div class=”ChapterContextInformation”>
5. Genetics of Cluster Headache and Other Trigeminal Autonomic Cephalalgias
Keywords
Cluster headacheTrigeminal autonomic cephalalgiaTwinsFamily studiesGenesPathways5.1 Why Study Genetics in Cluster Headache and Trigeminal Autonomic Cephalalgias?
Cluster headache belongs to a group of primary headache disorders, the trigeminal autonomic cephalalgias (TACs), all of which consist of disabling unilateral pain in trigeminal distribution associated with marked ipsilateral cranial autonomic features [1–4]. Cluster headache is the commonest TAC. The majority of TACs, including cluster headache, paroxysmal hemicranias, short-lasting unilateral neuralgiform headache attacks with conjunctival injection and tearing (SUNCT) and short-lasting unilateral neuralgiform headache attacks with cranial autonomic symptoms (SUNA), manifest as daily short-lived recurrent attacks and can be distinguished one from each other by the duration of the attacks [5]. Only hemicrania continua manifests as a continuous daily unilateral pain, so it is not presented in the form of attacks. Response to therapy is different among TACs. The neurobiological mechanisms underlying cluster headache and other TACs are complex and remain incompletely understood [2, 6, 7]. The leading hypothesis incriminates hypothalamic activation with secondary activation of the trigeminal-autonomic reflex, probably via a trigeminal-hypothalamic pathway [8–10]. In contrast to migraine, cluster headache has not been considered as a familial disorder until the last decades, and epidemiological studies suggested that this condition could be, at least in part, of genetic origin [11]. Thereafter, the genetics of cluster headache has become an emerging research field with the major goal to identify genes that confer disease risk. The identification of causal genes will give important clues to the molecular underpinning and insight in the pathogenesis of the disorder and may guide the development of new diagnostic and therapeutic strategies. At the moment, however, such genetic factors remain unknown for any of the TACs.
5.2 Why Is It Difficult to Identify Genes Involved in Cluster Headache and Other TACs?
Genetic studies in cluster headache and TACs encounter difficulties directly related to some of the characteristic features of the disorders, i.e. their low prevalence and the fact that diagnosis is not straightforward, assuming that the disorders are genetic in the first place.
With respect to cluster headache, the prevalence is only about 0.1% in the population, and the diagnosis is often missed. In the absence of a biological or radiological marker, the diagnosis is purely based on clinical criteria proposed by the International Classification of Headache Disorders (ICHD) [4]. The patient must have had at least five attacks of severe or very severe unilateral orbital or supraorbital and/or temporal pain lasting 15–180 min when untreated. The headache is accompanied by one ipsilateral autonomic symptom of the following: conjunctival injection and/or lacrimation, nasal congestion and/or rhinorrhoea, eyelid oedema, forehead and facial sweating, miosis and/or ptosis. Since the 2004 ICHD-2, cluster headache can be diagnosed in the absence of autonomic signs if headache is associated with restlessness or agitation [12]. Primary cluster headache is characterized by the repetition of such attacks in the absence of any underlying disorder causing secondary headaches. The clinical spectrum of cluster headache may be larger, as suggested by the description of several persons with painless attacks of unilateral autonomous signs who later on developed or who had previously suffered from typical painful cluster headache attacks [13, 14]. Recurrent cluster-like attacks may be associated with several organic cerebral disorders, including mainly pituitary tumours, other brain or cervical tumours, sinusitis and carotid artery dissection [15, 16]. Many reports have described that these structural lesions might cause symptoms that are indistinguishable from those of primary cluster headache. At present, whether the mechanisms of symptomatic cluster headache attacks are different from those of primary cluster headache attacks is an unresolved issue.
