Abstract
Background/Hypothesis
Cluster headache displays uniquely rhythmic patterns in its attack manifestation. This strong chronobiological influence suggests that part of the pathophysiology of cluster headache is distinctly different from migraine and has prompted genetic investigations probing these systems.
Methods
This is a narrative overview of the cluster headache chronobiological phenotype from the point of view of genetics covering existing knowledge, highlighting the specific challenges in cluster headache and suggesting novel research approaches to overcome these.
Results
The chronobiological features of cluster headache are a hallmark of the disorder and while discrepancies between study results do exist, the main findings are highly reproducible across populations and time. Particular findings in subgroups indicate that the heritability of the disorder is linked to chronobiological systems. Meanwhile, genetic markers of circadian rhythm genes have been implicated in cluster headache, but with conflicting results. However, in two recently published genome wide association studies two of the identified four loci include genes with an involvement in circadian rhythm, MER proto-oncogene, tyrosine kinase and four and a half LIM domains 5. These findings strengthen the involvement of circadian rhythm in cluster headache pathophysiology.
Conclusion/Interpretation
Studying chronobiology and genetics in cluster headache presents challenges unique to the disorder. Researchers are overcoming these challenges by pooling various data from different cohorts and performing meta-analyses providing novel insights into a classically enigmatic disorder. Further progress can likely be made by combining deep pheno- and genotyping.
Introduction
Among the primary headaches, cluster headache (CH) is distinctively predictable in its attack occurrence. The 15–180 min attacks of severe unilateral headache with accompanying autonomic features which may occur up to eight times a day (1–5), have a clear nocturnal preference. Apart from this circadian rhythmicity, CH also occurs, or worsens, at specific times of the year in clusters (circannual/infradian rhythmicity), hence the name (2,6–8). The presence and duration of the resultant intermittent remissions dichotomizes the disorder into an episodic and rarer chronic diagnostic entity (5).
Aside from the diagnostic criteria, it can be argued that the clinical CH phenotype is well encompassed by three core features including excruciating and unilateral head pain, autonomic involvement and/or restlessness, and a precise timing that includes predictability in the attack occurrence. Central structures, which are likely diencephalic and/or specifically hypothalamic, are understood to play a pivotal role (10). In part, the strong chronobiological influence likely seated in these central structures and which suggests a pathophysiology distinctly different from other headaches, has prompted genetic investigations also probing these systems. This assumes an a priori hypothesis that the chronobiology of CH is not just an epiphenomenon, but an essential component of the pathomechanism.
In a broad sense, these studies can be divided into those investigating heritability at the population level and those seeking to identify specific genes involved in CH pathology. Several studies have investigated the prevalence of a positive family history amongst CH patients with findings varying between 2–20% (2,4,11–14). Also, first-degree relatives have five- up to 39-times increased risk of CH (13,15). Despite the high variability, these findings indicate that CH is, at least in part, heritable.
Concerning genes specifically hypothesized to be involved in CH, we can establish that in health, the daily rhythms of physiology and behavior are driven by the endogenous oscillators called circadian clocks. They are built on a negative feedback loop in which a heterodimeric transcription factor of the gene products of circadian locomotor output cycles kaput (CLOCK) and brain, muscle Arnt-like 1 (BMAL1) drives the expression of its own inhibitors, Period Circadian Regulator 1, 2 and 3 (PER1/2/3) and Cryptochrome 1/and 2 (CRY1/2) (16,17). This mechanism produces a feedback loop which runs in circadian (∼24 h) cycles which allows for the variation of expression of circadian-regulated genes during the day (Figure 1). In CH there are, so far, several reports on genetic markers in circadian rhythm genes and links to sleep regulation.
Schematic overview of the molecular clock. Circadian locomotor output cycles kaput (CLOCK) and brain, muscle Arnt-like 1 (BMAL1) drives the expression of its own inhibitors, Period Circadian Regulator 1, 2 and 3 (PER1/2/3) and Cryptochrome 1/and 2 (CRY1/2) by binding to the DNA response element enhancer box (e-box).
