the ipsilateral posterior inferior hypothalamic gray besides structures of the general pain neuromatrix. Given that hypothalamic activation had not been observed in migraine or in experimental trigeminal pain induced by capsaicin injection into the forehead [12], the authors concluded that the hypothalamus is involved in the pain process of CH in a permissive or triggering manner rather than simply as a response to first division nociception per se. Hence, this study provided the first direct functional evidence for involvement of the hypothalamus in nitroglycerin-induced CH. Subsequently, hypothalamic activation was reported in a study employing H 152 O PET during a spontaneous CH attack in a patient treated with DBS [13]. A recent study employing functional magnetic resonance imaging (fMRI) in four episodic cluster headache (ECH) patients confirmed ipsilateral hypothalamic gray activation [14].

Matharu et al. [15] employed H 152 O PET to study seven PH patients and reported significant activation of the contralateral posterior hypothalamus and ventral midbrain, extending over the red nucleus and substantia nigra, in association with the pain of PH. Interestingly, a PET study in hemicrania continua (HC), which is an indometacin-responsive headache like PH, also showed significant activation of the contralateral posterior hypothalamus and the ventrolateral midbrain besides the dorsal rostral pons [16].

There are several reports of fMRI studies in SUNCT. May et al. [17] performed the first blood oxygen level dependent (BOLD)-fMRI study in a patient with SUNCT and reported significant activation in the region of the ipsilateral inferior posterior hypothalamic gray when comparing the pain attacks with the resting state. Sprenger et al. [18] performed a BOLD-fMRI study in a patient with right-sided SUNCT and reported bilateral hypothalamic activation during the pain attacks. Sprenger et al. [19] also reported a BOLD-fMRI study in a patient they considered to have an atypical TAC, who appears to have an SUNCT-like phenotype, and reported activation of the contralateral hypothalamus. Cohen [20] employed BOLD-fMRI to study nine SUNCT patients. There was activation in the region of the posterior hypothalamus, which was bilateral in five patients and contralateral in two patients.

These reports of hypothalamic activation in all the TACs and HC, taken together with the phenotype of these primary headache syndromes, have led to the suggestion that activation of this subcortical structure is a marker of prominent cranial autonomic features and may play a crucial role in the pathophysiology of these disorders.

Voxel-Based Morphometry

A fundamental tenet of primary headache syndromes is that they are due to abnormal brain function with completely normal brain structure [1]. This view was challenged in CH by a voxel-based morphometry (VBM) study. VBM is an automated, nonbiased, whole-brain technique that analyzes changes in brain structure. May et al. [21] compared regional differences in gray matter between 25 CH patients and 29 healthy volunteers using VBM analysis of the structural T1-weighted MRI scans. They reported a significant structural difference in the hypothalamic gray matter, consisting of an increase in volume in the CH cohort. The structural difference was bilaterally situated in the diencephalon, adjacent to the third ventricle and rostral to the aqueduct, and reported to be coinciding with the inferior posterior hypothalamus. An important criticism of this study is that the patients and controls were not age and sex matched. In this respect, it is noteworthy that this finding has not been reproduced hitherto.

A VBM study of 66 CH patients and 96 age- and sex-matched controls imaged with 1.5 Tesla volumetric MRI found no differences in global gray or white matter volumes between patients and controls [22]. Furthermore, regional gray or white matter volumes in the hypothalamus did not differ between patients and controls. These data suggest that CH is not associated with structural correlates and question the validity of the original study, especially given its methodological deficiencies. The issue of structural changes in CH needs further investigation with higher resolution volumetric MRI scans.

Magnetic Resonance Spectroscopy

Proton magnetic resonance spectroscopy (1H-MRS) has been reported to show evidence of altered hypothalamic metabolism in CH patients. Lodi et al. [23] studied 26 CH patients and 12 controls using a volume of interest (VOI) of 1 to 1.2 cm3 centered on the bilateral hypothalamic gray. They reported that the hypothalamic N-acetylaspartate/creatine ratio was significantly reduced in CH patients compared with controls. Wang et al. [24] studied 47 CH patients (35 in cluster bout, 12 in remission), 16 chronic migraine patients, and 21 controls using a VOI of 2.25 cm3. The authors reported that the hypothalamic N-acetylaspartate/creatine and choline/creatine ratios were significantly lower in patients with CH in comparison with either the control or chronic migraine groups. In these studies, the metabolite ratios in CH patients were altered in both the cluster and remission periods. These studies suggest that there is a persistent biochemical change in the region of the hypothalamus in CH patients, although it remains unclear whether these changes are causal or epiphenomenal.