Classification of individuals as definitely ‘affected’ or ‘unaffected’, which is essential for genetic studies, does not reflect the clinical heterogeneity of cluster headache. Indeed, the frequency and severity of attacks, as well as the duration of active periods show much variability from one patient to another, complicating genetic analysis. Depending on the long-term temporal course of attacks, cluster headache is divided by clinicians in episodic and chronic forms [2–4]. Most of the cluster headache patients are episodic and may present one or two active periods per year or even go in remission for years before the next bout starts. About 20% of cluster headache patients are chronic and have ongoing attacks for 1 year or more. The distinction between episodic and chronic forms is arbitrary, which further complicates genetic studies, also because it is not unknown whether the distinction reflects differences in underlying molecular pathways. The diagnostic criteria for chronic cluster headache have been modified in the latest ICHD-3 version, stipulating that attacks must have occurred without a remission period or with remissions lasting <3 months for at least 1 year [17]. In the previous versions of the classification, chronic cluster headache was diagnosed in patients having ongoing attacks for 1 year or more without more than 1 month (ICHD-2 and ICHD-3 beta) or 14 days of remission (ICHD-1) [4, 12, 18]. Some patients may evolve from one form to the other and reverse. Future genetic findings may shed light on whether episodic and chronic cluster headaches are in fact two clinical forms of the same disease or whether they represent distinct disorders. Finally, the age of onset is highly variable ranging from early childhood to more than 80 years, and the overall male to female ratio is 4.3:1, which also has to be taken into account in family and segregation studies [19].
Similar problems are related to studying the genetics of other TACs. In brief, these conditions are very rare, and their diagnosis is also purely clinical based on the ICHD criteria [4]. The prevalence of paroxysmal hemicrania is not known, but the relationship in comparison with cluster headache is reported to range from 1 to 15% [20, 21]. Paroxysmal hemicrania differs from cluster headache by a female preponderance, the shorter length and higher frequency of attacks and an absolute response to indomethacin [22], but distinguishing the two TACs may be difficult. SUNCT has an estimated prevalence around 6.6 per 100,000 [23] and manifests as very short attacks that can occur up to 300 times a day, which can be difficult to distinguish from trigeminal neuralgia [24]. The prevalence of hemicrania continua, another indomethacin-responsive TAC, is unknown, and only a few hundred cases have been published in the literature [25, 26].
5.3 Is Cluster Headache Really a Genetic Condition?
Initially, CH was not thought to be a genetic disease. However, with the official criteria for diagnosis being published, there came increasing recognition of the disease likely having a genetic basis. A logical first approach to investigate whether a trait is genetic is to perform twin studies by comparing co-occurrence of disease (concordance) in monozygotic twins versus dizygotic twins. However, because of the low prevalence of the disease, such studies are difficult to perform. Cluster headache has been reported in five concordant monozygotic twin pairs [27–31]. The first large twin survey based on the Swedish Twin Registry and the Swedish Inpatient Registry failed to identify any concordant pair; the two monozygotic and nine dizygotic twin pairs were all discordant for cluster headache and had been discordant for 10–31 years [32]. A subsequent larger Swedish study was conducted on 31,750 registered twins and showed a higher concordance rate for cluster headache in monozygotic twins (2/12 pairs) than dizygotic twins (0/25 pairs), presenting at least some indication that the trait is genetic. The fact that most monozygotic twins are discordant for CH shows that environmental factors also play a role [33].
There have been clinical reports of cluster headache in families [34, 35]. A genetic component in cluster headache pathogenesis was further confirmed by seven systematic family studies in large samples of probands with cluster headache [36–43]. Compared with the general population, the risk of cluster headache for first-degree relatives was found increased by 14- to 46-fold and for second-degree relatives by 2- to 8-fold [38–40, 42]. Differences between studies may be explained by methodological issues, for example, only the French study included a direct interview of all first-degree relatives by a headache specialist [40]. Estimating heritability in cluster headache, which is another often-used measure of the genetic component, is not feasible given the low prevalence of the disorder.