Several recent reviews exist on both the chronobiology (6,18) and genetics of CH (19,20). The aim of this narrative review is to approach the CH chronobiological phenotype from the point of view of genetics. We aim to cover existing knowledge to establish a model for integrating chronobiological and genetic findings in CH, highlight the specific challenges in CH and suggest novel research approaches to overcome these.
The chronobiological phenotype
Alcohol, histamine and nitroglycerin can trigger attacks with some reliability during the cluster period; however, no other specific and reproducible external triggers have been identified and therefore it can be said, in fairness, that attacks likely arise at predictable recurring times of increased susceptibility and are not triggered by an exogenous factor (21–23). This is a robust finding which is essentially reproducible across time and populations.
In its circadian attack manifestation, CH is perhaps one of the most predictable disorders as 60–80% of patients report some form of regularity in attack occurrence (1,3,24–27). A recently published meta-analysis, including almost 5000 participants, reported a circadian pattern of attacks in 70.5% of the participants from a total of 16 studies. In more detail, there were clear circadian and circannual peaks favoring 2 a.m. and the month of October, respectively (28). Concerning the evolution of CH over time, two Korean studies have found a decrease in the distinctiveness of the chronobiological features during the natural course of the disorder indicating that attack rhythmicity may not be a fixed factor (27,29). This has been confirmed in individual patients meticulously recording thousands of attacks (30) but may be highly individual as some also report a remarkable constancy of chronobiological features across impressive time periods (31).
In this manner, the characterization of CH chrono-patterns is becoming more detailed and robust. Presently, it seems there are three opportunities for developing our understanding further: 1) Identifying subgroups, 2) going beyond characterizing the time-wise occurrence of attacks and 3) validating findings in prospectively recorded data.
The most robust findings for subgroups come from comparison of males and females respectively episodic (eCH) and chronic cluster headache (cCH) patients. In a Swedish cohort more females (74%) than males (63%) with CH reported circadian rhythmicity of their attacks (32). In a Danish study the number was somewhat higher but similar for the sexes (83 and 82% respectively) likely reflecting methodological differences. Both sexes report a peak of attacks occurring at nighttime, but females may be phase-delayed, with no difference in bedtime (33). Circannual rhythmicity, possibly influenced by changes in daylight (1,26), is reported to occur in roughly 50% of both sexes (32,33).
Differences between episodic and chronic patients likely also exist (14,33,34). While superficial scrutiny of the time-wise 24-hour occurrence of attacks do not reveal major differences, using principles from signal analysis, a Danish study applied Fourier analysis to normalized attack patterns. This provided comparable measures of relative amplitude (termed chronorisk) and oscillation/frequency of chronopatterns (Figure 2). In this population-level analysis, clear differences between eCH and cCH patients were identified; episodic patients presented a strong circadian rhythm (24 hours, for example sleep-wake) whereas chronic patients had more prominent ultradian (<24 hours, for example heart rate, blood pressure, sleep phases) oscillations. The methodology of the study prevented detection of infradian rhythms (>24 hours, for example menstrual cycle, seasonal affective disorder, or indeed cluster bouts), however.
Adapted from reference (34). Representative Gaussian chronorisk analysis of episodic (eCH) and chronic (cCH) cluster headache patients showing peaks of chronorisk during the 24 hours of the day at the population level. Each color represents a particular oscillation period, for example blue is 24 hours. The more rapid, ultradian oscillations in cCH compared to eCH are also apparent. A detailed description of the methodology behind this analysis is available in the reference.
Looking at other subgroups, the analysis also revealed that in familial CH, nocturnal risk of attacks (specifically around 2 a.m.) is more than twice as high as in sporadic CH (14). Further, circadian, as opposed to ultradian, oscillations were much stronger in the familial subgroup. Together, this is circumstantial evidence that in familial CH particularly, attack occurrence is governed or at least strongly influenced by an oscillating mechanism with a 24-hour period, such as the suprachiasmatic nucleus. Extrapolating further, the genetic component may be found in genes involved in circadian rhythms, sleep regulation and/or homeostasis.