Biological Plausibility of Hypothalamic Dysfunction in TACs

A similarity in the clinical phenotype of TACs raises the possibility of a common pathophysiological basis. Any pathophysiological construct for TACs must account for the two major shared clinical features characteristic of the various conditions that comprises this group: trigeminal distribution pain and ipsilateral cranial autonomic features. The pain-producing innervation of the cranium projects through branches of the trigeminal and upper cervical nerves to the trigeminocervical complex from whence nociceptive pathways project to higher centers. This implies an integral role for the ipsilateral trigeminal nociceptive pathways in TACs. The ipsilateral autonomic features suggest cranial parasympathetic activation (lacrimation, rhinorrhea, nasal congestion, and eyelid edema) and sympathetic hypofunction (ptosis and miosis). It has been suggested that the pathophysiology of TACs revolves around the trigeminal autonomic reflex [25]. There is considerable experimental animal literature to document that stimulation of trigeminal afferents can result in cranial autonomic outflow, the trigeminal autonomic reflex [26].

How could hypothalamic dysfunction lead to the symptom complex of TACs? There is abundant evidence for a role of this structure in mediating antinociceptive [27, 28] and autonomic responses [29]. In fact, there is direct evidence from animal experimental studies for hypothalamic activation when intracranial pain structures are activated [30]. Moreover, the hypothalamic peptides orexin-A and orexin-B can elicit pronociceptive and antinociceptive effects in the trigeminal system [9]. Hence, dysfunction of the hypothalamus may be accompanied by a disruption of the balance between the trigeminal nociceptive and antinociceptive systems. One hypothesis posits that the pain and cranial autonomic symptoms in this syndrome may be due to a central disinhibition of the trigeminal parasympathetic reflex by the hypothalamus [31, 32], possibly via the hypothalamic trigeminal projection [33].

Locus of Dysfunction: Hypothalamus or Midbrain Tegmentum?

What is the Precise Location of the Diencephalic/Mesencephalic Activation Seen in TACs?

Although the neuroimaging data, taken together with clinical observations and neuroendocrine studies, strongly suggest a pivotal role for the hypothalamus in CH, Sanchez del Rio and Linera [2] have suggested that the diencephalic/mesencephalic activation in CH [11] lies in the midbrain tegmentum, in close proximity to the hypothalamus but not the hypothalamus per se. This proposition demands a review of the anatomy of the hypothalamus and localization of the diencephalic/mesencephalic activation seen in TACs and HC.

The anatomical boundaries of the hypothalamus are the lamina terminalis anteriorly, the posterior margin of the mamillary bodies posteriorly, the hypothalamic sulcus superiorly, the third ventricle medially, the subthalamus and internal capsule laterally, and the optic chiasm, median eminence, tuber cinereum, mammillary bodies, and posterior pituitary inferiorly [34]. The functional boundaries of the hypothalamus are less well defined and may extend beyond the anatomical boundaries. The hypothalamus has extensive connections with other brain regions involved in visceral activities. Among these connections are the midbrain tegmentum and periaqueductal gray, brain regions that are well established to participate in pain signal processing.

A careful re-examination of the statistical parametric maps and co-ordinates of the diencephalic/mesencephalic activation in PET studies in these primary headache disorders reveals that the diencephalic/mesencephalic activation in CH [11] is centered over the midbrain tegmentum but extends anteriorly to abut the posterior hypothalamus. The activations in PH [15] and HC [16] involve the posterior hypothalamic region with posterior extension to the midbrain tegmentum, straddling the red nucleus and substantial nigra. On the other hand, the functional imaging studies in CH and SUNCT that have been performed with BOLD-fMRI studies demonstrate activation of the posterior [17, 19, 35] and middle hypothalamus [18] rather than the mesencephalon. Table 2 shows that the co-ordinates of the diencephalic/mesencephalic activation in the PET studies (y co-ordinates = −14 to −18 mm, z co-ordinates = −6 to −8 mm) are posteroinferior to those in the BOLD-fMRI studies (y co-ordinates = −6 to −9 mm, z co-ordinates = 0 to −6 mm).

Table 2 Co-ordinates of hypothalamic activation reported in functional imaging studies of trigeminal autonomic cephalalgias
Full size table

What is the significance of these observations? It is difficult to be definitive about the localization of this activation due to the limitations of functional neuroimaging techniques. The spatial resolution of PET is limited to 2 to 7 mm, by virtue of the distance the positron travels through the tissue prior to the annihilation event. The spatial resolution is further constrained to approximately 5 to 10 mm by methodological issues, including warping of the brain to fit the standard atlas, smoothing and study design that are generic to functional imaging studies. Functional MRI has a better spatial resolution, though even this is constrained to the order of 4 to 5 mm. The diencephalic/mesencephalic structures under discussion lie close to each other and are beyond the limits of resolution of functional neuroimaging techniques, particularly the PET technique. Hence, localization of intimately placed structures on statistical parametric maps in PET and fMRI studies requires interpretation in the context of other knowledge of the disorder’s phenotype and pathophysiology rather than literal anatomic interpretation. It also appears that the location of activation is dependent on the imaging modality with the activation limited to the hypothalamus in the higher resolution fMRI studies, thereby raising the possibility of a methodological effect.