A precise transmission mode is not established for cluster headache. All types of transmission have been observed: from father to son, father to daughter, mother to son and mother to daughter. In a large Dutch study, some 1700 patients with cluster headache were asked whether they had relatives with the disease which revealed that of the 33 parent-to-child transmissions, 22 were from father to son, 6 from mother to son, 2 from mother to daughter and 3 from father to daughter [44]. The study identified 12 families with 3 affected (3 in 1 generation, 8 in 2 generations and 1 in 3 generations) and 58 families with 2 affected (33 in 1 generation, 20 in 2 generations and 5 included second-degree relatives).
A Danish complex segregation analysis [45] and two Italian family studies [37, 39] have suggested an autosomal dominant inheritance with incomplete penetrance. Such inheritance is further suggested by two studies that found a lower male/female ratio in familial cluster headache than in cluster headache in general [40, 43]. Conversely, an analysis of a single Italian pedigree suggested autosomal recessive inheritance [46]. The French family study identified 12 vertical and 8 horizontal patterns of transmissions consistent with autosomal dominant and recessive inheritance, respectively [40]. Still, a polygenic mode of inheritance should not be excluded, as in most families with cluster headache no clear segregation pattern is observed.
Altogether, these results have confirmed that cluster headache has a genetic component but that the mode of transmission may vary while the amount of heritability is unclear.
5.4 Are Other TACs Genetic Conditions?
There have been only few reports of other TACs running in families. Familial paroxysmal hemicrania was reported in a mother and a daughter [47]. In a series of 74 patients with paroxysmal hemicrania, 80% did not report a family history of any headaches, 15% had a family history of migraine and 5% had a family history of other types of headache and facial pain including cluster headache and trigeminal neuralgia, but none had paroxysmal hemicrania [48]. There has been one report of familial hemicrania continua in a single family with a mother and a daughter both affected by hemicrania continua and migraine [49]. Familial SUNCT was reported in two families each comprising of two affected first-degree relatives, namely, a brother and a sister [50] and a mother and a son [51]. One report showed, in a cohort of 117 SUNCT/SUNA and 107 hemiplegic migraine patients, co-occurrence of both disorders in 10 patients, which is not expected given the low prevalence of both disorders, suggesting that they may be brought about by a common mechanism, although this was not supported by evidence [52].
All in all, these reports suggest that the other TACs may have a genetic component, but rarity of these conditions renders epidemiological surveys difficult to conduct.
5.5 Are Other Primary Headache Types Genetic Conditions?
Of the other primary headache disorders, the genetics of migraine is the best studied. Information about the genetics of other primary headache disorders is scarce and not further discussed here. Twin and family studies in migraine have indicated a strong genetic component. The genetic component was higher in migraine with aura than migraine without aura [4], with concordance rates in monozygotic twins being 1.5- to 2-fold higher than in dizygotic twins [53–55] and an increased risk of migraine for first-degree relatives that is 1.4- to 4-fold higher, depending on the migraine type [55, 56]. The lower fold changes in migraine compared to those in cluster headache merely reflect the higher prevalence of migraine in the population (15–20%) and not that the genetic component in migraine would be lower than in cluster headache. In fact, migraine has an estimated heritability of 42%, indicating that about half of the risk for migraine is conferred through genetic factors and which is in the range of most other complex (polygenic) disorders [57].
5.6 Which Strategies Can Be Used to Identify Genes for CH?
The epidemiological evidence that cluster headache is rare, runs in families and is brought about by both genetic and environmental factors determines chances for success of the genetic approach applied to identify causal genes.
5.6.1 Gene Identification Strategy in Case of a Monogenic Inheritance
The fact that some extended Mendelian pedigrees have been identified with multiple patients with cluster headache, at least in theory, should make them suited to identify disease-causing mutations using linkage. In a linkage approach, several hundreds of genetic markers, equally spread over all chromosomes, are tested in affected and unaffected individuals of a single family or of multiple families, and markers that best segregate with the disease reveal the likely chromosomal location of the disease gene. Next, using a positional cloning approach, the pathogenetic mutation is identified in a disease locus. In recent years, next-generation sequencing (NGS) is used to speed up the process, as it allows massive parallel sequencing of all protein-coding regions (‘exome sequencing’), or in fact the whole genome (‘whole genome sequencing’), in a single experiment, and has been very successful in identifying disease genes for monogenic disorders [58].