Further circumstantial evidence supporting this claim exists in the form of treatment response where episodic patients may respond better to the well-known modifier of circadian rhythms, melatonin (35) whereas chronic patients may benefit more from lithium which is thought to lengthen the circadian period (36,37). Besides lithium, as the first-choice preventive treatment for CH, verapamil, is particularly interesting from a chronobiological point of view. It is a calcium channel blocker and similar medications have shown efficacy in other neurological and psychiatric disorders with a cyclic component including epilepsy, depression and bipolar disorder (38). Animal studies have shed some light on the possible effects of these medications in CH, albeit that the results are not completely in agreement (39,40). Speculatively however, it is congruent with clinical observation and the bulk of chronobiological studies in CH that verapamil seems to alter the circadian period and induces sex-specific sleep alterations as one of these studies showed (40).
Two other adjacent areas are relevant in the context of CH chronobiology and genetics, namely sleep and chronotype. Sleep is the most noticeable of the human chronorhythms, and chronobiology and sleep regulation are closely intertwined. Cluster headache may be classified as a sleep-associated headache but direct evidence of a specific sleep-related trigger, for example a particular sleep phase or pathological or physiological sleep phenomenon, is missing (41). There is no doubt that CH patients have abnormal sleep, however, as both polysomnographic, actigraphic and questionnaire data support a homeostatic disturbance of sleep patterns which likely extends beyond the cluster (1,42–44), but may be curtailed by successful treatment of the headache (45).
Chronotype, or circadian preference, has also been the target of some interest in CH especially since there are known genetic correlates to the phenomenon. Some of the earliest observations in CH, which have since been confirmed, were regarding a possible overrepresentation of shift work amongst patients (3,46). Results are somewhat inconsistent with a possible overrepresentation of the late/evening chronotype (4,46–48). This particular research area is difficult to approach as the relationship between CH and circadian preference may be bidirectional, non-static and subject to cultural influences.
One major limitation of all the retrospective clinical studies is that digit preference and recall bias may make it more difficult to detect infradian attack rhythms (for example those with phases of 36 hours). This may also apply to the cluster periods where prospective recording revealed five to seven month phases in one patient (49). With the exception of the cluster periods, this makes it difficult to identify infradian biorhythms (>24 hours). This is a sizable blind spot in our understanding of CH chronobiology especially since we have clear indications that CH chronobiology is indeed partly infradian (i.e., the clusters). Also, almost all available studies are reporting data from patients with ongoing preventive treatment. This may present a different picture than the untreated “true” CH-state. Another limitation is studies using self-reported data, which may introduce recall bias in relation to, for example, treatment and being in or out of bout. Complicating the matter further is that there may be differences between populations. For example, it has been discovered that rhythmicity may differ between European and Asian populations (50). It has been speculated that the distance from the equator (latitude) may influence rhythmicity but if and how remains undetermined, including the possible role of genetics.
The genetics of cluster headache
As stated above, between 2–20% of patients with CH report that they have a close relative with CH in epidemiological studies (4,13,51). First-degree relatives are reported to have a five- up to 39-fold increased risk of developing CH (13,15,52) and there are a few reports of concurrent CH in monozygotic twins (51,53,54). There are three main approaches to perform genetics studies: 1) linkage studies, 2) candidate gene studies and 3) genome wide association studies. These three techniques have different advantages and disadvantages. To perform linkage studies biological tissue from family members in several generations with and without CH is needed and any part of the genome can be screened. Linkage studies have been conducted on CH families of Dutch, Italian, Danish, Swedish and British origin, but so far with no significant findings (55–57). This could possibly be explained by the small sample sizes of the family pedigrees.