As the clinical, neuroendocrine, genetic, and experimental animal data suggest a pivotal role for the hypothalamus in the pathophysiology of CH, it is likely that the diencephalic/mesencephalic activation seen in TACs, particularly CH, reflects posterior hypothalamic involvement. However, it is possible, if not highly likely, that both the hypothalamus and midbrain tegmentum are activated in TACs, given the well-established role of these structures in nociceptive processing and the connections between them. Indeed, discrete activations of both the posterior hypothalamus and the ventral midbrain, straddling the red nucleus and the substantia nigra, have been observed in PH and HC.

Lateralization of Diencephalic/Mesencephalic Activation in TACs

A signature feature of these disorders is that the pain is generally strictly unilateral. This raises the question of whether the lateralization of pain in these syndromes is due to lateralized diencephalic/mesencephalic activation. It is interesting to note that there is an inconsistent relationship between the laterality of the pain and activation reported in functional imaging studies. The activation is ipsilateral to pain in CH and contralateral to pain in PH (and HC). In SUNCT, ipsilateral [17], bilateral [18, 35], and contralateral [19, 35] activation have been reported.

This poses some interesting issues about the basis of laterality of attacks in these syndromes. First, it is possible that the observed data arise from a methodological issue. In cases where unilateral hypothalamic activation is observed, the contralateral hypothalamus may also be activated but below the threshold for statistical significance. Secondly, these data suggest that unilateral as well as bilateral hypothalamic dysfunction may be able to induce these syndromes, or at least SUNCT. The mechanism by which bilateral hypothalamic dysfunction may exert a unilateral clinical phenotype demands further study. It is possible that, as one area of the hypothalamus is activated, another region is deactivated. Sprenger et al. [18] have also raised the possibility that the hypothalamic regions involved in nociceptive functions may not be somatotopically organized. Thirdly, bilateral hypothalamic dysfunction may be the basis for bilateral or side-variable unilateral attacks.

Conclusions

Implications for Deep Brain Stimulation in Trigeminal Autonomic Cephalalgias

Based upon the reports of ipsilateral posterior hypothalamic activation in CH, various centers have treated medically intractable CCH patients by electrode implantation and stimulation of this region. Leone et al. [36••] reviewed the results of hypothalamic DBS in 38 medically intractable CCH patients with a follow-up period ranging between 1 and 4 years; 23 patients (61%) were rendered pain-free or almost pain-free. There could be a delay of up to 3 months after initiation of stimulation before the beneficial effect emerges. Similarly, when the stimulation is switched off there can be a delay of 3 months before headache recurrence. There are no substantial changes in hypothalamus-controlled functions during stimulation. A randomized crossover study performed by this group did not support the efficacy of DBS in CH. However, the crossover period (1 month) was deemed too short given the delayed response to DBS described above. Indeed, in the open phase portion of this study long-term efficacy was achieved in over 50% of patients [37]. Ipsilateral hypothalamic DBS has also been reported to be effective in case reports of PH [38] and SUNCT [39, 40]. However, a note of caution needs to be struck about this exciting novel approach. One patient died soon after the operation due to implantation-induced intracerebral hemorrhage, and implantation was stopped intraoperatively in another patient who had a panic attack [41]. This underlines the importance of appreciating that DBS procedures are associated with a small risk of morbidity and mortality.

The use of this technique raises the issue of the precise anatomical location of the DBS target employed in these disorders, as the target is based on the co-ordinates derived from the PET study in CH. Franzini and Leone [42], who pioneered this procedure in CH, besides several other groups [43, 44], described the anatomical target as the posterior hypothalamus. However, Fontaine et al. [3], who implanted 11 electrodes in CCH patients using the same co-ordinates used by other investigators, accurately defined the anatomical location of the effective contacts using the Schaltenbrandt atlas and a stereotactic 3D MRI atlas of the diencephalon-mesencephalon junction. They reported that the electrodes were located in the posterior and ventral wall of the third ventricle, posterior to the mamillary body and the mamillothalamic tract, which mark the caudal border of the hypothalamus. They reported that the electrodes were embedded in two adjacent structures, substantia grisea mesencephalic and the crown of the red nucleus.

This raises some important issues regarding the utilization of DBS for the treatment of TACs. The DBS target has been derived from PET data of group studies with an inherent limitation of spatial resolution. However, the fMRI data, which has better spatial resolution, suggest that the locus of activation may be anterosuperior to that obtained from PET studies. This prompts the question whether the stereotactic stimulation in these disorders has been targeted at the appropriate locus. This may be the reason that approximately 40% of patients demonstrate a relatively poor response. Furthermore, group data do not account for the individual variations in anatomy and possibly activation patterns. To optimize the outcomes from DBS, it is likely that we will need to study patients individually using functional imaging techniques that have high spatial and temporal resolution to enable us to target the appropriate locus with DBS [45•].

The lateralization of the diencephalic/mesencephalic activation in PH and SUNCT poses another challenge. The published studies report contralateral activation in PH, whereas ipsilateral, contralateral, and bilateral activation patterns have been reported in SUNCT. Should these patients have ipsilateral or contralateral stereotactic stimulation? It is interesting to note that the cases reported hitherto have all performed DBS ipsilateral to pain. The merit of this approach will only become clear with long-term follow-up of larger cohorts.