Proof of causality of a mutation comes from the fact that (1) the mutation is present in most affected family members (in case of reduced penetrance, the mutation is also found in some cases who do not (yet) express the disease) and not present in non-affected family members (although phenocopies may occur, which are defined as patients that express disease because of an unlinked cause), (2) the mutation is not found in large cohorts of control individuals and (3) follow-up functional studies provide convincing support that the mutated gene affects a pathway implicated in the disease. At the moment, no pathogenic mutation for cluster headache or any of the other TACs was identified by a linkage or NGS study. For comparison, and as demonstration of how powerful such approach can be, in familial hemiplegic migraine, three genes and in them many different mutations have been identified in hundreds of families and sporadic patients, which had a profound impact on clinical care, already because the identification of a causal mutation confirms the clinical diagnosis [59].
5.6.2 Gene Identification Strategy in Case of a Polygenic Inheritance
However, as most cluster headache patients are not part of Mendelian families, but merely singletons, or at best two or three patients in a pedigree without a clear segregation pattern, alternative approaches that take into account that one genetic factor is not sufficient to bring about disease may be more appropriate to identify genetic factors. Association-based methods are particularly suited, because they can identify susceptibility genes and DNA variants with small relative risks. However, such studies only give convincing evidence for involvement of a gene when (1) a large sample size is used (preferably many hundreds to thousands); (2) when (well-phenotyped) patients and control samples have a comparable genetic architecture, so spurious association because of population stratification is prevented; and, most importantly, (3) that promising findings are replicated in other samples.
Until recently, a genetic association study was designed based on a specific hypothesis (candidate gene association study), which entails that a gene is selected based on prior knowledge that it acts in a presumed disease pathway. The frequency of alleles of polymorphic genetic markers, typically single nucleotide polymorphisms (SNPs), in such a gene is compared between patients and healthy controls. A significant difference suggests that the polymorphism (and the gene) is involved in disease pathology or that because of linkage disequilibrium, a gene in close proximity is the causal gene. Instead of testing one SNP at a time, nowadays, because of technological advances, it is feasible to cost-effectively genotype hundreds of thousands of SNPs in the entire genome (hypothesis-free approach) in thousands of patients and controls.
5.7 What Is Currently Known About the Genetics of CH?
Almost all genetic studies in cluster headache to date used the candidate gene association approach. Genes were selected because they are believed to be involved in molecular pathways that explain clinical features, in particular the characteristic periodicity and circadian rhythm of attacks, which suggests hypothalamic dysfunction, and the fact that patients are heavy smokers and drinkers [60, 61] although the latter is debated [62].
5.7.1 Candidate Gene Association Studies in Cluster Headache
A large number of candidate genes were tested, among others, CLOCK, PER3, HCRTR1 and HCRTR2 that affect functioning of the biological clock, which reside in the suprachiasmatic nucleus of the hypothalamus, ADH4 that is involved in alcohol metabolization and CACNA1A and NOS1-3 genes that are involved in pain processing (for comprehensive reviews, see [11] and [63]). Except for HCRTR2, which will be discussed separately below, none of the candidate genes showed evidence for association with cluster headache, mainly because investigated samples were much too small—most of them had below 230 cases, a number that often was subdivided into episodic and chronic cluster headaches which reduced power even more—and, without exception, no association result could be convincingly replicated. Therefore, with the present data, no conclusion can be drawn about the involvement of any of these genes in cluster headache.
The situation is slightly better for HCRTR2, which encodes the hypocretin receptor 2. The hypocretin (orexin) system is thought to play an important role in cluster headache as hypocretin-containing neurons almost exclusively are located in the posterolateral hypothalamus that generates rhythms [8]. Foremost, SNP rs2653349 (G1246A), which changes a valine residue to an isoleucine at position 308 of the receptor protein, was investigated, and the first study showed promising association in an Italian cohort of patients with cluster headache [64]. The association was replicated in one [65] but not in another study [66]. A recent, larger Dutch study that investigated 575 patients with cluster headache was also not able to show association, although a meta-analysis of 1167 cases and 1618 controls was again positive [67]. In light of the fact that cluster headache patients are heavy smokers, it certainly is interesting that the same SNP was linked to increased nicotine dependence because of increased rewarding effects due to changed activity of the orexin system [68].