The candidate gene approach has been applied on CH by conducting genetic screening on a case control material focusing on associations between genetic variation within pre-specified genes of interest e.g., circadian rhythm genes. These genes code for molecular clock proteins which are found in cells throughout the body and their expression levels control the circadian rhythm. Several genetic markers in these genes have been reported to be linked to sleep preferences and chronotype and screened in CH cohorts (see Table 1). In CLOCK the genetic marker rs1801260 is reported to regulate both sleep preference and chronotype and rs11932595 and rs12649507 are linked to sleep duration (58–60). These markers have been screened in CH cohorts, and rs12649507 has been reported to be more frequent in CH than controls and associated with higher CLOCK expression in human tissue in a Swedish CH cohort, but rs12649507 was not more frequent in Italian or Chinese CH cohorts (61–63). The link to rs12649507 in the Swedish cohort was strengthened when stratifying for reported diurnal rhythmicity of attacks. No association has been found with rs1801260 or rs11932595 in CH so far.
Genetic markers in circadian rhythm genes implicated in different traits screened for cluster headache.
SNP, Single nucleotide polymorphism; CLOCK; Circadian locomotor output cycles kaput, BMAL1, brain, muscle Arnt-like 1; PER1/2/3, Period Circadian Regulator 1, 2 and 3; CRY1/2, Cryptochrome 1/and 2.
CRY1 and 2 are key regulators of the circadian clock and genetic markers of CRY1 and 2 have previously been reported to be associated with neurological and psychiatric disorders and to have a potential effect on gene expression and/or function (64–68). Four CRY1 and 2 genetic markers have been screened in a Swedish CH cohort, rs2287161 and rs8192440 in CRY1 and rs10838524 and rs1554338 in CRY2. A strong association was found with rs8192440 and CH. Furthermore, CRY1 gene expression levels are significantly higher in tissue from CH patients compared to controls strengthening the hypothesis of circadian dysregulation in CH (69).
The PER1 and PER2 genes both work as light sensitive clock genes and their expression can be induced by pituitary adenylate cyclase-activating peptide-38 (PACAP-38) reported to have an altered expression in inter-bout CH patients (70–73). As stated above, CH can be successfully treated with lithium, which is known to alter PER3 mRNA expression (74). Genetic markers in all three PER genes, PER1, 2, and 3, have been associated with sleep disorders and diurnal preference. The minor allele of rs2735611 in PER1 has been linked to extreme morning preference and rs2304672 in the PER2 gene has been shown to be involved in circadian-related reward circuitry in the brain (75,76). Rs934945 another single nucleotide polymorphism (SNP) in PER2 has been associated with diurnal preference where the major allele was linked to morningness (77,78). The minor allele of rs10462020 in PER3 has been associated with altered morningness/eveningness score and delayed sleep phase syndrome while another PER3 SNP, rs228697 is significantly associated with evening preference (79–81). All these PER markers have been screened in a Swedish CH cohort, in addition to rs57875989 in PER3, linked to diurnal preference and screened in a Norwegian CH cohort, but no association has been found (47,82).
In addition to genes directly involved in circadian rhythm several other candidate genes of interest to CH pathology have been screened in CH cohorts such as; MT-TL1 (Mitochondrial transfer RNA Leu(UUR)), CACNA1 (Calcium channel alpha 1 subunit), HCRTR2 (Hypocretin (orexin) receptor 2), ADH4 (Alcohol dehydrogenase 4) and genes linked to CGRP signaling e.g. RAMP1 (Receptor activity-modifying protein) (83–87). Both CACNA1 and HCRTR2 had previously been screened in linkage studies, but with no significant findings, likely on account of small pedigrees (55,57). Results from replication studies of these candidate genes have in addition so far been conflicting, which could be due to potential geographical differences and small sample sizes and thereby low power (82,88,89). In a transcriptional study comparing lymphoblasts from patients with CH, respectively bipolar disorder selected for positive response to lithium and healthy controls dysregulation of two circadian genes was implicated; RNA binding motif protein 3 (RBM3) and nuclear receptor subfamily 1, group D, member 1 (NR1D1) (90). RBM3 and NR1D1 both encode for elements involved in circadian rhythm regulation, but so far, no screening for genetic markers in CH has been reported for these two genes.