Finally, an association of mtDNA abnormalities with cluster headache was suggested, even though transmission of disease from the father to offspring often occurs. Evidence for mitochondrial involvement came from the fact that (1) a sporadic Japanese patient with a 3243 point mutation in platelet mitochondrial tRNALeu(UUR), known to cause MELAS (mitochondrial myopathy, encephalopathy, lactacidosis and stroke-like episodes), also had cluster headache [69] and (2) cluster headache was also reported in a patient with multiple deletions in mitochondrial DNA. It remains unclear to what extent abnormal mitochondrial function actually is involved in cluster headache [70].
5.7.2 Candidate Gene Association Studies in Other Primary Headache Disorders
Candidate gene association studies have also been performed in other primary headache disorders, mainly in migraine with an equally disappointing outcome (for reviews, see [71] and [72]). Despite overwhelming evidence from clinical, pharmacological and neuroanatomical studies that, for instance, the serotonin and dopamine systems are involved in migraine, no convincing evidence for association was found for SNPs in genes of these pathways. Using a much larger large data set of 5175 patients with migraine with or without aura and 13,972 controls, a systematic re-evaluation was conducted of the most promising 21 genes, which included the MTHFR gene that encodes a key enzyme in folate and homocysteine metabolism that had surfaced as the best migraine candidate gene [71, 73] and also showed suggestive association in chronic cluster headache [74]. Neither the previously implicated variants nor any other variant in a region of 500 kb surrounding the respective genes provided evidence for association [73]. The most likely conclusion, therefore, is that all published associations from candidate gene association studies, in retrospect, are false positives. Proof that this scenario indeed applies comes from the consideration that in such a large sample, the associated T allele of the MTHFR C677T polymorphism, with reported effect sizes of ~1.5, when assuming a minor allele frequency of 31%, should have produced a p-value ~1.5 × 10−63 instead of the non-impressive observed 9.7 × 10−3 [73]. The effect size in these small studies is very much inflated, as evidenced by the small effect size (1.08), which is more in line with the larger genome-wide association studies (GWAS) discussed below. A sobering thought perhaps is the realization that a p-value cut-off of 0.05 simply is not reliable to obtain robust results when dealing with small data sets, which had already been recognized in other areas of science [75].
5.7.3 Genome-Wide Association Study in Cluster Headache
The low prevalence of cluster headache did not stop Italian researchers from performing the first, and thus far only, genome-wide association study (GWAS) in cluster headache [76]. Admittedly, their sample of 99 clinically well-defined patients with cluster headache and 360 healthy individuals is tiny compared to the thousands to tens of thousands of cases and controls that one nowadays investigates in a GWAS [77]. Also, the male-female distribution and smoking status was not appropriately matched in the cluster headache GWA study [76]. Still, by combining single-marker (testing common SNPs) and gene-based (focussing on rare protein-altering variants in 745 candidate genes with a putative role in CH) association analyses, they claim some suggestive hits, although the observed effect sizes (<0.5 or >2.0) were more extreme than those reported in other GWA studies (between ~0.8 and ~1.2). Also observed p-values, often by orders of magnitude, did not reach the commonly accepted threshold for a genome-wide significant association (p ≤ 5 × 10−8). Therefore, the findings can, at best, be taken as hypothesis-generating. Among their hits was an association with a variant in the PACAP receptor gene ADCYAP1R1, which is of interest given that PACAP induces activation of specific neurons that have been implicated in the pathophysiology of cluster headache and that higher plasmatic levels of PACAP have been detected in patients with cluster headache inside a period of attacks than outside such a period [78]. In addition, their gene-based analysis provided some evidence of association for a rare potentially damaging missense variant in MME that encodes the membrane metallo-endopeptidase neprilysin [76]. The fact, however, that these findings could not be replicated in a much larger sample set of over 500 Swedish cluster headache patients may suggest that the initial hits were false positives [79].