Genome wide association studies (GWAS) are hypothesis free in the sense that genetic markers all over the genome are screened in large case control material. In 2016, the first CH GWAS was published on a small Italian sample of 99 CH patients and 360 controls (91). This study suggested a role for the PACAP receptor 1 gene (ADCYAP1R1) and the membrane metalloendopeptidase neprilysin gene (MME) in CH. Both of these genes have a key role in pain processing, but these findings have so far not been replicated (92). Two GWAS on individuals of European ancestry have demonstrated genetic associations for CH (93,94). These two GWAS identified loci which reached genome wide significance close to genes which are indirectly involved with the circadian rhythm. One significant locus on chromosome 2 overlies an intronic region of the gene MER proto-oncogene, tyrosine kinase (MERTK), which in turn activates the cAMP-responsive element binding protein (CREB). The CREB pathway is crucial for regulating the timing and light entrainment of the suprachiasmatic nucleus, which is known as the master clock (95). This locus and a second locus on chromosome 2 was replicated by a third GWAS in a Han Chinese CH cohort in addition to a new locus on chromosome 1 (96). A second locus on chromosome 6 was located in the four and a half LIM domains 5 (FHL5) gene and FHL5 is also an activator of CREB and thereby strengthening the involvement of circadian rhythm in CH (97). A meta GWAS was recently published by the International Consortium for Cluster Headache Genetics including ten European CH cohorts and one East Asian CH cohort (98). The five previously associated loci were confirmed and three additional loci on chromosome 7, 10 and 12 were identified. The closest protein-coding genes in the associated chromosomal regions, excluding the loci which harbors MERTK and FHL5, have no known involvement in circadian rhythm regulation (Table 2). Interestingly none of the previous associations reported from candidate gene studies were replicated in the meta GWAS.
Summary of all key cluster headache associated genes (excluding circadian rhythm specific genes), method of identification and known circadian relevance.
SNP, Single nucleotide polymorphism; GWAS, Genome wide association study; MT-TL1, Mitochondrial transfer RNA Leu(UUR); HCRTR2, Hypocretin (orexin) receptor 2; ADH4, Alcohol dehydrogenase 4; RAMP1, Receptor activity-modifying protein; ADCYAP1R1, MME Membrane Metalloendopeptidase; MERTK, MER proto-oncogene, tyrosine kinase; FHL5, four and a half LIM domains 5; DUSP10, Dual Specificity Phosphatase 10; CAPN2, Calpain 2; WNT2, Wnt Family Member 2; FTCDNL1, Formiminotransferase Cyclodeaminase N-Terminal Lik; PLCE1, Phospholipase C Epsilon 1; LRP1, LDL Receptor Related Protein.
Discussion
Despite the advances described above, CH remains a scientific and clinical conundrum. Progress has been made in understanding the neuroanatomical and physiological underpinnings of the attacks, but much remains unresolved with regards to the central influences, including chronobiological and genetic influences. While not being as prevalent as some other primary headache disorders, the dreadfulness of the fulminant cluster bids scientists to seek a better understanding of these issues. There is ample evidence that CH is uniquely chronobiological and a better understanding of this, including the genetic substrate, holds great promise (99) in furthering our understanding of the disorder and in turn possibly leading to better treatment options.
The fundamental mechanism governing the painful attacks has so far not been identified. The challenge herein may be due to the attacks arising as a result of increased susceptibility combined with a threshold-mechanism inducing a permissive state. The latter scenario may be more reconcilable with a multi-factorial pathogenesis which better encompasses the totality of observations concerning the occurrence of CH attacks including geno- and phenotype. A framework for understanding the central, peripheral and genetic contributions to CH pathophysiology can be made by adopting the tripartite model of chronobiology (100). This model describes three main components of circadian systems: 1) inputs, 2) an internal oscillator and 3) outputs. In CH, a fourth component in the form of a mediator could be interjected between the oscillator and the output (Figure 3).
Adaptation of the chronobiological tripartite model to CH providing a theoretical framework for understanding how genetics and chronobiology contributes to CH pathophysiology (107). The primary oscillator of the system is the hypothalamic SCN which produces 24-hour oscillations. It is entrained by retinal cells via the retino-hypothalamic pathway (108). Further, compensatory mechanisms ensure that periodical lack of sleep is off-set by following increased sleep. This circadian oscillator and sleep-wave homeostat, govern our fundamental circadian rhythm of sleep and wake. The mediator represents circuits not involved in generating the rhythms of the disorder and could include the brainstem circuitry of the trigeminal-autonomic reflex as well as genetic markers (109).
Apart from the specified genes involved in chronobiology, some observations thought to be peculiarities of CH may also hint at an underlying mechanism. For example, some CH patients are reported to experience recurring awakenings at the times of their usual attacks ahead of the cluster before the attack manifests (101). Speculatively, this separates the pathomechanisms of the pain from the chronobiology. Indirectly this may be supported by the anecdotal possible (but not frequent) dissociation of pain from cranial autonomic symptoms which may indicate that involved mechanisms, and hence candidate genes, should be sought also in other homeostatic systems including sleep, arousal and autonomic regulation (102). Further, an interesting observation made by Naber et al. (6) is that of the possible role of the so-called ensemble peak (concerted activity of neurons) of the suprachiasmatic nucleus. The amplitude and duration of this peak may provide a neurophysiological correlate for a possible pathomechanism behind the circannual nature of CH.
Future directions
The clinical implications of discovering a possible strong involvement of genes associated with chronobiological/homeostatic regulation may change our approach to treating CH. For example, it may provide support for timing the administration of (preventive) treatments or direct manipulation of sleep (for example sleep hygiene, sleep deprivation) and lead to the identification of new potential drug targets. It is likely that better understanding of the genetics of CH may also aid us in understanding not only the chronobiological features but also the broader concepts of sleep, autonomic and homeostatic regulation. Unfortunately, these are all topics which are subject to the same knowledge gaps as CH. In any case, a cross-disciplinary approach, drawing upon both clinical and paraclinical expertise may facilitate development of these strategies which must rest on robust clinical trials.
In spite of the reported genetic influence of CH, the low prevalence and small pedigrees and cohorts have complicated the feasibility of genetic studies over the years (103,104). But recent collaborative multicenter studies on large clinically well-defined materials by the International Consortium for Cluster Headache Genetics has led to chances of a better understanding of the complex genetic background of CH. Data from the most recent genome-wide association meta-analysis by the consortium of >4,700 CH patients and >30,000 controls from ten cohorts of European ancestry and one of East Asian ancestry identified eight risk loci of which three were novel, but no new links to circadian rhythm (98). The next step will be to further characterize genes at these loci to investigate their potential role in CH pathophysiology and chronobiology. The GWAS loci will also be investigated in relation to CH subtypes such as circadian patterns of attacks, age of onset, treatment response etc. Ultimately, these studies may lead to identification of subgroups where coupling genetic studies with chronobiology may explain some of the observed differences in treatment response. So far, no genetic association on genes involved in circadian rhythm has been found with specific coupling to eCH vs cCH, which might explain some of the above differences, but the link to rs12649507 in CLOCK and CH was strengthened when stratifying for reported circadian rhythmicity of attacks (105). Ultimately, such findings may provide a substrate for developing new treatments or improve the effect of currently available ones by making application of chronotherapy in CH more tangible (106).
Article highlights
Cluster headache is in part inherited and also uniquely chronobiological in its manifestation. Candidate gene studies on genes involved in circadian rhythm points to a link to cluster headache. Pooling data across countries holds promise in identifying genetic markers.