5.7.4 Genome-Wide Association Studies in Other Primary Headache Disorders
Various GWAS were performed for various migraine types in increasingly large cohorts of patients and controls [80]. The most recent, and largest, migraine GWAS investigated 59,674 migraine cases and 316,078 healthy controls and identified 38 genome-wide significant loci with 45 independent SNPs that confer migraine risk [81]. Of the 29 migraine-associated SNPs that could directly be genotyped or captured by a tag SNP in the cluster headache GWAS, only SNP rs9349379 in PHACTR1 achieved a nominally significant p-value in cluster headache [76]. The same SNP associated with several vascular diseases (coronary heart disease, coronary artery calcification and cervical artery dissection), and through various genetic and molecular approaches, it was convincingly shown that the SNP in fact regulated endothelin-1, a potent vasoconstrictor encoded by EDN1 located 600 kb upstream of PHACTR1 [82]. The observation that mean endothelin-1 plasma levels were increased during attacks of cluster headache suggests that vascular or immune function, two known functions of endothelin-1, may be involved in its pathology [83].
5.8 Investigating the Genetics of Cluster Headache Through RNA-Based Approaches
RNA-based approaches have also been tried to identify genes and pathways involved in cluster headache by searching for differential gene expression in tissue from patients and controls. Microarray profiling of RNA from whole blood or immortalized lymphoblastoid cell lines from a few cluster headache patients revealed differentially expressed genes: i.e. 90 in blood [84] and 1100 in cell lines [85], albeit with very limited overlap. It remains unclear to what extent the suggested pathways (noninfectious inflammation or endoplasmic reticulum protein processing) are involved in the pathology. In a recent Dutch study, RNA-sequencing (RNA-seq), a deep sequencing-based technique that is more robust and detects a wider range of transcripts than microarray technology, was used to compare whole blood gene expression profiles of a much larger sample: 39 well-characterized patients with cluster headache (19 episodic, 20 chronic) and 20 matched controls [86]. No single cluster headache-associated gene survived false discovery rate multiple testing correction. So, unlike previous reports, differences in gene expression in cluster headache, at best, seem very modest. At the level of functional gene sets, associations were observed for genes involved in several brain-related mechanisms, such as GABA receptor function and voltage-gated channels. The analysis of genes and modules of co-expressed genes suggested a role for intracellular signalling cascades, mitochondria and inflammation [86]. A role for abnormal inflammation has been proposed before [85], although the genes identified in both studies did not overlap. In the Dutch study, no evidence was obtained for the involvement of hypocretin, by analysing custom hypocretin gene sets [86]. Clearly, even larger samples should be studied with carefully taking into account (1) the matching of patients and controls and (2) the time of blood withdrawal relative to the occurrence of attacks (i.e. for cluster headache inside or outside an attack period), to identify the full range of cluster headache-associated genes and pathways. Ideally, of course, gene expression profiling should be performed in well-characterized human postmortem brain samples, but these are extremely difficult to obtain.
5.9 Conclusion
There is compelling epidemiological evidence that cluster headache has a genetic component. Evidence is less convincing for the other TACs. Despite many efforts, it has not been possible to identify causal genetic factors for cluster headache that are undisputed. Lessons can be learnt from other primary headache disorders, foremost migraine, which has shown that it is possible to identify genetic factors. Future studies in cluster headache should perhaps take into account many of the mentioned factors for success, i.e. much larger sample sizes, less heterogeneity of phenotypes (not mixing episodic and chronic patients in the analyses), more adequately matching cases and controls and—for RNA-based studies—ideally analysing diseased brain tissue or in the case of blood optimally timing the moment of drawing blood in relation to the occurrence of attacks.
